From c7f515f6f5325c65f32dd64e1ad479660c751e99 Mon Sep 17 00:00:00 2001 From: tgrogers Date: Sun, 9 Jun 2019 22:20:15 -0400 Subject: A bunch of boilerbplate to get 10.1 to compile. Still does not yet run. The way CUDA calls kerenels (even on old code) has changed. --- libcuda/cuda_api.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'libcuda/cuda_api.h') diff --git a/libcuda/cuda_api.h b/libcuda/cuda_api.h index 3808e8a..7ee26dc 100644 --- a/libcuda/cuda_api.h +++ b/libcuda/cuda_api.h @@ -234,9 +234,11 @@ typedef struct CUgraphicsResource_st *CUgraphicsResource; /**< CUDA graphics int typedef unsigned long long CUtexObject; /**< An opaque value that represents a CUDA texture object */ typedef unsigned long long CUsurfObject; /**< An opaque value that represents a CUDA surface object */ +#if __CUDA_API_VERSION < 1010 typedef struct CUuuid_st { /**< CUDA definition of UUID */ char bytes[16]; } CUuuid; +#endif #if __CUDA_API_VERSION >= 4010 -- cgit v1.3 From 8a6d828b9adfe8bf690f671e9fc5902315b030b7 Mon Sep 17 00:00:00 2001 From: tgrogers Date: Sun, 9 Jun 2019 22:35:34 -0400 Subject: updating the APU --- libcuda/cuda_api.h | 5470 +++++++++++++++++++++++++++++++++++++++++----------- 1 file changed, 4371 insertions(+), 1099 deletions(-) (limited to 'libcuda/cuda_api.h') diff --git a/libcuda/cuda_api.h b/libcuda/cuda_api.h index 7ee26dc..27983b4 100644 --- a/libcuda/cuda_api.h +++ b/libcuda/cuda_api.h @@ -1,5 +1,5 @@ /* - * Copyright 1993-2014 NVIDIA Corporation. All rights reserved. + * Copyright 1993-2018 NVIDIA Corporation. All rights reserved. * * NOTICE TO LICENSEE: * @@ -63,6 +63,16 @@ typedef uint64_t cuuint64_t; /** * CUDA API versioning support */ +#if defined(__CUDA_API_VERSION_INTERNAL) || defined(__DOXYGEN_ONLY__) || defined(CUDA_ENABLE_DEPRECATED) +#define __CUDA_DEPRECATED +#elif defined(_MSC_VER) +#define __CUDA_DEPRECATED __declspec(deprecated) +#elif defined(__GNUC__) +#define __CUDA_DEPRECATED __attribute__((deprecated)) +#else +#define __CUDA_DEPRECATED +#endif + #if defined(CUDA_FORCE_API_VERSION) #if (CUDA_FORCE_API_VERSION == 3010) #define __CUDA_API_VERSION 3010 @@ -70,7 +80,7 @@ typedef uint64_t cuuint64_t; #error "Unsupported value of CUDA_FORCE_API_VERSION" #endif #else - #define __CUDA_API_VERSION 8000 + #define __CUDA_API_VERSION 10010 #endif /* CUDA_FORCE_API_VERSION */ #if defined(__CUDA_API_VERSION_INTERNAL) || defined(CUDA_API_PER_THREAD_DEFAULT_STREAM) @@ -135,10 +145,20 @@ typedef uint64_t cuuint64_t; #if defined(__CUDA_API_VERSION_INTERNAL) || __CUDA_API_VERSION >= 4010 #define cuTexRefSetAddress2D cuTexRefSetAddress2D_v3 #endif /* __CUDA_API_VERSION_INTERNAL || __CUDA_API_VERSION >= 4010 */ +#if defined(__CUDA_API_VERSION_INTERNAL) || __CUDA_API_VERSION >= 6050 + #define cuLinkCreate cuLinkCreate_v2 + #define cuLinkAddData cuLinkAddData_v2 + #define cuLinkAddFile cuLinkAddFile_v2 +#endif /* __CUDA_API_VERSION_INTERNAL || __CUDA_API_VERSION >= 6050 */ #if defined(__CUDA_API_VERSION_INTERNAL) || __CUDA_API_VERSION >= 6050 #define cuMemHostRegister cuMemHostRegister_v2 #define cuGraphicsResourceSetMapFlags cuGraphicsResourceSetMapFlags_v2 #endif /* __CUDA_API_VERSION_INTERNAL || __CUDA_API_VERSION >= 6050 */ +#if defined(__CUDA_API_VERSION_INTERNAL) || __CUDA_API_VERSION >= 10010 + #define cuStreamBeginCapture __CUDA_API_PTSZ(cuStreamBeginCapture_v2) +#elif defined(__CUDA_API_PER_THREAD_DEFAULT_STREAM) + #define cuStreamBeginCapture __CUDA_API_PTSZ(cuStreamBeginCapture) +#endif /* __CUDA_API_VERSION_INTERNAL || __CUDA_API_VERSION >= 10010 */ #if !defined(__CUDA_API_VERSION_INTERNAL) #if defined(__CUDA_API_VERSION) && __CUDA_API_VERSION >= 3020 && __CUDA_API_VERSION < 4010 @@ -164,19 +184,33 @@ typedef uint64_t cuuint64_t; #define cuStreamGetPriority __CUDA_API_PTSZ(cuStreamGetPriority) #define cuStreamGetFlags __CUDA_API_PTSZ(cuStreamGetFlags) + #define cuStreamGetCtx __CUDA_API_PTSZ(cuStreamGetCtx) #define cuStreamWaitEvent __CUDA_API_PTSZ(cuStreamWaitEvent) + #define cuStreamEndCapture __CUDA_API_PTSZ(cuStreamEndCapture) + #define cuStreamIsCapturing __CUDA_API_PTSZ(cuStreamIsCapturing) + #define cuStreamGetCaptureInfo __CUDA_API_PTSZ(cuStreamGetCaptureInfo) #define cuStreamAddCallback __CUDA_API_PTSZ(cuStreamAddCallback) #define cuStreamAttachMemAsync __CUDA_API_PTSZ(cuStreamAttachMemAsync) #define cuStreamQuery __CUDA_API_PTSZ(cuStreamQuery) #define cuStreamSynchronize __CUDA_API_PTSZ(cuStreamSynchronize) #define cuEventRecord __CUDA_API_PTSZ(cuEventRecord) #define cuLaunchKernel __CUDA_API_PTSZ(cuLaunchKernel) + #define cuLaunchHostFunc __CUDA_API_PTSZ(cuLaunchHostFunc) #define cuGraphicsMapResources __CUDA_API_PTSZ(cuGraphicsMapResources) #define cuGraphicsUnmapResources __CUDA_API_PTSZ(cuGraphicsUnmapResources) #define cuStreamWriteValue32 __CUDA_API_PTSZ(cuStreamWriteValue32) #define cuStreamWaitValue32 __CUDA_API_PTSZ(cuStreamWaitValue32) + #define cuStreamWriteValue64 __CUDA_API_PTSZ(cuStreamWriteValue64) + #define cuStreamWaitValue64 __CUDA_API_PTSZ(cuStreamWaitValue64) #define cuStreamBatchMemOp __CUDA_API_PTSZ(cuStreamBatchMemOp) + + #define cuLaunchCooperativeKernel __CUDA_API_PTSZ(cuLaunchCooperativeKernel) + + #define cuSignalExternalSemaphoresAsync __CUDA_API_PTSZ(cuSignalExternalSemaphoresAsync) + #define cuWaitExternalSemaphoresAsync __CUDA_API_PTSZ(cuWaitExternalSemaphoresAsync) + + #define cuGraphLaunch __CUDA_API_PTSZ(cuGraphLaunch) #endif /** @@ -200,7 +234,7 @@ typedef uint64_t cuuint64_t; /** * CUDA API version number */ -#define CUDA_VERSION 8000 +#define CUDA_VERSION 10010 #ifdef __cplusplus extern "C" { @@ -209,7 +243,7 @@ extern "C" { /** * CUDA device pointer * CUdeviceptr is defined as an unsigned integer type whose size matches the size of a pointer on the target platform. - */ + */ #if __CUDA_API_VERSION >= 3020 #if defined(_WIN64) || defined(__LP64__) @@ -233,18 +267,23 @@ typedef struct CUstream_st *CUstream; /**< CUDA stream */ typedef struct CUgraphicsResource_st *CUgraphicsResource; /**< CUDA graphics interop resource */ typedef unsigned long long CUtexObject; /**< An opaque value that represents a CUDA texture object */ typedef unsigned long long CUsurfObject; /**< An opaque value that represents a CUDA surface object */ - -#if __CUDA_API_VERSION < 1010 +typedef struct CUextMemory_st *CUexternalMemory; /**< CUDA external memory */ +typedef struct CUextSemaphore_st *CUexternalSemaphore; /**< CUDA external semaphore */ +typedef struct CUgraph_st *CUgraph; /**< CUDA graph */ +typedef struct CUgraphNode_st *CUgraphNode; /**< CUDA graph node */ +typedef struct CUgraphExec_st *CUgraphExec; /**< CUDA executable graph */ + +#ifndef CU_UUID_HAS_BEEN_DEFINED +#define CU_UUID_HAS_BEEN_DEFINED typedef struct CUuuid_st { /**< CUDA definition of UUID */ char bytes[16]; } CUuuid; #endif - #if __CUDA_API_VERSION >= 4010 /** - * CUDA IPC handle size + * CUDA IPC handle size */ #define CU_IPC_HANDLE_SIZE 64 @@ -291,7 +330,7 @@ typedef enum CUctx_flags_enum { CU_CTX_BLOCKING_SYNC = 0x04, /**< Set blocking synchronization as default scheduling * \deprecated This flag was deprecated as of CUDA 4.0 * and was replaced with ::CU_CTX_SCHED_BLOCKING_SYNC. */ - CU_CTX_SCHED_MASK = 0x07, + CU_CTX_SCHED_MASK = 0x07, CU_CTX_MAP_HOST = 0x08, /**< Support mapped pinned allocations */ CU_CTX_LMEM_RESIZE_TO_MAX = 0x10, /**< Keep local memory allocation after launch */ CU_CTX_FLAGS_MASK = 0x1f @@ -337,20 +376,26 @@ typedef enum CUevent_flags_enum { #if __CUDA_API_VERSION >= 8000 /** - * Flags for ::cuStreamWaitValue32 + * Flags for ::cuStreamWaitValue32 and ::cuStreamWaitValue64 */ typedef enum CUstreamWaitValue_flags_enum { - CU_STREAM_WAIT_VALUE_GEQ = 0x0, /**< Wait until (int32_t)(*addr - value) >= 0. Note this is a - cyclic comparison which ignores wraparound. (Default behavior.) */ + CU_STREAM_WAIT_VALUE_GEQ = 0x0, /**< Wait until (int32_t)(*addr - value) >= 0 (or int64_t for 64 bit + values). Note this is a cyclic comparison which ignores wraparound. + (Default behavior.) */ CU_STREAM_WAIT_VALUE_EQ = 0x1, /**< Wait until *addr == value. */ CU_STREAM_WAIT_VALUE_AND = 0x2, /**< Wait until (*addr & value) != 0. */ + CU_STREAM_WAIT_VALUE_NOR = 0x3, /**< Wait until ~(*addr | value) != 0. Support for this operation can be + queried with ::cuDeviceGetAttribute() and + ::CU_DEVICE_ATTRIBUTE_CAN_USE_STREAM_WAIT_VALUE_NOR.*/ CU_STREAM_WAIT_VALUE_FLUSH = 1<<30 /**< Follow the wait operation with a flush of outstanding remote writes. This means that, if a remote write operation is guaranteed to have reached the device before the wait can be satisfied, that write is guaranteed to be visible to downstream device work. The device is permitted to reorder remote writes internally. For example, this flag would be required if two remote writes arrive in a defined order, the wait is satisfied by the - second write, and downstream work needs to observe the first write. */ + second write, and downstream work needs to observe the first write. + Support for this operation is restricted to selected platforms and can be + queried with ::CU_DEVICE_ATTRIBUTE_CAN_USE_WAIT_VALUE_FLUSH.*/ } CUstreamWaitValue_flags; /** @@ -372,6 +417,8 @@ typedef enum CUstreamWriteValue_flags_enum { typedef enum CUstreamBatchMemOpType_enum { CU_STREAM_MEM_OP_WAIT_VALUE_32 = 1, /**< Represents a ::cuStreamWaitValue32 operation */ CU_STREAM_MEM_OP_WRITE_VALUE_32 = 2, /**< Represents a ::cuStreamWriteValue32 operation */ + CU_STREAM_MEM_OP_WAIT_VALUE_64 = 4, /**< Represents a ::cuStreamWaitValue64 operation */ + CU_STREAM_MEM_OP_WRITE_VALUE_64 = 5, /**< Represents a ::cuStreamWriteValue64 operation */ CU_STREAM_MEM_OP_FLUSH_REMOTE_WRITES = 3 /**< This has the same effect as ::CU_STREAM_WAIT_VALUE_FLUSH, but as a standalone operation. */ } CUstreamBatchMemOpType; @@ -386,7 +433,7 @@ typedef union CUstreamBatchMemOpParams_union { CUdeviceptr address; union { cuuint32_t value; - cuuint64_t pad; + cuuint64_t value64; }; unsigned int flags; CUdeviceptr alias; /**< For driver internal use. Initial value is unimportant. */ @@ -396,7 +443,7 @@ typedef union CUstreamBatchMemOpParams_union { CUdeviceptr address; union { cuuint32_t value; - cuuint64_t pad; + cuuint64_t value64; }; unsigned int flags; CUdeviceptr alias; /**< For driver internal use. Initial value is unimportant. */ @@ -498,7 +545,7 @@ typedef enum CUdevice_attribute_enum { CU_DEVICE_ATTRIBUTE_L2_CACHE_SIZE = 38, /**< Size of L2 cache in bytes */ CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_MULTIPROCESSOR = 39, /**< Maximum resident threads per multiprocessor */ CU_DEVICE_ATTRIBUTE_ASYNC_ENGINE_COUNT = 40, /**< Number of asynchronous engines */ - CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING = 41, /**< Device shares a unified address space with the host */ + CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING = 41, /**< Device shares a unified address space with the host */ CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LAYERED_WIDTH = 42, /**< Maximum 1D layered texture width */ CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LAYERED_LAYERS = 43, /**< Maximum layers in a 1D layered texture */ CU_DEVICE_ATTRIBUTE_CAN_TEX2D_GATHER = 44, /**< Deprecated, do not use. */ @@ -532,7 +579,7 @@ typedef enum CUdevice_attribute_enum { CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_PITCH = 72, /**< Maximum 2D linear texture pitch in bytes */ CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_MIPMAPPED_WIDTH = 73, /**< Maximum mipmapped 2D texture width */ CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_MIPMAPPED_HEIGHT = 74,/**< Maximum mipmapped 2D texture height */ - CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR = 75, /**< Major compute capability version number */ + CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR = 75, /**< Major compute capability version number */ CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR = 76, /**< Minor compute capability version number */ CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_MIPMAPPED_WIDTH = 77, /**< Maximum mipmapped 1D texture width */ CU_DEVICE_ATTRIBUTE_STREAM_PRIORITIES_SUPPORTED = 78, /**< Device supports stream priorities */ @@ -541,7 +588,7 @@ typedef enum CUdevice_attribute_enum { CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_MULTIPROCESSOR = 81, /**< Maximum shared memory available per multiprocessor in bytes */ CU_DEVICE_ATTRIBUTE_MAX_REGISTERS_PER_MULTIPROCESSOR = 82, /**< Maximum number of 32-bit registers available per multiprocessor */ CU_DEVICE_ATTRIBUTE_MANAGED_MEMORY = 83, /**< Device can allocate managed memory on this system */ - CU_DEVICE_ATTRIBUTE_MULTI_GPU_BOARD = 84, /**< Device is on a multi-GPU board */ + CU_DEVICE_ATTRIBUTE_MULTI_GPU_BOARD = 84, /**< Device is on a multi-GPU board */ CU_DEVICE_ATTRIBUTE_MULTI_GPU_BOARD_GROUP_ID = 85, /**< Unique id for a group of devices on the same multi-GPU board */ CU_DEVICE_ATTRIBUTE_HOST_NATIVE_ATOMIC_SUPPORTED = 86, /**< Link between the device and the host supports native atomic operations (this is a placeholder attribute, and is not supported on any current hardware)*/ CU_DEVICE_ATTRIBUTE_SINGLE_TO_DOUBLE_PRECISION_PERF_RATIO = 87, /**< Ratio of single precision performance (in floating-point operations per second) to double precision performance */ @@ -549,6 +596,16 @@ typedef enum CUdevice_attribute_enum { CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS = 89, /**< Device can coherently access managed memory concurrently with the CPU */ CU_DEVICE_ATTRIBUTE_COMPUTE_PREEMPTION_SUPPORTED = 90, /**< Device supports compute preemption. */ CU_DEVICE_ATTRIBUTE_CAN_USE_HOST_POINTER_FOR_REGISTERED_MEM = 91, /**< Device can access host registered memory at the same virtual address as the CPU */ + CU_DEVICE_ATTRIBUTE_CAN_USE_STREAM_MEM_OPS = 92, /**< ::cuStreamBatchMemOp and related APIs are supported. */ + CU_DEVICE_ATTRIBUTE_CAN_USE_64_BIT_STREAM_MEM_OPS = 93, /**< 64-bit operations are supported in ::cuStreamBatchMemOp and related APIs. */ + CU_DEVICE_ATTRIBUTE_CAN_USE_STREAM_WAIT_VALUE_NOR = 94, /**< ::CU_STREAM_WAIT_VALUE_NOR is supported. */ + CU_DEVICE_ATTRIBUTE_COOPERATIVE_LAUNCH = 95, /**< Device supports launching cooperative kernels via ::cuLaunchCooperativeKernel */ + CU_DEVICE_ATTRIBUTE_COOPERATIVE_MULTI_DEVICE_LAUNCH = 96, /**< Device can participate in cooperative kernels launched via ::cuLaunchCooperativeKernelMultiDevice */ + CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK_OPTIN = 97, /**< Maximum optin shared memory per block */ + CU_DEVICE_ATTRIBUTE_CAN_FLUSH_REMOTE_WRITES = 98, /**< Both the ::CU_STREAM_WAIT_VALUE_FLUSH flag and the ::CU_STREAM_MEM_OP_FLUSH_REMOTE_WRITES MemOp are supported on the device. See \ref CUDA_MEMOP for additional details. */ + CU_DEVICE_ATTRIBUTE_HOST_REGISTER_SUPPORTED = 99, /**< Device supports host memory registration via ::cudaHostRegister. */ + CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS_USES_HOST_PAGE_TABLES = 100, /**< Device accesses pageable memory via the host's page tables. */ + CU_DEVICE_ATTRIBUTE_DIRECT_MANAGED_MEM_ACCESS_FROM_HOST = 101, /**< The host can directly access managed memory on the device without migration. */ CU_DEVICE_ATTRIBUTE_MAX } CUdevice_attribute; @@ -579,7 +636,8 @@ typedef enum CUpointer_attribute_enum { CU_POINTER_ATTRIBUTE_P2P_TOKENS = 5, /**< A pair of tokens for use with the nv-p2p.h Linux kernel interface */ CU_POINTER_ATTRIBUTE_SYNC_MEMOPS = 6, /**< Synchronize every synchronous memory operation initiated on this region */ CU_POINTER_ATTRIBUTE_BUFFER_ID = 7, /**< A process-wide unique ID for an allocated memory region*/ - CU_POINTER_ATTRIBUTE_IS_MANAGED = 8 /**< Indicates if the pointer points to managed memory */ + CU_POINTER_ATTRIBUTE_IS_MANAGED = 8, /**< Indicates if the pointer points to managed memory */ + CU_POINTER_ATTRIBUTE_DEVICE_ORDINAL = 9 /**< A device ordinal of a device on which a pointer was allocated or registered */ } CUpointer_attribute; /** @@ -635,11 +693,28 @@ typedef enum CUfunction_attribute_enum { CU_FUNC_ATTRIBUTE_BINARY_VERSION = 6, /** - * The attribute to indicate whether the function has been compiled with + * The attribute to indicate whether the function has been compiled with * user specified option "-Xptxas --dlcm=ca" set . */ CU_FUNC_ATTRIBUTE_CACHE_MODE_CA = 7, + /** + * The maximum size in bytes of dynamically-allocated shared memory that can be used by + * this function. If the user-specified dynamic shared memory size is larger than this + * value, the launch will fail. + * See ::cuFuncSetAttribute + */ + CU_FUNC_ATTRIBUTE_MAX_DYNAMIC_SHARED_SIZE_BYTES = 8, + + /** + * On devices where the L1 cache and shared memory use the same hardware resources, + * this sets the shared memory carveout preference, in percent of the total shared memory. + * Refer to ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_MULTIPROCESSOR. + * This is only a hint, and the driver can choose a different ratio if required to execute the function. + * See ::cuFuncSetAttribute + */ + CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT = 9, + CU_FUNC_ATTRIBUTE_MAX } CUfunction_attribute; @@ -662,6 +737,15 @@ typedef enum CUsharedconfig_enum { CU_SHARED_MEM_CONFIG_EIGHT_BYTE_BANK_SIZE = 0x02 /**< set shared memory bank width to eight bytes */ } CUsharedconfig; +/** + * Shared memory carveout configurations. These may be passed to ::cuFuncSetAttribute + */ +typedef enum CUshared_carveout_enum { + CU_SHAREDMEM_CARVEOUT_DEFAULT = -1, /**< No preference for shared memory or L1 (default) */ + CU_SHAREDMEM_CARVEOUT_MAX_SHARED = 100, /**< Prefer maximum available shared memory, minimum L1 cache */ + CU_SHAREDMEM_CARVEOUT_MAX_L1 = 0 /**< Prefer maximum available L1 cache, minimum shared memory */ +} CUshared_carveout; + /** * Memory types */ @@ -840,6 +924,37 @@ typedef enum CUjit_option_enum CU_JIT_NEW_SM3X_OPT, CU_JIT_FAST_COMPILE, + /** + * Array of device symbol names that will be relocated to the corresponing + * host addresses stored in ::CU_JIT_GLOBAL_SYMBOL_ADDRESSES.\n + * Must contain ::CU_JIT_GLOBAL_SYMBOL_COUNT entries.\n + * When loding a device module, driver will relocate all encountered + * unresolved symbols to the host addresses.\n + * It is only allowed to register symbols that correspond to unresolved + * global variables.\n + * It is illegal to register the same device symbol at multiple addresses.\n + * Option type: const char **\n + * Applies to: dynamic linker only + */ + CU_JIT_GLOBAL_SYMBOL_NAMES, + + /** + * Array of host addresses that will be used to relocate corresponding + * device symbols stored in ::CU_JIT_GLOBAL_SYMBOL_NAMES.\n + * Must contain ::CU_JIT_GLOBAL_SYMBOL_COUNT entries.\n + * Option type: void **\n + * Applies to: dynamic linker only + */ + CU_JIT_GLOBAL_SYMBOL_ADDRESSES, + + /** + * Number of entries in ::CU_JIT_GLOBAL_SYMBOL_NAMES and + * ::CU_JIT_GLOBAL_SYMBOL_ADDRESSES arrays.\n + * Option type: unsigned int\n + * Applies to: dynamic linker only + */ + CU_JIT_GLOBAL_SYMBOL_COUNT, + CU_JIT_NUM_OPTIONS } CUjit_option; @@ -849,10 +964,6 @@ typedef enum CUjit_option_enum */ typedef enum CUjit_target_enum { - CU_TARGET_COMPUTE_10 = 10, /**< Compute device class 1.0 */ - CU_TARGET_COMPUTE_11 = 11, /**< Compute device class 1.1 */ - CU_TARGET_COMPUTE_12 = 12, /**< Compute device class 1.2 */ - CU_TARGET_COMPUTE_13 = 13, /**< Compute device class 1.3 */ CU_TARGET_COMPUTE_20 = 20, /**< Compute device class 2.0 */ CU_TARGET_COMPUTE_21 = 21, /**< Compute device class 2.1 */ CU_TARGET_COMPUTE_30 = 30, /**< Compute device class 3.0 */ @@ -862,9 +973,12 @@ typedef enum CUjit_target_enum CU_TARGET_COMPUTE_50 = 50, /**< Compute device class 5.0 */ CU_TARGET_COMPUTE_52 = 52, /**< Compute device class 5.2 */ CU_TARGET_COMPUTE_53 = 53, /**< Compute device class 5.3 */ - CU_TARGET_COMPUTE_60 = 60, /**< Compute device class 6.0. This must be removed for CUDA 7.0 toolkit. See bug 1518217. */ - CU_TARGET_COMPUTE_61 = 61, /**< Compute device class 6.1. This must be removed for CUDA 7.0 toolkit.*/ - CU_TARGET_COMPUTE_62 = 62 /**< Compute device class 6.2. This must be removed for CUDA 7.0 toolkit.*/ + CU_TARGET_COMPUTE_60 = 60, /**< Compute device class 6.0.*/ + CU_TARGET_COMPUTE_61 = 61, /**< Compute device class 6.1.*/ + CU_TARGET_COMPUTE_62 = 62, /**< Compute device class 6.2.*/ + CU_TARGET_COMPUTE_70 = 70, /**< Compute device class 7.0.*/ + CU_TARGET_COMPUTE_72 = 72, /**< Compute device class 7.2.*/ + CU_TARGET_COMPUTE_75 = 75 /**< Compute device class 7.5.*/ } CUjit_target; /** @@ -879,7 +993,7 @@ typedef enum CUjit_fallback_enum } CUjit_fallback; /** - * Caching modes for dlcm + * Caching modes for dlcm */ typedef enum CUjit_cacheMode_enum { @@ -971,6 +1085,7 @@ typedef enum CUlimit_enum { CU_LIMIT_MALLOC_HEAP_SIZE = 0x02, /**< GPU malloc heap size */ CU_LIMIT_DEV_RUNTIME_SYNC_DEPTH = 0x03, /**< GPU device runtime launch synchronize depth */ CU_LIMIT_DEV_RUNTIME_PENDING_LAUNCH_COUNT = 0x04, /**< GPU device runtime pending launch count */ + CU_LIMIT_MAX_L2_FETCH_GRANULARITY = 0x05, /**< A value between 0 and 128 that indicates the maximum fetch granularity of L2 (in Bytes). This is a hint */ CU_LIMIT_MAX } CUlimit; @@ -984,13 +1099,102 @@ typedef enum CUresourcetype_enum { CU_RESOURCE_TYPE_PITCH2D = 0x03 /**< Pitch 2D resource */ } CUresourcetype; +#ifdef _WIN32 +#define CUDA_CB __stdcall +#else +#define CUDA_CB +#endif + +#if __CUDA_API_VERSION >= 10000 + +/** + * CUDA host function + * \param userData Argument value passed to the function + */ +typedef void (CUDA_CB *CUhostFn)(void *userData); + +/** + * GPU kernel node parameters + */ +typedef struct CUDA_KERNEL_NODE_PARAMS_st { + CUfunction func; /**< Kernel to launch */ + unsigned int gridDimX; /**< Width of grid in blocks */ + unsigned int gridDimY; /**< Height of grid in blocks */ + unsigned int gridDimZ; /**< Depth of grid in blocks */ + unsigned int blockDimX; /**< X dimension of each thread block */ + unsigned int blockDimY; /**< Y dimension of each thread block */ + unsigned int blockDimZ; /**< Z dimension of each thread block */ + unsigned int sharedMemBytes; /**< Dynamic shared-memory size per thread block in bytes */ + void **kernelParams; /**< Array of pointers to kernel parameters */ + void **extra; /**< Extra options */ +} CUDA_KERNEL_NODE_PARAMS; + +/** + * Memset node parameters + */ +typedef struct CUDA_MEMSET_NODE_PARAMS_st { + CUdeviceptr dst; /**< Destination device pointer */ + size_t pitch; /**< Pitch of destination device pointer. Unused if height is 1 */ + unsigned int value; /**< Value to be set */ + unsigned int elementSize; /**< Size of each element in bytes. Must be 1, 2, or 4. */ + size_t width; /**< Width in bytes, of the row */ + size_t height; /**< Number of rows */ +} CUDA_MEMSET_NODE_PARAMS; + +/** + * Host node parameters + */ +typedef struct CUDA_HOST_NODE_PARAMS_st { + CUhostFn fn; /**< The function to call when the node executes */ + void* userData; /**< Argument to pass to the function */ +} CUDA_HOST_NODE_PARAMS; + +/** + * Graph node types + */ +typedef enum CUgraphNodeType_enum { + CU_GRAPH_NODE_TYPE_KERNEL = 0, /**< GPU kernel node */ + CU_GRAPH_NODE_TYPE_MEMCPY = 1, /**< Memcpy node */ + CU_GRAPH_NODE_TYPE_MEMSET = 2, /**< Memset node */ + CU_GRAPH_NODE_TYPE_HOST = 3, /**< Host (executable) node */ + CU_GRAPH_NODE_TYPE_GRAPH = 4, /**< Node which executes an embedded graph */ + CU_GRAPH_NODE_TYPE_EMPTY = 5, /**< Empty (no-op) node */ + CU_GRAPH_NODE_TYPE_COUNT +} CUgraphNodeType; + +/** + * Possible stream capture statuses returned by ::cuStreamIsCapturing + */ +typedef enum CUstreamCaptureStatus_enum { + CU_STREAM_CAPTURE_STATUS_NONE = 0, /**< Stream is not capturing */ + CU_STREAM_CAPTURE_STATUS_ACTIVE = 1, /**< Stream is actively capturing */ + CU_STREAM_CAPTURE_STATUS_INVALIDATED = 2 /**< Stream is part of a capture sequence that + has been invalidated, but not terminated */ +} CUstreamCaptureStatus; + +#endif /* __CUDA_API_VERSION >= 10000 */ + +#if __CUDA_API_VERSION >= 10010 + +/** + * Possible modes for stream capture thread interactions. For more details see + * ::cuStreamBeginCapture and ::cuThreadExchangeStreamCaptureMode + */ +typedef enum CUstreamCaptureMode_enum { + CU_STREAM_CAPTURE_MODE_GLOBAL = 0, + CU_STREAM_CAPTURE_MODE_THREAD_LOCAL = 1, + CU_STREAM_CAPTURE_MODE_RELAXED = 2 +} CUstreamCaptureMode; + +#endif /* __CUDA_API_VERSION >= 10010 */ + /** * Error codes */ typedef enum cudaError_enum { /** * The API call returned with no errors. In the case of query calls, this - * can also mean that the operation being queried is complete (see + * also means that the operation being queried is complete (see * ::cuEventQuery() and ::cuStreamQuery()). */ CUDA_SUCCESS = 0, @@ -1150,7 +1354,7 @@ typedef enum cudaError_enum { /** * This indicates that the ::CUcontext passed to the API call can - * only be bound to a single CPU thread at a time but is already + * only be bound to a single CPU thread at a time but is already * bound to a CPU thread. */ CUDA_ERROR_CONTEXT_ALREADY_IN_USE = 216, @@ -1177,6 +1381,11 @@ typedef enum cudaError_enum { */ CUDA_ERROR_NVLINK_UNCORRECTABLE = 220, + /** + * This indicates that the PTX JIT compiler library was not found. + */ + CUDA_ERROR_JIT_COMPILER_NOT_FOUND = 221, + /** * This indicates that the device kernel source is invalid. */ @@ -1208,6 +1417,12 @@ typedef enum cudaError_enum { */ CUDA_ERROR_INVALID_HANDLE = 400, + /** + * This indicates that a resource required by the API call is not in a + * valid state to perform the requested operation. + */ + CUDA_ERROR_ILLEGAL_STATE = 401, + /** * This indicates that a named symbol was not found. Examples of symbols * are global/constant variable names, texture names, and surface names. @@ -1257,7 +1472,7 @@ typedef enum cudaError_enum { * mode. */ CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING = 703, - + /** * This error indicates that a call to ::cuCtxEnablePeerAccess() is * trying to re-enable peer access to a context which has already @@ -1266,9 +1481,9 @@ typedef enum cudaError_enum { CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED = 704, /** - * This error indicates that ::cuCtxDisablePeerAccess() is - * trying to disable peer access which has not been enabled yet - * via ::cuCtxEnablePeerAccess(). + * This error indicates that ::cuCtxDisablePeerAccess() is + * trying to disable peer access which has not been enabled yet + * via ::cuCtxEnablePeerAccess(). */ CUDA_ERROR_PEER_ACCESS_NOT_ENABLED = 705, @@ -1287,15 +1502,15 @@ typedef enum cudaError_enum { /** * A device-side assert triggered during kernel execution. The context - * cannot be used anymore, and must be destroyed. All existing device - * memory allocations from this context are invalid and must be + * cannot be used anymore, and must be destroyed. All existing device + * memory allocations from this context are invalid and must be * reconstructed if the program is to continue using CUDA. */ CUDA_ERROR_ASSERT = 710, /** * This error indicates that the hardware resources required to enable - * peer access have been exhausted for one or more of the devices + * peer access have been exhausted for one or more of the devices * passed to ::cuCtxEnablePeerAccess(). */ CUDA_ERROR_TOO_MANY_PEERS = 711, @@ -1360,13 +1575,22 @@ typedef enum cudaError_enum { /** * An exception occurred on the device while executing a kernel. Common * causes include dereferencing an invalid device pointer and accessing - * out of bounds shared memory. + * out of bounds shared memory. Less common cases can be system specific - more + * information about these cases can be found in the system specific user guide. * This leaves the process in an inconsistent state and any further CUDA work * will return the same error. To continue using CUDA, the process must be terminated * and relaunched. */ CUDA_ERROR_LAUNCH_FAILED = 719, + /** + * This error indicates that the number of blocks launched per grid for a kernel that was + * launched via either ::cuLaunchCooperativeKernel or ::cuLaunchCooperativeKernelMultiDevice + * exceeds the maximum number of blocks as allowed by ::cuOccupancyMaxActiveBlocksPerMultiprocessor + * or ::cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags times the number of multiprocessors + * as specified by the device attribute ::CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT. + */ + CUDA_ERROR_COOPERATIVE_LAUNCH_TOO_LARGE = 720, /** * This error indicates that the attempted operation is not permitted. @@ -1379,6 +1603,86 @@ typedef enum cudaError_enum { */ CUDA_ERROR_NOT_SUPPORTED = 801, + /** + * This error indicates that the system is not yet ready to start any CUDA + * work. To continue using CUDA, verify the system configuration is in a + * valid state and all required driver daemons are actively running. + * More information about this error can be found in the system specific + * user guide. + */ + CUDA_ERROR_SYSTEM_NOT_READY = 802, + + /** + * This error indicates that there is a mismatch between the versions of + * the display driver and the CUDA driver. Refer to the compatibility documentation + * for supported versions. + */ + CUDA_ERROR_SYSTEM_DRIVER_MISMATCH = 803, + + /** + * This error indicates that the system was upgraded to run with forward compatibility + * but the visible hardware detected by CUDA does not support this configuration. + * Refer to the compatibility documentation for the supported hardware matrix or ensure + * that only supported hardware is visible during initialization via the CUDA_VISIBLE_DEVICES + * environment variable. + */ + CUDA_ERROR_COMPAT_NOT_SUPPORTED_ON_DEVICE = 804, + + /** + * This error indicates that the operation is not permitted when + * the stream is capturing. + */ + CUDA_ERROR_STREAM_CAPTURE_UNSUPPORTED = 900, + + /** + * This error indicates that the current capture sequence on the stream + * has been invalidated due to a previous error. + */ + CUDA_ERROR_STREAM_CAPTURE_INVALIDATED = 901, + + /** + * This error indicates that the operation would have resulted in a merge + * of two independent capture sequences. + */ + CUDA_ERROR_STREAM_CAPTURE_MERGE = 902, + + /** + * This error indicates that the capture was not initiated in this stream. + */ + CUDA_ERROR_STREAM_CAPTURE_UNMATCHED = 903, + + /** + * This error indicates that the capture sequence contains a fork that was + * not joined to the primary stream. + */ + CUDA_ERROR_STREAM_CAPTURE_UNJOINED = 904, + + /** + * This error indicates that a dependency would have been created which + * crosses the capture sequence boundary. Only implicit in-stream ordering + * dependencies are allowed to cross the boundary. + */ + CUDA_ERROR_STREAM_CAPTURE_ISOLATION = 905, + + /** + * This error indicates a disallowed implicit dependency on a current capture + * sequence from cudaStreamLegacy. + */ + CUDA_ERROR_STREAM_CAPTURE_IMPLICIT = 906, + + /** + * This error indicates that the operation is not permitted on an event which + * was last recorded in a capturing stream. + */ + CUDA_ERROR_CAPTURED_EVENT = 907, + + /** + * A stream capture sequence not initiated with the ::CU_STREAM_CAPTURE_MODE_RELAXED + * argument to ::cuStreamBeginCapture was passed to ::cuStreamEndCapture in a + * different thread. + */ + CUDA_ERROR_STREAM_CAPTURE_WRONG_THREAD = 908, + /** * This indicates that an unknown internal error has occurred. */ @@ -1389,17 +1693,13 @@ typedef enum cudaError_enum { * P2P Attributes */ typedef enum CUdevice_P2PAttribute_enum { - CU_DEVICE_P2P_ATTRIBUTE_PERFORMANCE_RANK = 0x01, /**< A relative value indicating the performance of the link between two devices */ - CU_DEVICE_P2P_ATTRIBUTE_ACCESS_SUPPORTED = 0x02, /**< P2P Access is enable */ - CU_DEVICE_P2P_ATTRIBUTE_NATIVE_ATOMIC_SUPPORTED = 0x03 /**< Atomic operation over the link supported */ + CU_DEVICE_P2P_ATTRIBUTE_PERFORMANCE_RANK = 0x01, /**< A relative value indicating the performance of the link between two devices */ + CU_DEVICE_P2P_ATTRIBUTE_ACCESS_SUPPORTED = 0x02, /**< P2P Access is enable */ + CU_DEVICE_P2P_ATTRIBUTE_NATIVE_ATOMIC_SUPPORTED = 0x03, /**< Atomic operation over the link supported */ + CU_DEVICE_P2P_ATTRIBUTE_ACCESS_ACCESS_SUPPORTED = 0x04, /**< \deprecated use CU_DEVICE_P2P_ATTRIBUTE_CUDA_ARRAY_ACCESS_SUPPORTED instead */ + CU_DEVICE_P2P_ATTRIBUTE_CUDA_ARRAY_ACCESS_SUPPORTED = 0x04 /**< Accessing CUDA arrays over the link supported */ } CUdevice_P2PAttribute; -#ifdef _WIN32 -#define CUDA_CB __stdcall -#else -#define CUDA_CB -#endif - /** * CUDA stream callback * \param hStream The stream the callback was added to, as passed to ::cuStreamAddCallback. May be NULL. @@ -1635,7 +1935,7 @@ typedef struct CUDA_TEXTURE_DESC_st { CUfilter_mode mipmapFilterMode; /**< Mipmap filter mode */ float mipmapLevelBias; /**< Mipmap level bias */ float minMipmapLevelClamp; /**< Mipmap minimum level clamp */ - float maxMipmapLevelClamp; /**< Mipmap maximum level clamp */ + float maxMipmapLevelClamp; /**< Mipmap maximum level clamp */ float borderColor[4]; /**< Border Color */ int reserved[12]; } CUDA_TEXTURE_DESC; @@ -1708,9 +2008,281 @@ typedef struct CUDA_POINTER_ATTRIBUTE_P2P_TOKENS_st { #endif /* __CUDA_API_VERSION >= 5000 */ +#if __CUDA_API_VERSION >= 9000 + +/** + * Kernel launch parameters + */ +typedef struct CUDA_LAUNCH_PARAMS_st { + CUfunction function; /**< Kernel to launch */ + unsigned int gridDimX; /**< Width of grid in blocks */ + unsigned int gridDimY; /**< Height of grid in blocks */ + unsigned int gridDimZ; /**< Depth of grid in blocks */ + unsigned int blockDimX; /**< X dimension of each thread block */ + unsigned int blockDimY; /**< Y dimension of each thread block */ + unsigned int blockDimZ; /**< Z dimension of each thread block */ + unsigned int sharedMemBytes; /**< Dynamic shared-memory size per thread block in bytes */ + CUstream hStream; /**< Stream identifier */ + void **kernelParams; /**< Array of pointers to kernel parameters */ +} CUDA_LAUNCH_PARAMS; + +#endif /* __CUDA_API_VERSION >= 9000 */ + +#if __CUDA_API_VERSION >= 10000 + +/** + * External memory handle types + */ +typedef enum CUexternalMemoryHandleType_enum { + /** + * Handle is an opaque file descriptor + */ + CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD = 1, + /** + * Handle is an opaque shared NT handle + */ + CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32 = 2, + /** + * Handle is an opaque, globally shared handle + */ + CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT = 3, + /** + * Handle is a D3D12 heap object + */ + CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_HEAP = 4, + /** + * Handle is a D3D12 committed resource + */ + CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_RESOURCE = 5 +} CUexternalMemoryHandleType; + +/** + * Indicates that the external memory object is a dedicated resource + */ +#define CUDA_EXTERNAL_MEMORY_DEDICATED 0x1 + +/** + * External memory handle descriptor + */ +typedef struct CUDA_EXTERNAL_MEMORY_HANDLE_DESC_st { + /** + * Type of the handle + */ + CUexternalMemoryHandleType type; + union { + /** + * File descriptor referencing the memory object. Valid + * when type is + * ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD + */ + int fd; + /** + * Win32 handle referencing the semaphore object. Valid when + * type is one of the following: + * - ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32 + * - ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT + * - ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_HEAP + * - ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_RESOURCE + * Exactly one of 'handle' and 'name' must be non-NULL. If + * type is + * ::CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32_KMT + * then 'name' must be NULL. + */ + struct { + /** + * Valid NT handle. Must be NULL if 'name' is non-NULL + */ + void *handle; + /** + * Name of a valid memory object. + * Must be NULL if 'handle' is non-NULL. + */ + const void *name; + } win32; + } handle; + /** + * Size of the memory allocation + */ + unsigned long long size; + /** + * Flags must either be zero or ::CUDA_EXTERNAL_MEMORY_DEDICATED + */ + unsigned int flags; + unsigned int reserved[16]; +} CUDA_EXTERNAL_MEMORY_HANDLE_DESC; + +/** + * External memory buffer descriptor + */ +typedef struct CUDA_EXTERNAL_MEMORY_BUFFER_DESC_st { + /** + * Offset into the memory object where the buffer's base is + */ + unsigned long long offset; + /** + * Size of the buffer + */ + unsigned long long size; + /** + * Flags reserved for future use. Must be zero. + */ + unsigned int flags; + unsigned int reserved[16]; +} CUDA_EXTERNAL_MEMORY_BUFFER_DESC; + +/** + * External memory mipmap descriptor + */ +typedef struct CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC_st { + /** + * Offset into the memory object where the base level of the + * mipmap chain is. + */ + unsigned long long offset; + /** + * Format, dimension and type of base level of the mipmap chain + */ + CUDA_ARRAY3D_DESCRIPTOR arrayDesc; + /** + * Total number of levels in the mipmap chain + */ + unsigned int numLevels; + unsigned int reserved[16]; +} CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC; + +/** + * External semaphore handle types + */ +typedef enum CUexternalSemaphoreHandleType_enum { + /** + * Handle is an opaque file descriptor + */ + CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD = 1, + /** + * Handle is an opaque shared NT handle + */ + CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32 = 2, + /** + * Handle is an opaque, globally shared handle + */ + CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32_KMT = 3, + /** + * Handle is a shared NT handle referencing a D3D12 fence object + */ + CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_D3D12_FENCE = 4 +} CUexternalSemaphoreHandleType; + +/** + * External semaphore handle descriptor + */ +typedef struct CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC_st { + /** + * Type of the handle + */ + CUexternalSemaphoreHandleType type; + union { + /** + * File descriptor referencing the semaphore object. Valid + * when type is + * ::CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD + */ + int fd; + /** + * Win32 handle referencing the semaphore object. Valid when + * type is one of the following: + * - ::CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32 + * - ::CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32_KMT + * - ::CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_D3D12_FENCE + * Exactly one of 'handle' and 'name' must be non-NULL. If + * type is + * ::CU_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_WIN32_KMT + * then 'name' must be NULL. + */ + struct { + /** + * Valid NT handle. Must be NULL if 'name' is non-NULL + */ + void *handle; + /** + * Name of a valid synchronization primitive. + * Must be NULL if 'handle' is non-NULL. + */ + const void *name; + } win32; + } handle; + /** + * Flags reserved for the future. Must be zero. + */ + unsigned int flags; + unsigned int reserved[16]; +} CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC; + +/** + * External semaphore signal parameters + */ +typedef struct CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS_st { + struct { + /** + * Parameters for fence objects + */ + struct { + /** + * Value of fence to be signaled + */ + unsigned long long value; + } fence; + unsigned int reserved[16]; + } params; + /** + * Flags reserved for the future. Must be zero. + */ + unsigned int flags; + unsigned int reserved[16]; +} CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS; + +/** + * External semaphore wait parameters + */ +typedef struct CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS_st { + struct { + /** + * Parameters for fence objects + */ + struct { + /** + * Value of fence to be waited on + */ + unsigned long long value; + } fence; + unsigned int reserved[16]; + } params; + /** + * Flags reserved for the future. Must be zero. + */ + unsigned int flags; + unsigned int reserved[16]; +} CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS; + + +#endif /* __CUDA_API_VERSION >= 10000 */ + +/** + * If set, each kernel launched as part of ::cuLaunchCooperativeKernelMultiDevice only + * waits for prior work in the stream corresponding to that GPU to complete before the + * kernel begins execution. + */ +#define CUDA_COOPERATIVE_LAUNCH_MULTI_DEVICE_NO_PRE_LAUNCH_SYNC 0x01 + +/** + * If set, any subsequent work pushed in a stream that participated in a call to + * ::cuLaunchCooperativeKernelMultiDevice will only wait for the kernel launched on + * the GPU corresponding to that stream to complete before it begins execution. + */ +#define CUDA_COOPERATIVE_LAUNCH_MULTI_DEVICE_NO_POST_LAUNCH_SYNC 0x02 + /** * If set, the CUDA array is a collection of layers, where each layer is either a 1D - * or a 2D array and the Depth member of CUDA_ARRAY3D_DESCRIPTOR specifies the number + * or a 2D array and the Depth member of CUDA_ARRAY3D_DESCRIPTOR specifies the number * of layers, not the depth of a 3D array. */ #define CUDA_ARRAY3D_LAYERED 0x01 @@ -1743,9 +2315,15 @@ typedef struct CUDA_POINTER_ATTRIBUTE_P2P_TOKENS_st { /** * This flag if set indicates that the CUDA * array is a DEPTH_TEXTURE. -*/ + */ #define CUDA_ARRAY3D_DEPTH_TEXTURE 0x10 +/** + * This flag indicates that the CUDA array may be bound as a color target + * in an external graphics API + */ +#define CUDA_ARRAY3D_COLOR_ATTACHMENT 0x20 + /** * Override the texref format with a format inferred from the array. * Flag for ::cuTexRefSetArray() @@ -1849,7 +2427,9 @@ typedef struct CUDA_POINTER_ATTRIBUTE_P2P_TOKENS_st { * ::CUDA_SUCCESS, * ::CUDA_ERROR_INVALID_VALUE * - * \sa ::CUresult + * \sa + * ::CUresult, + * ::cudaGetErrorString */ CUresult CUDAAPI cuGetErrorString(CUresult error, const char **pStr); @@ -1868,7 +2448,9 @@ CUresult CUDAAPI cuGetErrorString(CUresult error, const char **pStr); * ::CUDA_SUCCESS, * ::CUDA_ERROR_INVALID_VALUE * - * \sa ::CUresult + * \sa + * ::CUresult, + * ::cudaGetErrorName */ CUresult CUDAAPI cuGetErrorName(CUresult error, const char **pStr); @@ -1899,7 +2481,9 @@ CUresult CUDAAPI cuGetErrorName(CUresult error, const char **pStr); * \return * ::CUDA_SUCCESS, * ::CUDA_ERROR_INVALID_VALUE, - * ::CUDA_ERROR_INVALID_DEVICE + * ::CUDA_ERROR_INVALID_DEVICE, + * ::CUDA_ERROR_SYSTEM_DRIVER_MISMATCH, + * ::CUDA_ERROR_COMPAT_NOT_SUPPORTED_ON_DEVICE * \notefnerr */ CUresult CUDAAPI cuInit(unsigned int Flags); @@ -1919,11 +2503,15 @@ CUresult CUDAAPI cuInit(unsigned int Flags); */ /** - * \brief Returns the CUDA driver version + * \brief Returns the latest CUDA version supported by driver + * + * Returns in \p *driverVersion the version of CUDA supported by + * the driver. The version is returned as + * (1000 × major + 10 × minor). For example, CUDA 9.2 + * would be represented by 9020. * - * Returns in \p *driverVersion the version number of the installed CUDA - * driver. This function automatically returns ::CUDA_ERROR_INVALID_VALUE if - * the \p driverVersion argument is NULL. + * This function automatically returns ::CUDA_ERROR_INVALID_VALUE if + * \p driverVersion is NULL. * * \param driverVersion - Returns the CUDA driver version * @@ -1931,6 +2519,10 @@ CUresult CUDAAPI cuInit(unsigned int Flags); * ::CUDA_SUCCESS, * ::CUDA_ERROR_INVALID_VALUE * \notefnerr + * + * \sa + * ::cudaDriverGetVersion, + * ::cudaRuntimeGetVersion */ CUresult CUDAAPI cuDriverGetVersion(int *driverVersion); @@ -1970,6 +2562,8 @@ CUresult CUDAAPI cuDriverGetVersion(int *driverVersion); * ::cuDeviceGetAttribute, * ::cuDeviceGetCount, * ::cuDeviceGetName, + * ::cuDeviceGetUuid, + * ::cuDeviceGetLuid, * ::cuDeviceTotalMem */ CUresult CUDAAPI cuDeviceGet(CUdevice *device, int ordinal); @@ -1978,7 +2572,7 @@ CUresult CUDAAPI cuDeviceGet(CUdevice *device, int ordinal); * \brief Returns the number of compute-capable devices * * Returns in \p *count the number of devices with compute capability greater - * than or equal to 1.0 that are available for execution. If there is no such + * than or equal to 2.0 that are available for execution. If there is no such * device, ::cuDeviceGetCount() returns 0. * * \param count - Returned number of compute-capable devices @@ -1994,8 +2588,11 @@ CUresult CUDAAPI cuDeviceGet(CUdevice *device, int ordinal); * \sa * ::cuDeviceGetAttribute, * ::cuDeviceGetName, + * ::cuDeviceGetUuid, + * ::cuDeviceGetLuid, * ::cuDeviceGet, - * ::cuDeviceTotalMem + * ::cuDeviceTotalMem, + * ::cudaGetDeviceCount */ CUresult CUDAAPI cuDeviceGetCount(int *count); @@ -2021,27 +2618,29 @@ CUresult CUDAAPI cuDeviceGetCount(int *count); * * \sa * ::cuDeviceGetAttribute, + * ::cuDeviceGetUuid, + * ::cuDeviceGetLuid, * ::cuDeviceGetCount, * ::cuDeviceGet, - * ::cuDeviceTotalMem + * ::cuDeviceTotalMem, + * ::cudaGetDeviceProperties */ CUresult CUDAAPI cuDeviceGetName(char *name, int len, CUdevice dev); -#if __CUDA_API_VERSION >= 3020 +#if __CUDA_API_VERSION >= 9020 /** - * \brief Returns the total amount of memory on the device + * \brief Return an UUID for the device * - * Returns in \p *bytes the total amount of memory available on the device - * \p dev in bytes. + * Returns 16-octets identifing the device \p dev in the structure + * pointed by the \p uuid. * - * \param bytes - Returned memory available on device in bytes - * \param dev - Device handle + * \param uuid - Returned UUID + * \param dev - Device to get identifier string for * * \return * ::CUDA_SUCCESS, * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, - * ::CUDA_ERROR_INVALID_CONTEXT, * ::CUDA_ERROR_INVALID_VALUE, * ::CUDA_ERROR_INVALID_DEVICE * \notefnerr @@ -2050,41 +2649,104 @@ CUresult CUDAAPI cuDeviceGetName(char *name, int len, CUdevice dev); * ::cuDeviceGetAttribute, * ::cuDeviceGetCount, * ::cuDeviceGetName, + * ::cuDeviceGetLuid, * ::cuDeviceGet, + * ::cuDeviceTotalMem, + * ::cudaGetDeviceProperties */ -CUresult CUDAAPI cuDeviceTotalMem(size_t *bytes, CUdevice dev); -#endif /* __CUDA_API_VERSION >= 3020 */ +CUresult CUDAAPI cuDeviceGetUuid(CUuuid *uuid, CUdevice dev); +#endif +#if defined(_WIN32) && __CUDA_API_VERSION >= 10000 /** - * \brief Returns information about the device + * \brief Return an LUID and device node mask for the device * - * Returns in \p *pi the integer value of the attribute \p attrib on device - * \p dev. The supported attributes are: - * - ::CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK: Maximum number of threads per - * block; - * - ::CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X: Maximum x-dimension of a block; - * - ::CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y: Maximum y-dimension of a block; - * - ::CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z: Maximum z-dimension of a block; - * - ::CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X: Maximum x-dimension of a grid; - * - ::CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Y: Maximum y-dimension of a grid; - * - ::CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Z: Maximum z-dimension of a grid; - * - ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK: Maximum amount of - * shared memory available to a thread block in bytes; - * - ::CU_DEVICE_ATTRIBUTE_TOTAL_CONSTANT_MEMORY: Memory available on device for - * __constant__ variables in a CUDA C kernel in bytes; - * - ::CU_DEVICE_ATTRIBUTE_WARP_SIZE: Warp size in threads; - * - ::CU_DEVICE_ATTRIBUTE_MAX_PITCH: Maximum pitch in bytes allowed by the - * memory copy functions that involve memory regions allocated through - * ::cuMemAllocPitch(); - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_WIDTH: Maximum 1D - * texture width; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LINEAR_WIDTH: Maximum width - * for a 1D texture bound to linear memory; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_MIPMAPPED_WIDTH: Maximum - * mipmapped 1D texture width; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_WIDTH: Maximum 2D + * Return identifying information (\p luid and \p deviceNodeMask) to allow + * matching device with graphics APIs. + * + * \param luid - Returned LUID + * \param deviceNodeMask - Returned device node mask + * \param dev - Device to get identifier string for + * + * \return + * ::CUDA_SUCCESS, + * ::CUDA_ERROR_DEINITIALIZED, + * ::CUDA_ERROR_NOT_INITIALIZED, + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_DEVICE + * \notefnerr + * + * \sa + * ::cuDeviceGetAttribute, + * ::cuDeviceGetCount, + * ::cuDeviceGetName, + * ::cuDeviceGet, + * ::cuDeviceTotalMem, + * ::cudaGetDeviceProperties + */ +CUresult CUDAAPI cuDeviceGetLuid(char *luid, unsigned int *deviceNodeMask, CUdevice dev); +#endif + +#if __CUDA_API_VERSION >= 3020 +/** + * \brief Returns the total amount of memory on the device + * + * Returns in \p *bytes the total amount of memory available on the device + * \p dev in bytes. + * + * \param bytes - Returned memory available on device in bytes + * \param dev - Device handle + * + * \return + * ::CUDA_SUCCESS, + * ::CUDA_ERROR_DEINITIALIZED, + * ::CUDA_ERROR_NOT_INITIALIZED, + * ::CUDA_ERROR_INVALID_CONTEXT, + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_DEVICE + * \notefnerr + * + * \sa + * ::cuDeviceGetAttribute, + * ::cuDeviceGetCount, + * ::cuDeviceGetName, + * ::cuDeviceGetUuid, + * ::cuDeviceGet, + * ::cudaMemGetInfo + */ +CUresult CUDAAPI cuDeviceTotalMem(size_t *bytes, CUdevice dev); +#endif /* __CUDA_API_VERSION >= 3020 */ + +/** + * \brief Returns information about the device + * + * Returns in \p *pi the integer value of the attribute \p attrib on device + * \p dev. The supported attributes are: + * - ::CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK: Maximum number of threads per + * block; + * - ::CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_X: Maximum x-dimension of a block; + * - ::CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Y: Maximum y-dimension of a block; + * - ::CU_DEVICE_ATTRIBUTE_MAX_BLOCK_DIM_Z: Maximum z-dimension of a block; + * - ::CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X: Maximum x-dimension of a grid; + * - ::CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Y: Maximum y-dimension of a grid; + * - ::CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_Z: Maximum z-dimension of a grid; + * - ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK: Maximum amount of + * shared memory available to a thread block in bytes; + * - ::CU_DEVICE_ATTRIBUTE_TOTAL_CONSTANT_MEMORY: Memory available on device for + * __constant__ variables in a CUDA C kernel in bytes; + * - ::CU_DEVICE_ATTRIBUTE_WARP_SIZE: Warp size in threads; + * - ::CU_DEVICE_ATTRIBUTE_MAX_PITCH: Maximum pitch in bytes allowed by the + * memory copy functions that involve memory regions allocated through + * ::cuMemAllocPitch(); + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_WIDTH: Maximum 1D + * texture width; + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LINEAR_WIDTH: Maximum width + * for a 1D texture bound to linear memory; + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_MIPMAPPED_WIDTH: Maximum + * mipmapped 1D texture width; + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_WIDTH: Maximum 2D * texture width; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_HEIGHT: Maximum 2D + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_HEIGHT: Maximum 2D * texture height; * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_WIDTH: Maximum width * for a 2D texture bound to linear memory; @@ -2092,40 +2754,40 @@ CUresult CUDAAPI cuDeviceTotalMem(size_t *bytes, CUdevice dev); * for a 2D texture bound to linear memory; * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_PITCH: Maximum pitch * in bytes for a 2D texture bound to linear memory; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_MIPMAPPED_WIDTH: Maximum + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_MIPMAPPED_WIDTH: Maximum * mipmapped 2D texture width; * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_MIPMAPPED_HEIGHT: Maximum * mipmapped 2D texture height; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_WIDTH: Maximum 3D + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_WIDTH: Maximum 3D * texture width; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_HEIGHT: Maximum 3D + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_HEIGHT: Maximum 3D * texture height; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_DEPTH: Maximum 3D + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_DEPTH: Maximum 3D * texture depth; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_WIDTH_ALTERNATE: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_WIDTH_ALTERNATE: * Alternate maximum 3D texture width, 0 if no alternate * maximum 3D texture size is supported; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_HEIGHT_ALTERNATE: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_HEIGHT_ALTERNATE: * Alternate maximum 3D texture height, 0 if no alternate * maximum 3D texture size is supported; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_DEPTH_ALTERNATE: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE3D_DEPTH_ALTERNATE: * Alternate maximum 3D texture depth, 0 if no alternate * maximum 3D texture size is supported; * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURECUBEMAP_WIDTH: * Maximum cubemap texture width or height; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LAYERED_WIDTH: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LAYERED_WIDTH: * Maximum 1D layered texture width; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LAYERED_LAYERS: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LAYERED_LAYERS: * Maximum layers in a 1D layered texture; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_WIDTH: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_WIDTH: * Maximum 2D layered texture width; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_HEIGHT: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_HEIGHT: * Maximum 2D layered texture height; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_LAYERS: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LAYERED_LAYERS: * Maximum layers in a 2D layered texture; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURECUBEMAP_LAYERED_WIDTH: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURECUBEMAP_LAYERED_WIDTH: * Maximum cubemap layered texture width or height; - * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURECUBEMAP_LAYERED_LAYERS: + * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURECUBEMAP_LAYERED_LAYERS: * Maximum layers in a cubemap layered texture; * - ::CU_DEVICE_ATTRIBUTE_MAXIMUM_SURFACE1D_WIDTH: * Maximum 1D surface width; @@ -2191,19 +2853,20 @@ CUresult CUDAAPI cuDeviceTotalMem(size_t *bytes, CUdevice dev); * - ::CU_DEVICE_ATTRIBUTE_PCI_BUS_ID: PCI bus identifier of the device; * - ::CU_DEVICE_ATTRIBUTE_PCI_DEVICE_ID: PCI device (also known as slot) identifier * of the device; + * - ::CU_DEVICE_ATTRIBUTE_PCI_DOMAIN_ID: PCI domain identifier of the device * - ::CU_DEVICE_ATTRIBUTE_TCC_DRIVER: 1 if the device is using a TCC driver. TCC * is only available on Tesla hardware running Windows Vista or later; * - ::CU_DEVICE_ATTRIBUTE_MEMORY_CLOCK_RATE: Peak memory clock frequency in kilohertz; * - ::CU_DEVICE_ATTRIBUTE_GLOBAL_MEMORY_BUS_WIDTH: Global memory bus width in bits; * - ::CU_DEVICE_ATTRIBUTE_L2_CACHE_SIZE: Size of L2 cache in bytes. 0 if the device doesn't have L2 cache; * - ::CU_DEVICE_ATTRIBUTE_MAX_THREADS_PER_MULTIPROCESSOR: Maximum resident threads per multiprocessor; - * - ::CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING: 1 if the device shares a unified address space with + * - ::CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING: 1 if the device shares a unified address space with * the host, or 0 if not; * - ::CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR: Major compute capability version number; * - ::CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR: Minor compute capability version number; - * - ::CU_DEVICE_ATTRIBUTE_GLOBAL_L1_CACHE_SUPPORTED: 1 if device supports caching globals + * - ::CU_DEVICE_ATTRIBUTE_GLOBAL_L1_CACHE_SUPPORTED: 1 if device supports caching globals * in L1 cache, 0 if caching globals in L1 cache is not supported by the device; - * - ::CU_DEVICE_ATTRIBUTE_LOCAL_L1_CACHE_SUPPORTED: 1 if device supports caching locals + * - ::CU_DEVICE_ATTRIBUTE_LOCAL_L1_CACHE_SUPPORTED: 1 if device supports caching locals * in L1 cache, 0 if caching locals in L1 cache is not supported by the device; * - ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_MULTIPROCESSOR: Maximum amount of * shared memory available to a multiprocessor in bytes; this amount is shared @@ -2227,6 +2890,12 @@ CUresult CUDAAPI cuDeviceTotalMem(size_t *bytes, CUdevice dev); * - ::CU_DEVICE_ATTRIBUTE_COMPUTE_PREEMPTION_SUPPORTED: Device supports Compute Preemption. * - ::CU_DEVICE_ATTRIBUTE_CAN_USE_HOST_POINTER_FOR_REGISTERED_MEM: Device can access host registered * memory at the same virtual address as the CPU. + * - ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK_OPTIN: The maximum per block shared memory size + * suported on this device. This is the maximum value that can be opted into when using the cuFuncSetAttribute() call. + * For more details see ::CU_FUNC_ATTRIBUTE_MAX_DYNAMIC_SHARED_SIZE_BYTES + * - ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS_USES_HOST_PAGE_TABLES: Device accesses pageable memory via the host's + * page tables. + * - ::CU_DEVICE_ATTRIBUTE_DIRECT_MANAGED_MEM_ACCESS_FROM_HOST: The host can directly access managed memory on the device without migration. * * \param pi - Returned device attribute value * \param attrib - Device attribute to query @@ -2244,8 +2913,11 @@ CUresult CUDAAPI cuDeviceTotalMem(size_t *bytes, CUdevice dev); * \sa * ::cuDeviceGetCount, * ::cuDeviceGetName, + * ::cuDeviceGetUuid, * ::cuDeviceGet, - * ::cuDeviceTotalMem + * ::cuDeviceTotalMem, + * ::cudaDeviceGetAttribute, + * ::cudaGetDeviceProperties */ CUresult CUDAAPI cuDeviceGetAttribute(int *pi, CUdevice_attribute attrib, CUdevice dev); @@ -2321,10 +2993,11 @@ CUresult CUDAAPI cuDeviceGetAttribute(int *pi, CUdevice_attribute attrib, CUdevi * ::cuDeviceGetAttribute, * ::cuDeviceGetCount, * ::cuDeviceGetName, + * ::cuDeviceGetUuid, * ::cuDeviceGet, * ::cuDeviceTotalMem */ -CUresult CUDAAPI cuDeviceGetProperties(CUdevprop *prop, CUdevice dev); +__CUDA_DEPRECATED CUresult CUDAAPI cuDeviceGetProperties(CUdevprop *prop, CUdevice dev); /** * \brief Returns the compute capability of the device @@ -2332,7 +3005,7 @@ CUresult CUDAAPI cuDeviceGetProperties(CUdevprop *prop, CUdevice dev); * \deprecated * * This function was deprecated as of CUDA 5.0 and its functionality superceded - * by ::cuDeviceGetAttribute(). + * by ::cuDeviceGetAttribute(). * * Returns in \p *major and \p *minor the major and minor revision numbers that * define the compute capability of the device \p dev. @@ -2354,10 +3027,11 @@ CUresult CUDAAPI cuDeviceGetProperties(CUdevprop *prop, CUdevice dev); * ::cuDeviceGetAttribute, * ::cuDeviceGetCount, * ::cuDeviceGetName, + * ::cuDeviceGetUuid, * ::cuDeviceGet, * ::cuDeviceTotalMem */ -CUresult CUDAAPI cuDeviceComputeCapability(int *major, int *minor, CUdevice dev); +__CUDA_DEPRECATED CUresult CUDAAPI cuDeviceComputeCapability(int *major, int *minor, CUdevice dev); /** @} */ /* END CUDA_DEVICE_DEPRECATED */ @@ -2370,8 +3044,8 @@ CUresult CUDAAPI cuDeviceComputeCapability(int *major, int *minor, CUdevice dev) * This section describes the primary context management functions of the low-level * CUDA driver application programming interface. * - * The primary context unique per device and it's shared with CUDA runtime API. - * Those functions allows seemless integration with other libraries using CUDA. + * The primary context is unique per device and shared with the CUDA runtime API. + * These functions allow integration with other libraries using CUDA. * * @{ */ @@ -2387,9 +3061,9 @@ CUresult CUDAAPI cuDeviceComputeCapability(int *major, int *minor, CUdevice dev) * Unlike ::cuCtxCreate() the newly created context is not pushed onto the stack. * * Context creation will fail with ::CUDA_ERROR_UNKNOWN if the compute mode of - * the device is ::CU_COMPUTEMODE_PROHIBITED. The function ::cuDeviceGetAttribute() - * can be used with ::CU_DEVICE_ATTRIBUTE_COMPUTE_MODE to determine the compute mode - * of the device. + * the device is ::CU_COMPUTEMODE_PROHIBITED. The function ::cuDeviceGetAttribute() + * can be used with ::CU_DEVICE_ATTRIBUTE_COMPUTE_MODE to determine the compute mode + * of the device. * The nvidia-smi tool can be used to set the compute mode for * devices. Documentation for nvidia-smi can be obtained by passing a * -h option to it. @@ -2497,8 +3171,9 @@ CUresult CUDAAPI cuDevicePrimaryCtxRelease(CUdevice dev); * \e C > \e P, then CUDA will yield to other OS threads when waiting for * the GPU (::CU_CTX_SCHED_YIELD), otherwise CUDA will not yield while * waiting for results and actively spin on the processor (::CU_CTX_SCHED_SPIN). - * However, on low power devices like Tegra, it always defaults to - * ::CU_CTX_SCHED_BLOCKING_SYNC. + * Additionally, on Tegra devices, ::CU_CTX_SCHED_AUTO uses a heuristic based on + * the power profile of the platform and may choose ::CU_CTX_SCHED_BLOCKING_SYNC + * for low-powered devices. * * - ::CU_CTX_LMEM_RESIZE_TO_MAX: Instruct CUDA to not reduce local memory * after resizing local memory for a kernel. This can prevent thrashing by @@ -2520,7 +3195,8 @@ CUresult CUDAAPI cuDevicePrimaryCtxRelease(CUdevice dev); * \sa ::cuDevicePrimaryCtxRetain, * ::cuDevicePrimaryCtxGetState, * ::cuCtxCreate, - * ::cuCtxGetFlags + * ::cuCtxGetFlags, + * ::cudaSetDeviceFlags */ CUresult CUDAAPI cuDevicePrimaryCtxSetFlags(CUdevice dev, unsigned int flags); @@ -2543,8 +3219,10 @@ CUresult CUDAAPI cuDevicePrimaryCtxSetFlags(CUdevice dev, unsigned int flags); * ::CUDA_ERROR_INVALID_VALUE, * \notefnerr * - * \sa ::cuDevicePrimaryCtxSetFlags, - * ::cuCtxGetFlags + * \sa + * ::cuDevicePrimaryCtxSetFlags, + * ::cuCtxGetFlags, + * ::cudaGetDeviceFlags */ CUresult CUDAAPI cuDevicePrimaryCtxGetState(CUdevice dev, unsigned int *flags, int *active); @@ -2581,8 +3259,8 @@ CUresult CUDAAPI cuDevicePrimaryCtxGetState(CUdevice dev, unsigned int *flags, i * ::cuCtxPushCurrent, * ::cuCtxSetCacheConfig, * ::cuCtxSetLimit, - * ::cuCtxSynchronize - * + * ::cuCtxSynchronize, + * ::cudaDeviceReset */ CUresult CUDAAPI cuDevicePrimaryCtxReset(CUdevice dev); @@ -2600,6 +3278,9 @@ CUresult CUDAAPI cuDevicePrimaryCtxReset(CUdevice dev); * This section describes the context management functions of the low-level * CUDA driver application programming interface. * + * Please note that some functions are described in + * \ref CUDA_PRIMARY_CTX "Primary Context Management" section. + * * @{ */ @@ -2607,11 +3288,13 @@ CUresult CUDAAPI cuDevicePrimaryCtxReset(CUdevice dev); /** * \brief Create a CUDA context * + * \note In most cases it is recommended to use ::cuDevicePrimaryCtxRetain. + * * Creates a new CUDA context and associates it with the calling thread. The * \p flags parameter is described below. The context is created with a usage - * count of 1 and the caller of ::cuCtxCreate() must call ::cuCtxDestroy() or - * when done using the context. If a context is already current to the thread, - * it is supplanted by the newly created context and may be restored by a subsequent + * count of 1 and the caller of ::cuCtxCreate() must call ::cuCtxDestroy() + * when done using the context. If a context is already current to the thread, + * it is supplanted by the newly created context and may be restored by a subsequent * call to ::cuCtxPopCurrent(). * * The three LSBs of the \p flags parameter can be used to control how the OS @@ -2628,23 +3311,24 @@ CUresult CUDAAPI cuDevicePrimaryCtxReset(CUdevice dev); * results from the GPU. This can increase latency when waiting for the GPU, * but can increase the performance of CPU threads performing work in parallel * with the GPU. - * + * * - ::CU_CTX_SCHED_BLOCKING_SYNC: Instruct CUDA to block the CPU thread on a * synchronization primitive when waiting for the GPU to finish work. * * - ::CU_CTX_BLOCKING_SYNC: Instruct CUDA to block the CPU thread on a * synchronization primitive when waiting for the GPU to finish work.
* Deprecated: This flag was deprecated as of CUDA 4.0 and was - * replaced with ::CU_CTX_SCHED_BLOCKING_SYNC. + * replaced with ::CU_CTX_SCHED_BLOCKING_SYNC. * * - ::CU_CTX_SCHED_AUTO: The default value if the \p flags parameter is zero, * uses a heuristic based on the number of active CUDA contexts in the * process \e C and the number of logical processors in the system \e P. If - * \e C > \e P, then CUDA will yield to other OS threads when waiting for - * the GPU (::CU_CTX_SCHED_YIELD), otherwise CUDA will not yield while - * waiting for results and actively spin on the processor (::CU_CTX_SCHED_SPIN). - * However, on low power devices like Tegra, it always defaults to - * ::CU_CTX_SCHED_BLOCKING_SYNC. + * \e C > \e P, then CUDA will yield to other OS threads when waiting for + * the GPU (::CU_CTX_SCHED_YIELD), otherwise CUDA will not yield while + * waiting for results and actively spin on the processor (::CU_CTX_SCHED_SPIN). + * Additionally, on Tegra devices, ::CU_CTX_SCHED_AUTO uses a heuristic based on + * the power profile of the platform and may choose ::CU_CTX_SCHED_BLOCKING_SYNC + * for low-powered devices. * * - ::CU_CTX_MAP_HOST: Instruct CUDA to support mapped pinned allocations. * This flag must be set in order to allocate pinned host memory that is @@ -2656,10 +3340,10 @@ CUresult CUDAAPI cuDevicePrimaryCtxReset(CUdevice dev); * memory usage at the cost of potentially increased memory usage. * * Context creation will fail with ::CUDA_ERROR_UNKNOWN if the compute mode of - * the device is ::CU_COMPUTEMODE_PROHIBITED. The function ::cuDeviceGetAttribute() - * can be used with ::CU_DEVICE_ATTRIBUTE_COMPUTE_MODE to determine the - * compute mode of the device. The nvidia-smi tool can be used to set - * the compute mode for * devices. + * the device is ::CU_COMPUTEMODE_PROHIBITED. The function ::cuDeviceGetAttribute() + * can be used with ::CU_DEVICE_ATTRIBUTE_COMPUTE_MODE to determine the + * compute mode of the device. The nvidia-smi tool can be used to set + * the compute mode for * devices. * Documentation for nvidia-smi can be obtained by passing a * -h option to it. * @@ -2702,7 +3386,7 @@ CUresult CUDAAPI cuCtxCreate(CUcontext *pctx, unsigned int flags, CUdevice dev); * It is the responsibility of the calling function to ensure that no API * call issues using \p ctx while ::cuCtxDestroy() is executing. * - * If \p ctx is current to the calling thread then \p ctx will also be + * If \p ctx is current to the calling thread then \p ctx will also be * popped from the current thread's context stack (as though ::cuCtxPopCurrent() * were called). If \p ctx is current to other threads, then \p ctx will * remain current to those threads, and attempting to access \p ctx from @@ -2771,8 +3455,8 @@ CUresult CUDAAPI cuCtxPushCurrent(CUcontext ctx); /** * \brief Pops the current CUDA context from the current CPU thread. * - * Pops the current CUDA context from the CPU thread and passes back the - * old context handle in \p *pctx. That context may then be made current + * Pops the current CUDA context from the CPU thread and passes back the + * old context handle in \p *pctx. That context may then be made current * to a different CPU thread by calling ::cuCtxPushCurrent(). * * If a context was current to the CPU thread before ::cuCtxCreate() or @@ -2810,7 +3494,7 @@ CUresult CUDAAPI cuCtxPopCurrent(CUcontext *pctx); * calling CPU thread is unbound and ::CUDA_SUCCESS is returned. * * If there exists a CUDA context stack on the calling CPU thread, this - * will replace the top of that stack with \p ctx. + * will replace the top of that stack with \p ctx. * If \p ctx is NULL then this will be equivalent to popping the top * of the calling CPU thread's CUDA context stack (or a no-op if the * calling CPU thread's CUDA context stack is empty). @@ -2824,7 +3508,11 @@ CUresult CUDAAPI cuCtxPopCurrent(CUcontext *pctx); * ::CUDA_ERROR_INVALID_CONTEXT * \notefnerr * - * \sa ::cuCtxGetCurrent, ::cuCtxCreate, ::cuCtxDestroy + * \sa + * ::cuCtxGetCurrent, + * ::cuCtxCreate, + * ::cuCtxDestroy, + * ::cudaSetDevice */ CUresult CUDAAPI cuCtxSetCurrent(CUcontext ctx); @@ -2843,7 +3531,11 @@ CUresult CUDAAPI cuCtxSetCurrent(CUcontext ctx); * ::CUDA_ERROR_NOT_INITIALIZED, * \notefnerr * - * \sa ::cuCtxSetCurrent, ::cuCtxCreate, ::cuCtxDestroy + * \sa + * ::cuCtxSetCurrent, + * ::cuCtxCreate, + * ::cuCtxDestroy, + * ::cudaGetDevice */ CUresult CUDAAPI cuCtxGetCurrent(CUcontext *pctx); #endif /* __CUDA_API_VERSION >= 4000 */ @@ -2873,7 +3565,8 @@ CUresult CUDAAPI cuCtxGetCurrent(CUcontext *pctx); * ::cuCtxPushCurrent, * ::cuCtxSetCacheConfig, * ::cuCtxSetLimit, - * ::cuCtxSynchronize + * ::cuCtxSynchronize, + * ::cudaGetDevice */ CUresult CUDAAPI cuCtxGetDevice(CUdevice *device); @@ -2901,7 +3594,8 @@ CUresult CUDAAPI cuCtxGetDevice(CUdevice *device); * ::cuCtxGetDevice * ::cuCtxGetLimit, * ::cuCtxGetSharedMemConfig, - * ::cuCtxGetStreamPriorityRange + * ::cuCtxGetStreamPriorityRange, + * ::cudaGetDeviceFlags */ CUresult CUDAAPI cuCtxGetFlags(unsigned int *flags); #endif /* __CUDA_API_VERSION >= 7000 */ @@ -2911,7 +3605,7 @@ CUresult CUDAAPI cuCtxGetFlags(unsigned int *flags); * * Blocks until the device has completed all preceding requested tasks. * ::cuCtxSynchronize() returns an error if one of the preceding tasks failed. - * If the context was created with the ::CU_CTX_SCHED_BLOCKING_SYNC flag, the + * If the context was created with the ::CU_CTX_SCHED_BLOCKING_SYNC flag, the * CPU thread will block until the GPU context has finished its work. * * \return @@ -2931,7 +3625,8 @@ CUresult CUDAAPI cuCtxGetFlags(unsigned int *flags); * ::cuCtxPopCurrent, * ::cuCtxPushCurrent, * ::cuCtxSetCacheConfig, - * ::cuCtxSetLimit + * ::cuCtxSetLimit, + * ::cudaDeviceSynchronize */ CUresult CUDAAPI cuCtxSynchronize(void); @@ -2949,50 +3644,41 @@ CUresult CUDAAPI cuCtxSynchronize(void); * discussed here. * * - ::CU_LIMIT_STACK_SIZE controls the stack size in bytes of each GPU thread. - * This limit is only applicable to devices of compute capability 2.0 and - * higher. Attempting to set this limit on devices of compute capability - * less than 2.0 will result in the error ::CUDA_ERROR_UNSUPPORTED_LIMIT - * being returned. + * Note that the CUDA driver will set the \p limit to the maximum of \p value + * and what the kernel function requires. * * - ::CU_LIMIT_PRINTF_FIFO_SIZE controls the size in bytes of the FIFO used * by the ::printf() device system call. Setting ::CU_LIMIT_PRINTF_FIFO_SIZE * must be performed before launching any kernel that uses the ::printf() * device system call, otherwise ::CUDA_ERROR_INVALID_VALUE will be returned. - * This limit is only applicable to devices of compute capability 2.0 and - * higher. Attempting to set this limit on devices of compute capability - * less than 2.0 will result in the error ::CUDA_ERROR_UNSUPPORTED_LIMIT - * being returned. * * - ::CU_LIMIT_MALLOC_HEAP_SIZE controls the size in bytes of the heap used * by the ::malloc() and ::free() device system calls. Setting * ::CU_LIMIT_MALLOC_HEAP_SIZE must be performed before launching any kernel * that uses the ::malloc() or ::free() device system calls, otherwise - * ::CUDA_ERROR_INVALID_VALUE will be returned. This limit is only applicable - * to devices of compute capability 2.0 and higher. Attempting to set this - * limit on devices of compute capability less than 2.0 will result in the - * error ::CUDA_ERROR_UNSUPPORTED_LIMIT being returned. + * ::CUDA_ERROR_INVALID_VALUE will be returned. * * - ::CU_LIMIT_DEV_RUNTIME_SYNC_DEPTH controls the maximum nesting depth of * a grid at which a thread can safely call ::cudaDeviceSynchronize(). Setting - * this limit must be performed before any launch of a kernel that uses the + * this limit must be performed before any launch of a kernel that uses the * device runtime and calls ::cudaDeviceSynchronize() above the default sync - * depth, two levels of grids. Calls to ::cudaDeviceSynchronize() will fail - * with error code ::cudaErrorSyncDepthExceeded if the limitation is + * depth, two levels of grids. Calls to ::cudaDeviceSynchronize() will fail + * with error code ::cudaErrorSyncDepthExceeded if the limitation is * violated. This limit can be set smaller than the default or up the maximum * launch depth of 24. When setting this limit, keep in mind that additional * levels of sync depth require the driver to reserve large amounts of device - * memory which can no longer be used for user allocations. If these - * reservations of device memory fail, ::cuCtxSetLimit will return + * memory which can no longer be used for user allocations. If these + * reservations of device memory fail, ::cuCtxSetLimit will return * ::CUDA_ERROR_OUT_OF_MEMORY, and the limit can be reset to a lower value. * This limit is only applicable to devices of compute capability 3.5 and * higher. Attempting to set this limit on devices of compute capability less - * than 3.5 will result in the error ::CUDA_ERROR_UNSUPPORTED_LIMIT being + * than 3.5 will result in the error ::CUDA_ERROR_UNSUPPORTED_LIMIT being * returned. * * - ::CU_LIMIT_DEV_RUNTIME_PENDING_LAUNCH_COUNT controls the maximum number of * outstanding device runtime launches that can be made from the current * context. A grid is outstanding from the point of launch up until the grid - * is known to have been completed. Device runtime launches which violate + * is known to have been completed. Device runtime launches which violate * this limitation fail and return ::cudaErrorLaunchPendingCountExceeded when * ::cudaGetLastError() is called after launch. If more pending launches than * the default (2048 launches) are needed for a module using the device @@ -3006,6 +3692,10 @@ CUresult CUDAAPI cuCtxSynchronize(void); * than 3.5 will result in the error ::CUDA_ERROR_UNSUPPORTED_LIMIT being * returned. * + * - ::CU_LIMIT_MAX_L2_FETCH_GRANULARITY controls the L2 cache fetch granularity. + * Values can range from 0B to 128B. This is purely a performance hint and + * it can be ignored or clamped depending on the platform. + * * \param limit - Limit to set * \param value - Size of limit * @@ -3013,7 +3703,8 @@ CUresult CUDAAPI cuCtxSynchronize(void); * ::CUDA_SUCCESS, * ::CUDA_ERROR_INVALID_VALUE, * ::CUDA_ERROR_UNSUPPORTED_LIMIT, - * ::CUDA_ERROR_OUT_OF_MEMORY + * ::CUDA_ERROR_OUT_OF_MEMORY, + * ::CUDA_ERROR_INVALID_CONTEXT * \notefnerr * * \sa ::cuCtxCreate, @@ -3026,7 +3717,8 @@ CUresult CUDAAPI cuCtxSynchronize(void); * ::cuCtxPopCurrent, * ::cuCtxPushCurrent, * ::cuCtxSetCacheConfig, - * ::cuCtxSynchronize + * ::cuCtxSynchronize, + * ::cudaDeviceSetLimit */ CUresult CUDAAPI cuCtxSetLimit(CUlimit limit, size_t value); @@ -3045,6 +3737,7 @@ CUresult CUDAAPI cuCtxSetLimit(CUlimit limit, size_t value); * child grid launches to complete. * - ::CU_LIMIT_DEV_RUNTIME_PENDING_LAUNCH_COUNT: maximum number of outstanding * device runtime launches that can be made from this context. + * - ::CU_LIMIT_MAX_L2_FETCH_GRANULARITY: L2 cache fetch granularity. * * \param limit - Limit to query * \param pvalue - Returned size of limit @@ -3065,7 +3758,8 @@ CUresult CUDAAPI cuCtxSetLimit(CUlimit limit, size_t value); * ::cuCtxPushCurrent, * ::cuCtxSetCacheConfig, * ::cuCtxSetLimit, - * ::cuCtxSynchronize + * ::cuCtxSynchronize, + * ::cudaDeviceGetLimit */ CUresult CUDAAPI cuCtxGetLimit(size_t *pvalue, CUlimit limit); @@ -3108,7 +3802,8 @@ CUresult CUDAAPI cuCtxGetLimit(size_t *pvalue, CUlimit limit); * ::cuCtxSetCacheConfig, * ::cuCtxSetLimit, * ::cuCtxSynchronize, - * ::cuFuncSetCacheConfig + * ::cuFuncSetCacheConfig, + * ::cudaDeviceGetCacheConfig */ CUresult CUDAAPI cuCtxGetCacheConfig(CUfunc_cache *pconfig); @@ -3158,7 +3853,8 @@ CUresult CUDAAPI cuCtxGetCacheConfig(CUfunc_cache *pconfig); * ::cuCtxPushCurrent, * ::cuCtxSetLimit, * ::cuCtxSynchronize, - * ::cuFuncSetCacheConfig + * ::cuFuncSetCacheConfig, + * ::cudaDeviceSetCacheConfig */ CUresult CUDAAPI cuCtxSetCacheConfig(CUfunc_cache config); @@ -3167,20 +3863,20 @@ CUresult CUDAAPI cuCtxSetCacheConfig(CUfunc_cache config); * \brief Returns the current shared memory configuration for the current context. * * This function will return in \p pConfig the current size of shared memory banks - * in the current context. On devices with configurable shared memory banks, - * ::cuCtxSetSharedMemConfig can be used to change this setting, so that all - * subsequent kernel launches will by default use the new bank size. When - * ::cuCtxGetSharedMemConfig is called on devices without configurable shared + * in the current context. On devices with configurable shared memory banks, + * ::cuCtxSetSharedMemConfig can be used to change this setting, so that all + * subsequent kernel launches will by default use the new bank size. When + * ::cuCtxGetSharedMemConfig is called on devices without configurable shared * memory, it will return the fixed bank size of the hardware. * * The returned bank configurations can be either: - * - ::CU_SHARED_MEM_CONFIG_FOUR_BYTE_BANK_SIZE: shared memory bank width is + * - ::CU_SHARED_MEM_CONFIG_FOUR_BYTE_BANK_SIZE: shared memory bank width is * four bytes. * - ::CU_SHARED_MEM_CONFIG_EIGHT_BYTE_BANK_SIZE: shared memory bank width will * eight bytes. * * \param pConfig - returned shared memory configuration - * \return + * \return * ::CUDA_SUCCESS, * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, @@ -3201,6 +3897,7 @@ CUresult CUDAAPI cuCtxSetCacheConfig(CUfunc_cache config); * ::cuCtxSynchronize, * ::cuCtxGetSharedMemConfig, * ::cuFuncSetCacheConfig, + * ::cudaDeviceGetSharedMemConfig */ CUresult CUDAAPI cuCtxGetSharedMemConfig(CUsharedconfig *pConfig); @@ -3208,16 +3905,16 @@ CUresult CUDAAPI cuCtxGetSharedMemConfig(CUsharedconfig *pConfig); * \brief Sets the shared memory configuration for the current context. * * On devices with configurable shared memory banks, this function will set - * the context's shared memory bank size which is used for subsequent kernel - * launches. + * the context's shared memory bank size which is used for subsequent kernel + * launches. * * Changed the shared memory configuration between launches may insert a device * side synchronization point between those launches. * * Changing the shared memory bank size will not increase shared memory usage - * or affect occupancy of kernels, but may have major effects on performance. + * or affect occupancy of kernels, but may have major effects on performance. * Larger bank sizes will allow for greater potential bandwidth to shared memory, - * but will change what kinds of accesses to shared memory will result in bank + * but will change what kinds of accesses to shared memory will result in bank * conflicts. * * This function will do nothing on devices with fixed shared memory bank size. @@ -3253,6 +3950,7 @@ CUresult CUDAAPI cuCtxGetSharedMemConfig(CUsharedconfig *pConfig); * ::cuCtxSynchronize, * ::cuCtxGetSharedMemConfig, * ::cuFuncSetCacheConfig, + * ::cudaDeviceSetSharedMemConfig */ CUresult CUDAAPI cuCtxSetSharedMemConfig(CUsharedconfig config); #endif @@ -3278,6 +3976,7 @@ CUresult CUDAAPI cuCtxSetSharedMemConfig(CUsharedconfig config); * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, + * ::CUDA_ERROR_INVALID_VALUE, * ::CUDA_ERROR_UNKNOWN * \notefnerr * @@ -3329,7 +4028,8 @@ CUresult CUDAAPI cuCtxGetApiVersion(CUcontext ctx, unsigned int *version); * ::cuCtxGetDevice, * ::cuCtxGetFlags, * ::cuCtxSetLimit, - * ::cuCtxSynchronize + * ::cuCtxSynchronize, + * ::cudaDeviceGetStreamPriorityRange */ CUresult CUDAAPI cuCtxGetStreamPriorityRange(int *leastPriority, int *greatestPriority); @@ -3386,7 +4086,7 @@ CUresult CUDAAPI cuCtxGetStreamPriorityRange(int *leastPriority, int *greatestPr * ::cuCtxSetLimit, * ::cuCtxSynchronize */ -CUresult CUDAAPI cuCtxAttach(CUcontext *pctx, unsigned int flags); +__CUDA_DEPRECATED CUresult CUDAAPI cuCtxAttach(CUcontext *pctx, unsigned int flags); /** * \brief Decrement a context's usage-count @@ -3422,7 +4122,7 @@ CUresult CUDAAPI cuCtxAttach(CUcontext *pctx, unsigned int flags); * ::cuCtxSetLimit, * ::cuCtxSynchronize */ -CUresult CUDAAPI cuCtxDetach(CUcontext ctx); +__CUDA_DEPRECATED CUresult CUDAAPI cuCtxDetach(CUcontext ctx); /** @} */ /* END CUDA_CTX_DEPRECATED */ @@ -3465,7 +4165,8 @@ CUresult CUDAAPI cuCtxDetach(CUcontext ctx); * ::CUDA_ERROR_FILE_NOT_FOUND, * ::CUDA_ERROR_NO_BINARY_FOR_GPU, * ::CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND, - * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED + * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED, + * ::CUDA_ERROR_JIT_COMPILER_NOT_FOUND * \notefnerr * * \sa ::cuModuleGetFunction, @@ -3501,7 +4202,8 @@ CUresult CUDAAPI cuModuleLoad(CUmodule *module, const char *fname); * ::CUDA_ERROR_OUT_OF_MEMORY, * ::CUDA_ERROR_NO_BINARY_FOR_GPU, * ::CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND, - * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED + * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED, + * ::CUDA_ERROR_JIT_COMPILER_NOT_FOUND * \notefnerr * * \sa ::cuModuleGetFunction, @@ -3525,7 +4227,7 @@ CUresult CUDAAPI cuModuleLoadData(CUmodule *module, const void *image); * as Windows \c FindResource() to obtain the pointer. Options are passed as * an array via \p options and any corresponding parameters are passed in * \p optionValues. The number of total options is supplied via \p numOptions. - * Any outputs will be returned via \p optionValues. + * Any outputs will be returned via \p optionValues. * * \param module - Returned module * \param image - Module data to load @@ -3543,7 +4245,8 @@ CUresult CUDAAPI cuModuleLoadData(CUmodule *module, const void *image); * ::CUDA_ERROR_OUT_OF_MEMORY, * ::CUDA_ERROR_NO_BINARY_FOR_GPU, * ::CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND, - * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED + * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED, + * ::CUDA_ERROR_JIT_COMPILER_NOT_FOUND * \notefnerr * * \sa ::cuModuleGetFunction, @@ -3584,7 +4287,8 @@ CUresult CUDAAPI cuModuleLoadDataEx(CUmodule *module, const void *image, unsigne * ::CUDA_ERROR_OUT_OF_MEMORY, * ::CUDA_ERROR_NO_BINARY_FOR_GPU, * ::CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND, - * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED + * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED, + * ::CUDA_ERROR_JIT_COMPILER_NOT_FOUND * \notefnerr * * \sa ::cuModuleGetFunction, @@ -3682,7 +4386,9 @@ CUresult CUDAAPI cuModuleGetFunction(CUfunction *hfunc, CUmodule hmod, const cha * ::cuModuleLoadData, * ::cuModuleLoadDataEx, * ::cuModuleLoadFatBinary, - * ::cuModuleUnload + * ::cuModuleUnload, + * ::cudaGetSymbolAddress, + * ::cudaGetSymbolSize */ CUresult CUDAAPI cuModuleGetGlobal(CUdeviceptr *dptr, size_t *bytes, CUmodule hmod, const char *name); #endif /* __CUDA_API_VERSION >= 3020 */ @@ -3716,7 +4422,8 @@ CUresult CUDAAPI cuModuleGetGlobal(CUdeviceptr *dptr, size_t *bytes, CUmodule hm * ::cuModuleLoadData, * ::cuModuleLoadDataEx, * ::cuModuleLoadFatBinary, - * ::cuModuleUnload + * ::cuModuleUnload, + * ::cudaGetTextureReference */ CUresult CUDAAPI cuModuleGetTexRef(CUtexref *pTexRef, CUmodule hmod, const char *name); @@ -3747,7 +4454,8 @@ CUresult CUDAAPI cuModuleGetTexRef(CUtexref *pTexRef, CUmodule hmod, const char * ::cuModuleLoadData, * ::cuModuleLoadDataEx, * ::cuModuleLoadFatBinary, - * ::cuModuleUnload + * ::cuModuleUnload, + * ::cudaGetSurfaceReference */ CUresult CUDAAPI cuModuleGetSurfRef(CUsurfref *pSurfRef, CUmodule hmod, const char *name); @@ -3782,7 +4490,8 @@ CUresult CUDAAPI cuModuleGetSurfRef(CUsurfref *pSurfRef, CUmodule hmod, const ch * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, * ::CUDA_ERROR_INVALID_VALUE, - * ::CUDA_ERROR_OUT_OF_MEMORY + * ::CUDA_ERROR_OUT_OF_MEMORY, + * ::CUDA_ERROR_JIT_COMPILER_NOT_FOUND * \notefnerr * * \sa ::cuLinkAddData, @@ -3954,7 +4663,8 @@ cuLinkDestroy(CUlinkState state); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemGetInfo */ CUresult CUDAAPI cuMemGetInfo(size_t *free, size_t *total); @@ -3987,7 +4697,8 @@ CUresult CUDAAPI cuMemGetInfo(size_t *free, size_t *total); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMalloc */ CUresult CUDAAPI cuMemAlloc(CUdeviceptr *dptr, size_t bytesize); @@ -4048,7 +4759,8 @@ CUresult CUDAAPI cuMemAlloc(CUdeviceptr *dptr, size_t bytesize); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMallocPitch */ CUresult CUDAAPI cuMemAllocPitch(CUdeviceptr *dptr, size_t *pPitch, size_t WidthInBytes, size_t Height, unsigned int ElementSizeBytes); @@ -4077,7 +4789,8 @@ CUresult CUDAAPI cuMemAllocPitch(CUdeviceptr *dptr, size_t *pPitch, size_t Width * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaFree */ CUresult CUDAAPI cuMemFree(CUdeviceptr dptr); @@ -4098,6 +4811,7 @@ CUresult CUDAAPI cuMemFree(CUdeviceptr dptr); * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, + * ::CUDA_ERROR_NOT_FOUND, * ::CUDA_ERROR_INVALID_VALUE * \notefnerr * @@ -4131,7 +4845,7 @@ CUresult CUDAAPI cuMemGetAddressRange(CUdeviceptr *pbase, size_t *psize, CUdevic * Note all host memory allocated using ::cuMemHostAlloc() will automatically * be immediately accessible to all contexts on all devices which support unified * addressing (as may be queried using ::CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING). - * The device pointer that may be used to access this host memory from those + * The device pointer that may be used to access this host memory from those * contexts is always equal to the returned host pointer \p *pp. * See \ref CUDA_UNIFIED for additional details. * @@ -4156,7 +4870,8 @@ CUresult CUDAAPI cuMemGetAddressRange(CUdeviceptr *pbase, size_t *psize, CUdevic * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMallocHost */ CUresult CUDAAPI cuMemAllocHost(void **pp, size_t bytesize); #endif /* __CUDA_API_VERSION >= 3020 */ @@ -4186,7 +4901,8 @@ CUresult CUDAAPI cuMemAllocHost(void **pp, size_t bytesize); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaFreeHost */ CUresult CUDAAPI cuMemFreeHost(void *p); @@ -4213,8 +4929,7 @@ CUresult CUDAAPI cuMemFreeHost(void *p); * * - ::CU_MEMHOSTALLOC_DEVICEMAP: Maps the allocation into the CUDA address * space. The device pointer to the memory may be obtained by calling - * ::cuMemHostGetDevicePointer(). This feature is available only on GPUs - * with compute capability greater than or equal to 1.1. + * ::cuMemHostGetDevicePointer(). * * - ::CU_MEMHOSTALLOC_WRITECOMBINED: Allocates the memory as write-combined * (WC). WC memory can be transferred across the PCI Express bus more @@ -4239,8 +4954,8 @@ CUresult CUDAAPI cuMemFreeHost(void *p); * Note all host memory allocated using ::cuMemHostAlloc() will automatically * be immediately accessible to all contexts on all devices which support unified * addressing (as may be queried using ::CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING). - * Unless the flag ::CU_MEMHOSTALLOC_WRITECOMBINED is specified, the device pointer - * that may be used to access this host memory from those contexts is always equal + * Unless the flag ::CU_MEMHOSTALLOC_WRITECOMBINED is specified, the device pointer + * that may be used to access this host memory from those contexts is always equal * to the returned host pointer \p *pp. If the flag ::CU_MEMHOSTALLOC_WRITECOMBINED * is specified, then the function ::cuMemHostGetDevicePointer() must be used * to query the device pointer, even if the context supports unified addressing. @@ -4268,7 +4983,8 @@ CUresult CUDAAPI cuMemFreeHost(void *p); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaHostAlloc */ CUresult CUDAAPI cuMemHostAlloc(void **pp, size_t bytesize, unsigned int Flags); @@ -4321,7 +5037,8 @@ CUresult CUDAAPI cuMemHostAlloc(void **pp, size_t bytesize, unsigned int Flags); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaHostGetDevicePointer */ CUresult CUDAAPI cuMemHostGetDevicePointer(CUdeviceptr *pdptr, void *p, unsigned int Flags); #endif /* __CUDA_API_VERSION >= 3020 */ @@ -4346,7 +5063,10 @@ CUresult CUDAAPI cuMemHostGetDevicePointer(CUdeviceptr *pdptr, void *p, unsigned * ::CUDA_ERROR_INVALID_VALUE * \notefnerr * - * \sa ::cuMemAllocHost, ::cuMemHostAlloc + * \sa + * ::cuMemAllocHost, + * ::cuMemHostAlloc, + * ::cudaHostGetFlags */ CUresult CUDAAPI cuMemHostGetFlags(unsigned int *pFlags, void *p); @@ -4431,6 +5151,7 @@ CUresult CUDAAPI cuMemHostGetFlags(unsigned int *pFlags, void *p); * managed memory supporting devices on which contexts were previously created, even if * those contexts have been destroyed. These environment variables are described * in the CUDA programming guide under the "CUDA environment variables" section. + * - On ARM, managed memory is not available on discrete gpu with Drive PX-2. * * \param dptr - Returned device pointer * \param bytesize - Requested allocation size in bytes @@ -4456,7 +5177,8 @@ CUresult CUDAAPI cuMemHostGetFlags(unsigned int *pFlags, void *p); * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, - * ::cuDeviceGetAttribute, ::cuStreamAttachMemAsync + * ::cuDeviceGetAttribute, ::cuStreamAttachMemAsync, + * ::cudaMallocManaged */ CUresult CUDAAPI cuMemAllocManaged(CUdeviceptr *dptr, size_t bytesize, unsigned int flags); @@ -4471,7 +5193,7 @@ CUresult CUDAAPI cuMemAllocManaged(CUdeviceptr *dptr, size_t bytesize, unsigned * * \param dev - Returned device handle * - * \param pciBusId - String in one of the following forms: + * \param pciBusId - String in one of the following forms: * [domain]:[bus]:[device].[function] * [domain]:[bus]:[device] * [bus]:[device].[function] @@ -4485,7 +5207,11 @@ CUresult CUDAAPI cuMemAllocManaged(CUdeviceptr *dptr, size_t bytesize, unsigned * ::CUDA_ERROR_INVALID_DEVICE * \notefnerr * - * \sa ::cuDeviceGet, ::cuDeviceGetAttribute, ::cuDeviceGetPCIBusId + * \sa + * ::cuDeviceGet, + * ::cuDeviceGetAttribute, + * ::cuDeviceGetPCIBusId, + * ::cudaDeviceGetByPCIBusId */ CUresult CUDAAPI cuDeviceGetByPCIBusId(CUdevice *dev, const char *pciBusId); @@ -4513,65 +5239,73 @@ CUresult CUDAAPI cuDeviceGetByPCIBusId(CUdevice *dev, const char *pciBusId); * ::CUDA_ERROR_INVALID_DEVICE * \notefnerr * - * \sa ::cuDeviceGet, ::cuDeviceGetAttribute, ::cuDeviceGetByPCIBusId + * \sa + * ::cuDeviceGet, + * ::cuDeviceGetAttribute, + * ::cuDeviceGetByPCIBusId, + * ::cudaDeviceGetPCIBusId */ CUresult CUDAAPI cuDeviceGetPCIBusId(char *pciBusId, int len, CUdevice dev); /** * \brief Gets an interprocess handle for a previously allocated event * - * Takes as input a previously allocated event. This event must have been - * created with the ::CU_EVENT_INTERPROCESS and ::CU_EVENT_DISABLE_TIMING + * Takes as input a previously allocated event. This event must have been + * created with the ::CU_EVENT_INTERPROCESS and ::CU_EVENT_DISABLE_TIMING * flags set. This opaque handle may be copied into other processes and * opened with ::cuIpcOpenEventHandle to allow efficient hardware * synchronization between GPU work in different processes. * - * After the event has been opened in the importing process, - * ::cuEventRecord, ::cuEventSynchronize, ::cuStreamWaitEvent and - * ::cuEventQuery may be used in either process. Performing operations - * on the imported event after the exported event has been freed + * After the event has been opened in the importing process, + * ::cuEventRecord, ::cuEventSynchronize, ::cuStreamWaitEvent and + * ::cuEventQuery may be used in either process. Performing operations + * on the imported event after the exported event has been freed * with ::cuEventDestroy will result in undefined behavior. * - * IPC functionality is restricted to devices with support for unified - * addressing on Linux operating systems. + * IPC functionality is restricted to devices with support for unified + * addressing on Linux and Windows operating systems. + * IPC functionality on Windows is restricted to GPUs in TCC mode * * \param pHandle - Pointer to a user allocated CUipcEventHandle * in which to return the opaque event handle - * \param event - Event allocated with ::CU_EVENT_INTERPROCESS and + * \param event - Event allocated with ::CU_EVENT_INTERPROCESS and * ::CU_EVENT_DISABLE_TIMING flags. * * \return * ::CUDA_SUCCESS, * ::CUDA_ERROR_INVALID_HANDLE, * ::CUDA_ERROR_OUT_OF_MEMORY, - * ::CUDA_ERROR_MAP_FAILED + * ::CUDA_ERROR_MAP_FAILED, + * ::CUDA_ERROR_INVALID_VALUE * - * \sa - * ::cuEventCreate, - * ::cuEventDestroy, + * \sa + * ::cuEventCreate, + * ::cuEventDestroy, * ::cuEventSynchronize, * ::cuEventQuery, * ::cuStreamWaitEvent, * ::cuIpcOpenEventHandle, * ::cuIpcGetMemHandle, * ::cuIpcOpenMemHandle, - * ::cuIpcCloseMemHandle + * ::cuIpcCloseMemHandle, + * ::cudaIpcGetEventHandle */ CUresult CUDAAPI cuIpcGetEventHandle(CUipcEventHandle *pHandle, CUevent event); /** * \brief Opens an interprocess event handle for use in the current process * - * Opens an interprocess event handle exported from another process with - * ::cuIpcGetEventHandle. This function returns a ::CUevent that behaves like - * a locally created event with the ::CU_EVENT_DISABLE_TIMING flag specified. + * Opens an interprocess event handle exported from another process with + * ::cuIpcGetEventHandle. This function returns a ::CUevent that behaves like + * a locally created event with the ::CU_EVENT_DISABLE_TIMING flag specified. * This event must be freed with ::cuEventDestroy. * - * Performing operations on the imported event after the exported event has + * Performing operations on the imported event after the exported event has * been freed with ::cuEventDestroy will result in undefined behavior. * - * IPC functionality is restricted to devices with support for unified - * addressing on Linux operating systems. + * IPC functionality is restricted to devices with support for unified + * addressing on Linux and Windows operating systems. + * IPC functionality on Windows is restricted to GPUs in TCC mode * * \param phEvent - Returns the imported event * \param handle - Interprocess handle to open @@ -4581,18 +5315,20 @@ CUresult CUDAAPI cuIpcGetEventHandle(CUipcEventHandle *pHandle, CUevent event); * ::CUDA_ERROR_INVALID_CONTEXT, * ::CUDA_ERROR_MAP_FAILED, * ::CUDA_ERROR_PEER_ACCESS_UNSUPPORTED, - * ::CUDA_ERROR_INVALID_HANDLE + * ::CUDA_ERROR_INVALID_HANDLE, + * ::CUDA_ERROR_INVALID_VALUE * * \sa - * ::cuEventCreate, - * ::cuEventDestroy, + * ::cuEventCreate, + * ::cuEventDestroy, * ::cuEventSynchronize, * ::cuEventQuery, * ::cuStreamWaitEvent, * ::cuIpcGetEventHandle, * ::cuIpcGetMemHandle, * ::cuIpcOpenMemHandle, - * ::cuIpcCloseMemHandle + * ::cuIpcCloseMemHandle, + * ::cudaIpcOpenEventHandle */ CUresult CUDAAPI cuIpcOpenEventHandle(CUevent *phEvent, CUipcEventHandle handle); @@ -4600,36 +5336,39 @@ CUresult CUDAAPI cuIpcOpenEventHandle(CUevent *phEvent, CUipcEventHandle handle) * \brief Gets an interprocess memory handle for an existing device memory * allocation * - * Takes a pointer to the base of an existing device memory allocation created - * with ::cuMemAlloc and exports it for use in another process. This is a + * Takes a pointer to the base of an existing device memory allocation created + * with ::cuMemAlloc and exports it for use in another process. This is a * lightweight operation and may be called multiple times on an allocation - * without adverse effects. + * without adverse effects. * * If a region of memory is freed with ::cuMemFree and a subsequent call * to ::cuMemAlloc returns memory with the same device address, * ::cuIpcGetMemHandle will return a unique handle for the - * new memory. + * new memory. * - * IPC functionality is restricted to devices with support for unified - * addressing on Linux operating systems. + * IPC functionality is restricted to devices with support for unified + * addressing on Linux and Windows operating systems. + * IPC functionality on Windows is restricted to GPUs in TCC mode * * \param pHandle - Pointer to user allocated ::CUipcMemHandle to return * the handle in. - * \param dptr - Base pointer to previously allocated device memory + * \param dptr - Base pointer to previously allocated device memory * * \returns * ::CUDA_SUCCESS, * ::CUDA_ERROR_INVALID_HANDLE, * ::CUDA_ERROR_OUT_OF_MEMORY, * ::CUDA_ERROR_MAP_FAILED, - * + * ::CUDA_ERROR_INVALID_VALUE + * * \sa * ::cuMemAlloc, * ::cuMemFree, * ::cuIpcGetEventHandle, * ::cuIpcOpenEventHandle, * ::cuIpcOpenMemHandle, - * ::cuIpcCloseMemHandle + * ::cuIpcCloseMemHandle, + * ::cudaIpcGetMemHandle */ CUresult CUDAAPI cuIpcGetMemHandle(CUipcMemHandle *pHandle, CUdeviceptr dptr); @@ -4638,14 +5377,17 @@ CUresult CUDAAPI cuIpcGetMemHandle(CUipcMemHandle *pHandle, CUdeviceptr dptr); * and returns a device pointer usable in the local process. * * Maps memory exported from another process with ::cuIpcGetMemHandle into - * the current device address space. For contexts on different devices + * the current device address space. For contexts on different devices * ::cuIpcOpenMemHandle can attempt to enable peer access between the - * devices as if the user called ::cuCtxEnablePeerAccess. This behavior is - * controlled by the ::CU_IPC_MEM_LAZY_ENABLE_PEER_ACCESS flag. + * devices as if the user called ::cuCtxEnablePeerAccess. This behavior is + * controlled by the ::CU_IPC_MEM_LAZY_ENABLE_PEER_ACCESS flag. * ::cuDeviceCanAccessPeer can determine if a mapping is possible. * + * ::cuIpcOpenMemHandle can open handles to devices that may not be visible + * in the process calling the API. + * * Contexts that may open ::CUipcMemHandles are restricted in the following way. - * ::CUipcMemHandles from each ::CUdevice in a given process may only be opened + * ::CUipcMemHandles from each ::CUdevice in a given process may only be opened * by one ::CUcontext per ::CUdevice per other process. * * Memory returned from ::cuIpcOpenMemHandle must be freed with @@ -4655,9 +5397,10 @@ CUresult CUDAAPI cuIpcGetMemHandle(CUipcMemHandle *pHandle, CUdeviceptr dptr); * ::cuIpcCloseMemHandle in the importing context will result in undefined * behavior. * - * IPC functionality is restricted to devices with support for unified - * addressing on Linux operating systems. - * + * IPC functionality is restricted to devices with support for unified + * addressing on Linux and Windows operating systems. + * IPC functionality on Windows is restricted to GPUs in TCC mode + * * \param pdptr - Returned device pointer * \param handle - ::CUipcMemHandle to open * \param Flags - Flags for this operation. Must be specified as ::CU_IPC_MEM_LAZY_ENABLE_PEER_ACCESS @@ -4667,9 +5410,10 @@ CUresult CUDAAPI cuIpcGetMemHandle(CUipcMemHandle *pHandle, CUdeviceptr dptr); * ::CUDA_ERROR_INVALID_CONTEXT, * ::CUDA_ERROR_MAP_FAILED, * ::CUDA_ERROR_INVALID_HANDLE, - * ::CUDA_ERROR_TOO_MANY_PEERS + * ::CUDA_ERROR_TOO_MANY_PEERS, + * ::CUDA_ERROR_INVALID_VALUE * - * \note No guarantees are made about the address returned in \p *pdptr. + * \note No guarantees are made about the address returned in \p *pdptr. * In particular, multiple processes may not receive the same address for the same \p handle. * * \sa @@ -4681,12 +5425,13 @@ CUresult CUDAAPI cuIpcGetMemHandle(CUipcMemHandle *pHandle, CUdeviceptr dptr); * ::cuIpcCloseMemHandle, * ::cuCtxEnablePeerAccess, * ::cuDeviceCanAccessPeer, + * ::cudaIpcOpenMemHandle */ CUresult CUDAAPI cuIpcOpenMemHandle(CUdeviceptr *pdptr, CUipcMemHandle handle, unsigned int Flags); /** * \brief Close memory mapped with ::cuIpcOpenMemHandle - * + * * Unmaps memory returnd by ::cuIpcOpenMemHandle. The original allocation * in the exporting process as well as imported mappings in other processes * will be unaffected. @@ -4694,17 +5439,18 @@ CUresult CUDAAPI cuIpcOpenMemHandle(CUdeviceptr *pdptr, CUipcMemHandle handle, u * Any resources used to enable peer access will be freed if this is the * last mapping using them. * - * IPC functionality is restricted to devices with support for unified - * addressing on Linux operating systems. + * IPC functionality is restricted to devices with support for unified + * addressing on Linux and Windows operating systems. + * IPC functionality on Windows is restricted to GPUs in TCC mode * * \param dptr - Device pointer returned by ::cuIpcOpenMemHandle - * + * * \returns * ::CUDA_SUCCESS, * ::CUDA_ERROR_INVALID_CONTEXT, * ::CUDA_ERROR_MAP_FAILED, * ::CUDA_ERROR_INVALID_HANDLE, - * + * ::CUDA_ERROR_INVALID_VALUE * \sa * ::cuMemAlloc, * ::cuMemFree, @@ -4712,6 +5458,7 @@ CUresult CUDAAPI cuIpcOpenMemHandle(CUdeviceptr *pdptr, CUipcMemHandle handle, u * ::cuIpcOpenEventHandle, * ::cuIpcGetMemHandle, * ::cuIpcOpenMemHandle, + * ::cudaIpcCloseMemHandle */ CUresult CUDAAPI cuIpcCloseMemHandle(CUdeviceptr dptr); @@ -4724,8 +5471,8 @@ CUresult CUDAAPI cuIpcCloseMemHandle(CUdeviceptr dptr); * Page-locks the memory range specified by \p p and \p bytesize and maps it * for the device(s) as specified by \p Flags. This memory range also is added * to the same tracking mechanism as ::cuMemHostAlloc to automatically accelerate - * calls to functions such as ::cuMemcpyHtoD(). Since the memory can be accessed - * directly by the device, it can be read or written with much higher bandwidth + * calls to functions such as ::cuMemcpyHtoD(). Since the memory can be accessed + * directly by the device, it can be read or written with much higher bandwidth * than pageable memory that has not been registered. Page-locking excessive * amounts of memory may degrade system performance, since it reduces the amount * of memory available to the system for paging. As a result, this function is @@ -4743,8 +5490,7 @@ CUresult CUDAAPI cuIpcCloseMemHandle(CUdeviceptr dptr); * * - ::CU_MEMHOSTREGISTER_DEVICEMAP: Maps the allocation into the CUDA address * space. The device pointer to the memory may be obtained by calling - * ::cuMemHostGetDevicePointer(). This feature is available only on GPUs - * with compute capability greater than or equal to 1.1. + * ::cuMemHostGetDevicePointer(). * * - ::CU_MEMHOSTREGISTER_IOMEMORY: The pointer is treated as pointing to some * I/O memory space, e.g. the PCI Express resource of a 3rd party device. @@ -4775,7 +5521,7 @@ CUresult CUDAAPI cuIpcCloseMemHandle(CUdeviceptr dptr); * on devices that have a non-zero value for the device attribute. Note however that * such devices should access the memory using only of the two pointers and not both. * - * The memory page-locked by this function must be unregistered with + * The memory page-locked by this function must be unregistered with * ::cuMemHostUnregister(). * * \param p - Host pointer to memory to page-lock @@ -4794,7 +5540,11 @@ CUresult CUDAAPI cuIpcCloseMemHandle(CUdeviceptr dptr); * ::CUDA_ERROR_NOT_SUPPORTED * \notefnerr * - * \sa ::cuMemHostUnregister, ::cuMemHostGetFlags, ::cuMemHostGetDevicePointer + * \sa + * ::cuMemHostUnregister, + * ::cuMemHostGetFlags, + * ::cuMemHostGetDevicePointer, + * ::cudaHostRegister */ CUresult CUDAAPI cuMemHostRegister(void *p, size_t bytesize, unsigned int Flags); @@ -4818,17 +5568,19 @@ CUresult CUDAAPI cuMemHostRegister(void *p, size_t bytesize, unsigned int Flags) * ::CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED, * \notefnerr * - * \sa ::cuMemHostRegister + * \sa + * ::cuMemHostRegister, + * ::cudaHostUnregister */ CUresult CUDAAPI cuMemHostUnregister(void *p); /** * \brief Copies memory * - * Copies data between two pointers. - * \p dst and \p src are base pointers of the destination and source, respectively. + * Copies data between two pointers. + * \p dst and \p src are base pointers of the destination and source, respectively. * \p ByteCount specifies the number of bytes to copy. - * Note that this function infers the type of the transfer (host to host, host to + * Note that this function infers the type of the transfer (host to host, host to * device, device to device, or device to host) from the pointer values. This * function is only allowed in contexts which support unified addressing. * @@ -4854,7 +5606,10 @@ CUresult CUDAAPI cuMemHostUnregister(void *p); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpy, + * ::cudaMemcpyToSymbol, + * ::cudaMemcpyFromSymbol */ CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount); @@ -4862,9 +5617,9 @@ CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount); * \brief Copies device memory between two contexts * * Copies from device memory in one context to device memory in another - * context. \p dstDevice is the base device pointer of the destination memory - * and \p dstContext is the destination context. \p srcDevice is the base - * device pointer of the source memory and \p srcContext is the source pointer. + * context. \p dstDevice is the base device pointer of the destination memory + * and \p dstContext is the destination context. \p srcDevice is the base + * device pointer of the source memory and \p srcContext is the source pointer. * \p ByteCount specifies the number of bytes to copy. * * \param dstDevice - Destination device pointer @@ -4883,7 +5638,8 @@ CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount); * \note_sync * * \sa ::cuMemcpyDtoD, ::cuMemcpy3DPeer, ::cuMemcpyDtoDAsync, ::cuMemcpyPeerAsync, - * ::cuMemcpy3DPeerAsync + * ::cuMemcpy3DPeerAsync, + * ::cudaMemcpyPeer */ CUresult CUDAAPI cuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount); @@ -4919,7 +5675,9 @@ CUresult CUDAAPI cuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, CUdev * ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpy, + * ::cudaMemcpyToSymbol */ CUresult CUDAAPI cuMemcpyHtoD(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount); @@ -4952,7 +5710,9 @@ CUresult CUDAAPI cuMemcpyHtoD(CUdeviceptr dstDevice, const void *srcHost, size_t * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpy, + * ::cudaMemcpyFromSymbol */ CUresult CUDAAPI cuMemcpyDtoH(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount); @@ -4985,7 +5745,10 @@ CUresult CUDAAPI cuMemcpyDtoH(void *dstHost, CUdeviceptr srcDevice, size_t ByteC * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpy, + * ::cudaMemcpyToSymbol, + * ::cudaMemcpyFromSymbol */ CUresult CUDAAPI cuMemcpyDtoD(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount); @@ -5020,7 +5783,8 @@ CUresult CUDAAPI cuMemcpyDtoD(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpyToArray */ CUresult CUDAAPI cuMemcpyDtoA(CUarray dstArray, size_t dstOffset, CUdeviceptr srcDevice, size_t ByteCount); @@ -5057,7 +5821,8 @@ CUresult CUDAAPI cuMemcpyDtoA(CUarray dstArray, size_t dstOffset, CUdeviceptr sr * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpyFromArray */ CUresult CUDAAPI cuMemcpyAtoD(CUdeviceptr dstDevice, CUarray srcArray, size_t srcOffset, size_t ByteCount); @@ -5092,7 +5857,8 @@ CUresult CUDAAPI cuMemcpyAtoD(CUdeviceptr dstDevice, CUarray srcArray, size_t sr * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpyToArray */ CUresult CUDAAPI cuMemcpyHtoA(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount); @@ -5127,7 +5893,8 @@ CUresult CUDAAPI cuMemcpyHtoA(CUarray dstArray, size_t dstOffset, const void *sr * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpyFromArray */ CUresult CUDAAPI cuMemcpyAtoH(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount); @@ -5166,7 +5933,8 @@ CUresult CUDAAPI cuMemcpyAtoH(void *dstHost, CUarray srcArray, size_t srcOffset, * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpyArrayToArray */ CUresult CUDAAPI cuMemcpyAtoA(CUarray dstArray, size_t dstOffset, CUarray srcArray, size_t srcOffset, size_t ByteCount); @@ -5211,9 +5979,9 @@ CUresult CUDAAPI cuMemcpyAtoA(CUarray dstArray, size_t dstOffset, CUarray srcArr * * \par * If ::srcMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::srcDevice and ::srcPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::srcArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::srcArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -5238,9 +6006,9 @@ CUresult CUDAAPI cuMemcpyAtoA(CUarray dstArray, size_t dstOffset, CUarray srcArr * * \par * If ::dstMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::dstDevice and ::dstPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::dstArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::dstArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -5327,7 +6095,10 @@ CUresult CUDAAPI cuMemcpyAtoA(CUarray dstArray, size_t dstOffset, CUarray srcArr * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpy2D, + * ::cudaMemcpy2DToArray, + * ::cudaMemcpy2DFromArray */ CUresult CUDAAPI cuMemcpy2D(const CUDA_MEMCPY2D *pCopy); @@ -5370,9 +6141,9 @@ CUresult CUDAAPI cuMemcpy2D(const CUDA_MEMCPY2D *pCopy); * * \par * If ::srcMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::srcDevice and ::srcPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::srcArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::srcArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -5392,9 +6163,9 @@ CUresult CUDAAPI cuMemcpy2D(const CUDA_MEMCPY2D *pCopy); * * \par * If ::dstMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::dstDevice and ::dstPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::dstArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::dstArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -5486,7 +6257,10 @@ CUresult CUDAAPI cuMemcpy2D(const CUDA_MEMCPY2D *pCopy); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpy2D, + * ::cudaMemcpy2DToArray, + * ::cudaMemcpy2DFromArray */ CUresult CUDAAPI cuMemcpy2DUnaligned(const CUDA_MEMCPY2D *pCopy); @@ -5537,9 +6311,9 @@ CUresult CUDAAPI cuMemcpy2DUnaligned(const CUDA_MEMCPY2D *pCopy); * * \par * If ::srcMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::srcDevice and ::srcPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::srcArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::srcArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -5561,9 +6335,9 @@ CUresult CUDAAPI cuMemcpy2DUnaligned(const CUDA_MEMCPY2D *pCopy); * * \par * If ::dstMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::dstDevice and ::dstPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::dstArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::dstArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -5654,7 +6428,8 @@ CUresult CUDAAPI cuMemcpy2DUnaligned(const CUDA_MEMCPY2D *pCopy); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMemcpy3D */ CUresult CUDAAPI cuMemcpy3D(const CUDA_MEMCPY3D *pCopy); #endif /* __CUDA_API_VERSION >= 3020 */ @@ -5679,17 +6454,18 @@ CUresult CUDAAPI cuMemcpy3D(const CUDA_MEMCPY3D *pCopy); * \note_sync * * \sa ::cuMemcpyDtoD, ::cuMemcpyPeer, ::cuMemcpyDtoDAsync, ::cuMemcpyPeerAsync, - * ::cuMemcpy3DPeerAsync + * ::cuMemcpy3DPeerAsync, + * ::cudaMemcpy3DPeer */ CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy); /** * \brief Copies memory asynchronously * - * Copies data between two pointers. - * \p dst and \p src are base pointers of the destination and source, respectively. + * Copies data between two pointers. + * \p dst and \p src are base pointers of the destination and source, respectively. * \p ByteCount specifies the number of bytes to copy. - * Note that this function infers the type of the transfer (host to host, host to + * Note that this function infers the type of the transfer (host to host, host to * device, device to device, or device to host) from the pointer values. This * function is only allowed in contexts which support unified addressing. * @@ -5703,7 +6479,8 @@ CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy); * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream @@ -5719,7 +6496,10 @@ CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy); * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemcpyAsync, + * ::cudaMemcpyToSymbolAsync, + * ::cudaMemcpyFromSymbolAsync */ CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount, CUstream hStream); @@ -5727,9 +6507,9 @@ CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, size_t ByteCoun * \brief Copies device memory between two contexts asynchronously. * * Copies from device memory in one context to device memory in another - * context. \p dstDevice is the base device pointer of the destination memory - * and \p dstContext is the destination context. \p srcDevice is the base - * device pointer of the source memory and \p srcContext is the source pointer. + * context. \p dstDevice is the base device pointer of the destination memory + * and \p dstContext is the destination context. \p srcDevice is the base + * device pointer of the source memory and \p srcContext is the source pointer. * \p ByteCount specifies the number of bytes to copy. * * \param dstDevice - Destination device pointer @@ -5744,13 +6524,15 @@ CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, size_t ByteCoun * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream * - * \sa ::cuMemcpyDtoD, ::cuMemcpyPeer, ::cuMemcpy3DPeer, ::cuMemcpyDtoDAsync, - * ::cuMemcpy3DPeerAsync + * \sa ::cuMemcpyDtoD, ::cuMemcpyPeer, ::cuMemcpy3DPeer, ::cuMemcpyDtoDAsync, + * ::cuMemcpy3DPeerAsync, + * ::cudaMemcpyPeerAsync */ CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount, CUstream hStream); #endif /* __CUDA_API_VERSION >= 4000 */ @@ -5773,7 +6555,8 @@ CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream @@ -5789,7 +6572,9 @@ CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemcpyAsync, + * ::cudaMemcpyToSymbolAsync */ CUresult CUDAAPI cuMemcpyHtoDAsync(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount, CUstream hStream); @@ -5810,7 +6595,8 @@ CUresult CUDAAPI cuMemcpyHtoDAsync(CUdeviceptr dstDevice, const void *srcHost, s * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream @@ -5826,7 +6612,9 @@ CUresult CUDAAPI cuMemcpyHtoDAsync(CUdeviceptr dstDevice, const void *srcHost, s * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemcpyAsync, + * ::cudaMemcpyFromSymbolAsync */ CUresult CUDAAPI cuMemcpyDtoHAsync(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream); @@ -5847,7 +6635,8 @@ CUresult CUDAAPI cuMemcpyDtoHAsync(void *dstHost, CUdeviceptr srcDevice, size_t * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream @@ -5863,7 +6652,10 @@ CUresult CUDAAPI cuMemcpyDtoHAsync(void *dstHost, CUdeviceptr srcDevice, size_t * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemcpyAsync, + * ::cudaMemcpyToSymbolAsync, + * ::cudaMemcpyFromSymbolAsync */ CUresult CUDAAPI cuMemcpyDtoDAsync(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream); @@ -5886,7 +6678,8 @@ CUresult CUDAAPI cuMemcpyDtoDAsync(CUdeviceptr dstDevice, CUdeviceptr srcDevice, * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream @@ -5902,7 +6695,8 @@ CUresult CUDAAPI cuMemcpyDtoDAsync(CUdeviceptr dstDevice, CUdeviceptr srcDevice, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemcpyToArrayAsync */ CUresult CUDAAPI cuMemcpyHtoAAsync(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount, CUstream hStream); @@ -5925,7 +6719,8 @@ CUresult CUDAAPI cuMemcpyHtoAAsync(CUarray dstArray, size_t dstOffset, const voi * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream @@ -5941,7 +6736,8 @@ CUresult CUDAAPI cuMemcpyHtoAAsync(CUarray dstArray, size_t dstOffset, const voi * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemcpyFromArrayAsync */ CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount, CUstream hStream); @@ -5989,9 +6785,9 @@ CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, size_t srcOf * * \par * If ::srcMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::srcDevice and ::srcPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::srcArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::srcArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -6006,9 +6802,9 @@ CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, size_t srcOf * * \par * If ::dstMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::dstDevice and ::dstPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::dstArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::dstArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -6090,7 +6886,8 @@ CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, size_t srcOf * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream @@ -6106,7 +6903,10 @@ CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, size_t srcOf * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemcpy2DAsync, + * ::cudaMemcpy2DToArrayAsync, + * ::cudaMemcpy2DFromArrayAsync */ CUresult CUDAAPI cuMemcpy2DAsync(const CUDA_MEMCPY2D *pCopy, CUstream hStream); @@ -6157,9 +6957,9 @@ CUresult CUDAAPI cuMemcpy2DAsync(const CUDA_MEMCPY2D *pCopy, CUstream hStream); * * \par * If ::srcMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::srcDevice and ::srcPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::srcArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::srcArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -6181,9 +6981,9 @@ CUresult CUDAAPI cuMemcpy2DAsync(const CUDA_MEMCPY2D *pCopy, CUstream hStream); * * \par * If ::dstMemoryType is ::CU_MEMORYTYPE_UNIFIED, ::dstDevice and ::dstPitch - * specify the (unified virtual address space) base address of the source data - * and the bytes per row to apply. ::dstArray is ignored. - * This value may be used only if unified addressing is supported in the calling + * specify the (unified virtual address space) base address of the source data + * and the bytes per row to apply. ::dstArray is ignored. + * This value may be used only if unified addressing is supported in the calling * context. * * \par @@ -6262,7 +7062,8 @@ CUresult CUDAAPI cuMemcpy2DAsync(const CUDA_MEMCPY2D *pCopy, CUstream hStream); * ::CUDA_ERROR_DEINITIALIZED, * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, - * ::CUDA_ERROR_INVALID_VALUE + * ::CUDA_ERROR_INVALID_VALUE, + * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * \note_async * \note_null_stream @@ -6278,7 +7079,8 @@ CUresult CUDAAPI cuMemcpy2DAsync(const CUDA_MEMCPY2D *pCopy, CUstream hStream); * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemcpy3DAsync */ CUresult CUDAAPI cuMemcpy3DAsync(const CUDA_MEMCPY3D *pCopy, CUstream hStream); #endif /* __CUDA_API_VERSION >= 3020 */ @@ -6305,7 +7107,8 @@ CUresult CUDAAPI cuMemcpy3DAsync(const CUDA_MEMCPY3D *pCopy, CUstream hStream); * \note_null_stream * * \sa ::cuMemcpyDtoD, ::cuMemcpyPeer, ::cuMemcpyDtoDAsync, ::cuMemcpyPeerAsync, - * ::cuMemcpy3DPeerAsync + * ::cuMemcpy3DPeerAsync, + * ::cudaMemcpy3DPeerAsync */ CUresult CUDAAPI cuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER *pCopy, CUstream hStream); #endif /* __CUDA_API_VERSION >= 4000 */ @@ -6341,7 +7144,8 @@ CUresult CUDAAPI cuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER *pCopy, CUstream h * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemset */ CUresult CUDAAPI cuMemsetD8(CUdeviceptr dstDevice, unsigned char uc, size_t N); @@ -6375,7 +7179,8 @@ CUresult CUDAAPI cuMemsetD8(CUdeviceptr dstDevice, unsigned char uc, size_t N); * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemset */ CUresult CUDAAPI cuMemsetD16(CUdeviceptr dstDevice, unsigned short us, size_t N); @@ -6409,7 +7214,8 @@ CUresult CUDAAPI cuMemsetD16(CUdeviceptr dstDevice, unsigned short us, size_t N) * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32Async + * ::cuMemsetD32Async, + * ::cudaMemset */ CUresult CUDAAPI cuMemsetD32(CUdeviceptr dstDevice, unsigned int ui, size_t N); @@ -6448,7 +7254,8 @@ CUresult CUDAAPI cuMemsetD32(CUdeviceptr dstDevice, unsigned int ui, size_t N); * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemset2D */ CUresult CUDAAPI cuMemsetD2D8(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height); @@ -6488,7 +7295,8 @@ CUresult CUDAAPI cuMemsetD2D8(CUdeviceptr dstDevice, size_t dstPitch, unsigned c * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemset2D */ CUresult CUDAAPI cuMemsetD2D16(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height); @@ -6528,7 +7336,8 @@ CUresult CUDAAPI cuMemsetD2D16(CUdeviceptr dstDevice, size_t dstPitch, unsigned * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemset2D */ CUresult CUDAAPI cuMemsetD2D32(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height); @@ -6564,7 +7373,8 @@ CUresult CUDAAPI cuMemsetD2D32(CUdeviceptr dstDevice, size_t dstPitch, unsigned * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemsetAsync */ CUresult CUDAAPI cuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, size_t N, CUstream hStream); @@ -6600,7 +7410,8 @@ CUresult CUDAAPI cuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, size_t * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemsetAsync */ CUresult CUDAAPI cuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, size_t N, CUstream hStream); @@ -6635,7 +7446,8 @@ CUresult CUDAAPI cuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, size * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, - * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, ::cuMemsetD32 + * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, ::cuMemsetD32, + * ::cudaMemsetAsync */ CUresult CUDAAPI cuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, size_t N, CUstream hStream); @@ -6676,7 +7488,8 @@ CUresult CUDAAPI cuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, size_t * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemset2DAsync */ CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height, CUstream hStream); @@ -6685,7 +7498,7 @@ CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, unsig * * Sets the 2D memory range of \p Width 16-bit values to the specified value * \p us. \p Height specifies the number of rows to set, and \p dstPitch - * specifies the number of bytes between each row. The \p dstDevice pointer + * specifies the number of bytes between each row. The \p dstDevice pointer * and \p dstPitch offset must be two byte aligned. This function performs * fastest when the pitch is one that has been passed back by * ::cuMemAllocPitch(). @@ -6718,7 +7531,8 @@ CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, unsig * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D32, ::cuMemsetD2D32Async, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemset2DAsync */ CUresult CUDAAPI cuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height, CUstream hStream); @@ -6760,7 +7574,8 @@ CUresult CUDAAPI cuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, unsi * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D8Async, * ::cuMemsetD2D16, ::cuMemsetD2D16Async, ::cuMemsetD2D32, * ::cuMemsetD8, ::cuMemsetD8Async, ::cuMemsetD16, ::cuMemsetD16Async, - * ::cuMemsetD32, ::cuMemsetD32Async + * ::cuMemsetD32, ::cuMemsetD32Async, + * ::cudaMemset2DAsync */ CUresult CUDAAPI cuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height, CUstream hStream); @@ -6863,7 +7678,8 @@ CUresult CUDAAPI cuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, unsi * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMallocArray */ CUresult CUDAAPI cuArrayCreate(CUarray *pHandle, const CUDA_ARRAY_DESCRIPTOR *pAllocateArray); @@ -6896,7 +7712,8 @@ CUresult CUDAAPI cuArrayCreate(CUarray *pHandle, const CUDA_ARRAY_DESCRIPTOR *pA * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaArrayGetInfo */ CUresult CUDAAPI cuArrayGetDescriptor(CUDA_ARRAY_DESCRIPTOR *pArrayDescriptor, CUarray hArray); #endif /* __CUDA_API_VERSION >= 3020 */ @@ -6915,7 +7732,8 @@ CUresult CUDAAPI cuArrayGetDescriptor(CUDA_ARRAY_DESCRIPTOR *pArrayDescriptor, C * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, * ::CUDA_ERROR_INVALID_HANDLE, - * ::CUDA_ERROR_ARRAY_IS_MAPPED + * ::CUDA_ERROR_ARRAY_IS_MAPPED, + * ::CUDA_ERROR_CONTEXT_IS_DESTROYED * \notefnerr * * \sa ::cuArray3DCreate, ::cuArray3DGetDescriptor, ::cuArrayCreate, @@ -6927,7 +7745,8 @@ CUresult CUDAAPI cuArrayGetDescriptor(CUDA_ARRAY_DESCRIPTOR *pArrayDescriptor, C * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaFreeArray */ CUresult CUDAAPI cuArrayDestroy(CUarray hArray); @@ -6956,22 +7775,22 @@ CUresult CUDAAPI cuArrayDestroy(CUarray hArray); * - A 1D array is allocated if \p Height and \p Depth extents are both zero. * - A 2D array is allocated if only \p Depth extent is zero. * - A 3D array is allocated if all three extents are non-zero. - * - A 1D layered CUDA array is allocated if only \p Height is zero and the - * ::CUDA_ARRAY3D_LAYERED flag is set. Each layer is a 1D array. The number + * - A 1D layered CUDA array is allocated if only \p Height is zero and the + * ::CUDA_ARRAY3D_LAYERED flag is set. Each layer is a 1D array. The number * of layers is determined by the depth extent. - * - A 2D layered CUDA array is allocated if all three extents are non-zero and - * the ::CUDA_ARRAY3D_LAYERED flag is set. Each layer is a 2D array. The number + * - A 2D layered CUDA array is allocated if all three extents are non-zero and + * the ::CUDA_ARRAY3D_LAYERED flag is set. Each layer is a 2D array. The number * of layers is determined by the depth extent. * - A cubemap CUDA array is allocated if all three extents are non-zero and the - * ::CUDA_ARRAY3D_CUBEMAP flag is set. \p Width must be equal to \p Height, and - * \p Depth must be six. A cubemap is a special type of 2D layered CUDA array, - * where the six layers represent the six faces of a cube. The order of the six + * ::CUDA_ARRAY3D_CUBEMAP flag is set. \p Width must be equal to \p Height, and + * \p Depth must be six. A cubemap is a special type of 2D layered CUDA array, + * where the six layers represent the six faces of a cube. The order of the six * layers in memory is the same as that listed in ::CUarray_cubemap_face. - * - A cubemap layered CUDA array is allocated if all three extents are non-zero, - * and both, ::CUDA_ARRAY3D_CUBEMAP and ::CUDA_ARRAY3D_LAYERED flags are set. - * \p Width must be equal to \p Height, and \p Depth must be a multiple of six. - * A cubemap layered CUDA array is a special type of 2D layered CUDA array that - * consists of a collection of cubemaps. The first six layers represent the first + * - A cubemap layered CUDA array is allocated if all three extents are non-zero, + * and both, ::CUDA_ARRAY3D_CUBEMAP and ::CUDA_ARRAY3D_LAYERED flags are set. + * \p Width must be equal to \p Height, and \p Depth must be a multiple of six. + * A cubemap layered CUDA array is a special type of 2D layered CUDA array that + * consists of a collection of cubemaps. The first six layers represent the first * cubemap, the next six layers form the second cubemap, and so on. * * - ::Format specifies the format of the elements; ::CUarray_format is @@ -6992,11 +7811,11 @@ CUresult CUDAAPI cuArrayDestroy(CUarray hArray); * - \p NumChannels specifies the number of packed components per CUDA array * element; it may be 1, 2, or 4; * - * - ::Flags may be set to - * - ::CUDA_ARRAY3D_LAYERED to enable creation of layered CUDA arrays. If this flag is set, + * - ::Flags may be set to + * - ::CUDA_ARRAY3D_LAYERED to enable creation of layered CUDA arrays. If this flag is set, * \p Depth specifies the number of layers, not the depth of a 3D array. - * - ::CUDA_ARRAY3D_SURFACE_LDST to enable surface references to be bound to the CUDA array. - * If this flag is not set, ::cuSurfRefSetArray will fail when attempting to bind the CUDA array + * - ::CUDA_ARRAY3D_SURFACE_LDST to enable surface references to be bound to the CUDA array. + * If this flag is not set, ::cuSurfRefSetArray will fail when attempting to bind the CUDA array * to a surface reference. * - ::CUDA_ARRAY3D_CUBEMAP to enable creation of cubemaps. If this flag is set, \p Width must be * equal to \p Height, and \p Depth must be six. If the ::CUDA_ARRAY3D_LAYERED flag is also set, @@ -7004,20 +7823,20 @@ CUresult CUDAAPI cuArrayDestroy(CUarray hArray); * - ::CUDA_ARRAY3D_TEXTURE_GATHER to indicate that the CUDA array will be used for texture gather. * Texture gather can only be performed on 2D CUDA arrays. * - * \p Width, \p Height and \p Depth must meet certain size requirements as listed in the following table. - * All values are specified in elements. Note that for brevity's sake, the full name of the device attribute - * is not specified. For ex., TEXTURE1D_WIDTH refers to the device attribute + * \p Width, \p Height and \p Depth must meet certain size requirements as listed in the following table. + * All values are specified in elements. Note that for brevity's sake, the full name of the device attribute + * is not specified. For ex., TEXTURE1D_WIDTH refers to the device attribute * ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_WIDTH. * - * Note that 2D CUDA arrays have different size requirements if the ::CUDA_ARRAY3D_TEXTURE_GATHER flag - * is set. \p Width and \p Height must not be greater than ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_GATHER_WIDTH + * Note that 2D CUDA arrays have different size requirements if the ::CUDA_ARRAY3D_TEXTURE_GATHER flag + * is set. \p Width and \p Height must not be greater than ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_GATHER_WIDTH * and ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_GATHER_HEIGHT respectively, in that case. * * * - * - * * * @@ -7027,28 +7846,28 @@ CUresult CUDAAPI cuArrayDestroy(CUarray hArray); * * * - * * - * - * * - * - * * * - * * - * - * *
CUDA array typeValid extents that must always be met
{(width range in elements), (height range), + * 		Valid extents that must always be met
{(width range in elements), (height range), * (depth range)}		Valid extents with CUDA_ARRAY3D_SURFACE_LDST set
+ * 		Valid extents with CUDA_ARRAY3D_SURFACE_LDST set
* {(width range in elements), (height range), (depth range)}
1D{ (1,TEXTURE1D_WIDTH), 0, 0 }{ (1,SURFACE2D_WIDTH), (1,SURFACE2D_HEIGHT), 0 }
3D{ (1,TEXTURE3D_WIDTH), (1,TEXTURE3D_HEIGHT), (1,TEXTURE3D_DEPTH) } - *
OR
{ (1,TEXTURE3D_WIDTH_ALTERNATE), (1,TEXTURE3D_HEIGHT_ALTERNATE), + *
OR
{ (1,TEXTURE3D_WIDTH_ALTERNATE), (1,TEXTURE3D_HEIGHT_ALTERNATE), * (1,TEXTURE3D_DEPTH_ALTERNATE) }		{ (1,SURFACE3D_WIDTH), (1,SURFACE3D_HEIGHT), + * { (1,SURFACE3D_WIDTH), (1,SURFACE3D_HEIGHT), * (1,SURFACE3D_DEPTH) }
1D Layered{ (1,TEXTURE1D_LAYERED_WIDTH), 0, + * { (1,TEXTURE1D_LAYERED_WIDTH), 0, * (1,TEXTURE1D_LAYERED_LAYERS) }{ (1,SURFACE1D_LAYERED_WIDTH), 0, + * { (1,SURFACE1D_LAYERED_WIDTH), 0, * (1,SURFACE1D_LAYERED_LAYERS) }
2D Layered{ (1,TEXTURE2D_LAYERED_WIDTH), (1,TEXTURE2D_LAYERED_HEIGHT), + * { (1,TEXTURE2D_LAYERED_WIDTH), (1,TEXTURE2D_LAYERED_HEIGHT), * (1,TEXTURE2D_LAYERED_LAYERS) }{ (1,SURFACE2D_LAYERED_WIDTH), (1,SURFACE2D_LAYERED_HEIGHT), + * { (1,SURFACE2D_LAYERED_WIDTH), (1,SURFACE2D_LAYERED_HEIGHT), * (1,SURFACE2D_LAYERED_LAYERS) }
Cubemap{ (1,TEXTURECUBEMAP_WIDTH), (1,TEXTURECUBEMAP_WIDTH), 6 }{ (1,SURFACECUBEMAP_WIDTH), + * { (1,SURFACECUBEMAP_WIDTH), * (1,SURFACECUBEMAP_WIDTH), 6 }
Cubemap Layered{ (1,TEXTURECUBEMAP_LAYERED_WIDTH), (1,TEXTURECUBEMAP_LAYERED_WIDTH), + * { (1,TEXTURECUBEMAP_LAYERED_WIDTH), (1,TEXTURECUBEMAP_LAYERED_WIDTH), * (1,TEXTURECUBEMAP_LAYERED_LAYERS) }{ (1,SURFACECUBEMAP_LAYERED_WIDTH), (1,SURFACECUBEMAP_LAYERED_WIDTH), + * { (1,SURFACECUBEMAP_LAYERED_WIDTH), (1,SURFACECUBEMAP_LAYERED_WIDTH), * (1,SURFACECUBEMAP_LAYERED_LAYERS) }
* @@ -7107,7 +7926,8 @@ CUresult CUDAAPI cuArrayDestroy(CUarray hArray); * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaMalloc3DArray */ CUresult CUDAAPI cuArray3DCreate(CUarray *pHandle, const CUDA_ARRAY3D_DESCRIPTOR *pAllocateArray); @@ -7131,7 +7951,8 @@ CUresult CUDAAPI cuArray3DCreate(CUarray *pHandle, const CUDA_ARRAY3D_DESCRIPTOR * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, * ::CUDA_ERROR_INVALID_VALUE, - * ::CUDA_ERROR_INVALID_HANDLE + * ::CUDA_ERROR_INVALID_HANDLE, + * ::CUDA_ERROR_CONTEXT_IS_DESTROYED * \notefnerr * * \sa ::cuArray3DCreate, ::cuArrayCreate, @@ -7143,7 +7964,8 @@ CUresult CUDAAPI cuArray3DCreate(CUarray *pHandle, const CUDA_ARRAY3D_DESCRIPTOR * ::cuMemcpyHtoD, ::cuMemcpyHtoDAsync, ::cuMemFree, ::cuMemFreeHost, * ::cuMemGetAddressRange, ::cuMemGetInfo, ::cuMemHostAlloc, * ::cuMemHostGetDevicePointer, ::cuMemsetD2D8, ::cuMemsetD2D16, - * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32 + * ::cuMemsetD2D32, ::cuMemsetD8, ::cuMemsetD16, ::cuMemsetD32, + * ::cudaArrayGetInfo */ CUresult CUDAAPI cuArray3DGetDescriptor(CUDA_ARRAY3D_DESCRIPTOR *pArrayDescriptor, CUarray hArray); #endif /* __CUDA_API_VERSION >= 3020 */ @@ -7177,22 +7999,22 @@ CUresult CUDAAPI cuArray3DGetDescriptor(CUDA_ARRAY3D_DESCRIPTOR *pArrayDescripto * - A 1D mipmapped array is allocated if \p Height and \p Depth extents are both zero. * - A 2D mipmapped array is allocated if only \p Depth extent is zero. * - A 3D mipmapped array is allocated if all three extents are non-zero. - * - A 1D layered CUDA mipmapped array is allocated if only \p Height is zero and the - * ::CUDA_ARRAY3D_LAYERED flag is set. Each layer is a 1D array. The number + * - A 1D layered CUDA mipmapped array is allocated if only \p Height is zero and the + * ::CUDA_ARRAY3D_LAYERED flag is set. Each layer is a 1D array. The number * of layers is determined by the depth extent. - * - A 2D layered CUDA mipmapped array is allocated if all three extents are non-zero and - * the ::CUDA_ARRAY3D_LAYERED flag is set. Each layer is a 2D array. The number + * - A 2D layered CUDA mipmapped array is allocated if all three extents are non-zero and + * the ::CUDA_ARRAY3D_LAYERED flag is set. Each layer is a 2D array. The number * of layers is determined by the depth extent. * - A cubemap CUDA mipmapped array is allocated if all three extents are non-zero and the - * ::CUDA_ARRAY3D_CUBEMAP flag is set. \p Width must be equal to \p Height, and - * \p Depth must be six. A cubemap is a special type of 2D layered CUDA array, - * where the six layers represent the six faces of a cube. The order of the six + * ::CUDA_ARRAY3D_CUBEMAP flag is set. \p Width must be equal to \p Height, and + * \p Depth must be six. A cubemap is a special type of 2D layered CUDA array, + * where the six layers represent the six faces of a cube. The order of the six * layers in memory is the same as that listed in ::CUarray_cubemap_face. - * - A cubemap layered CUDA mipmapped array is allocated if all three extents are non-zero, - * and both, ::CUDA_ARRAY3D_CUBEMAP and ::CUDA_ARRAY3D_LAYERED flags are set. - * \p Width must be equal to \p Height, and \p Depth must be a multiple of six. - * A cubemap layered CUDA array is a special type of 2D layered CUDA array that - * consists of a collection of cubemaps. The first six layers represent the first + * - A cubemap layered CUDA mipmapped array is allocated if all three extents are non-zero, + * and both, ::CUDA_ARRAY3D_CUBEMAP and ::CUDA_ARRAY3D_LAYERED flags are set. + * \p Width must be equal to \p Height, and \p Depth must be a multiple of six. + * A cubemap layered CUDA array is a special type of 2D layered CUDA array that + * consists of a collection of cubemaps. The first six layers represent the first * cubemap, the next six layers form the second cubemap, and so on. * * - ::Format specifies the format of the elements; ::CUarray_format is @@ -7213,11 +8035,11 @@ CUresult CUDAAPI cuArray3DGetDescriptor(CUDA_ARRAY3D_DESCRIPTOR *pArrayDescripto * - \p NumChannels specifies the number of packed components per CUDA array * element; it may be 1, 2, or 4; * - * - ::Flags may be set to - * - ::CUDA_ARRAY3D_LAYERED to enable creation of layered CUDA mipmapped arrays. If this flag is set, + * - ::Flags may be set to + * - ::CUDA_ARRAY3D_LAYERED to enable creation of layered CUDA mipmapped arrays. If this flag is set, * \p Depth specifies the number of layers, not the depth of a 3D array. * - ::CUDA_ARRAY3D_SURFACE_LDST to enable surface references to be bound to individual mipmap levels of - * the CUDA mipmapped array. If this flag is not set, ::cuSurfRefSetArray will fail when attempting to + * the CUDA mipmapped array. If this flag is not set, ::cuSurfRefSetArray will fail when attempting to * bind a mipmap level of the CUDA mipmapped array to a surface reference. * - ::CUDA_ARRAY3D_CUBEMAP to enable creation of mipmapped cubemaps. If this flag is set, \p Width must be * equal to \p Height, and \p Depth must be six. If the ::CUDA_ARRAY3D_LAYERED flag is also set, @@ -7225,34 +8047,48 @@ CUresult CUDAAPI cuArray3DGetDescriptor(CUDA_ARRAY3D_DESCRIPTOR *pArrayDescripto * - ::CUDA_ARRAY3D_TEXTURE_GATHER to indicate that the CUDA mipmapped array will be used for texture gather. * Texture gather can only be performed on 2D CUDA mipmapped arrays. * - * \p Width, \p Height and \p Depth must meet certain size requirements as listed in the following table. - * All values are specified in elements. Note that for brevity's sake, the full name of the device attribute - * is not specified. For ex., TEXTURE1D_MIPMAPPED_WIDTH refers to the device attribute + * \p Width, \p Height and \p Depth must meet certain size requirements as listed in the following table. + * All values are specified in elements. Note that for brevity's sake, the full name of the device attribute + * is not specified. For ex., TEXTURE1D_MIPMAPPED_WIDTH refers to the device attribute * ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_MIPMAPPED_WIDTH. * * * - * + * + * * - * + * + * * - * + * + * * * + *
OR
{ (1,TEXTURE3D_WIDTH_ALTERNATE), (1,TEXTURE3D_HEIGHT_ALTERNATE), + * (1,TEXTURE3D_DEPTH_ALTERNATE) } + * * - * + * + * * - * + * + * * - * + * + * * - * + * + * *
CUDA array typeValid extents that must always be met
{(width range in elements), (height range), - * (depth range)}
Valid extents that must always be met
{(width range in elements), (height range), + * (depth range)}		Valid extents with CUDA_ARRAY3D_SURFACE_LDST set
+ * {(width range in elements), (height range), (depth range)}
1D{ (1,TEXTURE1D_MIPMAPPED_WIDTH), 0, 0 }
{ (1,TEXTURE1D_MIPMAPPED_WIDTH), 0, 0 }{ (1,SURFACE1D_WIDTH), 0, 0 }
2D{ (1,TEXTURE2D_MIPMAPPED_WIDTH), (1,TEXTURE2D_MIPMAPPED_HEIGHT), 0 }
{ (1,TEXTURE2D_MIPMAPPED_WIDTH), (1,TEXTURE2D_MIPMAPPED_HEIGHT), 0 }{ (1,SURFACE2D_WIDTH), (1,SURFACE2D_HEIGHT), 0 }
3D{ (1,TEXTURE3D_WIDTH), (1,TEXTURE3D_HEIGHT), (1,TEXTURE3D_DEPTH) } - *
OR
{ (1,TEXTURE3D_WIDTH_ALTERNATE), (1,TEXTURE3D_HEIGHT_ALTERNATE), - * (1,TEXTURE3D_DEPTH_ALTERNATE) }
{ (1,SURFACE3D_WIDTH), (1,SURFACE3D_HEIGHT), + * (1,SURFACE3D_DEPTH) }
1D Layered{ (1,TEXTURE1D_LAYERED_WIDTH), 0, - * (1,TEXTURE1D_LAYERED_LAYERS) }
{ (1,TEXTURE1D_LAYERED_WIDTH), 0, + * (1,TEXTURE1D_LAYERED_LAYERS) }{ (1,SURFACE1D_LAYERED_WIDTH), 0, + * (1,SURFACE1D_LAYERED_LAYERS) }
2D Layered{ (1,TEXTURE2D_LAYERED_WIDTH), (1,TEXTURE2D_LAYERED_HEIGHT), - * (1,TEXTURE2D_LAYERED_LAYERS) }
{ (1,TEXTURE2D_LAYERED_WIDTH), (1,TEXTURE2D_LAYERED_HEIGHT), + * (1,TEXTURE2D_LAYERED_LAYERS) }{ (1,SURFACE2D_LAYERED_WIDTH), (1,SURFACE2D_LAYERED_HEIGHT), + * (1,SURFACE2D_LAYERED_LAYERS) }
Cubemap{ (1,TEXTURECUBEMAP_WIDTH), (1,TEXTURECUBEMAP_WIDTH), 6 }
{ (1,TEXTURECUBEMAP_WIDTH), (1,TEXTURECUBEMAP_WIDTH), 6 }{ (1,SURFACECUBEMAP_WIDTH), + * (1,SURFACECUBEMAP_WIDTH), 6 }
Cubemap Layered{ (1,TEXTURECUBEMAP_LAYERED_WIDTH), (1,TEXTURECUBEMAP_LAYERED_WIDTH), - * (1,TEXTURECUBEMAP_LAYERED_LAYERS) }
{ (1,TEXTURECUBEMAP_LAYERED_WIDTH), (1,TEXTURECUBEMAP_LAYERED_WIDTH), + * (1,TEXTURECUBEMAP_LAYERED_LAYERS) }{ (1,SURFACECUBEMAP_LAYERED_WIDTH), (1,SURFACECUBEMAP_LAYERED_WIDTH), + * (1,SURFACECUBEMAP_LAYERED_LAYERS) }
* * @@ -7270,7 +8106,11 @@ CUresult CUDAAPI cuArray3DGetDescriptor(CUDA_ARRAY3D_DESCRIPTOR *pArrayDescripto * ::CUDA_ERROR_UNKNOWN * \notefnerr * - * \sa ::cuMipmappedArrayDestroy, ::cuMipmappedArrayGetLevel, ::cuArrayCreate, + * \sa + * ::cuMipmappedArrayDestroy, + * ::cuMipmappedArrayGetLevel, + * ::cuArrayCreate, + * ::cudaMallocMipmappedArray */ CUresult CUDAAPI cuMipmappedArrayCreate(CUmipmappedArray *pHandle, const CUDA_ARRAY3D_DESCRIPTOR *pMipmappedArrayDesc, unsigned int numMipmapLevels); @@ -7296,7 +8136,11 @@ CUresult CUDAAPI cuMipmappedArrayCreate(CUmipmappedArray *pHandle, const CUDA_AR * ::CUDA_ERROR_INVALID_HANDLE * \notefnerr * - * \sa ::cuMipmappedArrayCreate, ::cuMipmappedArrayDestroy, ::cuArrayCreate, + * \sa + * ::cuMipmappedArrayCreate, + * ::cuMipmappedArrayDestroy, + * ::cuArrayCreate, + * ::cudaGetMipmappedArrayLevel */ CUresult CUDAAPI cuMipmappedArrayGetLevel(CUarray *pLevelArray, CUmipmappedArray hMipmappedArray, unsigned int level); @@ -7313,10 +8157,15 @@ CUresult CUDAAPI cuMipmappedArrayGetLevel(CUarray *pLevelArray, CUmipmappedArray * ::CUDA_ERROR_NOT_INITIALIZED, * ::CUDA_ERROR_INVALID_CONTEXT, * ::CUDA_ERROR_INVALID_HANDLE, - * ::CUDA_ERROR_ARRAY_IS_MAPPED + * ::CUDA_ERROR_ARRAY_IS_MAPPED, + * ::CUDA_ERROR_CONTEXT_IS_DESTROYED * \notefnerr * - * \sa ::cuMipmappedArrayCreate, ::cuMipmappedArrayGetLevel, ::cuArrayCreate, + * \sa + * ::cuMipmappedArrayCreate, + * ::cuMipmappedArrayGetLevel, + * ::cuArrayCreate, + * ::cudaFreeMipmappedArray */ CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray); @@ -7330,43 +8179,42 @@ CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray); * ___MANBRIEF___ unified addressing functions of the low-level CUDA driver * API (___CURRENT_FILE___) ___ENDMANBRIEF___ * - * This section describes the unified addressing functions of the + * This section describes the unified addressing functions of the * low-level CUDA driver application programming interface. * * @{ * * \section CUDA_UNIFIED_overview Overview * - * CUDA devices can share a unified address space with the host. + * CUDA devices can share a unified address space with the host. * For these devices there is no distinction between a device - * pointer and a host pointer -- the same pointer value may be - * used to access memory from the host program and from a kernel + * pointer and a host pointer -- the same pointer value may be + * used to access memory from the host program and from a kernel * running on the device (with exceptions enumerated below). * * \section CUDA_UNIFIED_support Supported Platforms - * - * Whether or not a device supports unified addressing may be - * queried by calling ::cuDeviceGetAttribute() with the device + * + * Whether or not a device supports unified addressing may be + * queried by calling ::cuDeviceGetAttribute() with the device * attribute ::CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING. * - * Unified addressing is automatically enabled in 64-bit processes - * on devices with compute capability greater than or equal to 2.0. + * Unified addressing is automatically enabled in 64-bit processes * * \section CUDA_UNIFIED_lookup Looking Up Information from Pointer Values * - * It is possible to look up information about the memory which backs a + * It is possible to look up information about the memory which backs a * pointer value. For instance, one may want to know if a pointer points - * to host or device memory. As another example, in the case of device - * memory, one may want to know on which CUDA device the memory - * resides. These properties may be queried using the function + * to host or device memory. As another example, in the case of device + * memory, one may want to know on which CUDA device the memory + * resides. These properties may be queried using the function * ::cuPointerGetAttribute() * * Since pointers are unique, it is not necessary to specify information - * about the pointers specified to the various copy functions in the + * about the pointers specified to the various copy functions in the * CUDA API. The function ::cuMemcpy() may be used to perform a copy * between two pointers, ignoring whether they point to host or device * memory (making ::cuMemcpyHtoD(), ::cuMemcpyDtoD(), and ::cuMemcpyDtoH() - * unnecessary for devices supporting unified addressing). For + * unnecessary for devices supporting unified addressing). For * multidimensional copies, the memory type ::CU_MEMORYTYPE_UNIFIED may be * used to specify that the CUDA driver should infer the location of the * pointer from its value. @@ -7375,45 +8223,45 @@ CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray); * * All host memory allocated in all contexts using ::cuMemAllocHost() and * ::cuMemHostAlloc() is always directly accessible from all contexts on - * all devices that support unified addressing. This is the case regardless + * all devices that support unified addressing. This is the case regardless * of whether or not the flags ::CU_MEMHOSTALLOC_PORTABLE and * ::CU_MEMHOSTALLOC_DEVICEMAP are specified. * - * The pointer value through which allocated host memory may be accessed - * in kernels on all devices that support unified addressing is the same + * The pointer value through which allocated host memory may be accessed + * in kernels on all devices that support unified addressing is the same * as the pointer value through which that memory is accessed on the host, - * so it is not necessary to call ::cuMemHostGetDevicePointer() to get the device + * so it is not necessary to call ::cuMemHostGetDevicePointer() to get the device * pointer for these allocations. - * + * * Note that this is not the case for memory allocated using the flag * ::CU_MEMHOSTALLOC_WRITECOMBINED, as discussed below. * * \section CUDA_UNIFIED_autopeerregister Automatic Registration of Peer Memory * - * Upon enabling direct access from a context that supports unified addressing - * to another peer context that supports unified addressing using - * ::cuCtxEnablePeerAccess() all memory allocated in the peer context using - * ::cuMemAlloc() and ::cuMemAllocPitch() will immediately be accessible + * Upon enabling direct access from a context that supports unified addressing + * to another peer context that supports unified addressing using + * ::cuCtxEnablePeerAccess() all memory allocated in the peer context using + * ::cuMemAlloc() and ::cuMemAllocPitch() will immediately be accessible * by the current context. The device pointer value through * which any peer memory may be accessed in the current context * is the same pointer value through which that memory may be * accessed in the peer context. * * \section CUDA_UNIFIED_exceptions Exceptions, Disjoint Addressing - * + * * Not all memory may be accessed on devices through the same pointer * value through which they are accessed on the host. These exceptions * are host memory registered using ::cuMemHostRegister() and host memory - * allocated using the flag ::CU_MEMHOSTALLOC_WRITECOMBINED. For these + * allocated using the flag ::CU_MEMHOSTALLOC_WRITECOMBINED. For these * exceptions, there exists a distinct host and device address for the * memory. The device address is guaranteed to not overlap any valid host - * pointer range and is guaranteed to have the same value across all - * contexts that support unified addressing. - * - * This device address may be queried using ::cuMemHostGetDevicePointer() - * when a context using unified addressing is current. Either the host - * or the unified device pointer value may be used to refer to this memory - * through ::cuMemcpy() and similar functions using the + * pointer range and is guaranteed to have the same value across all + * contexts that support unified addressing. + * + * This device address may be queried using ::cuMemHostGetDevicePointer() + * when a context using unified addressing is current. Either the host + * or the unified device pointer value may be used to refer to this memory + * through ::cuMemcpy() and similar functions using the * ::CU_MEMORYTYPE_UNIFIED memory type. * */ @@ -7421,69 +8269,69 @@ CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray); #if __CUDA_API_VERSION >= 4000 /** * \brief Returns information about a pointer - * + * * The supported attributes are: - * - * - ::CU_POINTER_ATTRIBUTE_CONTEXT: - * - * Returns in \p *data the ::CUcontext in which \p ptr was allocated or - * registered. - * The type of \p data must be ::CUcontext *. - * + * + * - ::CU_POINTER_ATTRIBUTE_CONTEXT: + * + * Returns in \p *data the ::CUcontext in which \p ptr was allocated or + * registered. + * The type of \p data must be ::CUcontext *. + * * If \p ptr was not allocated by, mapped by, or registered with - * a ::CUcontext which uses unified virtual addressing then + * a ::CUcontext which uses unified virtual addressing then * ::CUDA_ERROR_INVALID_VALUE is returned. - * - * - ::CU_POINTER_ATTRIBUTE_MEMORY_TYPE: - * - * Returns in \p *data the physical memory type of the memory that + * + * - ::CU_POINTER_ATTRIBUTE_MEMORY_TYPE: + * + * Returns in \p *data the physical memory type of the memory that * \p ptr addresses as a ::CUmemorytype enumerated value. * The type of \p data must be unsigned int. - * - * If \p ptr addresses device memory then \p *data is set to - * ::CU_MEMORYTYPE_DEVICE. The particular ::CUdevice on which the - * memory resides is the ::CUdevice of the ::CUcontext returned by the + * + * If \p ptr addresses device memory then \p *data is set to + * ::CU_MEMORYTYPE_DEVICE. The particular ::CUdevice on which the + * memory resides is the ::CUdevice of the ::CUcontext returned by the * ::CU_POINTER_ATTRIBUTE_CONTEXT attribute of \p ptr. - * - * If \p ptr addresses host memory then \p *data is set to + * + * If \p ptr addresses host memory then \p *data is set to * ::CU_MEMORYTYPE_HOST. - * + * * If \p ptr was not allocated by, mapped by, or registered with - * a ::CUcontext which uses unified virtual addressing then + * a ::CUcontext which uses unified virtual addressing then * ::CUDA_ERROR_INVALID_VALUE is returned. * - * If the current ::CUcontext does not support unified virtual + * If the current ::CUcontext does not support unified virtual * addressing then ::CUDA_ERROR_INVALID_CONTEXT is returned. - * + * * - ::CU_POINTER_ATTRIBUTE_DEVICE_POINTER: - * + * * Returns in \p *data the device pointer value through which - * \p ptr may be accessed by kernels running in the current + * \p ptr may be accessed by kernels running in the current * ::CUcontext. * The type of \p data must be CUdeviceptr *. - * + * * If there exists no device pointer value through which * kernels running in the current ::CUcontext may access * \p ptr then ::CUDA_ERROR_INVALID_VALUE is returned. - * - * If there is no current ::CUcontext then + * + * If there is no current ::CUcontext then * ::CUDA_ERROR_INVALID_CONTEXT is returned. - * - * Except in the exceptional disjoint addressing cases discussed - * below, the value returned in \p *data will equal the input + * + * Except in the exceptional disjoint addressing cases discussed + * below, the value returned in \p *data will equal the input * value \p ptr. - * + * * - ::CU_POINTER_ATTRIBUTE_HOST_POINTER: - * - * Returns in \p *data the host pointer value through which + * + * Returns in \p *data the host pointer value through which * \p ptr may be accessed by by the host program. * The type of \p data must be void **. * If there exists no host pointer value through which - * the host program may directly access \p ptr then + * the host program may directly access \p ptr then * ::CUDA_ERROR_INVALID_VALUE is returned. - * - * Except in the exceptional disjoint addressing cases discussed - * below, the value returned in \p *data will equal the input + * + * Except in the exceptional disjoint addressing cases discussed + * below, the value returned in \p *data will equal the input * value \p ptr. * * - ::CU_POINTER_ATTRIBUTE_P2P_TOKENS: @@ -7499,7 +8347,7 @@ CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray); * Querying this attribute has a side effect of setting the attribute * ::CU_POINTER_ATTRIBUTE_SYNC_MEMOPS for the region of memory that * \p ptr points to. - * + * * - ::CU_POINTER_ATTRIBUTE_SYNC_MEMOPS: * * A boolean attribute which when set, ensures that synchronous memory operations @@ -7524,22 +8372,27 @@ CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray); * Returns in \p *data a boolean that indicates whether the pointer points to * managed memory or not. * + * - ::CU_POINTER_ATTRIBUTE_DEVICE_ORDINAL: + * + * Returns in \p *data an integer representing a device ordinal of a device against + * which the memory was allocated or registered. + * * \par * * Note that for most allocations in the unified virtual address space - * the host and device pointer for accessing the allocation will be the + * the host and device pointer for accessing the allocation will be the * same. The exceptions to this are - * - user memory registered using ::cuMemHostRegister - * - host memory allocated using ::cuMemHostAlloc with the + * - user memory registered using ::cuMemHostRegister + * - host memory allocated using ::cuMemHostAlloc with the * ::CU_MEMHOSTALLOC_WRITECOMBINED flag - * For these types of allocation there will exist separate, disjoint host - * and device addresses for accessing the allocation. In particular - * - The host address will correspond to an invalid unmapped device address - * (which will result in an exception if accessed from the device) - * - The device address will correspond to an invalid unmapped host address + * For these types of allocation there will exist separate, disjoint host + * and device addresses for accessing the allocation. In particular + * - The host address will correspond to an invalid unmapped device address + * (which will result in an exception if accessed from the device) + * - The device address will correspond to an invalid unmapped host address * (which will result in an exception if accessed from the host). - * For these types of allocations, querying ::CU_POINTER_ATTRIBUTE_HOST_POINTER - * and ::CU_POINTER_ATTRIBUTE_DEVICE_POINTER may be used to retrieve the host + * For these types of allocations, querying ::CU_POINTER_ATTRIBUTE_HOST_POINTER + * and ::CU_POINTER_ATTRIBUTE_DEVICE_POINTER may be used to retrieve the host * and device addresses from either address. * * \param data - Returned pointer attribute value @@ -7555,14 +8408,16 @@ CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray); * ::CUDA_ERROR_INVALID_DEVICE * \notefnerr * - * \sa cuPointerSetAttribute, + * \sa + * ::cuPointerSetAttribute, * ::cuMemAlloc, * ::cuMemFree, * ::cuMemAllocHost, * ::cuMemFreeHost, * ::cuMemHostAlloc, * ::cuMemHostRegister, - * ::cuMemHostUnregister + * ::cuMemHostUnregister, + * ::cudaPointerGetAttributes */ CUresult CUDAAPI cuPointerGetAttribute(void *data, CUpointer_attribute attribute, CUdeviceptr ptr); #endif /* __CUDA_API_VERSION >= 4000 */ @@ -7571,8 +8426,8 @@ CUresult CUDAAPI cuPointerGetAttribute(void *data, CUpointer_attribute attribute /** * \brief Prefetches memory to the specified destination device * - * Prefetches memory to the specified destination device. \p devPtr is the - * base device pointer of the memory to be prefetched and \p dstDevice is the + * Prefetches memory to the specified destination device. \p devPtr is the + * base device pointer of the memory to be prefetched and \p dstDevice is the * destination device. \p count specifies the number of bytes to copy. \p hStream * is the stream in which the operation is enqueued. The memory range must refer * to managed memory allocated via ::cuMemAllocManaged or declared via __managed__ variables. @@ -7631,7 +8486,8 @@ CUresult CUDAAPI cuPointerGetAttribute(void *data, CUpointer_attribute attribute * \note_null_stream * * \sa ::cuMemcpy, ::cuMemcpyPeer, ::cuMemcpyAsync, - * ::cuMemcpy3DPeerAsync, ::cuMemAdvise + * ::cuMemcpy3DPeerAsync, ::cuMemAdvise, + * ::cudaMemPrefetchAsync */ CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice dstDevice, CUstream hStream); @@ -7642,7 +8498,10 @@ CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice d * starting at \p devPtr with a size of \p count bytes. The start address and end address of the memory * range will be rounded down and rounded up respectively to be aligned to CPU page size before the * advice is applied. The memory range must refer to managed memory allocated via ::cuMemAllocManaged - * or declared via __managed__ variables. + * or declared via __managed__ variables. The memory range could also refer to system-allocated pageable + * memory provided it represents a valid, host-accessible region of memory and all additional constraints + * imposed by \p advice as outlined below are also satisfied. Specifying an invalid system-allocated pageable + * memory range results in an error being returned. * * The \p advice parameter can take the following values: * - ::CU_MEM_ADVISE_SET_READ_MOSTLY: This implies that the data is mostly going to be read @@ -7656,11 +8515,18 @@ CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice d * Also, if a context is created on a device that does not have the device attribute * ::CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS set, then read-duplication will not occur until * all such contexts are destroyed. + * If the memory region refers to valid system-allocated pageable memory, then the accessing device must + * have a non-zero value for the device attribute ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS for a read-only + * copy to be created on that device. Note however that if the accessing device also has a non-zero value for the + * device attribute ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS_USES_HOST_PAGE_TABLES, then setting this advice + * will not create a read-only copy when that device accesses this memory region. + * * - ::CU_MEM_ADVISE_UNSET_READ_MOSTLY: Undoes the effect of ::CU_MEM_ADVISE_SET_READ_MOSTLY and also prevents the * Unified Memory driver from attempting heuristic read-duplication on the memory range. Any read-duplicated * copies of the data will be collapsed into a single copy. The location for the collapsed * copy will be the preferred location if the page has a preferred location and one of the read-duplicated * copies was resident at that location. Otherwise, the location chosen is arbitrary. + * * - ::CU_MEM_ADVISE_SET_PREFERRED_LOCATION: This advice sets the preferred location for the * data to be the memory belonging to \p device. Passing in CU_DEVICE_CPU for \p device sets the * preferred location as host memory. If \p device is a GPU, then it must have a non-zero value for the @@ -7677,9 +8543,17 @@ CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice d * memory, the page may eventually be pinned to host memory by the Unified Memory driver. But * if the preferred location is set as device memory, then the page will continue to thrash indefinitely. * If ::CU_MEM_ADVISE_SET_READ_MOSTLY is also set on this memory region or any subset of it, then the - * policies associated with that advice will override the policies of this advice. + * policies associated with that advice will override the policies of this advice, unless read accesses from + * \p device will not result in a read-only copy being created on that device as outlined in description for + * the advice ::CU_MEM_ADVISE_SET_READ_MOSTLY. + * If the memory region refers to valid system-allocated pageable memory, then \p device must have a non-zero + * value for the device attribute ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS. Additionally, if \p device has + * a non-zero value for the device attribute ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS_USES_HOST_PAGE_TABLES, + * then this call has no effect. Note however that this behavior may change in the future. + * * - ::CU_MEM_ADVISE_UNSET_PREFERRED_LOCATION: Undoes the effect of ::CU_MEM_ADVISE_SET_PREFERRED_LOCATION * and changes the preferred location to none. + * * - ::CU_MEM_ADVISE_SET_ACCESSED_BY: This advice implies that the data will be accessed by \p device. * Passing in ::CU_DEVICE_CPU for \p device will set the advice for the CPU. If \p device is a GPU, then * the device attribute ::CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS must be non-zero. @@ -7700,8 +8574,17 @@ CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice d * policies associated with that advice will override the policies of this advice. Additionally, if the * preferred location of this memory region or any subset of it is also \p device, then the policies * associated with ::CU_MEM_ADVISE_SET_PREFERRED_LOCATION will override the policies of this advice. + * If the memory region refers to valid system-allocated pageable memory, then \p device must have a non-zero + * value for the device attribute ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS. Additionally, if \p device has + * a non-zero value for the device attribute ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS_USES_HOST_PAGE_TABLES, + * then this call has no effect. + * * - ::CU_MEM_ADVISE_UNSET_ACCESSED_BY: Undoes the effect of ::CU_MEM_ADVISE_SET_ACCESSED_BY. Any mappings to * the data from \p device may be removed at any time causing accesses to result in non-fatal page faults. + * If the memory region refers to valid system-allocated pageable memory, then \p device must have a non-zero + * value for the device attribute ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS. Additionally, if \p device has + * a non-zero value for the device attribute ::CU_DEVICE_ATTRIBUTE_PAGEABLE_MEMORY_ACCESS_USES_HOST_PAGE_TABLES, + * then this call has no effect. * * \param devPtr - Pointer to memory to set the advice for * \param count - Size in bytes of the memory range @@ -7717,13 +8600,14 @@ CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice d * \note_null_stream * * \sa ::cuMemcpy, ::cuMemcpyPeer, ::cuMemcpyAsync, - * ::cuMemcpy3DPeerAsync, ::cuMemPrefetchAsync + * ::cuMemcpy3DPeerAsync, ::cuMemPrefetchAsync, + * ::cudaMemAdvise */ CUresult CUDAAPI cuMemAdvise(CUdeviceptr devPtr, size_t count, CUmem_advise advice, CUdevice device); /** * \brief Query an attribute of a given memory range - * + * * Query an attribute about the memory range starting at \p devPtr with a size of \p count bytes. The * memory range must refer to managed memory allocated via ::cuMemAllocManaged or declared via * __managed__ variables. @@ -7738,7 +8622,7 @@ CUresult CUDAAPI cuMemAdvise(CUdeviceptr devPtr, size_t count, CUmem_advise advi * if all pages in the memory range have the CPU as their preferred location, or it will be CU_DEVICE_INVALID * if either all the pages don't have the same preferred location or some of the pages don't have a * preferred location at all. Note that the actual location of the pages in the memory range at the time of - * the query may be different from the preferred location. + * the query may be different from the preferred location. * - ::CU_MEM_RANGE_ATTRIBUTE_ACCESSED_BY: If this attribute is specified, \p data will be interpreted * as an array of 32-bit integers, and \p dataSize must be a non-zero multiple of 4. The result returned * will be a list of device ids that had ::CU_MEM_ADVISE_SET_ACCESSED_BY set for that entire memory range. @@ -7774,7 +8658,8 @@ CUresult CUDAAPI cuMemAdvise(CUdeviceptr devPtr, size_t count, CUmem_advise advi * \note_null_stream * * \sa ::cuMemRangeGetAttributes, ::cuMemPrefetchAsync, - * ::cuMemAdvise + * ::cuMemAdvise, + * ::cudaMemRangeGetAttribute */ CUresult CUDAAPI cuMemRangeGetAttribute(void *data, size_t dataSize, CUmem_range_attribute attribute, CUdeviceptr devPtr, size_t count); @@ -7813,7 +8698,8 @@ CUresult CUDAAPI cuMemRangeGetAttribute(void *data, size_t dataSize, CUmem_range * \notefnerr * * \sa ::cuMemRangeGetAttribute, ::cuMemAdvise - * ::cuMemPrefetchAsync + * ::cuMemPrefetchAsync, + * ::cudaMemRangeGetAttributes */ CUresult CUDAAPI cuMemRangeGetAttributes(void **data, size_t *dataSizes, CUmem_range_attribute *attributes, size_t numAttributes, CUdeviceptr devPtr, size_t count); #endif /* __CUDA_API_VERSION >= 8000 */ @@ -7875,6 +8761,7 @@ CUresult CUDAAPI cuPointerSetAttribute(const void *value, CUpointer_attribute at * - ::CU_POINTER_ATTRIBUTE_SYNC_MEMOPS * - ::CU_POINTER_ATTRIBUTE_BUFFER_ID * - ::CU_POINTER_ATTRIBUTE_IS_MANAGED + * - ::CU_POINTER_ATTRIBUTE_DEVICE_ORDINAL * * \param numAttributes - Number of attributes to query * \param attributes - An array of attributes to query @@ -7898,8 +8785,10 @@ CUresult CUDAAPI cuPointerSetAttribute(const void *value, CUpointer_attribute at * ::CUDA_ERROR_INVALID_DEVICE * \notefnerr * - * \sa ::cuPointerGetAttribute, - * ::cuPointerSetAttribute + * \sa + * ::cuPointerGetAttribute, + * ::cuPointerSetAttribute, + * ::cudaPointerGetAttributes */ CUresult CUDAAPI cuPointerGetAttributes(unsigned int numAttributes, CUpointer_attribute *attributes, void **data, CUdeviceptr ptr); #endif /* __CUDA_API_VERSION >= 7000 */ @@ -7924,7 +8813,7 @@ CUresult CUDAAPI cuPointerGetAttributes(unsigned int numAttributes, CUpointer_at * Creates a stream and returns a handle in \p phStream. The \p Flags argument * determines behaviors of the stream. Valid values for \p Flags are: * - ::CU_STREAM_DEFAULT: Default stream creation flag. - * - ::CU_STREAM_NON_BLOCKING: Specifies that work running in the created + * - ::CU_STREAM_NON_BLOCKING: Specifies that work running in the created * stream may run concurrently with work in stream 0 (the NULL stream), and that * the created stream should perform no implicit synchronization with stream 0. * @@ -7947,7 +8836,9 @@ CUresult CUDAAPI cuPointerGetAttributes(unsigned int numAttributes, CUpointer_at * ::cuStreamWaitEvent, * ::cuStreamQuery, * ::cuStreamSynchronize, - * ::cuStreamAddCallback + * ::cuStreamAddCallback, + * ::cudaStreamCreate, + * ::cudaStreamCreateWithFlags */ CUresult CUDAAPI cuStreamCreate(CUstream *phStream, unsigned int Flags); @@ -7980,7 +8871,7 @@ CUresult CUDAAPI cuStreamCreate(CUstream *phStream, unsigned int Flags); * ::CUDA_ERROR_OUT_OF_MEMORY * \notefnerr * - * \note Stream priorities are supported only on Quadro and Tesla GPUs + * \note Stream priorities are supported only on GPUs * with compute capability 3.5 or higher. * * \note In the current implementation, only compute kernels launched in @@ -7995,7 +8886,8 @@ CUresult CUDAAPI cuStreamCreate(CUstream *phStream, unsigned int Flags); * ::cuStreamWaitEvent, * ::cuStreamQuery, * ::cuStreamSynchronize, - * ::cuStreamAddCallback + * ::cuStreamAddCallback, + * ::cudaStreamCreateWithPriority */ CUresult CUDAAPI cuStreamCreateWithPriority(CUstream *phStream, unsigned int flags, int priority); @@ -8025,7 +8917,8 @@ CUresult CUDAAPI cuStreamCreateWithPriority(CUstream *phStream, unsigned int fla * ::cuStreamCreate, * ::cuStreamCreateWithPriority, * ::cuCtxGetStreamPriorityRange, - * ::cuStreamGetFlags + * ::cuStreamGetFlags, + * ::cudaStreamGetPriority */ CUresult CUDAAPI cuStreamGetPriority(CUstream hStream, int *priority); @@ -8052,28 +8945,66 @@ CUresult CUDAAPI cuStreamGetPriority(CUstream hStream, int *priority); * * \sa ::cuStreamDestroy, * ::cuStreamCreate, - * ::cuStreamGetPriority + * ::cuStreamGetPriority, + * ::cudaStreamGetFlags */ CUresult CUDAAPI cuStreamGetFlags(CUstream hStream, unsigned int *flags); +#if __CUDA_API_VERSION >= 9020 /** - * \brief Make a compute stream wait on an event + * \brief Query the context associated with a stream + * + * Returns the CUDA context that the stream is associated with. + * + * The stream handle \p hStream can refer to any of the following: + * * - * Makes all future work submitted to \p hStream wait until \p hEvent - * reports completion before beginning execution. This synchronization - * will be performed efficiently on the device. The event \p hEvent may - * be from a different context than \p hStream, in which case this function - * will perform cross-device synchronization. + * \param hStream - Handle to the stream to be queried + * \param pctx - Returned context associated with the stream + * + * \return + * ::CUDA_SUCCESS, + * ::CUDA_ERROR_DEINITIALIZED, + * ::CUDA_ERROR_NOT_INITIALIZED, + * ::CUDA_ERROR_INVALID_CONTEXT, + * ::CUDA_ERROR_INVALID_HANDLE, + * \notefnerr * - * The stream \p hStream will wait only for the completion of the most recent - * host call to ::cuEventRecord() on \p hEvent. Once this call has returned, - * any functions (including ::cuEventRecord() and ::cuEventDestroy()) may be - * called on \p hEvent again, and subsequent calls will not have any - * effect on \p hStream. + * \sa ::cuStreamDestroy, + * ::cuStreamCreateWithPriority, + * ::cuStreamGetPriority, + * ::cuStreamGetFlags, + * ::cuStreamWaitEvent, + * ::cuStreamQuery, + * ::cuStreamSynchronize, + * ::cuStreamAddCallback, + * ::cudaStreamCreate, + * ::cudaStreamCreateWithFlags + */ +CUresult CUDAAPI cuStreamGetCtx(CUstream hStream, CUcontext *pctx); + +#endif /* __CUDA_API_VERSION >= 9020 */ + +/** + * \brief Make a compute stream wait on an event * - * If ::cuEventRecord() has not been called on \p hEvent, this call acts as if - * the record has already completed, and so is a functional no-op. + * Makes all future work submitted to \p hStream wait for all work captured in + * \p hEvent. See ::cuEventRecord() for details on what is captured by an event. + * The synchronization will be performed efficiently on the device when applicable. + * \p hEvent may be from a different context or device than \p hStream. * * \param hStream - Stream to wait * \param hEvent - Event to wait on (may not be NULL) @@ -8093,15 +9024,22 @@ CUresult CUDAAPI cuStreamGetFlags(CUstream hStream, unsigned int *flags); * ::cuStreamQuery, * ::cuStreamSynchronize, * ::cuStreamAddCallback, - * ::cuStreamDestroy + * ::cuStreamDestroy, + * ::cudaStreamWaitEvent */ CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, unsigned int Flags); /** * \brief Add a callback to a compute stream * + * \note This function is slated for eventual deprecation and removal. If + * you do not require the callback to execute in case of a device error, + * consider using ::cuLaunchHostFunc. Additionally, this function is not + * supported with ::cuStreamBeginCapture and ::cuStreamEndCapture, unlike + * ::cuLaunchHostFunc. + * * Adds a callback to be called on the host after all currently enqueued - * items in the stream have completed. For each + * items in the stream have completed. For each * cuStreamAddCallback call, the callback will be executed exactly once. * The callback will block later work in the stream until it is finished. * @@ -8115,11 +9053,6 @@ CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, unsigned in * that are not mandated to run earlier. Callbacks without a mandated order * (in independent streams) execute in undefined order and may be serialized. * - * This API requires compute capability 1.1 or greater. See - * ::cuDeviceGetAttribute or ::cuDeviceGetProperties to query compute - * capability. Attempting to use this API with earlier compute versions will - * return ::CUDA_ERROR_NOT_SUPPORTED. - * * For the purposes of Unified Memory, callback execution makes a number of * guarantees: * * * \param hStream - Stream to add callback to - * \param callback - The function to call once preceding stream operations are complete - * \param userData - User specified data to be passed to the callback function - * \param flags - Reserved for future use, must be 0 + * \param callback - The function to call once preceding stream operations are + * complete \param userData - User specified data to be passed to the callback + * function \param flags - Reserved for future use, must be 0 * * \return * ::CUDA_SUCCESS, @@ -9102,32 +9616,36 @@ CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, unsigned in * ::cuStreamLaunchHostFunc, * ::cudaStreamAddCallback */ -CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, CUstreamCallback callback, void *userData, unsigned int flags); +CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, + CUstreamCallback callback, void *userData, + unsigned int flags); #if __CUDA_API_VERSION >= 10000 /** * \brief Begins graph capture on a stream * - * Begin graph capture on \p hStream. When a stream is in capture mode, all operations - * pushed into the stream will not be executed, but will instead be captured into - * a graph, which will be returned via ::cuStreamEndCapture. Capture may not be initiated - * if \p stream is CU_STREAM_LEGACY. Capture must be ended on the same stream in which - * it was initiated, and it may only be initiated if the stream is not already in capture - * mode. The capture mode may be queried via ::cuStreamIsCapturing. A unique id - * representing the capture sequence may be queried via ::cuStreamGetCaptureInfo. + * Begin graph capture on \p hStream. When a stream is in capture mode, all + * operations pushed into the stream will not be executed, but will instead be + * captured into a graph, which will be returned via ::cuStreamEndCapture. + * Capture may not be initiated if \p stream is CU_STREAM_LEGACY. Capture must + * be ended on the same stream in which it was initiated, and it may only be + * initiated if the stream is not already in capture mode. The capture mode may + * be queried via ::cuStreamIsCapturing. A unique id representing the capture + * sequence may be queried via ::cuStreamGetCaptureInfo. * - * If \p mode is not ::CU_STREAM_CAPTURE_MODE_RELAXED, ::cuStreamEndCapture must be - * called on this stream from the same thread. + * If \p mode is not ::CU_STREAM_CAPTURE_MODE_RELAXED, ::cuStreamEndCapture must + * be called on this stream from the same thread. * * \param hStream - Stream in which to initiate capture - * \param mode - Controls the interaction of this capture sequence with other API - * calls that are potentially unsafe. For more details see + * \param mode - Controls the interaction of this capture sequence with other + * API calls that are potentially unsafe. For more details see * ::cuThreadExchangeStreamCaptureMode. * - * \note Kernels captured using this API must not use texture and surface references. - * Reading or writing through any texture or surface reference is undefined - * behavior. This restriction does not apply to texture and surface objects. + * \note Kernels captured using this API must not use texture and surface + * references. Reading or writing through any texture or surface reference is + * undefined behavior. This restriction does not apply to texture and surface + * objects. * * \return * ::CUDA_SUCCESS, @@ -9142,7 +9660,8 @@ CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, CUstreamCallback callback * ::cuStreamEndCapture, * ::cuThreadExchangeStreamCaptureMode */ -CUresult CUDAAPI cuStreamBeginCapture(CUstream hStream, CUstreamCaptureMode mode); +CUresult CUDAAPI cuStreamBeginCapture(CUstream hStream, + CUstreamCaptureMode mode); #endif /* __CUDA_API_VERSION >= 10000 */ #if __CUDA_API_VERSION >= 10010 @@ -9150,9 +9669,12 @@ CUresult CUDAAPI cuStreamBeginCapture(CUstream hStream, CUstreamCaptureMode mode /** * \brief Swaps the stream capture interaction mode for a thread * - * Sets the calling thread's stream capture interaction mode to the value contained - * in \p *mode, and overwrites \p *mode with the previous mode for the thread. To - * facilitate deterministic behavior across function or module boundaries, callers + * Sets the calling thread's stream capture interaction mode to the value + contained + * in \p *mode, and overwrites \p *mode with the previous mode for the thread. + To + * facilitate deterministic behavior across function or module boundaries, + callers * are encouraged to use this API in a push-pop fashion: \code CUstreamCaptureMode mode = desiredMode; cuThreadExchangeStreamCaptureMode(&mode); @@ -9160,30 +9682,43 @@ CUresult CUDAAPI cuStreamBeginCapture(CUstream hStream, CUstreamCaptureMode mode cuThreadExchangeStreamCaptureMode(&mode); // restore previous mode * \endcode * - * During stream capture (see ::cuStreamBeginCapture), some actions, such as a call + * During stream capture (see ::cuStreamBeginCapture), some actions, such as a + call * to ::cudaMalloc, may be unsafe. In the case of ::cudaMalloc, the operation is - * not enqueued asynchronously to a stream, and is not observed by stream capture. + * not enqueued asynchronously to a stream, and is not observed by stream + capture. * Therefore, if the sequence of operations captured via ::cuStreamBeginCapture * depended on the allocation being replayed whenever the graph is launched, the * captured graph would be invalid. * - * Therefore, stream capture places restrictions on API calls that can be made within - * or concurrently to a ::cuStreamBeginCapture-::cuStreamEndCapture sequence. This + * Therefore, stream capture places restrictions on API calls that can be made + within + * or concurrently to a ::cuStreamBeginCapture-::cuStreamEndCapture sequence. + This * behavior can be controlled via this API and flags to ::cuStreamBeginCapture. * * A thread's mode is one of the following: - * - \p CU_STREAM_CAPTURE_MODE_GLOBAL: This is the default mode. If the local thread has + * - \p CU_STREAM_CAPTURE_MODE_GLOBAL: This is the default mode. If the local + thread has * an ongoing capture sequence that was not initiated with - * \p CU_STREAM_CAPTURE_MODE_RELAXED at \p cuStreamBeginCapture, or if any other thread - * has a concurrent capture sequence initiated with \p CU_STREAM_CAPTURE_MODE_GLOBAL, + * \p CU_STREAM_CAPTURE_MODE_RELAXED at \p cuStreamBeginCapture, or if any + other thread + * has a concurrent capture sequence initiated with \p + CU_STREAM_CAPTURE_MODE_GLOBAL, * this thread is prohibited from potentially unsafe API calls. - * - \p CU_STREAM_CAPTURE_MODE_THREAD_LOCAL: If the local thread has an ongoing capture - * sequence not initiated with \p CU_STREAM_CAPTURE_MODE_RELAXED, it is prohibited - * from potentially unsafe API calls. Concurrent capture sequences in other threads + * - \p CU_STREAM_CAPTURE_MODE_THREAD_LOCAL: If the local thread has an ongoing + capture + * sequence not initiated with \p CU_STREAM_CAPTURE_MODE_RELAXED, it is + prohibited + * from potentially unsafe API calls. Concurrent capture sequences in other + threads * are ignored. - * - \p CU_STREAM_CAPTURE_MODE_RELAXED: The local thread is not prohibited from potentially - * unsafe API calls. Note that the thread is still prohibited from API calls which - * necessarily conflict with stream capture, for example, attempting ::cuEventQuery + * - \p CU_STREAM_CAPTURE_MODE_RELAXED: The local thread is not prohibited from + potentially + * unsafe API calls. Note that the thread is still prohibited from API calls + which + * necessarily conflict with stream capture, for example, attempting + ::cuEventQuery * on an event that was last recorded inside a capture sequence. * * \param mode - Pointer to mode value to swap with the current mode @@ -9207,9 +9742,9 @@ CUresult CUDAAPI cuThreadExchangeStreamCaptureMode(CUstreamCaptureMode *mode); * \brief Ends capture on a stream, returning the captured graph * * End capture on \p hStream, returning the captured graph via \p phGraph. - * Capture must have been initiated on \p hStream via a call to ::cuStreamBeginCapture. - * If capture was invalidated, due to a violation of the rules of stream capture, then - * a NULL graph will be returned. + * Capture must have been initiated on \p hStream via a call to + * ::cuStreamBeginCapture. If capture was invalidated, due to a violation of the + * rules of stream capture, then a NULL graph will be returned. * * If the \p mode argument to ::cuStreamBeginCapture was not * ::CU_STREAM_CAPTURE_MODE_RELAXED, this call must be from the same thread as @@ -9236,14 +9771,14 @@ CUresult CUDAAPI cuStreamEndCapture(CUstream hStream, CUgraph *phGraph); /** * \brief Returns a stream's capture status * - * Return the capture status of \p hStream via \p captureStatus. After a successful - * call, \p *captureStatus will contain one of the following: + * Return the capture status of \p hStream via \p captureStatus. After a + * successful call, \p *captureStatus will contain one of the following: * - ::CU_STREAM_CAPTURE_STATUS_NONE: The stream is not capturing. * - ::CU_STREAM_CAPTURE_STATUS_ACTIVE: The stream is capturing. - * - ::CU_STREAM_CAPTURE_STATUS_INVALIDATED: The stream was capturing but an error - * has invalidated the capture sequence. The capture sequence must be terminated - * with ::cuStreamEndCapture on the stream where it was initiated in order to - * continue using \p hStream. + * - ::CU_STREAM_CAPTURE_STATUS_INVALIDATED: The stream was capturing but an + * error has invalidated the capture sequence. The capture sequence must be + * terminated with ::cuStreamEndCapture on the stream where it was initiated in + * order to continue using \p hStream. * * Note that, if this is called on ::CU_STREAM_LEGACY (the "null stream") while * a blocking stream in the same context is capturing, it will return @@ -9271,7 +9806,8 @@ CUresult CUDAAPI cuStreamEndCapture(CUstream hStream, CUgraph *phGraph); * ::cuStreamBeginCapture, * ::cuStreamEndCapture */ -CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, CUstreamCaptureStatus *captureStatus); +CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, + CUstreamCaptureStatus *captureStatus); #endif /* __CUDA_API_VERSION >= 10000 */ @@ -9283,8 +9819,9 @@ CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, CUstreamCaptureStatus *ca * Query the capture status of a stream and and get an id for * the capture sequence, which is unique over the lifetime of the process. * - * If called on ::CU_STREAM_LEGACY (the "null stream") while a stream not created - * with ::CU_STREAM_NON_BLOCKING is capturing, returns ::CUDA_ERROR_STREAM_CAPTURE_IMPLICIT. + * If called on ::CU_STREAM_LEGACY (the "null stream") while a stream not + * created with ::CU_STREAM_NON_BLOCKING is capturing, returns + * ::CUDA_ERROR_STREAM_CAPTURE_IMPLICIT. * * A valid id is returned only if both of the following are true: * - the call returns CUDA_SUCCESS @@ -9299,7 +9836,9 @@ CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, CUstreamCaptureStatus *ca * ::cuStreamBeginCapture, * ::cuStreamIsCapturing */ - CUresult CUDAAPI cuStreamGetCaptureInfo(CUstream hStream, CUstreamCaptureStatus *captureStatus, cuuint64_t *id); +CUresult CUDAAPI cuStreamGetCaptureInfo(CUstream hStream, + CUstreamCaptureStatus *captureStatus, + cuuint64_t *id); #endif /* __CUDA_API_VERSION >= 10010 */ @@ -9334,19 +9873,21 @@ CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, CUstreamCaptureStatus *ca * If the ::CU_MEM_ATTACH_GLOBAL flag is specified, the memory can be accessed * by any stream on any device. * If the ::CU_MEM_ATTACH_HOST flag is specified, the program makes a guarantee - * that it won't access the memory on the device from any stream on a device that - * has a zero value for the device attribute ::CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS. - * If the ::CU_MEM_ATTACH_SINGLE flag is specified and \p hStream is associated with - * a device that has a zero value for the device attribute ::CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS, - * the program makes a guarantee that it will only access the memory on the device - * from \p hStream. It is illegal to attach singly to the NULL stream, because the - * NULL stream is a virtual global stream and not a specific stream. An error will - * be returned in this case. - * - * When memory is associated with a single stream, the Unified Memory system will - * allow CPU access to this memory region so long as all operations in \p hStream - * have completed, regardless of whether other streams are active. In effect, - * this constrains exclusive ownership of the managed memory region by + * that it won't access the memory on the device from any stream on a device + * that has a zero value for the device attribute + * ::CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS. If the + * ::CU_MEM_ATTACH_SINGLE flag is specified and \p hStream is associated with a + * device that has a zero value for the device attribute + * ::CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS, the program makes a + * guarantee that it will only access the memory on the device from \p hStream. + * It is illegal to attach singly to the NULL stream, because the NULL stream is + * a virtual global stream and not a specific stream. An error will be returned + * in this case. + * + * When memory is associated with a single stream, the Unified Memory system + * will allow CPU access to this memory region so long as all operations in \p + * hStream have completed, regardless of whether other streams are active. In + * effect, this constrains exclusive ownership of the managed memory region by * an active GPU to per-stream activity instead of whole-GPU activity. * * Accessing memory on the device from streams that are not associated with @@ -9359,12 +9900,13 @@ CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, CUstreamCaptureStatus *ca * at all times. Data visibility and coherency will be changed appropriately * for all kernels which follow a stream-association change. * - * If \p hStream is destroyed while data is associated with it, the association is - * removed and the association reverts to the default visibility of the allocation - * as specified at ::cuMemAllocManaged. For __managed__ variables, the default - * association is always ::CU_MEM_ATTACH_GLOBAL. Note that destroying a stream is an - * asynchronous operation, and as a result, the change to default association won't - * happen until all work in the stream has completed. + * If \p hStream is destroyed while data is associated with it, the association + * is removed and the association reverts to the default visibility of the + * allocation as specified at ::cuMemAllocManaged. For __managed__ variables, + * the default association is always ::CU_MEM_ATTACH_GLOBAL. Note that + * destroying a stream is an asynchronous operation, and as a result, the change + * to default association won't happen until all work in the stream has + * completed. * * \param hStream - Stream in which to enqueue the attach operation * \param dptr - Pointer to memory (must be a pointer to managed memory or @@ -9391,7 +9933,8 @@ CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, CUstreamCaptureStatus *ca * ::cuMemAllocManaged, * ::cudaStreamAttachMemAsync */ -CUresult CUDAAPI cuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, size_t length, unsigned int flags); +CUresult CUDAAPI cuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, + size_t length, unsigned int flags); #endif /* __CUDA_API_VERSION >= 6000 */ @@ -9488,7 +10031,6 @@ CUresult CUDAAPI cuStreamDestroy(CUstream hStream); /** @} */ /* END CUDA_STREAM */ - /** * \defgroup CUDA_EVENT Event Management * @@ -9504,13 +10046,13 @@ CUresult CUDAAPI cuStreamDestroy(CUstream hStream); /** * \brief Creates an event * - * Creates an event *phEvent for the current context with the flags specified via - * \p Flags. Valid flags include: + * Creates an event *phEvent for the current context with the flags specified + * via \p Flags. Valid flags include: * - ::CU_EVENT_DEFAULT: Default event creation flag. - * - ::CU_EVENT_BLOCKING_SYNC: Specifies that the created event should use blocking - * synchronization. A CPU thread that uses ::cuEventSynchronize() to wait on - * an event created with this flag will block until the event has actually - * been recorded. + * - ::CU_EVENT_BLOCKING_SYNC: Specifies that the created event should use + * blocking synchronization. A CPU thread that uses ::cuEventSynchronize() to + * wait on an event created with this flag will block until the event has + * actually been recorded. * - ::CU_EVENT_DISABLE_TIMING: Specifies that the created event does not need * to record timing data. Events created with this flag specified and * the ::CU_EVENT_BLOCKING_SYNC flag not specified will provide the best @@ -9719,147 +10261,152 @@ CUresult CUDAAPI cuEventDestroy(CUevent hEvent); * ::cuEventDestroy, * ::cudaEventElapsedTime */ -CUresult CUDAAPI cuEventElapsedTime(float *pMilliseconds, CUevent hStart, CUevent hEnd); +CUresult CUDAAPI cuEventElapsedTime(float *pMilliseconds, CUevent hStart, + CUevent hEnd); /** @} */ /* END CUDA_EVENT */ /** * \defgroup CUDA_EXTRES_INTEROP External Resource Interoperability * - * ___MANBRIEF___ External resource interoperability functions of the low-level CUDA driver API + * ___MANBRIEF___ External resource interoperability functions of the low-level + * CUDA driver API * (___CURRENT_FILE___) ___ENDMANBRIEF___ * - * This section describes the external resource interoperability functions of the low-level CUDA - * driver application programming interface. + * This section describes the external resource interoperability functions of + * the low-level CUDA driver application programming interface. * * @{ */ #if __CUDA_API_VERSION >= 10000 - /** - * \brief Imports an external memory object - * - * Imports an externally allocated memory object and returns - * a handle to that in \p extMem_out. - * - * The properties of the handle being imported must be described in - * \p memHandleDesc. The ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC structure - * is defined as follows: - * - * \code - typedef struct CUDA_EXTERNAL_MEMORY_HANDLE_DESC_st { - CUexternalMemoryHandleType type; - union { - int fd; - struct { - void *handle; - const void *name; - } win32; - } handle; - unsigned long long size; - unsigned int flags; - } CUDA_EXTERNAL_MEMORY_HANDLE_DESC; - * \endcode - * - * where ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type specifies the type - * of handle being imported. ::CUexternalMemoryHandleType is - * defined as: - * - * \code - typedef enum CUexternalMemoryHandleType_enum { - CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD = 1, - CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32 = 2, - CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT = 3, - CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_HEAP = 4, - CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_RESOURCE = 5 - } CUexternalMemoryHandleType; - * \endcode - * - * If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is - * ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD, then - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::fd must be a valid - * file descriptor referencing a memory object. Ownership of - * the file descriptor is transferred to the CUDA driver when the - * handle is imported successfully. Performing any operations on the - * file descriptor after it is imported results in undefined behavior. - * - * If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is - * ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32, then exactly one - * of ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle and - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name must not be - * NULL. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle - * is not NULL, then it must represent a valid shared NT handle that - * references a memory object. Ownership of this handle is - * not transferred to CUDA after the import operation, so the - * application must release the handle using the appropriate system - * call. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name - * is not NULL, then it must point to a NULL-terminated array of - * UTF-16 characters that refers to a memory object. - * - * If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is - * ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT, then - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle must - * be non-NULL and - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name - * must be NULL. The handle specified must be a globally shared KMT - * handle. This handle does not hold a reference to the underlying - * object, and thus will be invalid when all references to the - * memory object are destroyed. - * - * If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is - * ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_HEAP, then exactly one - * of ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle and - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name must not be - * NULL. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle - * is not NULL, then it must represent a valid shared NT handle that - * is returned by ID3DDevice::CreateSharedHandle when referring to a - * ID3D12Heap object. This handle holds a reference to the underlying - * object. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name - * is not NULL, then it must point to a NULL-terminated array of - * UTF-16 characters that refers to a ID3D12Heap object. - * - * If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is - * ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_RESOURCE, then exactly one - * of ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle and - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name must not be - * NULL. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle - * is not NULL, then it must represent a valid shared NT handle that - * is returned by ID3DDevice::CreateSharedHandle when referring to a - * ID3D12Resource object. This handle holds a reference to the - * underlying object. If - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name - * is not NULL, then it must point to a NULL-terminated array of - * UTF-16 characters that refers to a ID3D12Resource object. - * - * The size of the memory object must be specified in - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::size. - * - * Specifying the flag ::CUDA_EXTERNAL_MEMORY_DEDICATED in - * ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::flags indicates that the - * resource is a dedicated resource. The definition of what a - * dedicated resource is outside the scope of this extension. - * - * \param extMem_out - Returned handle to an external memory object - * \param memHandleDesc - Memory import handle descriptor - * - * \return - * ::CUDA_SUCCESS, - * ::CUDA_ERROR_NOT_INITIALIZED, - * ::CUDA_ERROR_INVALID_HANDLE - * \notefnerr - * - * \note If the Vulkan memory imported into CUDA is mapped on the CPU then the - * application must use vkInvalidateMappedMemoryRanges/vkFlushMappedMemoryRanges - * as well as appropriate Vulkan pipeline barriers to maintain coherence between - * CPU and GPU. For more information on these APIs, please refer to "Synchronization - * and Cache Control" chapter from Vulkan specification. - * - * \sa ::cuDestroyExternalMemory, - * ::cuExternalMemoryGetMappedBuffer, - * ::cuExternalMemoryGetMappedMipmappedArray - */ -CUresult CUDAAPI cuImportExternalMemory(CUexternalMemory *extMem_out, const CUDA_EXTERNAL_MEMORY_HANDLE_DESC *memHandleDesc); +/** +* \brief Imports an external memory object +* +* Imports an externally allocated memory object and returns +* a handle to that in \p extMem_out. +* +* The properties of the handle being imported must be described in +* \p memHandleDesc. The ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC structure +* is defined as follows: +* +* \code + typedef struct CUDA_EXTERNAL_MEMORY_HANDLE_DESC_st { + CUexternalMemoryHandleType type; + union { + int fd; + struct { + void *handle; + const void *name; + } win32; + } handle; + unsigned long long size; + unsigned int flags; + } CUDA_EXTERNAL_MEMORY_HANDLE_DESC; +* \endcode +* +* where ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type specifies the type +* of handle being imported. ::CUexternalMemoryHandleType is +* defined as: +* +* \code + typedef enum CUexternalMemoryHandleType_enum { + CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD = 1, + CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32 = 2, + CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT = 3, + CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_HEAP = 4, + CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_RESOURCE = 5 + } CUexternalMemoryHandleType; +* \endcode +* +* If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is +* ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD, then +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::fd must be a valid +* file descriptor referencing a memory object. Ownership of +* the file descriptor is transferred to the CUDA driver when the +* handle is imported successfully. Performing any operations on the +* file descriptor after it is imported results in undefined behavior. +* +* If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is +* ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32, then exactly one +* of ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle and +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name must not be +* NULL. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle +* is not NULL, then it must represent a valid shared NT handle that +* references a memory object. Ownership of this handle is +* not transferred to CUDA after the import operation, so the +* application must release the handle using the appropriate system +* call. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name +* is not NULL, then it must point to a NULL-terminated array of +* UTF-16 characters that refers to a memory object. +* +* If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is +* ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT, then +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle must +* be non-NULL and +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name +* must be NULL. The handle specified must be a globally shared KMT +* handle. This handle does not hold a reference to the underlying +* object, and thus will be invalid when all references to the +* memory object are destroyed. +* +* If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is +* ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_HEAP, then exactly one +* of ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle and +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name must not be +* NULL. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle +* is not NULL, then it must represent a valid shared NT handle that +* is returned by ID3DDevice::CreateSharedHandle when referring to a +* ID3D12Heap object. This handle holds a reference to the underlying +* object. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name +* is not NULL, then it must point to a NULL-terminated array of +* UTF-16 characters that refers to a ID3D12Heap object. +* +* If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::type is +* ::CU_EXTERNAL_MEMORY_HANDLE_TYPE_D3D12_RESOURCE, then exactly one +* of ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle and +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name must not be +* NULL. If ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::handle +* is not NULL, then it must represent a valid shared NT handle that +* is returned by ID3DDevice::CreateSharedHandle when referring to a +* ID3D12Resource object. This handle holds a reference to the +* underlying object. If +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::handle::win32::name +* is not NULL, then it must point to a NULL-terminated array of +* UTF-16 characters that refers to a ID3D12Resource object. +* +* The size of the memory object must be specified in +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::size. +* +* Specifying the flag ::CUDA_EXTERNAL_MEMORY_DEDICATED in +* ::CUDA_EXTERNAL_MEMORY_HANDLE_DESC::flags indicates that the +* resource is a dedicated resource. The definition of what a +* dedicated resource is outside the scope of this extension. +* +* \param extMem_out - Returned handle to an external memory object +* \param memHandleDesc - Memory import handle descriptor +* +* \return +* ::CUDA_SUCCESS, +* ::CUDA_ERROR_NOT_INITIALIZED, +* ::CUDA_ERROR_INVALID_HANDLE +* \notefnerr +* +* \note If the Vulkan memory imported into CUDA is mapped on the CPU then the +* application must use vkInvalidateMappedMemoryRanges/vkFlushMappedMemoryRanges +* as well as appropriate Vulkan pipeline barriers to maintain coherence between +* CPU and GPU. For more information on these APIs, please refer to +"Synchronization +* and Cache Control" chapter from Vulkan specification. +* +* \sa ::cuDestroyExternalMemory, +* ::cuExternalMemoryGetMappedBuffer, +* ::cuExternalMemoryGetMappedMipmappedArray +*/ +CUresult CUDAAPI +cuImportExternalMemory(CUexternalMemory *extMem_out, + const CUDA_EXTERNAL_MEMORY_HANDLE_DESC *memHandleDesc); /** * \brief Maps a buffer onto an imported memory object @@ -9912,7 +10459,9 @@ CUresult CUDAAPI cuImportExternalMemory(CUexternalMemory *extMem_out, const CUDA * ::cuDestroyExternalMemory, * ::cuExternalMemoryGetMappedMipmappedArray */ -CUresult CUDAAPI cuExternalMemoryGetMappedBuffer(CUdeviceptr *devPtr, CUexternalMemory extMem, const CUDA_EXTERNAL_MEMORY_BUFFER_DESC *bufferDesc); +CUresult CUDAAPI cuExternalMemoryGetMappedBuffer( + CUdeviceptr *devPtr, CUexternalMemory extMem, + const CUDA_EXTERNAL_MEMORY_BUFFER_DESC *bufferDesc); /** * \brief Maps a CUDA mipmapped array onto an external memory object @@ -9945,7 +10494,8 @@ CUresult CUDAAPI cuExternalMemoryGetMappedBuffer(CUdeviceptr *devPtr, CUexternal * ::CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC::numLevels specifies * the total number of levels in the mipmap chain. * - * The returned CUDA mipmapped array must be freed using ::cuMipmappedArrayDestroy. + * The returned CUDA mipmapped array must be freed using + ::cuMipmappedArrayDestroy. * * \param mipmap - Returned CUDA mipmapped array * \param extMem - Handle to external memory object @@ -9961,7 +10511,9 @@ CUresult CUDAAPI cuExternalMemoryGetMappedBuffer(CUdeviceptr *devPtr, CUexternal * ::cuDestroyExternalMemory, * ::cuExternalMemoryGetMappedBuffer */ -CUresult CUDAAPI cuExternalMemoryGetMappedMipmappedArray(CUmipmappedArray *mipmap, CUexternalMemory extMem, const CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC *mipmapDesc); +CUresult CUDAAPI cuExternalMemoryGetMappedMipmappedArray( + CUmipmappedArray *mipmap, CUexternalMemory extMem, + const CUDA_EXTERNAL_MEMORY_MIPMAPPED_ARRAY_DESC *mipmapDesc); /** * \brief Destroys an external memory object. @@ -10080,7 +10632,9 @@ CUresult CUDAAPI cuDestroyExternalMemory(CUexternalMemory extMem); * ::cuSignalExternalSemaphoresAsync, * ::cuWaitExternalSemaphoresAsync */ -CUresult CUDAAPI cuImportExternalSemaphore(CUexternalSemaphore *extSem_out, const CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC *semHandleDesc); +CUresult CUDAAPI cuImportExternalSemaphore( + CUexternalSemaphore *extSem_out, + const CUDA_EXTERNAL_SEMAPHORE_HANDLE_DESC *semHandleDesc); /** * \brief Signals a set of external semaphore objects @@ -10118,7 +10672,10 @@ CUresult CUDAAPI cuImportExternalSemaphore(CUexternalSemaphore *extSem_out, cons * ::cuDestroyExternalSemaphore, * ::cuWaitExternalSemaphoresAsync */ -CUresult CUDAAPI cuSignalExternalSemaphoresAsync(const CUexternalSemaphore *extSemArray, const CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS *paramsArray, unsigned int numExtSems, CUstream stream); +CUresult CUDAAPI cuSignalExternalSemaphoresAsync( + const CUexternalSemaphore *extSemArray, + const CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS *paramsArray, + unsigned int numExtSems, CUstream stream); /** * \brief Waits on a set of external semaphore objects @@ -10160,7 +10717,10 @@ CUresult CUDAAPI cuSignalExternalSemaphoresAsync(const CUexternalSemaphore *extS * ::cuDestroyExternalSemaphore, * ::cuSignalExternalSemaphoresAsync */ -CUresult CUDAAPI cuWaitExternalSemaphoresAsync(const CUexternalSemaphore *extSemArray, const CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS *paramsArray, unsigned int numExtSems, CUstream stream); +CUresult CUDAAPI cuWaitExternalSemaphoresAsync( + const CUexternalSemaphore *extSemArray, + const CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS *paramsArray, + unsigned int numExtSems, CUstream stream); /** * \brief Destroys an external semaphore @@ -10247,7 +10807,8 @@ CUresult CUDAAPI cuDestroyExternalSemaphore(CUexternalSemaphore extSem); * Support for this can be queried with ::cuDeviceGetAttribute() and * ::CU_DEVICE_ATTRIBUTE_CAN_USE_STREAM_MEM_OPS. * - * Support for CU_STREAM_WAIT_VALUE_NOR can be queried with ::cuDeviceGetAttribute() and + * Support for CU_STREAM_WAIT_VALUE_NOR can be queried with + * ::cuDeviceGetAttribute() and * ::CU_DEVICE_ATTRIBUTE_CAN_USE_STREAM_WAIT_VALUE_NOR. * * \param stream The stream to synchronize on the memory location. @@ -10268,7 +10829,8 @@ CUresult CUDAAPI cuDestroyExternalSemaphore(CUexternalSemaphore extSem); * ::cuMemHostRegister, * ::cuStreamWaitEvent */ -CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags); +CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, + cuuint32_t value, unsigned int flags); /** * \brief Wait on a memory location @@ -10303,7 +10865,8 @@ CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, cuuint32 * ::cuMemHostRegister, * ::cuStreamWaitEvent */ -CUresult CUDAAPI cuStreamWaitValue64(CUstream stream, CUdeviceptr addr, cuuint64_t value, unsigned int flags); +CUresult CUDAAPI cuStreamWaitValue64(CUstream stream, CUdeviceptr addr, + cuuint64_t value, unsigned int flags); /** * \brief Write a value to memory @@ -10338,7 +10901,8 @@ CUresult CUDAAPI cuStreamWaitValue64(CUstream stream, CUdeviceptr addr, cuuint64 * ::cuMemHostRegister, * ::cuEventRecord */ -CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags); +CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, + cuuint32_t value, unsigned int flags); /** * \brief Write a value to memory @@ -10372,15 +10936,17 @@ CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, cuuint3 * ::cuMemHostRegister, * ::cuEventRecord */ -CUresult CUDAAPI cuStreamWriteValue64(CUstream stream, CUdeviceptr addr, cuuint64_t value, unsigned int flags); +CUresult CUDAAPI cuStreamWriteValue64(CUstream stream, CUdeviceptr addr, + cuuint64_t value, unsigned int flags); /** * \brief Batch operations to synchronize the stream via memory operations * - * This is a batch version of ::cuStreamWaitValue32() and ::cuStreamWriteValue32(). - * Batching operations may avoid some performance overhead in both the API call - * and the device execution versus adding them to the stream in separate API - * calls. The operations are enqueued in the order they appear in the array. + * This is a batch version of ::cuStreamWaitValue32() and + * ::cuStreamWriteValue32(). Batching operations may avoid some performance + * overhead in both the API call and the device execution versus adding them to + * the stream in separate API calls. The operations are enqueued in the order + * they appear in the array. * * See ::CUstreamBatchMemOpType for the full set of supported operations, and * ::cuStreamWaitValue32(), ::cuStreamWaitValue64(), ::cuStreamWriteValue32(), @@ -10407,7 +10973,9 @@ CUresult CUDAAPI cuStreamWriteValue64(CUstream stream, CUdeviceptr addr, cuuint6 * ::cuStreamWriteValue64, * ::cuMemHostRegister */ -CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, CUstreamBatchMemOpParams *paramArray, unsigned int flags); +CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, + CUstreamBatchMemOpParams *paramArray, + unsigned int flags); #endif /* __CUDA_API_VERSION >= 8000 */ /** @} */ /* END CUDA_MEMOP */ @@ -10456,10 +11024,10 @@ CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, CUstrea * version. * - ::CU_FUNC_CACHE_MODE_CA: The attribute to indicate whether the function has * been compiled with user specified option "-Xptxas --dlcm=ca" set . - * - ::CU_FUNC_ATTRIBUTE_MAX_DYNAMIC_SHARED_SIZE_BYTES: The maximum size in bytes of - * dynamically-allocated shared memory. - * - ::CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT: Preferred shared memory-L1 - * cache split ratio in percent of total shared memory. + * - ::CU_FUNC_ATTRIBUTE_MAX_DYNAMIC_SHARED_SIZE_BYTES: The maximum size in + * bytes of dynamically-allocated shared memory. + * - ::CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT: Preferred shared + * memory-L1 cache split ratio in percent of total shared memory. * * \param pi - Returned attribute value * \param attrib - Attribute requested @@ -10481,33 +11049,35 @@ CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, CUstrea * ::cudaFuncGetAttributes * ::cudaFuncSetAttribute */ -CUresult CUDAAPI cuFuncGetAttribute(int *pi, CUfunction_attribute attrib, CUfunction hfunc); +CUresult CUDAAPI cuFuncGetAttribute(int *pi, CUfunction_attribute attrib, + CUfunction hfunc); #if __CUDA_API_VERSION >= 9000 /** * \brief Sets information about a function * - * This call sets the value of a specified attribute \p attrib on the kernel given - * by \p hfunc to an integer value specified by \p val - * This function returns CUDA_SUCCESS if the new value of the attribute could be - * successfully set. If the set fails, this call will return an error. - * Not all attributes can have values set. Attempting to set a value on a read-only - * attribute will result in an error (CUDA_ERROR_INVALID_VALUE) + * This call sets the value of a specified attribute \p attrib on the kernel + * given by \p hfunc to an integer value specified by \p val This function + * returns CUDA_SUCCESS if the new value of the attribute could be successfully + * set. If the set fails, this call will return an error. Not all attributes can + * have values set. Attempting to set a value on a read-only attribute will + * result in an error (CUDA_ERROR_INVALID_VALUE) * * Supported attributes for the cuFuncSetAttribute call are: - * - ::CU_FUNC_ATTRIBUTE_MAX_DYNAMIC_SHARED_SIZE_BYTES: This maximum size in bytes of - * dynamically-allocated shared memory. The value should contain the requested - * maximum size of dynamically-allocated shared memory. The sum of this value and - * the function attribute ::CU_FUNC_ATTRIBUTE_SHARED_SIZE_BYTES cannot exceed the - * device attribute ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK_OPTIN. - * The maximal size of requestable dynamic shared memory may differ by GPU - * architecture. - * - ::CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT: On devices where the L1 - * cache and shared memory use the same hardware resources, this sets the shared memory - * carveout preference, in percent of the total shared memory. - * See ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_MULTIPROCESSOR - * This is only a hint, and the driver can choose a different ratio if required to execute the function. + * - ::CU_FUNC_ATTRIBUTE_MAX_DYNAMIC_SHARED_SIZE_BYTES: This maximum size in + * bytes of dynamically-allocated shared memory. The value should contain the + * requested maximum size of dynamically-allocated shared memory. The sum of + * this value and the function attribute ::CU_FUNC_ATTRIBUTE_SHARED_SIZE_BYTES + * cannot exceed the device attribute + * ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_BLOCK_OPTIN. The maximal size of + * requestable dynamic shared memory may differ by GPU architecture. + * - ::CU_FUNC_ATTRIBUTE_PREFERRED_SHARED_MEMORY_CARVEOUT: On devices where the + * L1 cache and shared memory use the same hardware resources, this sets the + * shared memory carveout preference, in percent of the total shared memory. See + * ::CU_DEVICE_ATTRIBUTE_MAX_SHARED_MEMORY_PER_MULTIPROCESSOR This is only a + * hint, and the driver can choose a different ratio if required to execute the + * function. * * \param hfunc - Function to query attribute of * \param attrib - Attribute requested @@ -10529,8 +11099,9 @@ CUresult CUDAAPI cuFuncGetAttribute(int *pi, CUfunction_attribute attrib, CUfunc * ::cudaFuncGetAttributes * ::cudaFuncSetAttribute */ -CUresult CUDAAPI cuFuncSetAttribute(CUfunction hfunc, CUfunction_attribute attrib, int value); -#endif // __CUDA_API_VERSION >= 9000 +CUresult CUDAAPI cuFuncSetAttribute(CUfunction hfunc, + CUfunction_attribute attrib, int value); +#endif // __CUDA_API_VERSION >= 9000 /** * \brief Sets the preferred cache configuration for a device function @@ -10552,8 +11123,10 @@ CUresult CUDAAPI cuFuncSetAttribute(CUfunction hfunc, CUfunction_attribute attri * * * The supported cache configurations are: - * - ::CU_FUNC_CACHE_PREFER_NONE: no preference for shared memory or L1 (default) - * - ::CU_FUNC_CACHE_PREFER_SHARED: prefer larger shared memory and smaller L1 cache + * - ::CU_FUNC_CACHE_PREFER_NONE: no preference for shared memory or L1 + * (default) + * - ::CU_FUNC_CACHE_PREFER_SHARED: prefer larger shared memory and smaller L1 + * cache * - ::CU_FUNC_CACHE_PREFER_L1: prefer larger L1 cache and smaller shared memory * - ::CU_FUNC_CACHE_PREFER_EQUAL: prefer equal sized L1 cache and shared memory * @@ -10594,9 +11167,9 @@ CUresult CUDAAPI cuFuncSetCacheConfig(CUfunction hfunc, CUfunc_cache config); * * Changing the shared memory bank size will not increase shared memory usage * or affect occupancy of kernels, but may have major effects on performance. - * Larger bank sizes will allow for greater potential bandwidth to shared memory, - * but will change what kinds of accesses to shared memory will result in bank - * conflicts. + * Larger bank sizes will allow for greater potential bandwidth to shared + * memory, but will change what kinds of accesses to shared memory will result + * in bank conflicts. * * This function will do nothing on devices with fixed shared memory bank size. * @@ -10605,8 +11178,8 @@ CUresult CUDAAPI cuFuncSetCacheConfig(CUfunction hfunc, CUfunc_cache config); * configuration when launching this function. * - ::CU_SHARED_MEM_CONFIG_FOUR_BYTE_BANK_SIZE: set shared memory bank width to * be natively four bytes when launching this function. - * - ::CU_SHARED_MEM_CONFIG_EIGHT_BYTE_BANK_SIZE: set shared memory bank width to - * be natively eight bytes when launching this function. + * - ::CU_SHARED_MEM_CONFIG_EIGHT_BYTE_BANK_SIZE: set shared memory bank width + * to be natively eight bytes when launching this function. * * \param hfunc - kernel to be given a shared memory config * \param config - requested shared memory configuration @@ -10627,7 +11200,8 @@ CUresult CUDAAPI cuFuncSetCacheConfig(CUfunction hfunc, CUfunc_cache config); * ::cuLaunchKernel, * ::cudaFuncSetSharedMemConfig */ -CUresult CUDAAPI cuFuncSetSharedMemConfig(CUfunction hfunc, CUsharedconfig config); +CUresult CUDAAPI cuFuncSetSharedMemConfig(CUfunction hfunc, + CUsharedconfig config); #endif #if __CUDA_API_VERSION >= 4000 @@ -10742,21 +11316,17 @@ CUresult CUDAAPI cuFuncSetSharedMemConfig(CUfunction hfunc, CUsharedconfig confi * ::cuFuncGetAttribute, * ::cudaLaunchKernel */ -CUresult CUDAAPI cuLaunchKernel(CUfunction f, - unsigned int gridDimX, - unsigned int gridDimY, - unsigned int gridDimZ, - unsigned int blockDimX, - unsigned int blockDimY, +CUresult CUDAAPI cuLaunchKernel(CUfunction f, unsigned int gridDimX, + unsigned int gridDimY, unsigned int gridDimZ, + unsigned int blockDimX, unsigned int blockDimY, unsigned int blockDimZ, - unsigned int sharedMemBytes, - CUstream hStream, - void **kernelParams, - void **extra); + unsigned int sharedMemBytes, CUstream hStream, + void **kernelParams, void **extra); #endif /* __CUDA_API_VERSION >= 4000 */ #if __CUDA_API_VERSION >= 9000 /** - * \brief Launches a CUDA function where thread blocks can cooperate and synchronize as they execute + * \brief Launches a CUDA function where thread blocks can cooperate and + * synchronize as they execute * * Invokes the kernel \p f on a \p gridDimX x \p gridDimY x \p gridDimZ * grid of blocks. Each block contains \p blockDimX x \p blockDimY x @@ -10768,10 +11338,12 @@ CUresult CUDAAPI cuLaunchKernel(CUfunction f, * The device on which this kernel is invoked must have a non-zero value for * the device attribute ::CU_DEVICE_ATTRIBUTE_COOPERATIVE_LAUNCH. * - * The total number of blocks launched cannot exceed the maximum number of blocks per - * multiprocessor as returned by ::cuOccupancyMaxActiveBlocksPerMultiprocessor (or - * ::cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags) times the number of multiprocessors - * as specified by the device attribute ::CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT. + * The total number of blocks launched cannot exceed the maximum number of + * blocks per multiprocessor as returned by + * ::cuOccupancyMaxActiveBlocksPerMultiprocessor (or + * ::cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags) times the number of + * multiprocessors as specified by the device attribute + * ::CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT. * * The kernel cannot make use of CUDA dynamic parallelism. * @@ -10786,15 +11358,15 @@ CUresult CUDAAPI cuLaunchKernel(CUfunction f, * Calling ::cuLaunchCooperativeKernel() sets persistent function state that is * the same as function state set through ::cuLaunchKernel API * - * When the kernel \p f is launched via ::cuLaunchCooperativeKernel(), the previous - * block shape, shared size and parameter info associated with \p f - * is overwritten. + * When the kernel \p f is launched via ::cuLaunchCooperativeKernel(), the + * previous block shape, shared size and parameter info associated with \p f is + * overwritten. * - * Note that to use ::cuLaunchCooperativeKernel(), the kernel \p f must either have - * been compiled with toolchain version 3.2 or later so that it will + * Note that to use ::cuLaunchCooperativeKernel(), the kernel \p f must either + * have been compiled with toolchain version 3.2 or later so that it will * contain kernel parameter information, or have no kernel parameters. - * If either of these conditions is not met, then ::cuLaunchCooperativeKernel() will - * return ::CUDA_ERROR_INVALID_IMAGE. + * If either of these conditions is not met, then ::cuLaunchCooperativeKernel() + * will return ::CUDA_ERROR_INVALID_IMAGE. * * \param f - Kernel to launch * \param gridDimX - Width of grid in blocks @@ -10831,49 +11403,64 @@ CUresult CUDAAPI cuLaunchKernel(CUfunction f, * ::cuLaunchCooperativeKernelMultiDevice, * ::cudaLaunchCooperativeKernel */ -CUresult CUDAAPI cuLaunchCooperativeKernel(CUfunction f, - unsigned int gridDimX, - unsigned int gridDimY, - unsigned int gridDimZ, - unsigned int blockDimX, - unsigned int blockDimY, - unsigned int blockDimZ, - unsigned int sharedMemBytes, - CUstream hStream, - void **kernelParams); +CUresult CUDAAPI cuLaunchCooperativeKernel( + CUfunction f, unsigned int gridDimX, unsigned int gridDimY, + unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, + unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, + void **kernelParams); /** - * \brief Launches CUDA functions on multiple devices where thread blocks can cooperate and synchronize as they execute + * \brief Launches CUDA functions on multiple devices where thread blocks can + cooperate and synchronize as they execute * - * Invokes kernels as specified in the \p launchParamsList array where each element - * of the array specifies all the parameters required to perform a single kernel launch. - * These kernels can cooperate and synchronize as they execute. The size of the array is + * Invokes kernels as specified in the \p launchParamsList array where each + element + * of the array specifies all the parameters required to perform a single kernel + launch. + * These kernels can cooperate and synchronize as they execute. The size of the + array is * specified by \p numDevices. * - * No two kernels can be launched on the same device. All the devices targeted by this - * multi-device launch must be identical. All devices must have a non-zero value for the + * No two kernels can be launched on the same device. All the devices targeted + by this + * multi-device launch must be identical. All devices must have a non-zero value + for the * device attribute ::CU_DEVICE_ATTRIBUTE_COOPERATIVE_MULTI_DEVICE_LAUNCH. * - * All kernels launched must be identical with respect to the compiled code. Note that - * any __device__, __constant__ or __managed__ variables present in the module that owns - * the kernel launched on each device, are independently instantiated on every device. - * It is the application's responsiblity to ensure these variables are initialized and + * All kernels launched must be identical with respect to the compiled code. + Note that + * any __device__, __constant__ or __managed__ variables present in the module + that owns + * the kernel launched on each device, are independently instantiated on every + device. + * It is the application's responsiblity to ensure these variables are + initialized and * used appropriately. * - * The size of the grids as specified in blocks, the size of the blocks themselves - * and the amount of shared memory used by each thread block must also match across + * The size of the grids as specified in blocks, the size of the blocks + themselves + * and the amount of shared memory used by each thread block must also match + across * all launched kernels. * - * The streams used to launch these kernels must have been created via either ::cuStreamCreate - * or ::cuStreamCreateWithPriority. The NULL stream or ::CU_STREAM_LEGACY or ::CU_STREAM_PER_THREAD + * The streams used to launch these kernels must have been created via either + ::cuStreamCreate + * or ::cuStreamCreateWithPriority. The NULL stream or ::CU_STREAM_LEGACY or + ::CU_STREAM_PER_THREAD * cannot be used. * - * The total number of blocks launched per kernel cannot exceed the maximum number of blocks - * per multiprocessor as returned by ::cuOccupancyMaxActiveBlocksPerMultiprocessor (or - * ::cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags) times the number of multiprocessors - * as specified by the device attribute ::CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT. Since the - * total number of blocks launched per device has to match across all devices, the maximum - * number of blocks that can be launched per device will be limited by the device with the + * The total number of blocks launched per kernel cannot exceed the maximum + number of blocks + * per multiprocessor as returned by + ::cuOccupancyMaxActiveBlocksPerMultiprocessor (or + * ::cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags) times the number of + multiprocessors + * as specified by the device attribute + ::CU_DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT. Since the + * total number of blocks launched per device has to match across all devices, + the maximum + * number of blocks that can be launched per device will be limited by the + device with the * least number of multiprocessors. * * The kernels cannot make use of CUDA dynamic parallelism. @@ -10895,56 +11482,87 @@ CUresult CUDAAPI cuLaunchCooperativeKernel(CUfunction f, } CUDA_LAUNCH_PARAMS; * \endcode * where: - * - ::CUDA_LAUNCH_PARAMS::function specifies the kernel to be launched. All functions must + * - ::CUDA_LAUNCH_PARAMS::function specifies the kernel to be launched. All + functions must * be identical with respect to the compiled code. - * - ::CUDA_LAUNCH_PARAMS::gridDimX is the width of the grid in blocks. This must match across + * - ::CUDA_LAUNCH_PARAMS::gridDimX is the width of the grid in blocks. This + must match across * all kernels launched. - * - ::CUDA_LAUNCH_PARAMS::gridDimY is the height of the grid in blocks. This must match across + * - ::CUDA_LAUNCH_PARAMS::gridDimY is the height of the grid in blocks. This + must match across * all kernels launched. - * - ::CUDA_LAUNCH_PARAMS::gridDimZ is the depth of the grid in blocks. This must match across + * - ::CUDA_LAUNCH_PARAMS::gridDimZ is the depth of the grid in blocks. This + must match across * all kernels launched. - * - ::CUDA_LAUNCH_PARAMS::blockDimX is the X dimension of each thread block. This must match across + * - ::CUDA_LAUNCH_PARAMS::blockDimX is the X dimension of each thread block. + This must match across * all kernels launched. - * - ::CUDA_LAUNCH_PARAMS::blockDimX is the Y dimension of each thread block. This must match across + * - ::CUDA_LAUNCH_PARAMS::blockDimX is the Y dimension of each thread block. + This must match across * all kernels launched. - * - ::CUDA_LAUNCH_PARAMS::blockDimZ is the Z dimension of each thread block. This must match across + * - ::CUDA_LAUNCH_PARAMS::blockDimZ is the Z dimension of each thread block. + This must match across * all kernels launched. - * - ::CUDA_LAUNCH_PARAMS::sharedMemBytes is the dynamic shared-memory size per thread block in bytes. + * - ::CUDA_LAUNCH_PARAMS::sharedMemBytes is the dynamic shared-memory size per + thread block in bytes. * This must match across all kernels launched. - * - ::CUDA_LAUNCH_PARAMS::hStream is the handle to the stream to perform the launch in. This cannot - * be the NULL stream or ::CU_STREAM_LEGACY or ::CU_STREAM_PER_THREAD. The CUDA context associated - * with this stream must match that associated with ::CUDA_LAUNCH_PARAMS::function. - * - ::CUDA_LAUNCH_PARAMS::kernelParams is an array of pointers to kernel parameters. If - * ::CUDA_LAUNCH_PARAMS::function has N parameters, then ::CUDA_LAUNCH_PARAMS::kernelParams - * needs to be an array of N pointers. Each of ::CUDA_LAUNCH_PARAMS::kernelParams[0] through - * ::CUDA_LAUNCH_PARAMS::kernelParams[N-1] must point to a region of memory from which the actual - * kernel parameter will be copied. The number of kernel parameters and their offsets and sizes - * do not need to be specified as that information is retrieved directly from the kernel's image. - * - * By default, the kernel won't begin execution on any GPU until all prior work in all the specified + * - ::CUDA_LAUNCH_PARAMS::hStream is the handle to the stream to perform the + launch in. This cannot + * be the NULL stream or ::CU_STREAM_LEGACY or ::CU_STREAM_PER_THREAD. The + CUDA context associated + * with this stream must match that associated with + ::CUDA_LAUNCH_PARAMS::function. + * - ::CUDA_LAUNCH_PARAMS::kernelParams is an array of pointers to kernel + parameters. If + * ::CUDA_LAUNCH_PARAMS::function has N parameters, then + ::CUDA_LAUNCH_PARAMS::kernelParams + * needs to be an array of N pointers. Each of + ::CUDA_LAUNCH_PARAMS::kernelParams[0] through + * ::CUDA_LAUNCH_PARAMS::kernelParams[N-1] must point to a region of memory + from which the actual + * kernel parameter will be copied. The number of kernel parameters and their + offsets and sizes + * do not need to be specified as that information is retrieved directly from + the kernel's image. + * + * By default, the kernel won't begin execution on any GPU until all prior work + in all the specified * streams has completed. This behavior can be overridden by specifying the flag - * ::CUDA_COOPERATIVE_LAUNCH_MULTI_DEVICE_NO_PRE_LAUNCH_SYNC. When this flag is specified, each kernel - * will only wait for prior work in the stream corresponding to that GPU to complete before it begins + * ::CUDA_COOPERATIVE_LAUNCH_MULTI_DEVICE_NO_PRE_LAUNCH_SYNC. When this flag is + specified, each kernel + * will only wait for prior work in the stream corresponding to that GPU to + complete before it begins * execution. * - * Similarly, by default, any subsequent work pushed in any of the specified streams will not begin - * execution until the kernels on all GPUs have completed. This behavior can be overridden by specifying - * the flag ::CUDA_COOPERATIVE_LAUNCH_MULTI_DEVICE_NO_POST_LAUNCH_SYNC. When this flag is specified, - * any subsequent work pushed in any of the specified streams will only wait for the kernel launched - * on the GPU corresponding to that stream to complete before it begins execution. - * - * Calling ::cuLaunchCooperativeKernelMultiDevice() sets persistent function state that is - * the same as function state set through ::cuLaunchKernel API when called individually for each + * Similarly, by default, any subsequent work pushed in any of the specified + streams will not begin + * execution until the kernels on all GPUs have completed. This behavior can be + overridden by specifying + * the flag ::CUDA_COOPERATIVE_LAUNCH_MULTI_DEVICE_NO_POST_LAUNCH_SYNC. When + this flag is specified, + * any subsequent work pushed in any of the specified streams will only wait for + the kernel launched + * on the GPU corresponding to that stream to complete before it begins + execution. + * + * Calling ::cuLaunchCooperativeKernelMultiDevice() sets persistent function + state that is + * the same as function state set through ::cuLaunchKernel API when called + individually for each * element in \p launchParamsList. * - * When kernels are launched via ::cuLaunchCooperativeKernelMultiDevice(), the previous - * block shape, shared size and parameter info associated with each ::CUDA_LAUNCH_PARAMS::function + * When kernels are launched via ::cuLaunchCooperativeKernelMultiDevice(), the + previous + * block shape, shared size and parameter info associated with each + ::CUDA_LAUNCH_PARAMS::function * in \p launchParamsList is overwritten. * - * Note that to use ::cuLaunchCooperativeKernelMultiDevice(), the kernels must either have + * Note that to use ::cuLaunchCooperativeKernelMultiDevice(), the kernels must + either have * been compiled with toolchain version 3.2 or later so that it will * contain kernel parameter information, or have no kernel parameters. - * If either of these conditions is not met, then ::cuLaunchCooperativeKernelMultiDevice() will + * If either of these conditions is not met, then + ::cuLaunchCooperativeKernelMultiDevice() will * return ::CUDA_ERROR_INVALID_IMAGE. * * \param launchParamsList - List of launch parameters, one per device @@ -10975,7 +11593,9 @@ CUresult CUDAAPI cuLaunchCooperativeKernel(CUfunction f, * ::cuLaunchCooperativeKernel, * ::cudaLaunchCooperativeKernelMultiDevice */ -CUresult CUDAAPI cuLaunchCooperativeKernelMultiDevice(CUDA_LAUNCH_PARAMS *launchParamsList, unsigned int numDevices, unsigned int flags); +CUresult CUDAAPI cuLaunchCooperativeKernelMultiDevice( + CUDA_LAUNCH_PARAMS *launchParamsList, unsigned int numDevices, + unsigned int flags); #endif /* __CUDA_API_VERSION >= 9000 */ @@ -11022,8 +11642,8 @@ CUresult CUDAAPI cuLaunchCooperativeKernelMultiDevice(CUDA_LAUNCH_PARAMS *launch * called in the event of an error in the CUDA context. * * \param hStream - Stream to enqueue function call in - * \param fn - The function to call once preceding stream operations are complete - * \param userData - User-specified data to be passed to the function + * \param fn - The function to call once preceding stream operations are + * complete \param userData - User-specified data to be passed to the function * * \return * ::CUDA_SUCCESS, @@ -11044,7 +11664,8 @@ CUresult CUDAAPI cuLaunchCooperativeKernelMultiDevice(CUDA_LAUNCH_PARAMS *launch * ::cuStreamAttachMemAsync, * ::cuStreamAddCallback */ -CUresult CUDAAPI cuLaunchHostFunc(CUstream hStream, CUhostFn fn, void *userData); +CUresult CUDAAPI cuLaunchHostFunc(CUstream hStream, CUhostFn fn, + void *userData); #endif /* __CUDA_API_VERSION >= 10000 */ @@ -11096,7 +11717,8 @@ CUresult CUDAAPI cuLaunchHostFunc(CUstream hStream, CUhostFn fn, void *userData) * ::cuLaunchGridAsync, * ::cuLaunchKernel */ -__CUDA_DEPRECATED CUresult CUDAAPI cuFuncSetBlockShape(CUfunction hfunc, int x, int y, int z); +__CUDA_DEPRECATED CUresult CUDAAPI cuFuncSetBlockShape(CUfunction hfunc, int x, + int y, int z); /** * \brief Sets the dynamic shared-memory size for the function @@ -11130,7 +11752,8 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuFuncSetBlockShape(CUfunction hfunc, int x, * ::cuLaunchGridAsync, * ::cuLaunchKernel */ -__CUDA_DEPRECATED CUresult CUDAAPI cuFuncSetSharedSize(CUfunction hfunc, unsigned int bytes); +__CUDA_DEPRECATED CUresult CUDAAPI cuFuncSetSharedSize(CUfunction hfunc, + unsigned int bytes); /** * \brief Sets the parameter size for the function @@ -11162,7 +11785,8 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuFuncSetSharedSize(CUfunction hfunc, unsigne * ::cuLaunchGridAsync, * ::cuLaunchKernel */ -__CUDA_DEPRECATED CUresult CUDAAPI cuParamSetSize(CUfunction hfunc, unsigned int numbytes); +__CUDA_DEPRECATED CUresult CUDAAPI cuParamSetSize(CUfunction hfunc, + unsigned int numbytes); /** * \brief Adds an integer parameter to the function's argument list @@ -11195,7 +11819,8 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuParamSetSize(CUfunction hfunc, unsigned int * ::cuLaunchGridAsync, * ::cuLaunchKernel */ -__CUDA_DEPRECATED CUresult CUDAAPI cuParamSeti(CUfunction hfunc, int offset, unsigned int value); +__CUDA_DEPRECATED CUresult CUDAAPI cuParamSeti(CUfunction hfunc, int offset, + unsigned int value); /** * \brief Adds a floating-point parameter to the function's argument list @@ -11228,7 +11853,8 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuParamSeti(CUfunction hfunc, int offset, uns * ::cuLaunchGridAsync, * ::cuLaunchKernel */ -__CUDA_DEPRECATED CUresult CUDAAPI cuParamSetf(CUfunction hfunc, int offset, float value); +__CUDA_DEPRECATED CUresult CUDAAPI cuParamSetf(CUfunction hfunc, int offset, + float value); /** * \brief Adds arbitrary data to the function's argument list @@ -11263,7 +11889,9 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuParamSetf(CUfunction hfunc, int offset, flo * ::cuLaunchGridAsync, * ::cuLaunchKernel */ -__CUDA_DEPRECATED CUresult CUDAAPI cuParamSetv(CUfunction hfunc, int offset, void *ptr, unsigned int numbytes); +__CUDA_DEPRECATED CUresult CUDAAPI cuParamSetv(CUfunction hfunc, int offset, + void *ptr, + unsigned int numbytes); /** * \brief Launches a CUDA function @@ -11339,7 +11967,8 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuLaunch(CUfunction f); * ::cuLaunchGridAsync, * ::cuLaunchKernel */ -__CUDA_DEPRECATED CUresult CUDAAPI cuLaunchGrid(CUfunction f, int grid_width, int grid_height); +__CUDA_DEPRECATED CUresult CUDAAPI cuLaunchGrid(CUfunction f, int grid_width, + int grid_height); /** * \brief Launches a CUDA function @@ -11368,9 +11997,10 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuLaunchGrid(CUfunction f, int grid_width, in * ::CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING, * ::CUDA_ERROR_SHARED_OBJECT_INIT_FAILED * - * \note In certain cases where cubins are created with no ABI (i.e., using \p ptxas \p --abi-compile \p no), - * this function may serialize kernel launches. In order to force the CUDA driver to retain - * asynchronous behavior, set the ::CU_CTX_LMEM_RESIZE_TO_MAX flag during context creation (see ::cuCtxCreate). + * \note In certain cases where cubins are created with no ABI (i.e., using \p + * ptxas \p --abi-compile \p no), this function may serialize kernel launches. + * In order to force the CUDA driver to retain asynchronous behavior, set the + * ::CU_CTX_LMEM_RESIZE_TO_MAX flag during context creation (see ::cuCtxCreate). * * \note_null_stream * \notefnerr @@ -11386,8 +12016,10 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuLaunchGrid(CUfunction f, int grid_width, in * ::cuLaunchGrid, * ::cuLaunchKernel */ -__CUDA_DEPRECATED CUresult CUDAAPI cuLaunchGridAsync(CUfunction f, int grid_width, int grid_height, CUstream hStream); - +__CUDA_DEPRECATED CUresult CUDAAPI cuLaunchGridAsync(CUfunction f, + int grid_width, + int grid_height, + CUstream hStream); /** * \brief Adds a texture-reference to the function's argument list @@ -11411,7 +12043,9 @@ __CUDA_DEPRECATED CUresult CUDAAPI cuLaunchGridAsync(CUfunction f, int grid_widt * ::CUDA_ERROR_INVALID_VALUE * \notefnerr */ -__CUDA_DEPRECATED CUresult CUDAAPI cuParamSetTexRef(CUfunction hfunc, int texunit, CUtexref hTexRef); +__CUDA_DEPRECATED CUresult CUDAAPI cuParamSetTexRef(CUfunction hfunc, + int texunit, + CUtexref hTexRef); /** @} */ /* END CUDA_EXEC_DEPRECATED */ #if __CUDA_API_VERSION >= 10000 @@ -11463,11 +12097,12 @@ CUresult CUDAAPI cuGraphCreate(CUgraph *phGraph, unsigned int flags); /** * \brief Creates a kernel execution node and adds it to a graph * - * Creates a new kernel execution node and adds it to \p hGraph with \p numDependencies - * dependencies specified via \p dependencies and arguments specified in \p nodeParams. - * It is possible for \p numDependencies to be 0, in which case the node will be placed - * at the root of the graph. \p dependencies may not have any duplicate entries. - * A handle to the new node will be returned in \p phGraphNode. + * Creates a new kernel execution node and adds it to \p hGraph with \p + * numDependencies dependencies specified via \p dependencies and arguments + * specified in \p nodeParams. It is possible for \p numDependencies to be 0, in + * which case the node will be placed at the root of the graph. \p dependencies + * may not have any duplicate entries. A handle to the new node will be returned + * in \p phGraphNode. * * The CUDA_KERNEL_NODE_PARAMS structure is defined as: * @@ -11486,8 +12121,8 @@ CUresult CUDAAPI cuGraphCreate(CUgraph *phGraph, unsigned int flags); * } CUDA_KERNEL_NODE_PARAMS; * \endcode * - * When the graph is launched, the node will invoke kernel \p func on a (\p gridDimX x - * \p gridDimY x \p gridDimZ) grid of blocks. Each block contains + * When the graph is launched, the node will invoke kernel \p func on a (\p + * gridDimX x \p gridDimY x \p gridDimZ) grid of blocks. Each block contains * (\p blockDimX x \p blockDimY x \p blockDimZ) threads. * * \p sharedMemBytes sets the amount of dynamic shared memory that will be @@ -11495,19 +12130,21 @@ CUresult CUDAAPI cuGraphCreate(CUgraph *phGraph, unsigned int flags); * * Kernel parameters to \p func can be specified in one of two ways: * - * 1) Kernel parameters can be specified via \p kernelParams. If the kernel has N - * parameters, then \p kernelParams needs to be an array of N pointers. Each pointer, - * from \p kernelParams[0] to \p kernelParams[N-1], points to the region of memory from which the actual - * parameter will be copied. The number of kernel parameters and their offsets and sizes do not need - * to be specified as that information is retrieved directly from the kernel's image. - * - * 2) Kernel parameters can also be packaged by the application into a single buffer that is passed in - * via \p extra. This places the burden on the application of knowing each kernel - * parameter's size and alignment/padding within the buffer. The \p extra parameter exists - * to allow this function to take additional less commonly used arguments. \p extra specifies - * a list of names of extra settings and their corresponding values. Each extra setting name is - * immediately followed by the corresponding value. The list must be terminated with either NULL or - * CU_LAUNCH_PARAM_END. + * 1) Kernel parameters can be specified via \p kernelParams. If the kernel has + * N parameters, then \p kernelParams needs to be an array of N pointers. Each + * pointer, from \p kernelParams[0] to \p kernelParams[N-1], points to the + * region of memory from which the actual parameter will be copied. The number + * of kernel parameters and their offsets and sizes do not need to be specified + * as that information is retrieved directly from the kernel's image. + * + * 2) Kernel parameters can also be packaged by the application into a single + * buffer that is passed in via \p extra. This places the burden on the + * application of knowing each kernel parameter's size and alignment/padding + * within the buffer. The \p extra parameter exists to allow this function to + * take additional less commonly used arguments. \p extra specifies a list of + * names of extra settings and their corresponding values. Each extra setting + * name is immediately followed by the corresponding value. The list must be + * terminated with either NULL or CU_LAUNCH_PARAM_END. * * - ::CU_LAUNCH_PARAM_END, which indicates the end of the \p extra * array; @@ -11520,16 +12157,17 @@ CUresult CUDAAPI cuGraphCreate(CUgraph *phGraph, unsigned int flags); * containing the size of the buffer specified with * ::CU_LAUNCH_PARAM_BUFFER_POINTER; * - * The error ::CUDA_ERROR_INVALID_VALUE will be returned if kernel parameters are specified with both - * \p kernelParams and \p extra (i.e. both \p kernelParams and - * \p extra are non-NULL). + * The error ::CUDA_ERROR_INVALID_VALUE will be returned if kernel parameters + * are specified with both \p kernelParams and \p extra (i.e. both \p + * kernelParams and \p extra are non-NULL). * - * The \p kernelParams or \p extra array, as well as the argument values it points to, - * are copied during this call. + * The \p kernelParams or \p extra array, as well as the argument values it + * points to, are copied during this call. * - * \note Kernels launched using graphs must not use texture and surface references. Reading or - * writing through any texture or surface reference is undefined behavior. - * This restriction does not apply to texture and surface objects. + * \note Kernels launched using graphs must not use texture and surface + * references. Reading or writing through any texture or surface reference is + * undefined behavior. This restriction does not apply to texture and surface + * objects. * * \param phGraphNode - Returns newly created node * \param hGraph - Graph to which to add the node @@ -11557,7 +12195,9 @@ CUresult CUDAAPI cuGraphCreate(CUgraph *phGraph, unsigned int flags); * ::cuGraphAddMemcpyNode, * ::cuGraphAddMemsetNode */ -CUresult CUDAAPI cuGraphAddKernelNode(CUgraphNode *phGraphNode, CUgraph hGraph, const CUgraphNode *dependencies, size_t numDependencies, const CUDA_KERNEL_NODE_PARAMS *nodeParams); +CUresult CUDAAPI cuGraphAddKernelNode( + CUgraphNode *phGraphNode, CUgraph hGraph, const CUgraphNode *dependencies, + size_t numDependencies, const CUDA_KERNEL_NODE_PARAMS *nodeParams); /** * \brief Returns a kernel node's parameters @@ -11589,7 +12229,8 @@ CUresult CUDAAPI cuGraphAddKernelNode(CUgraphNode *phGraphNode, CUgraph hGraph, * ::cuGraphAddKernelNode, * ::cuGraphKernelNodeSetParams */ -CUresult CUDAAPI cuGraphKernelNodeGetParams(CUgraphNode hNode, CUDA_KERNEL_NODE_PARAMS *nodeParams); +CUresult CUDAAPI cuGraphKernelNodeGetParams( + CUgraphNode hNode, CUDA_KERNEL_NODE_PARAMS *nodeParams); /** * \brief Sets a kernel node's parameters @@ -11612,26 +12253,30 @@ CUresult CUDAAPI cuGraphKernelNodeGetParams(CUgraphNode hNode, CUDA_KERNEL_NODE_ * ::cuGraphAddKernelNode, * ::cuGraphKernelNodeGetParams */ -CUresult CUDAAPI cuGraphKernelNodeSetParams(CUgraphNode hNode, const CUDA_KERNEL_NODE_PARAMS *nodeParams); +CUresult CUDAAPI cuGraphKernelNodeSetParams( + CUgraphNode hNode, const CUDA_KERNEL_NODE_PARAMS *nodeParams); /** * \brief Creates a memcpy node and adds it to a graph * * Creates a new memcpy node and adds it to \p hGraph with \p numDependencies * dependencies specified via \p dependencies. - * It is possible for \p numDependencies to be 0, in which case the node will be placed - * at the root of the graph. \p dependencies may not have any duplicate entries. - * A handle to the new node will be returned in \p phGraphNode. - * - * When the graph is launched, the node will perform the memcpy described by \p copyParams. - * See ::cuMemcpy3D() for a description of the structure and its restrictions. - * - * Memcpy nodes have some additional restrictions with regards to managed memory, if the - * system contains at least one device which has a zero value for the device attribute - * ::CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS. If one or more of the operands refer - * to managed memory, then using the memory type ::CU_MEMORYTYPE_UNIFIED is disallowed - * for those operand(s). The managed memory will be treated as residing on either the - * host or the device, depending on which memory type is specified. + * It is possible for \p numDependencies to be 0, in which case the node will be + * placed at the root of the graph. \p dependencies may not have any duplicate + * entries. A handle to the new node will be returned in \p phGraphNode. + * + * When the graph is launched, the node will perform the memcpy described by \p + * copyParams. See ::cuMemcpy3D() for a description of the structure and its + * restrictions. + * + * Memcpy nodes have some additional restrictions with regards to managed + * memory, if the system contains at least one device which has a zero value for + * the device attribute + * ::CU_DEVICE_ATTRIBUTE_CONCURRENT_MANAGED_ACCESS. If one or more of the + * operands refer to managed memory, then using the memory type + * ::CU_MEMORYTYPE_UNIFIED is disallowed for those operand(s). The managed + * memory will be treated as residing on either the host or the device, + * depending on which memory type is specified. * * \param phGraphNode - Returns newly created node * \param hGraph - Graph to which to add the node @@ -11660,7 +12305,11 @@ CUresult CUDAAPI cuGraphKernelNodeSetParams(CUgraphNode hNode, const CUDA_KERNEL * ::cuGraphAddHostNode, * ::cuGraphAddMemsetNode */ -CUresult CUDAAPI cuGraphAddMemcpyNode(CUgraphNode *phGraphNode, CUgraph hGraph, const CUgraphNode *dependencies, size_t numDependencies, const CUDA_MEMCPY3D *copyParams, CUcontext ctx); +CUresult CUDAAPI cuGraphAddMemcpyNode(CUgraphNode *phGraphNode, CUgraph hGraph, + const CUgraphNode *dependencies, + size_t numDependencies, + const CUDA_MEMCPY3D *copyParams, + CUcontext ctx); /** * \brief Returns a memcpy node's parameters @@ -11683,7 +12332,8 @@ CUresult CUDAAPI cuGraphAddMemcpyNode(CUgraphNode *phGraphNode, CUgraph hGraph, * ::cuGraphAddMemcpyNode, * ::cuGraphMemcpyNodeSetParams */ -CUresult CUDAAPI cuGraphMemcpyNodeGetParams(CUgraphNode hNode, CUDA_MEMCPY3D *nodeParams); +CUresult CUDAAPI cuGraphMemcpyNodeGetParams(CUgraphNode hNode, + CUDA_MEMCPY3D *nodeParams); /** * \brief Sets a memcpy node's parameters @@ -11706,19 +12356,21 @@ CUresult CUDAAPI cuGraphMemcpyNodeGetParams(CUgraphNode hNode, CUDA_MEMCPY3D *no * ::cuGraphAddMemcpyNode, * ::cuGraphMemcpyNodeGetParams */ -CUresult CUDAAPI cuGraphMemcpyNodeSetParams(CUgraphNode hNode, const CUDA_MEMCPY3D *nodeParams); +CUresult CUDAAPI cuGraphMemcpyNodeSetParams(CUgraphNode hNode, + const CUDA_MEMCPY3D *nodeParams); /** * \brief Creates a memset node and adds it to a graph * * Creates a new memset node and adds it to \p hGraph with \p numDependencies * dependencies specified via \p dependencies. - * It is possible for \p numDependencies to be 0, in which case the node will be placed - * at the root of the graph. \p dependencies may not have any duplicate entries. - * A handle to the new node will be returned in \p phGraphNode. + * It is possible for \p numDependencies to be 0, in which case the node will be + * placed at the root of the graph. \p dependencies may not have any duplicate + * entries. A handle to the new node will be returned in \p phGraphNode. * * The element size must be 1, 2, or 4 bytes. - * When the graph is launched, the node will perform the memset described by \p memsetParams. + * When the graph is launched, the node will perform the memset described by \p + * memsetParams. * * \param phGraphNode - Returns newly created node * \param hGraph - Graph to which to add the node @@ -11748,7 +12400,10 @@ CUresult CUDAAPI cuGraphMemcpyNodeSetParams(CUgraphNode hNode, const CUDA_MEMCPY * ::cuGraphAddHostNode, * ::cuGraphAddMemcpyNode */ -CUresult CUDAAPI cuGraphAddMemsetNode(CUgraphNode *phGraphNode, CUgraph hGraph, const CUgraphNode *dependencies, size_t numDependencies, const CUDA_MEMSET_NODE_PARAMS *memsetParams, CUcontext ctx); +CUresult CUDAAPI cuGraphAddMemsetNode( + CUgraphNode *phGraphNode, CUgraph hGraph, const CUgraphNode *dependencies, + size_t numDependencies, const CUDA_MEMSET_NODE_PARAMS *memsetParams, + CUcontext ctx); /** * \brief Returns a memset node's parameters @@ -11771,7 +12426,8 @@ CUresult CUDAAPI cuGraphAddMemsetNode(CUgraphNode *phGraphNode, CUgraph hGraph, * ::cuGraphAddMemsetNode, * ::cuGraphMemsetNodeSetParams */ -CUresult CUDAAPI cuGraphMemsetNodeGetParams(CUgraphNode hNode, CUDA_MEMSET_NODE_PARAMS *nodeParams); +CUresult CUDAAPI cuGraphMemsetNodeGetParams( + CUgraphNode hNode, CUDA_MEMSET_NODE_PARAMS *nodeParams); /** * \brief Sets a memset node's parameters @@ -11794,16 +12450,18 @@ CUresult CUDAAPI cuGraphMemsetNodeGetParams(CUgraphNode hNode, CUDA_MEMSET_NODE_ * ::cuGraphAddMemsetNode, * ::cuGraphMemsetNodeGetParams */ -CUresult CUDAAPI cuGraphMemsetNodeSetParams(CUgraphNode hNode, const CUDA_MEMSET_NODE_PARAMS *nodeParams); +CUresult CUDAAPI cuGraphMemsetNodeSetParams( + CUgraphNode hNode, const CUDA_MEMSET_NODE_PARAMS *nodeParams); /** * \brief Creates a host execution node and adds it to a graph * - * Creates a new CPU execution node and adds it to \p hGraph with \p numDependencies - * dependencies specified via \p dependencies and arguments specified in \p nodeParams. - * It is possible for \p numDependencies to be 0, in which case the node will be placed - * at the root of the graph. \p dependencies may not have any duplicate entries. - * A handle to the new node will be returned in \p phGraphNode. + * Creates a new CPU execution node and adds it to \p hGraph with \p + * numDependencies dependencies specified via \p dependencies and arguments + * specified in \p nodeParams. It is possible for \p numDependencies to be 0, in + * which case the node will be placed at the root of the graph. \p dependencies + * may not have any duplicate entries. A handle to the new node will be returned + * in \p phGraphNode. * * When the graph is launched, the node will invoke the specified CPU function. * Host nodes are not supported under MPS with pre-Volta GPUs. @@ -11835,7 +12493,10 @@ CUresult CUDAAPI cuGraphMemsetNodeSetParams(CUgraphNode hNode, const CUDA_MEMSET * ::cuGraphAddMemcpyNode, * ::cuGraphAddMemsetNode */ -CUresult CUDAAPI cuGraphAddHostNode(CUgraphNode *phGraphNode, CUgraph hGraph, const CUgraphNode *dependencies, size_t numDependencies, const CUDA_HOST_NODE_PARAMS *nodeParams); +CUresult CUDAAPI cuGraphAddHostNode(CUgraphNode *phGraphNode, CUgraph hGraph, + const CUgraphNode *dependencies, + size_t numDependencies, + const CUDA_HOST_NODE_PARAMS *nodeParams); /** * \brief Returns a host node's parameters @@ -11858,7 +12519,8 @@ CUresult CUDAAPI cuGraphAddHostNode(CUgraphNode *phGraphNode, CUgraph hGraph, co * ::cuGraphAddHostNode, * ::cuGraphHostNodeSetParams */ -CUresult CUDAAPI cuGraphHostNodeGetParams(CUgraphNode hNode, CUDA_HOST_NODE_PARAMS *nodeParams); +CUresult CUDAAPI cuGraphHostNodeGetParams(CUgraphNode hNode, + CUDA_HOST_NODE_PARAMS *nodeParams); /** * \brief Sets a host node's parameters @@ -11881,18 +12543,20 @@ CUresult CUDAAPI cuGraphHostNodeGetParams(CUgraphNode hNode, CUDA_HOST_NODE_PARA * ::cuGraphAddHostNode, * ::cuGraphHostNodeGetParams */ -CUresult CUDAAPI cuGraphHostNodeSetParams(CUgraphNode hNode, const CUDA_HOST_NODE_PARAMS *nodeParams); +CUresult CUDAAPI cuGraphHostNodeSetParams( + CUgraphNode hNode, const CUDA_HOST_NODE_PARAMS *nodeParams); /** * \brief Creates a child graph node and adds it to a graph * - * Creates a new node which executes an embedded graph, and adds it to \p hGraph with - * \p numDependencies dependencies specified via \p dependencies. - * It is possible for \p numDependencies to be 0, in which case the node will be placed - * at the root of the graph. \p dependencies may not have any duplicate entries. - * A handle to the new node will be returned in \p phGraphNode. + * Creates a new node which executes an embedded graph, and adds it to \p hGraph + * with \p numDependencies dependencies specified via \p dependencies. It is + * possible for \p numDependencies to be 0, in which case the node will be + * placed at the root of the graph. \p dependencies may not have any duplicate + * entries. A handle to the new node will be returned in \p phGraphNode. * - * The node executes an embedded child graph. The child graph is cloned in this call. + * The node executes an embedded child graph. The child graph is cloned in this + * call. * * \param phGraphNode - Returns newly created node * \param hGraph - Graph to which to add the node @@ -11919,7 +12583,11 @@ CUresult CUDAAPI cuGraphHostNodeSetParams(CUgraphNode hNode, const CUDA_HOST_NOD * ::cuGraphAddMemsetNode, * ::cuGraphClone */ -CUresult CUDAAPI cuGraphAddChildGraphNode(CUgraphNode *phGraphNode, CUgraph hGraph, const CUgraphNode *dependencies, size_t numDependencies, CUgraph childGraph); +CUresult CUDAAPI cuGraphAddChildGraphNode(CUgraphNode *phGraphNode, + CUgraph hGraph, + const CUgraphNode *dependencies, + size_t numDependencies, + CUgraph childGraph); /** * \brief Gets a handle to the embedded graph of a child graph node @@ -11943,16 +12611,17 @@ CUresult CUDAAPI cuGraphAddChildGraphNode(CUgraphNode *phGraphNode, CUgraph hGra * ::cuGraphAddChildGraphNode, * ::cuGraphNodeFindInClone */ -CUresult CUDAAPI cuGraphChildGraphNodeGetGraph(CUgraphNode hNode, CUgraph *phGraph); +CUresult CUDAAPI cuGraphChildGraphNodeGetGraph(CUgraphNode hNode, + CUgraph *phGraph); /** * \brief Creates an empty node and adds it to a graph * * Creates a new node which performs no operation, and adds it to \p hGraph with * \p numDependencies dependencies specified via \p dependencies. - * It is possible for \p numDependencies to be 0, in which case the node will be placed - * at the root of the graph. \p dependencies may not have any duplicate entries. - * A handle to the new node will be returned in \p phGraphNode. + * It is possible for \p numDependencies to be 0, in which case the node will be + * placed at the root of the graph. \p dependencies may not have any duplicate + * entries. A handle to the new node will be returned in \p phGraphNode. * * An empty node performs no operation during execution, but can be used for * transitive ordering. For example, a phased execution graph with 2 groups of n @@ -11981,16 +12650,19 @@ CUresult CUDAAPI cuGraphChildGraphNodeGetGraph(CUgraphNode hNode, CUgraph *phGra * ::cuGraphAddMemcpyNode, * ::cuGraphAddMemsetNode */ -CUresult CUDAAPI cuGraphAddEmptyNode(CUgraphNode *phGraphNode, CUgraph hGraph, const CUgraphNode *dependencies, size_t numDependencies); +CUresult CUDAAPI cuGraphAddEmptyNode(CUgraphNode *phGraphNode, CUgraph hGraph, + const CUgraphNode *dependencies, + size_t numDependencies); /** * \brief Clones a graph * - * This function creates a copy of \p originalGraph and returns it in \p * phGraphClone. - * All parameters are copied into the cloned graph. The original graph may be modified - * after this call without affecting the clone. + * This function creates a copy of \p originalGraph and returns it in \p * + * phGraphClone. All parameters are copied into the cloned graph. The original + * graph may be modified after this call without affecting the clone. * - * Child graph nodes in the original graph are recursively copied into the clone. + * Child graph nodes in the original graph are recursively copied into the + * clone. * * \param phGraphClone - Returns newly created cloned graph * \param originalGraph - Graph to clone @@ -12011,13 +12683,14 @@ CUresult CUDAAPI cuGraphClone(CUgraph *phGraphClone, CUgraph originalGraph); /** * \brief Finds a cloned version of a node * - * This function returns the node in \p hClonedGraph corresponding to \p hOriginalNode - * in the original graph. + * This function returns the node in \p hClonedGraph corresponding to \p + * hOriginalNode in the original graph. * - * \p hClonedGraph must have been cloned from \p hOriginalGraph via ::cuGraphClone. - * \p hOriginalNode must have been in \p hOriginalGraph at the time of the call to - * ::cuGraphClone, and the corresponding cloned node in \p hClonedGraph must not have - * been removed. The cloned node is then returned via \p phClonedNode. + * \p hClonedGraph must have been cloned from \p hOriginalGraph via + * ::cuGraphClone. \p hOriginalNode must have been in \p hOriginalGraph at the + * time of the call to + * ::cuGraphClone, and the corresponding cloned node in \p hClonedGraph must not + * have been removed. The cloned node is then returned via \p phClonedNode. * * \param phNode - Returns handle to the cloned node * \param hOriginalNode - Handle to the original node @@ -12032,7 +12705,9 @@ CUresult CUDAAPI cuGraphClone(CUgraph *phGraphClone, CUgraph originalGraph); * \sa * ::cuGraphClone */ -CUresult CUDAAPI cuGraphNodeFindInClone(CUgraphNode *phNode, CUgraphNode hOriginalNode, CUgraph hClonedGraph); +CUresult CUDAAPI cuGraphNodeFindInClone(CUgraphNode *phNode, + CUgraphNode hOriginalNode, + CUgraph hClonedGraph); /** * \brief Returns a node's type @@ -12070,9 +12745,10 @@ CUresult CUDAAPI cuGraphNodeGetType(CUgraphNode hNode, CUgraphNodeType *type); * * Returns a list of \p hGraph's nodes. \p nodes may be NULL, in which case this * function will return the number of nodes in \p numNodes. Otherwise, - * \p numNodes entries will be filled in. If \p numNodes is higher than the actual - * number of nodes, the remaining entries in \p nodes will be set to NULL, and the - * number of nodes actually obtained will be returned in \p numNodes. + * \p numNodes entries will be filled in. If \p numNodes is higher than the + * actual number of nodes, the remaining entries in \p nodes will be set to + * NULL, and the number of nodes actually obtained will be returned in \p + * numNodes. * * \param hGraph - Graph to query * \param nodes - Pointer to return the nodes @@ -12094,16 +12770,18 @@ CUresult CUDAAPI cuGraphNodeGetType(CUgraphNode hNode, CUgraphNodeType *type); * ::cuGraphNodeGetDependencies, * ::cuGraphNodeGetDependentNodes */ -CUresult CUDAAPI cuGraphGetNodes(CUgraph hGraph, CUgraphNode *nodes, size_t *numNodes); +CUresult CUDAAPI cuGraphGetNodes(CUgraph hGraph, CUgraphNode *nodes, + size_t *numNodes); /** * \brief Returns a graph's root nodes * - * Returns a list of \p hGraph's root nodes. \p rootNodes may be NULL, in which case this - * function will return the number of root nodes in \p numRootNodes. Otherwise, - * \p numRootNodes entries will be filled in. If \p numRootNodes is higher than the actual - * number of root nodes, the remaining entries in \p rootNodes will be set to NULL, and the - * number of nodes actually obtained will be returned in \p numRootNodes. + * Returns a list of \p hGraph's root nodes. \p rootNodes may be NULL, in which + * case this function will return the number of root nodes in \p numRootNodes. + * Otherwise, \p numRootNodes entries will be filled in. If \p numRootNodes is + * higher than the actual number of root nodes, the remaining entries in \p + * rootNodes will be set to NULL, and the number of nodes actually obtained will + * be returned in \p numRootNodes. * * \param hGraph - Graph to query * \param rootNodes - Pointer to return the root nodes @@ -12125,18 +12803,20 @@ CUresult CUDAAPI cuGraphGetNodes(CUgraph hGraph, CUgraphNode *nodes, size_t *num * ::cuGraphNodeGetDependencies, * ::cuGraphNodeGetDependentNodes */ -CUresult CUDAAPI cuGraphGetRootNodes(CUgraph hGraph, CUgraphNode *rootNodes, size_t *numRootNodes); +CUresult CUDAAPI cuGraphGetRootNodes(CUgraph hGraph, CUgraphNode *rootNodes, + size_t *numRootNodes); /** * \brief Returns a graph's dependency edges * - * Returns a list of \p hGraph's dependency edges. Edges are returned via corresponding - * indices in \p from and \p to; that is, the node in \p to[i] has a dependency on the - * node in \p from[i]. \p from and \p to may both be NULL, in which - * case this function only returns the number of edges in \p numEdges. Otherwise, - * \p numEdges entries will be filled in. If \p numEdges is higher than the actual - * number of edges, the remaining entries in \p from and \p to will be set to NULL, and - * the number of edges actually returned will be written to \p numEdges. + * Returns a list of \p hGraph's dependency edges. Edges are returned via + * corresponding indices in \p from and \p to; that is, the node in \p to[i] has + * a dependency on the node in \p from[i]. \p from and \p to may both be NULL, + * in which case this function only returns the number of edges in \p numEdges. + * Otherwise, \p numEdges entries will be filled in. If \p numEdges is higher + * than the actual number of edges, the remaining entries in \p from and \p to + * will be set to NULL, and the number of edges actually returned will be + * written to \p numEdges. * * \param hGraph - Graph to get the edges from * \param from - Location to return edge endpoints @@ -12159,16 +12839,18 @@ CUresult CUDAAPI cuGraphGetRootNodes(CUgraph hGraph, CUgraphNode *rootNodes, siz * ::cuGraphNodeGetDependencies, * ::cuGraphNodeGetDependentNodes */ -CUresult CUDAAPI cuGraphGetEdges(CUgraph hGraph, CUgraphNode *from, CUgraphNode *to, size_t *numEdges); +CUresult CUDAAPI cuGraphGetEdges(CUgraph hGraph, CUgraphNode *from, + CUgraphNode *to, size_t *numEdges); /** * \brief Returns a node's dependencies * - * Returns a list of \p node's dependencies. \p dependencies may be NULL, in which case this - * function will return the number of dependencies in \p numDependencies. Otherwise, - * \p numDependencies entries will be filled in. If \p numDependencies is higher than the actual - * number of dependencies, the remaining entries in \p dependencies will be set to NULL, and the - * number of nodes actually obtained will be returned in \p numDependencies. + * Returns a list of \p node's dependencies. \p dependencies may be NULL, in + * which case this function will return the number of dependencies in \p + * numDependencies. Otherwise, \p numDependencies entries will be filled in. If + * \p numDependencies is higher than the actual number of dependencies, the + * remaining entries in \p dependencies will be set to NULL, and the number of + * nodes actually obtained will be returned in \p numDependencies. * * \param hNode - Node to query * \param dependencies - Pointer to return the dependencies @@ -12190,17 +12872,19 @@ CUresult CUDAAPI cuGraphGetEdges(CUgraph hGraph, CUgraphNode *from, CUgraphNode * ::cuGraphAddDependencies, * ::cuGraphRemoveDependencies */ -CUresult CUDAAPI cuGraphNodeGetDependencies(CUgraphNode hNode, CUgraphNode *dependencies, size_t *numDependencies); +CUresult CUDAAPI cuGraphNodeGetDependencies(CUgraphNode hNode, + CUgraphNode *dependencies, + size_t *numDependencies); /** * \brief Returns a node's dependent nodes * - * Returns a list of \p node's dependent nodes. \p dependentNodes may be NULL, in which - * case this function will return the number of dependent nodes in \p numDependentNodes. - * Otherwise, \p numDependentNodes entries will be filled in. If \p numDependentNodes is - * higher than the actual number of dependent nodes, the remaining entries in - * \p dependentNodes will be set to NULL, and the number of nodes actually obtained will - * be returned in \p numDependentNodes. + * Returns a list of \p node's dependent nodes. \p dependentNodes may be NULL, + * in which case this function will return the number of dependent nodes in \p + * numDependentNodes. Otherwise, \p numDependentNodes entries will be filled in. + * If \p numDependentNodes is higher than the actual number of dependent nodes, + * the remaining entries in \p dependentNodes will be set to NULL, and the + * number of nodes actually obtained will be returned in \p numDependentNodes. * * \param hNode - Node to query * \param dependentNodes - Pointer to return the dependent nodes @@ -12222,7 +12906,9 @@ CUresult CUDAAPI cuGraphNodeGetDependencies(CUgraphNode hNode, CUgraphNode *depe * ::cuGraphAddDependencies, * ::cuGraphRemoveDependencies */ -CUresult CUDAAPI cuGraphNodeGetDependentNodes(CUgraphNode hNode, CUgraphNode *dependentNodes, size_t *numDependentNodes); +CUresult CUDAAPI cuGraphNodeGetDependentNodes(CUgraphNode hNode, + CUgraphNode *dependentNodes, + size_t *numDependentNodes); /** * \brief Adds dependency edges to a graph @@ -12251,7 +12937,9 @@ CUresult CUDAAPI cuGraphNodeGetDependentNodes(CUgraphNode hNode, CUgraphNode *de * ::cuGraphNodeGetDependencies, * ::cuGraphNodeGetDependentNodes */ -CUresult CUDAAPI cuGraphAddDependencies(CUgraph hGraph, const CUgraphNode *from, const CUgraphNode *to, size_t numDependencies); +CUresult CUDAAPI cuGraphAddDependencies(CUgraph hGraph, const CUgraphNode *from, + const CUgraphNode *to, + size_t numDependencies); /** * \brief Removes dependency edges from a graph @@ -12280,13 +12968,16 @@ CUresult CUDAAPI cuGraphAddDependencies(CUgraph hGraph, const CUgraphNode *from, * ::cuGraphNodeGetDependencies, * ::cuGraphNodeGetDependentNodes */ -CUresult CUDAAPI cuGraphRemoveDependencies(CUgraph hGraph, const CUgraphNode *from, const CUgraphNode *to, size_t numDependencies); +CUresult CUDAAPI cuGraphRemoveDependencies(CUgraph hGraph, + const CUgraphNode *from, + const CUgraphNode *to, + size_t numDependencies); /** * \brief Remove a node from the graph * - * Removes \p hNode from its graph. This operation also severs any dependencies of other nodes - * on \p hNode and vice versa. + * Removes \p hNode from its graph. This operation also severs any dependencies + * of other nodes on \p hNode and vice versa. * * \param hNode - Node to remove * @@ -12314,18 +13005,18 @@ CUresult CUDAAPI cuGraphDestroyNode(CUgraphNode hNode); * validated. If instantiation is successful, a handle to the instantiated graph * is returned in \p graphExec. * - * If there are any errors, diagnostic information may be returned in \p errorNode and - * \p logBuffer. This is the primary way to inspect instantiation errors. The output - * will be null terminated unless the diagnostics overflow + * If there are any errors, diagnostic information may be returned in \p + * errorNode and \p logBuffer. This is the primary way to inspect instantiation + * errors. The output will be null terminated unless the diagnostics overflow * the buffer. In this case, they will be truncated, and the last byte can be * inspected to determine if truncation occurred. * * \param phGraphExec - Returns instantiated graph * \param hGraph - Graph to instantiate - * \param phErrorNode - In case of an instantiation error, this may be modified to - * indicate a node contributing to the error - * \param logBuffer - A character buffer to store diagnostic messages - * \param bufferSize - Size of the log buffer in bytes + * \param phErrorNode - In case of an instantiation error, this may be modified + * to indicate a node contributing to the error \param logBuffer - A character + * buffer to store diagnostic messages \param bufferSize - Size of the log + * buffer in bytes * * \return * ::CUDA_SUCCESS, @@ -12340,8 +13031,9 @@ CUresult CUDAAPI cuGraphDestroyNode(CUgraphNode hNode); * ::cuGraphLaunch, * ::cuGraphExecDestroy */ -CUresult CUDAAPI cuGraphInstantiate(CUgraphExec *phGraphExec, CUgraph hGraph, CUgraphNode *phErrorNode, char *logBuffer, size_t bufferSize); - +CUresult CUDAAPI cuGraphInstantiate(CUgraphExec *phGraphExec, CUgraph hGraph, + CUgraphNode *phErrorNode, char *logBuffer, + size_t bufferSize); #if __CUDA_API_VERSION >= 10010 /** @@ -12351,8 +13043,8 @@ CUresult CUDAAPI cuGraphInstantiate(CUgraphExec *phGraphExec, CUgraph hGraph, CU * The node is identified by the corresponding node \p hNode in the * non-executable graph, from which the executable graph was instantiated. * - * \p hNode must not have been removed from the original graph. The \p func field - * of \p nodeParams cannot be modified and must match the original value. + * \p hNode must not have been removed from the original graph. The \p func + * field of \p nodeParams cannot be modified and must match the original value. * All other values can be modified. * * The modifications take effect at the next launch of \p hGraphExec. Already @@ -12360,8 +13052,8 @@ CUresult CUDAAPI cuGraphInstantiate(CUgraphExec *phGraphExec, CUgraph hGraph, CU * \p hNode is also not modified by this call. * * \param hGraphExec - The executable graph in which to set the specified node - * \param hNode - kernel node from the graph from which graphExec was instantiated - * \param nodeParams - Updated Parameters to set + * \param hNode - kernel node from the graph from which graphExec was + * instantiated \param nodeParams - Updated Parameters to set * * \return * ::CUDA_SUCCESS, @@ -12374,17 +13066,20 @@ CUresult CUDAAPI cuGraphInstantiate(CUgraphExec *phGraphExec, CUgraph hGraph, CU * ::cuGraphKernelNodeSetParams, * ::cuGraphInstantiate */ - CUresult CUDAAPI cuGraphExecKernelNodeSetParams(CUgraphExec hGraphExec, CUgraphNode hNode, const CUDA_KERNEL_NODE_PARAMS *nodeParams); +CUresult CUDAAPI +cuGraphExecKernelNodeSetParams(CUgraphExec hGraphExec, CUgraphNode hNode, + const CUDA_KERNEL_NODE_PARAMS *nodeParams); #endif /* __CUDA_API_VERSION >= 10010 */ /** * \brief Launches an executable graph in a stream * - * Executes \p hGraphExec in \p hStream. Only one instance of \p hGraphExec may be executing - * at a time. Each launch is ordered behind both any previous work in \p hStream - * and any previous launches of \p hGraphExec. To execute a graph concurrently, it must be - * instantiated multiple times into multiple executable graphs. + * Executes \p hGraphExec in \p hStream. Only one instance of \p hGraphExec may + * be executing at a time. Each launch is ordered behind both any previous work + * in \p hStream and any previous launches of \p hGraphExec. To execute a graph + * concurrently, it must be instantiated multiple times into multiple executable + * graphs. * * \param hGraphExec - Executable graph to launch * \param hStream - Stream in which to launch the graph @@ -12447,7 +13142,7 @@ CUresult CUDAAPI cuGraphExecDestroy(CUgraphExec hGraphExec); */ CUresult CUDAAPI cuGraphDestroy(CUgraph hGraph); /** @} */ /* END CUDA_GRAPH */ -#endif /* __CUDA_API_VERSION >= 10000 */ +#endif /* __CUDA_API_VERSION >= 10000 */ #if __CUDA_API_VERSION >= 6050 /** @@ -12456,8 +13151,8 @@ CUresult CUDAAPI cuGraphDestroy(CUgraph hGraph); * ___MANBRIEF___ occupancy calculation functions of the low-level CUDA driver * API (___CURRENT_FILE___) ___ENDMANBRIEF___ * - * This section describes the occupancy calculation functions of the low-level CUDA - * driver application programming interface. + * This section describes the occupancy calculation functions of the low-level + * CUDA driver application programming interface. * * @{ */ @@ -12470,8 +13165,9 @@ CUresult CUDAAPI cuGraphDestroy(CUgraph hGraph); * * \param numBlocks - Returned occupancy * \param func - Kernel for which occupancy is calculated - * \param blockSize - Block size the kernel is intended to be launched with - * \param dynamicSMemSize - Per-block dynamic shared memory usage intended, in bytes + * \param blockSize - Block size the kernel is intended to be launched + * with \param dynamicSMemSize - Per-block dynamic shared memory usage intended, + * in bytes * * \return * ::CUDA_SUCCESS, @@ -12485,7 +13181,8 @@ CUresult CUDAAPI cuGraphDestroy(CUgraph hGraph); * \sa * ::cudaOccupancyMaxActiveBlocksPerMultiprocessor */ -CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize); +CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessor( + int *numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize); /** * \brief Returns occupancy of a function @@ -12511,9 +13208,10 @@ CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, CUf * * \param numBlocks - Returned occupancy * \param func - Kernel for which occupancy is calculated - * \param blockSize - Block size the kernel is intended to be launched with - * \param dynamicSMemSize - Per-block dynamic shared memory usage intended, in bytes - * \param flags - Requested behavior for the occupancy calculator + * \param blockSize - Block size the kernel is intended to be launched + * with \param dynamicSMemSize - Per-block dynamic shared memory usage intended, + * in bytes \param flags - Requested behavior for the occupancy + * calculator * * \return * ::CUDA_SUCCESS, @@ -12527,7 +13225,9 @@ CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, CUf * \sa * ::cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags */ -CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize, unsigned int flags); +CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags( + int *numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize, + unsigned int flags); /** * \brief Suggest a launch configuration with reasonable occupancy @@ -12560,12 +13260,14 @@ CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBl * size_t blockToSmem(int blockSize); * \endcode * - * \param minGridSize - Returned minimum grid size needed to achieve the maximum occupancy - * \param blockSize - Returned maximum block size that can achieve the maximum occupancy - * \param func - Kernel for which launch configuration is calculated - * \param blockSizeToDynamicSMemSize - A function that calculates how much per-block dynamic shared memory \p func uses based on the block size - * \param dynamicSMemSize - Dynamic shared memory usage intended, in bytes - * \param blockSizeLimit - The maximum block size \p func is designed to handle + * \param minGridSize - Returned minimum grid size needed to achieve the maximum + * occupancy \param blockSize - Returned maximum block size that can achieve + * the maximum occupancy \param func - Kernel for which launch + * configuration is calculated \param blockSizeToDynamicSMemSize - A function + * that calculates how much per-block dynamic shared memory \p func uses based + * on the block size \param dynamicSMemSize - Dynamic shared memory usage + * intended, in bytes \param blockSizeLimit - The maximum block size \p func is + * designed to handle * * \return * ::CUDA_SUCCESS, @@ -12579,7 +13281,10 @@ CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBl * \sa * ::cudaOccupancyMaxPotentialBlockSize */ -CUresult CUDAAPI cuOccupancyMaxPotentialBlockSize(int *minGridSize, int *blockSize, CUfunction func, CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, int blockSizeLimit); +CUresult CUDAAPI cuOccupancyMaxPotentialBlockSize( + int *minGridSize, int *blockSize, CUfunction func, + CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, + int blockSizeLimit); /** * \brief Suggest a launch configuration with reasonable occupancy @@ -12605,13 +13310,14 @@ CUresult CUDAAPI cuOccupancyMaxPotentialBlockSize(int *minGridSize, int *blockSi * can be found about this feature in the "Unified L1/Texture Cache" * section of the Maxwell tuning guide. * - * \param minGridSize - Returned minimum grid size needed to achieve the maximum occupancy - * \param blockSize - Returned maximum block size that can achieve the maximum occupancy - * \param func - Kernel for which launch configuration is calculated - * \param blockSizeToDynamicSMemSize - A function that calculates how much per-block dynamic shared memory \p func uses based on the block size - * \param dynamicSMemSize - Dynamic shared memory usage intended, in bytes - * \param blockSizeLimit - The maximum block size \p func is designed to handle - * \param flags - Options + * \param minGridSize - Returned minimum grid size needed to achieve the maximum + * occupancy \param blockSize - Returned maximum block size that can achieve + * the maximum occupancy \param func - Kernel for which launch + * configuration is calculated \param blockSizeToDynamicSMemSize - A function + * that calculates how much per-block dynamic shared memory \p func uses based + * on the block size \param dynamicSMemSize - Dynamic shared memory usage + * intended, in bytes \param blockSizeLimit - The maximum block size \p func is + * designed to handle \param flags - Options * * \return * ::CUDA_SUCCESS, @@ -12625,10 +13331,13 @@ CUresult CUDAAPI cuOccupancyMaxPotentialBlockSize(int *minGridSize, int *blockSi * \sa * ::cudaOccupancyMaxPotentialBlockSizeWithFlags */ -CUresult CUDAAPI cuOccupancyMaxPotentialBlockSizeWithFlags(int *minGridSize, int *blockSize, CUfunction func, CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, int blockSizeLimit, unsigned int flags); +CUresult CUDAAPI cuOccupancyMaxPotentialBlockSizeWithFlags( + int *minGridSize, int *blockSize, CUfunction func, + CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, + int blockSizeLimit, unsigned int flags); /** @} */ /* END CUDA_OCCUPANCY */ -#endif /* __CUDA_API_VERSION >= 6050 */ +#endif /* __CUDA_API_VERSION >= 6050 */ /** * \defgroup CUDA_TEXREF_DEPRECATED Texture Reference Management [DEPRECATED] @@ -12671,17 +13380,19 @@ CUresult CUDAAPI cuOccupancyMaxPotentialBlockSizeWithFlags(int *minGridSize, int * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat, * ::cudaBindTextureToArray */ -CUresult CUDAAPI cuTexRefSetArray(CUtexref hTexRef, CUarray hArray, unsigned int Flags); +CUresult CUDAAPI cuTexRefSetArray(CUtexref hTexRef, CUarray hArray, + unsigned int Flags); /** * \brief Binds a mipmapped array to a texture reference * * \deprecated * - * Binds the CUDA mipmapped array \p hMipmappedArray to the texture reference \p hTexRef. - * Any previous address or CUDA array state associated with the texture reference - * is superseded by this function. \p Flags must be set to ::CU_TRSA_OVERRIDE_FORMAT. - * Any CUDA array previously bound to \p hTexRef is unbound. + * Binds the CUDA mipmapped array \p hMipmappedArray to the texture reference \p + * hTexRef. Any previous address or CUDA array state associated with the texture + * reference is superseded by this function. \p Flags must be set to + * ::CU_TRSA_OVERRIDE_FORMAT. Any CUDA array previously bound to \p hTexRef is + * unbound. * * \param hTexRef - Texture reference to bind * \param hMipmappedArray - Mipmapped array to bind @@ -12701,7 +13412,9 @@ CUresult CUDAAPI cuTexRefSetArray(CUtexref hTexRef, CUarray hArray, unsigned int * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat, * ::cudaBindTextureToMipmappedArray */ -CUresult CUDAAPI cuTexRefSetMipmappedArray(CUtexref hTexRef, CUmipmappedArray hMipmappedArray, unsigned int Flags); +CUresult CUDAAPI cuTexRefSetMipmappedArray(CUtexref hTexRef, + CUmipmappedArray hMipmappedArray, + unsigned int Flags); #if __CUDA_API_VERSION >= 3020 /** @@ -12748,7 +13461,8 @@ CUresult CUDAAPI cuTexRefSetMipmappedArray(CUtexref hTexRef, CUmipmappedArray hM * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat, * ::cudaBindTexture */ -CUresult CUDAAPI cuTexRefSetAddress(size_t *ByteOffset, CUtexref hTexRef, CUdeviceptr dptr, size_t bytes); +CUresult CUDAAPI cuTexRefSetAddress(size_t *ByteOffset, CUtexref hTexRef, + CUdeviceptr dptr, size_t bytes); /** * \brief Binds an address as a 2D texture reference @@ -12803,7 +13517,9 @@ CUresult CUDAAPI cuTexRefSetAddress(size_t *ByteOffset, CUtexref hTexRef, CUdevi * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat, * ::cudaBindTexture2D */ -CUresult CUDAAPI cuTexRefSetAddress2D(CUtexref hTexRef, const CUDA_ARRAY_DESCRIPTOR *desc, CUdeviceptr dptr, size_t Pitch); +CUresult CUDAAPI cuTexRefSetAddress2D(CUtexref hTexRef, + const CUDA_ARRAY_DESCRIPTOR *desc, + CUdeviceptr dptr, size_t Pitch); #endif /* __CUDA_API_VERSION >= 3020 */ /** @@ -12839,7 +13555,8 @@ CUresult CUDAAPI cuTexRefSetAddress2D(CUtexref hTexRef, const CUDA_ARRAY_DESCRIP * ::cudaBindTextureToArray, * ::cudaBindTextureToMipmappedArray */ -CUresult CUDAAPI cuTexRefSetFormat(CUtexref hTexRef, CUarray_format fmt, int NumPackedComponents); +CUresult CUDAAPI cuTexRefSetFormat(CUtexref hTexRef, CUarray_format fmt, + int NumPackedComponents); /** * \brief Sets the addressing mode for a texture reference @@ -12885,7 +13602,8 @@ CUresult CUDAAPI cuTexRefSetFormat(CUtexref hTexRef, CUarray_format fmt, int Num * ::cudaBindTextureToArray, * ::cudaBindTextureToMipmappedArray */ -CUresult CUDAAPI cuTexRefSetAddressMode(CUtexref hTexRef, int dim, CUaddress_mode am); +CUresult CUDAAPI cuTexRefSetAddressMode(CUtexref hTexRef, int dim, + CUaddress_mode am); /** * \brief Sets the filtering mode for a texture reference @@ -12928,7 +13646,8 @@ CUresult CUDAAPI cuTexRefSetFilterMode(CUtexref hTexRef, CUfilter_mode fm); * * \deprecated * - * Specifies the mipmap filtering mode \p fm to be used when reading memory through + * Specifies the mipmap filtering mode \p fm to be used when reading memory + through * the texture reference \p hTexRef. ::CUfilter_mode_enum is defined as: * * \code @@ -12938,7 +13657,8 @@ CUresult CUDAAPI cuTexRefSetFilterMode(CUtexref hTexRef, CUfilter_mode fm); } CUfilter_mode; * \endcode * - * Note that this call has no effect if \p hTexRef is not bound to a mipmapped array. + * Note that this call has no effect if \p hTexRef is not bound to a mipmapped + array. * * \param hTexRef - Texture reference * \param fm - Filtering mode to set @@ -12957,17 +13677,19 @@ CUresult CUDAAPI cuTexRefSetFilterMode(CUtexref hTexRef, CUfilter_mode fm); * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat, * ::cudaBindTextureToMipmappedArray */ -CUresult CUDAAPI cuTexRefSetMipmapFilterMode(CUtexref hTexRef, CUfilter_mode fm); +CUresult CUDAAPI cuTexRefSetMipmapFilterMode(CUtexref hTexRef, + CUfilter_mode fm); /** * \brief Sets the mipmap level bias for a texture reference * * \deprecated * - * Specifies the mipmap level bias \p bias to be added to the specified mipmap level when - * reading memory through the texture reference \p hTexRef. + * Specifies the mipmap level bias \p bias to be added to the specified mipmap + * level when reading memory through the texture reference \p hTexRef. * - * Note that this call has no effect if \p hTexRef is not bound to a mipmapped array. + * Note that this call has no effect if \p hTexRef is not bound to a mipmapped + * array. * * \param hTexRef - Texture reference * \param bias - Mipmap level bias @@ -12993,11 +13715,12 @@ CUresult CUDAAPI cuTexRefSetMipmapLevelBias(CUtexref hTexRef, float bias); * * \deprecated * - * Specifies the min/max mipmap level clamps, \p minMipmapLevelClamp and \p maxMipmapLevelClamp - * respectively, to be used when reading memory through the texture reference - * \p hTexRef. + * Specifies the min/max mipmap level clamps, \p minMipmapLevelClamp and \p + * maxMipmapLevelClamp respectively, to be used when reading memory through the + * texture reference \p hTexRef. * - * Note that this call has no effect if \p hTexRef is not bound to a mipmapped array. + * Note that this call has no effect if \p hTexRef is not bound to a mipmapped + * array. * * \param hTexRef - Texture reference * \param minMipmapLevelClamp - Mipmap min level clamp @@ -13017,15 +13740,17 @@ CUresult CUDAAPI cuTexRefSetMipmapLevelBias(CUtexref hTexRef, float bias); * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat, * ::cudaBindTextureToMipmappedArray */ -CUresult CUDAAPI cuTexRefSetMipmapLevelClamp(CUtexref hTexRef, float minMipmapLevelClamp, float maxMipmapLevelClamp); +CUresult CUDAAPI cuTexRefSetMipmapLevelClamp(CUtexref hTexRef, + float minMipmapLevelClamp, + float maxMipmapLevelClamp); /** * \brief Sets the maximum anisotropy for a texture reference * * \deprecated * - * Specifies the maximum anisotropy \p maxAniso to be used when reading memory through - * the texture reference \p hTexRef. + * Specifies the maximum anisotropy \p maxAniso to be used when reading memory + * through the texture reference \p hTexRef. * * Note that this call has no effect if \p hTexRef is bound to linear memory. * @@ -13047,24 +13772,24 @@ CUresult CUDAAPI cuTexRefSetMipmapLevelClamp(CUtexref hTexRef, float minMipmapLe * ::cudaBindTextureToArray, * ::cudaBindTextureToMipmappedArray */ -CUresult CUDAAPI cuTexRefSetMaxAnisotropy(CUtexref hTexRef, unsigned int maxAniso); +CUresult CUDAAPI cuTexRefSetMaxAnisotropy(CUtexref hTexRef, + unsigned int maxAniso); /** * \brief Sets the border color for a texture reference * * \deprecated * - * Specifies the value of the RGBA color via the \p pBorderColor to the texture reference - * \p hTexRef. The color value supports only float type and holds color components in - * the following sequence: - * pBorderColor[0] holds 'R' component - * pBorderColor[1] holds 'G' component - * pBorderColor[2] holds 'B' component - * pBorderColor[3] holds 'A' component + * Specifies the value of the RGBA color via the \p pBorderColor to the texture + * reference \p hTexRef. The color value supports only float type and holds + * color components in the following sequence: pBorderColor[0] holds 'R' + * component pBorderColor[1] holds 'G' component pBorderColor[2] holds 'B' + * component pBorderColor[3] holds 'A' component * * Note that the color values can be set only when the Address mode is set to * CU_TR_ADDRESS_MODE_BORDER using ::cuTexRefSetAddressMode. - * Applications using integer border color values have to "reinterpret_cast" their values to float. + * Applications using integer border color values have to "reinterpret_cast" + * their values to float. * * \param hTexRef - Texture reference * \param pBorderColor - RGBA color @@ -13188,8 +13913,8 @@ CUresult CUDAAPI cuTexRefGetArray(CUarray *phArray, CUtexref hTexRef); * \deprecated * * Returns in \p *phMipmappedArray the CUDA mipmapped array bound to the texture - * reference \p hTexRef, or returns ::CUDA_ERROR_INVALID_VALUE if the texture reference - * is not bound to any CUDA mipmapped array. + * reference \p hTexRef, or returns ::CUDA_ERROR_INVALID_VALUE if the texture + * reference is not bound to any CUDA mipmapped array. * * \param phMipmappedArray - Returned mipmapped array * \param hTexRef - Texture reference @@ -13207,7 +13932,8 @@ CUresult CUDAAPI cuTexRefGetArray(CUarray *phArray, CUtexref hTexRef); * ::cuTexRefGetAddress, ::cuTexRefGetAddressMode, * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat */ -CUresult CUDAAPI cuTexRefGetMipmappedArray(CUmipmappedArray *phMipmappedArray, CUtexref hTexRef); +CUresult CUDAAPI cuTexRefGetMipmappedArray(CUmipmappedArray *phMipmappedArray, + CUtexref hTexRef); /** * \brief Gets the addressing mode used by a texture reference @@ -13235,7 +13961,8 @@ CUresult CUDAAPI cuTexRefGetMipmappedArray(CUmipmappedArray *phMipmappedArray, C * ::cuTexRefGetAddress, ::cuTexRefGetArray, * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat */ -CUresult CUDAAPI cuTexRefGetAddressMode(CUaddress_mode *pam, CUtexref hTexRef, int dim); +CUresult CUDAAPI cuTexRefGetAddressMode(CUaddress_mode *pam, CUtexref hTexRef, + int dim); /** * \brief Gets the filter-mode used by a texture reference @@ -13289,15 +14016,16 @@ CUresult CUDAAPI cuTexRefGetFilterMode(CUfilter_mode *pfm, CUtexref hTexRef); * ::cuTexRefGetAddress, ::cuTexRefGetAddressMode, ::cuTexRefGetArray, * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags */ -CUresult CUDAAPI cuTexRefGetFormat(CUarray_format *pFormat, int *pNumChannels, CUtexref hTexRef); +CUresult CUDAAPI cuTexRefGetFormat(CUarray_format *pFormat, int *pNumChannels, + CUtexref hTexRef); /** * \brief Gets the mipmap filtering mode for a texture reference * * \deprecated * - * Returns the mipmap filtering mode in \p pfm that's used when reading memory through - * the texture reference \p hTexRef. + * Returns the mipmap filtering mode in \p pfm that's used when reading memory + * through the texture reference \p hTexRef. * * \param pfm - Returned mipmap filtering mode * \param hTexRef - Texture reference @@ -13315,15 +14043,16 @@ CUresult CUDAAPI cuTexRefGetFormat(CUarray_format *pFormat, int *pNumChannels, C * ::cuTexRefGetAddress, ::cuTexRefGetAddressMode, ::cuTexRefGetArray, * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat */ -CUresult CUDAAPI cuTexRefGetMipmapFilterMode(CUfilter_mode *pfm, CUtexref hTexRef); +CUresult CUDAAPI cuTexRefGetMipmapFilterMode(CUfilter_mode *pfm, + CUtexref hTexRef); /** * \brief Gets the mipmap level bias for a texture reference * * \deprecated * - * Returns the mipmap level bias in \p pBias that's added to the specified mipmap - * level when reading memory through the texture reference \p hTexRef. + * Returns the mipmap level bias in \p pBias that's added to the specified + * mipmap level when reading memory through the texture reference \p hTexRef. * * \param pbias - Returned mipmap level bias * \param hTexRef - Texture reference @@ -13348,8 +14077,9 @@ CUresult CUDAAPI cuTexRefGetMipmapLevelBias(float *pbias, CUtexref hTexRef); * * \deprecated * - * Returns the min/max mipmap level clamps in \p pminMipmapLevelClamp and \p pmaxMipmapLevelClamp - * that's used when reading memory through the texture reference \p hTexRef. + * Returns the min/max mipmap level clamps in \p pminMipmapLevelClamp and \p + * pmaxMipmapLevelClamp that's used when reading memory through the texture + * reference \p hTexRef. * * \param pminMipmapLevelClamp - Returned mipmap min level clamp * \param pmaxMipmapLevelClamp - Returned mipmap max level clamp @@ -13368,15 +14098,17 @@ CUresult CUDAAPI cuTexRefGetMipmapLevelBias(float *pbias, CUtexref hTexRef); * ::cuTexRefGetAddress, ::cuTexRefGetAddressMode, ::cuTexRefGetArray, * ::cuTexRefGetFilterMode, ::cuTexRefGetFlags, ::cuTexRefGetFormat */ -CUresult CUDAAPI cuTexRefGetMipmapLevelClamp(float *pminMipmapLevelClamp, float *pmaxMipmapLevelClamp, CUtexref hTexRef); +CUresult CUDAAPI cuTexRefGetMipmapLevelClamp(float *pminMipmapLevelClamp, + float *pmaxMipmapLevelClamp, + CUtexref hTexRef); /** * \brief Gets the maximum anisotropy for a texture reference * * \deprecated * - * Returns the maximum anisotropy in \p pmaxAniso that's used when reading memory through - * the texture reference \p hTexRef. + * Returns the maximum anisotropy in \p pmaxAniso that's used when reading + * memory through the texture reference \p hTexRef. * * \param pmaxAniso - Returned maximum anisotropy * \param hTexRef - Texture reference @@ -13497,7 +14229,6 @@ CUresult CUDAAPI cuTexRefDestroy(CUtexref hTexRef); /** @} */ /* END CUDA_TEXREF_DEPRECATED */ - /** * \defgroup CUDA_SURFREF_DEPRECATED Surface Reference Management [DEPRECATED] * @@ -13537,7 +14268,8 @@ CUresult CUDAAPI cuTexRefDestroy(CUtexref hTexRef); * ::cuSurfRefGetArray, * ::cudaBindSurfaceToArray */ -CUresult CUDAAPI cuSurfRefSetArray(CUsurfref hSurfRef, CUarray hArray, unsigned int Flags); +CUresult CUDAAPI cuSurfRefSetArray(CUsurfref hSurfRef, CUarray hArray, + unsigned int Flags); /** * \brief Passes back the CUDA array bound to a surface reference. @@ -13581,15 +14313,21 @@ CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef); /** * \brief Creates a texture object * - * Creates a texture object and returns it in \p pTexObject. \p pResDesc describes - * the data to texture from. \p pTexDesc describes how the data should be sampled. - * \p pResViewDesc is an optional argument that specifies an alternate format for + * Creates a texture object and returns it in \p pTexObject. \p pResDesc + describes + * the data to texture from. \p pTexDesc describes how the data should be + sampled. + * \p pResViewDesc is an optional argument that specifies an alternate format + for * the data described by \p pResDesc, and also describes the subresource region - * to restrict access to when texturing. \p pResViewDesc can only be specified if + * to restrict access to when texturing. \p pResViewDesc can only be specified + if * the type of resource is a CUDA array or a CUDA mipmapped array. * - * Texture objects are only supported on devices of compute capability 3.0 or higher. - * Additionally, a texture object is an opaque value, and, as such, should only be + * Texture objects are only supported on devices of compute capability 3.0 or + higher. + * Additionally, a texture object is an opaque value, and, as such, should only + be * accessed through CUDA API calls. * * The ::CUDA_RESOURCE_DESC structure is defined as: @@ -13626,7 +14364,8 @@ CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef); * \endcode * where: - * - ::CUDA_RESOURCE_DESC::resType specifies the type of resource to texture from. + * - ::CUDA_RESOURCE_DESC::resType specifies the type of resource to texture + from. * CUresourceType is defined as: * \code typedef enum CUresourcetype_enum { @@ -13638,30 +14377,47 @@ CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef); * \endcode * * \par - * If ::CUDA_RESOURCE_DESC::resType is set to ::CU_RESOURCE_TYPE_ARRAY, ::CUDA_RESOURCE_DESC::res::array::hArray + * If ::CUDA_RESOURCE_DESC::resType is set to ::CU_RESOURCE_TYPE_ARRAY, + ::CUDA_RESOURCE_DESC::res::array::hArray * must be set to a valid CUDA array handle. * * \par - * If ::CUDA_RESOURCE_DESC::resType is set to ::CU_RESOURCE_TYPE_MIPMAPPED_ARRAY, ::CUDA_RESOURCE_DESC::res::mipmap::hMipmappedArray + * If ::CUDA_RESOURCE_DESC::resType is set to + ::CU_RESOURCE_TYPE_MIPMAPPED_ARRAY, + ::CUDA_RESOURCE_DESC::res::mipmap::hMipmappedArray * must be set to a valid CUDA mipmapped array handle. * * \par - * If ::CUDA_RESOURCE_DESC::resType is set to ::CU_RESOURCE_TYPE_LINEAR, ::CUDA_RESOURCE_DESC::res::linear::devPtr - * must be set to a valid device pointer, that is aligned to ::CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT. - * ::CUDA_RESOURCE_DESC::res::linear::format and ::CUDA_RESOURCE_DESC::res::linear::numChannels - * describe the format of each component and the number of components per array element. ::CUDA_RESOURCE_DESC::res::linear::sizeInBytes - * specifies the size of the array in bytes. The total number of elements in the linear address range cannot exceed - * ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LINEAR_WIDTH. The number of elements is computed as (sizeInBytes / (sizeof(format) * numChannels)). + * If ::CUDA_RESOURCE_DESC::resType is set to ::CU_RESOURCE_TYPE_LINEAR, + ::CUDA_RESOURCE_DESC::res::linear::devPtr + * must be set to a valid device pointer, that is aligned to + ::CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT. + * ::CUDA_RESOURCE_DESC::res::linear::format and + ::CUDA_RESOURCE_DESC::res::linear::numChannels + * describe the format of each component and the number of components per array + element. ::CUDA_RESOURCE_DESC::res::linear::sizeInBytes + * specifies the size of the array in bytes. The total number of elements in the + linear address range cannot exceed + * ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE1D_LINEAR_WIDTH. The number of elements + is computed as (sizeInBytes / (sizeof(format) * numChannels)). * * \par - * If ::CUDA_RESOURCE_DESC::resType is set to ::CU_RESOURCE_TYPE_PITCH2D, ::CUDA_RESOURCE_DESC::res::pitch2D::devPtr - * must be set to a valid device pointer, that is aligned to ::CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT. - * ::CUDA_RESOURCE_DESC::res::pitch2D::format and ::CUDA_RESOURCE_DESC::res::pitch2D::numChannels - * describe the format of each component and the number of components per array element. ::CUDA_RESOURCE_DESC::res::pitch2D::width - * and ::CUDA_RESOURCE_DESC::res::pitch2D::height specify the width and height of the array in elements, and cannot exceed - * ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_WIDTH and ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_HEIGHT respectively. - * ::CUDA_RESOURCE_DESC::res::pitch2D::pitchInBytes specifies the pitch between two rows in bytes and has to be aligned to - * ::CU_DEVICE_ATTRIBUTE_TEXTURE_PITCH_ALIGNMENT. Pitch cannot exceed ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_PITCH. + * If ::CUDA_RESOURCE_DESC::resType is set to ::CU_RESOURCE_TYPE_PITCH2D, + ::CUDA_RESOURCE_DESC::res::pitch2D::devPtr + * must be set to a valid device pointer, that is aligned to + ::CU_DEVICE_ATTRIBUTE_TEXTURE_ALIGNMENT. + * ::CUDA_RESOURCE_DESC::res::pitch2D::format and + ::CUDA_RESOURCE_DESC::res::pitch2D::numChannels + * describe the format of each component and the number of components per array + element. ::CUDA_RESOURCE_DESC::res::pitch2D::width + * and ::CUDA_RESOURCE_DESC::res::pitch2D::height specify the width and height + of the array in elements, and cannot exceed + * ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_WIDTH and + ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_HEIGHT respectively. + * ::CUDA_RESOURCE_DESC::res::pitch2D::pitchInBytes specifies the pitch between + two rows in bytes and has to be aligned to + * ::CU_DEVICE_ATTRIBUTE_TEXTURE_PITCH_ALIGNMENT. Pitch cannot exceed + ::CU_DEVICE_ATTRIBUTE_MAXIMUM_TEXTURE2D_LINEAR_PITCH. * * - ::flags must be set to zero. * @@ -13680,7 +14436,8 @@ CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef); } CUDA_TEXTURE_DESC; * \endcode * where - * - ::CUDA_TEXTURE_DESC::addressMode specifies the addressing mode for each dimension of the texture data. ::CUaddress_mode is defined as: + * - ::CUDA_TEXTURE_DESC::addressMode specifies the addressing mode for each + dimension of the texture data. ::CUaddress_mode is defined as: * \code typedef enum CUaddress_mode_enum { CU_TR_ADDRESS_MODE_WRAP = 0, @@ -13689,35 +14446,47 @@ CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef); CU_TR_ADDRESS_MODE_BORDER = 3 } CUaddress_mode; * \endcode - * This is ignored if ::CUDA_RESOURCE_DESC::resType is ::CU_RESOURCE_TYPE_LINEAR. Also, if the flag, ::CU_TRSF_NORMALIZED_COORDINATES + * This is ignored if ::CUDA_RESOURCE_DESC::resType is + ::CU_RESOURCE_TYPE_LINEAR. Also, if the flag, ::CU_TRSF_NORMALIZED_COORDINATES * is not set, the only supported address mode is ::CU_TR_ADDRESS_MODE_CLAMP. * - * - ::CUDA_TEXTURE_DESC::filterMode specifies the filtering mode to be used when fetching from the texture. CUfilter_mode is defined as: + * - ::CUDA_TEXTURE_DESC::filterMode specifies the filtering mode to be used + when fetching from the texture. CUfilter_mode is defined as: * \code typedef enum CUfilter_mode_enum { CU_TR_FILTER_MODE_POINT = 0, CU_TR_FILTER_MODE_LINEAR = 1 } CUfilter_mode; * \endcode - * This is ignored if ::CUDA_RESOURCE_DESC::resType is ::CU_RESOURCE_TYPE_LINEAR. + * This is ignored if ::CUDA_RESOURCE_DESC::resType is + ::CU_RESOURCE_TYPE_LINEAR. * * - ::CUDA_TEXTURE_DESC::flags can be any combination of the following: - * - ::CU_TRSF_READ_AS_INTEGER, which suppresses the default behavior of having the texture promote integer data to floating point data in the - * range [0, 1]. Note that texture with 32-bit integer format would not be promoted, regardless of whether or not this flag is specified. - * - ::CU_TRSF_NORMALIZED_COORDINATES, which suppresses the default behavior of having the texture coordinates range from [0, Dim) where Dim is - * the width or height of the CUDA array. Instead, the texture coordinates [0, 1.0) reference the entire breadth of the array dimension; Note + * - ::CU_TRSF_READ_AS_INTEGER, which suppresses the default behavior of + having the texture promote integer data to floating point data in the + * range [0, 1]. Note that texture with 32-bit integer format would not be + promoted, regardless of whether or not this flag is specified. + * - ::CU_TRSF_NORMALIZED_COORDINATES, which suppresses the default behavior + of having the texture coordinates range from [0, Dim) where Dim is + * the width or height of the CUDA array. Instead, the texture coordinates + [0, 1.0) reference the entire breadth of the array dimension; Note * that for CUDA mipmapped arrays, this flag has to be set. * - * - ::CUDA_TEXTURE_DESC::maxAnisotropy specifies the maximum anisotropy ratio to be used when doing anisotropic filtering. This value will be + * - ::CUDA_TEXTURE_DESC::maxAnisotropy specifies the maximum anisotropy ratio + to be used when doing anisotropic filtering. This value will be * clamped to the range [1,16]. * - * - ::CUDA_TEXTURE_DESC::mipmapFilterMode specifies the filter mode when the calculated mipmap level lies between two defined mipmap levels. + * - ::CUDA_TEXTURE_DESC::mipmapFilterMode specifies the filter mode when the + calculated mipmap level lies between two defined mipmap levels. * - * - ::CUDA_TEXTURE_DESC::mipmapLevelBias specifies the offset to be applied to the calculated mipmap level. + * - ::CUDA_TEXTURE_DESC::mipmapLevelBias specifies the offset to be applied to + the calculated mipmap level. * - * - ::CUDA_TEXTURE_DESC::minMipmapLevelClamp specifies the lower end of the mipmap level range to clamp access to. + * - ::CUDA_TEXTURE_DESC::minMipmapLevelClamp specifies the lower end of the + mipmap level range to clamp access to. * - * - ::CUDA_TEXTURE_DESC::maxMipmapLevelClamp specifies the upper end of the mipmap level range to clamp access to. + * - ::CUDA_TEXTURE_DESC::maxMipmapLevelClamp specifies the upper end of the + mipmap level range to clamp access to. * * * The ::CUDA_RESOURCE_VIEW_DESC struct is defined as @@ -13735,36 +14504,53 @@ CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef); } CUDA_RESOURCE_VIEW_DESC; * \endcode * where: - * - ::CUDA_RESOURCE_VIEW_DESC::format specifies how the data contained in the CUDA array or CUDA mipmapped array should - * be interpreted. Note that this can incur a change in size of the texture data. If the resource view format is a block - * compressed format, then the underlying CUDA array or CUDA mipmapped array has to have a base of format ::CU_AD_FORMAT_UNSIGNED_INT32. - * with 2 or 4 channels, depending on the block compressed format. For ex., BC1 and BC4 require the underlying CUDA array to have - * a format of ::CU_AD_FORMAT_UNSIGNED_INT32 with 2 channels. The other BC formats require the underlying resource to have the same base + * - ::CUDA_RESOURCE_VIEW_DESC::format specifies how the data contained in the + CUDA array or CUDA mipmapped array should + * be interpreted. Note that this can incur a change in size of the texture + data. If the resource view format is a block + * compressed format, then the underlying CUDA array or CUDA mipmapped array + has to have a base of format ::CU_AD_FORMAT_UNSIGNED_INT32. + * with 2 or 4 channels, depending on the block compressed format. For ex., + BC1 and BC4 require the underlying CUDA array to have + * a format of ::CU_AD_FORMAT_UNSIGNED_INT32 with 2 channels. The other BC + formats require the underlying resource to have the same base * format but with 4 channels. * - * - ::CUDA_RESOURCE_VIEW_DESC::width specifies the new width of the texture data. If the resource view format is a block - * compressed format, this value has to be 4 times the original width of the resource. For non block compressed formats, + * - ::CUDA_RESOURCE_VIEW_DESC::width specifies the new width of the texture + data. If the resource view format is a block + * compressed format, this value has to be 4 times the original width of the + resource. For non block compressed formats, * this value has to be equal to that of the original resource. * - * - ::CUDA_RESOURCE_VIEW_DESC::height specifies the new height of the texture data. If the resource view format is a block - * compressed format, this value has to be 4 times the original height of the resource. For non block compressed formats, + * - ::CUDA_RESOURCE_VIEW_DESC::height specifies the new height of the texture + data. If the resource view format is a block + * compressed format, this value has to be 4 times the original height of the + resource. For non block compressed formats, * this value has to be equal to that of the original resource. * - * - ::CUDA_RESOURCE_VIEW_DESC::depth specifies the new depth of the texture data. This value has to be equal to that of the + * - ::CUDA_RESOURCE_VIEW_DESC::depth specifies the new depth of the texture + data. This value has to be equal to that of the * original resource. * - * - ::CUDA_RESOURCE_VIEW_DESC::firstMipmapLevel specifies the most detailed mipmap level. This will be the new mipmap level zero. - * For non-mipmapped resources, this value has to be zero.::CUDA_TEXTURE_DESC::minMipmapLevelClamp and ::CUDA_TEXTURE_DESC::maxMipmapLevelClamp - * will be relative to this value. For ex., if the firstMipmapLevel is set to 2, and a minMipmapLevelClamp of 1.2 is specified, + * - ::CUDA_RESOURCE_VIEW_DESC::firstMipmapLevel specifies the most detailed + mipmap level. This will be the new mipmap level zero. + * For non-mipmapped resources, this value has to be + zero.::CUDA_TEXTURE_DESC::minMipmapLevelClamp and + ::CUDA_TEXTURE_DESC::maxMipmapLevelClamp + * will be relative to this value. For ex., if the firstMipmapLevel is set to + 2, and a minMipmapLevelClamp of 1.2 is specified, * then the actual minimum mipmap level clamp will be 3.2. * - * - ::CUDA_RESOURCE_VIEW_DESC::lastMipmapLevel specifies the least detailed mipmap level. For non-mipmapped resources, this value + * - ::CUDA_RESOURCE_VIEW_DESC::lastMipmapLevel specifies the least detailed + mipmap level. For non-mipmapped resources, this value * has to be zero. * - * - ::CUDA_RESOURCE_VIEW_DESC::firstLayer specifies the first layer index for layered textures. This will be the new layer zero. + * - ::CUDA_RESOURCE_VIEW_DESC::firstLayer specifies the first layer index for + layered textures. This will be the new layer zero. * For non-layered resources, this value has to be zero. * - * - ::CUDA_RESOURCE_VIEW_DESC::lastLayer specifies the last layer index for layered textures. For non-layered resources, + * - ::CUDA_RESOURCE_VIEW_DESC::lastLayer specifies the last layer index for + layered textures. For non-layered resources, * this value has to be zero. * * @@ -13784,7 +14570,10 @@ CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef); * ::cuTexObjectDestroy, * ::cudaCreateTextureObject */ -CUresult CUDAAPI cuTexObjectCreate(CUtexObject *pTexObject, const CUDA_RESOURCE_DESC *pResDesc, const CUDA_TEXTURE_DESC *pTexDesc, const CUDA_RESOURCE_VIEW_DESC *pResViewDesc); +CUresult CUDAAPI cuTexObjectCreate(CUtexObject *pTexObject, + const CUDA_RESOURCE_DESC *pResDesc, + const CUDA_TEXTURE_DESC *pTexDesc, + const CUDA_RESOURCE_VIEW_DESC *pResViewDesc); /** * \brief Destroys a texture object @@ -13809,7 +14598,8 @@ CUresult CUDAAPI cuTexObjectDestroy(CUtexObject texObject); /** * \brief Returns a texture object's resource descriptor * - * Returns the resource descriptor for the texture object specified by \p texObject. + * Returns the resource descriptor for the texture object specified by \p + * texObject. * * \param pResDesc - Resource descriptor * \param texObject - Texture object @@ -13825,12 +14615,14 @@ CUresult CUDAAPI cuTexObjectDestroy(CUtexObject texObject); * ::cuTexObjectCreate, * ::cudaGetTextureObjectResourceDesc, */ -CUresult CUDAAPI cuTexObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, CUtexObject texObject); +CUresult CUDAAPI cuTexObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, + CUtexObject texObject); /** * \brief Returns a texture object's texture descriptor * - * Returns the texture descriptor for the texture object specified by \p texObject. + * Returns the texture descriptor for the texture object specified by \p + * texObject. * * \param pTexDesc - Texture descriptor * \param texObject - Texture object @@ -13846,13 +14638,15 @@ CUresult CUDAAPI cuTexObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, CUtexO * ::cuTexObjectCreate, * ::cudaGetTextureObjectTextureDesc */ -CUresult CUDAAPI cuTexObjectGetTextureDesc(CUDA_TEXTURE_DESC *pTexDesc, CUtexObject texObject); +CUresult CUDAAPI cuTexObjectGetTextureDesc(CUDA_TEXTURE_DESC *pTexDesc, + CUtexObject texObject); /** * \brief Returns a texture object's resource view descriptor * - * Returns the resource view descriptor for the texture object specified by \p texObject. - * If no resource view was set for \p texObject, the ::CUDA_ERROR_INVALID_VALUE is returned. + * Returns the resource view descriptor for the texture object specified by \p + * texObject. If no resource view was set for \p texObject, the + * ::CUDA_ERROR_INVALID_VALUE is returned. * * \param pResViewDesc - Resource view descriptor * \param texObject - Texture object @@ -13868,7 +14662,8 @@ CUresult CUDAAPI cuTexObjectGetTextureDesc(CUDA_TEXTURE_DESC *pTexDesc, CUtexObj * ::cuTexObjectCreate, * ::cudaGetTextureObjectResourceViewDesc */ -CUresult CUDAAPI cuTexObjectGetResourceViewDesc(CUDA_RESOURCE_VIEW_DESC *pResViewDesc, CUtexObject texObject); +CUresult CUDAAPI cuTexObjectGetResourceViewDesc( + CUDA_RESOURCE_VIEW_DESC *pResViewDesc, CUtexObject texObject); /** @} */ /* END CUDA_TEXOBJECT */ @@ -13888,14 +14683,16 @@ CUresult CUDAAPI cuTexObjectGetResourceViewDesc(CUDA_RESOURCE_VIEW_DESC *pResVie /** * \brief Creates a surface object * - * Creates a surface object and returns it in \p pSurfObject. \p pResDesc describes - * the data to perform surface load/stores on. ::CUDA_RESOURCE_DESC::resType must be + * Creates a surface object and returns it in \p pSurfObject. \p pResDesc + * describes the data to perform surface load/stores on. + * ::CUDA_RESOURCE_DESC::resType must be * ::CU_RESOURCE_TYPE_ARRAY and ::CUDA_RESOURCE_DESC::res::array::hArray - * must be set to a valid CUDA array handle. ::CUDA_RESOURCE_DESC::flags must be set to zero. + * must be set to a valid CUDA array handle. ::CUDA_RESOURCE_DESC::flags must be + * set to zero. * - * Surface objects are only supported on devices of compute capability 3.0 or higher. - * Additionally, a surface object is an opaque value, and, as such, should only be - * accessed through CUDA API calls. + * Surface objects are only supported on devices of compute capability 3.0 or + * higher. Additionally, a surface object is an opaque value, and, as such, + * should only be accessed through CUDA API calls. * * \param pSurfObject - Surface object to create * \param pResDesc - Resource descriptor @@ -13911,7 +14708,8 @@ CUresult CUDAAPI cuTexObjectGetResourceViewDesc(CUDA_RESOURCE_VIEW_DESC *pResVie * ::cuSurfObjectDestroy, * ::cudaCreateSurfaceObject */ -CUresult CUDAAPI cuSurfObjectCreate(CUsurfObject *pSurfObject, const CUDA_RESOURCE_DESC *pResDesc); +CUresult CUDAAPI cuSurfObjectCreate(CUsurfObject *pSurfObject, + const CUDA_RESOURCE_DESC *pResDesc); /** * \brief Destroys a surface object @@ -13936,7 +14734,8 @@ CUresult CUDAAPI cuSurfObjectDestroy(CUsurfObject surfObject); /** * \brief Returns a surface object's resource descriptor * - * Returns the resource descriptor for the surface object specified by \p surfObject. + * Returns the resource descriptor for the surface object specified by \p + * surfObject. * * \param pResDesc - Resource descriptor * \param surfObject - Surface object @@ -13952,10 +14751,11 @@ CUresult CUDAAPI cuSurfObjectDestroy(CUsurfObject surfObject); * ::cuSurfObjectCreate, * ::cudaGetSurfaceObjectResourceDesc */ -CUresult CUDAAPI cuSurfObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, CUsurfObject surfObject); +CUresult CUDAAPI cuSurfObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, + CUsurfObject surfObject); /** @} */ /* END CUDA_SURFOBJECT */ -#endif /* __CUDA_API_VERSION >= 5000 */ +#endif /* __CUDA_API_VERSION >= 5000 */ /** * \defgroup CUDA_PEER_ACCESS Peer Context Memory Access @@ -13974,16 +14774,16 @@ CUresult CUDAAPI cuSurfObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, CUsur /** * \brief Queries if a device may directly access a peer device's memory. * - * Returns in \p *canAccessPeer a value of 1 if contexts on \p dev are capable of - * directly accessing memory from contexts on \p peerDev and 0 otherwise. - * If direct access of \p peerDev from \p dev is possible, then access may be + * Returns in \p *canAccessPeer a value of 1 if contexts on \p dev are capable + * of directly accessing memory from contexts on \p peerDev and 0 otherwise. If + * direct access of \p peerDev from \p dev is possible, then access may be * enabled on two specific contexts by calling ::cuCtxEnablePeerAccess(). * * \param canAccessPeer - Returned access capability * \param dev - Device from which allocations on \p peerDev are to * be directly accessed. - * \param peerDev - Device on which the allocations to be directly accessed - * by \p dev reside. + * \param peerDev - Device on which the allocations to be directly + * accessed by \p dev reside. * * \return * ::CUDA_SUCCESS, @@ -13997,26 +14797,28 @@ CUresult CUDAAPI cuSurfObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, CUsur * ::cuCtxDisablePeerAccess, * ::cudaDeviceCanAccessPeer */ -CUresult CUDAAPI cuDeviceCanAccessPeer(int *canAccessPeer, CUdevice dev, CUdevice peerDev); +CUresult CUDAAPI cuDeviceCanAccessPeer(int *canAccessPeer, CUdevice dev, + CUdevice peerDev); /** * \brief Enables direct access to memory allocations in a peer context. * - * If both the current context and \p peerContext are on devices which support unified - * addressing (as may be queried using ::CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING) and same - * major compute capability, then on success all allocations from \p peerContext will - * immediately be accessible by the current context. See \ref CUDA_UNIFIED for additional + * If both the current context and \p peerContext are on devices which support + * unified addressing (as may be queried using + * ::CU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING) and same major compute capability, + * then on success all allocations from \p peerContext will immediately be + * accessible by the current context. See \ref CUDA_UNIFIED for additional * details. * - * Note that access granted by this call is unidirectional and that in order to access - * memory from the current context in \p peerContext, a separate symmetric call - * to ::cuCtxEnablePeerAccess() is required. + * Note that access granted by this call is unidirectional and that in order to + * access memory from the current context in \p peerContext, a separate + * symmetric call to ::cuCtxEnablePeerAccess() is required. * * There is a system-wide maximum of eight peer connections per device. * - * Returns ::CUDA_ERROR_PEER_ACCESS_UNSUPPORTED if ::cuDeviceCanAccessPeer() indicates - * that the ::CUdevice of the current context cannot directly access memory - * from the ::CUdevice of \p peerContext. + * Returns ::CUDA_ERROR_PEER_ACCESS_UNSUPPORTED if ::cuDeviceCanAccessPeer() + * indicates that the ::CUdevice of the current context cannot directly access + * memory from the ::CUdevice of \p peerContext. * * Returns ::CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED if direct access of * \p peerContext from the current context has already been enabled. @@ -14024,13 +14826,14 @@ CUresult CUDAAPI cuDeviceCanAccessPeer(int *canAccessPeer, CUdevice dev, CUdevic * Returns ::CUDA_ERROR_TOO_MANY_PEERS if direct peer access is not possible * because hardware resources required for peer access have been exhausted. * - * Returns ::CUDA_ERROR_INVALID_CONTEXT if there is no current context, \p peerContext - * is not a valid context, or if the current context is \p peerContext. + * Returns ::CUDA_ERROR_INVALID_CONTEXT if there is no current context, \p + * peerContext is not a valid context, or if the current context is \p + * peerContext. * * Returns ::CUDA_ERROR_INVALID_VALUE if \p Flags is not 0. * - * \param peerContext - Peer context to enable direct access to from the current context - * \param Flags - Reserved for future use and must be set to 0 + * \param peerContext - Peer context to enable direct access to from the current + * context \param Flags - Reserved for future use and must be set to 0 * * \return * ::CUDA_SUCCESS, @@ -14048,7 +14851,8 @@ CUresult CUDAAPI cuDeviceCanAccessPeer(int *canAccessPeer, CUdevice dev, CUdevic * ::cuCtxDisablePeerAccess, * ::cudaDeviceEnablePeerAccess */ -CUresult CUDAAPI cuCtxEnablePeerAccess(CUcontext peerContext, unsigned int Flags); +CUresult CUDAAPI cuCtxEnablePeerAccess(CUcontext peerContext, + unsigned int Flags); /** * \brief Disables direct access to memory allocations in a peer context and @@ -14089,21 +14893,22 @@ CUresult CUDAAPI cuCtxDisablePeerAccess(CUcontext peerContext); * - ::CU_DEVICE_P2P_ATTRIBUTE_PERFORMANCE_RANK: A relative value indicating the * performance of the link between two devices. * - ::CU_DEVICE_P2P_ATTRIBUTE_ACCESS_SUPPORTED P2P: 1 if P2P Access is enable. - * - ::CU_DEVICE_P2P_ATTRIBUTE_NATIVE_ATOMIC_SUPPORTED: 1 if Atomic operations over - * the link are supported. + * - ::CU_DEVICE_P2P_ATTRIBUTE_NATIVE_ATOMIC_SUPPORTED: 1 if Atomic operations + * over the link are supported. * - ::CU_DEVICE_P2P_ATTRIBUTE_CUDA_ARRAY_ACCESS_SUPPORTED: 1 if cudaArray can * be accessed over the link. * - * Returns ::CUDA_ERROR_INVALID_DEVICE if \p srcDevice or \p dstDevice are not valid - * or if they represent the same device. + * Returns ::CUDA_ERROR_INVALID_DEVICE if \p srcDevice or \p dstDevice are not + * valid or if they represent the same device. * - * Returns ::CUDA_ERROR_INVALID_VALUE if \p attrib is not valid or if \p value is - * a null pointer. + * Returns ::CUDA_ERROR_INVALID_VALUE if \p attrib is not valid or if \p value + * is a null pointer. * * \param value - Returned value of the requested attribute - * \param attrib - The requested attribute of the link between \p srcDevice and \p dstDevice. - * \param srcDevice - The source device of the target link. - * \param dstDevice - The destination device of the target link. + * \param attrib - The requested attribute of the link between \p + * srcDevice and \p dstDevice. \param srcDevice - The source device of the + * target link. \param dstDevice - The destination device of the target + * link. * * \return * ::CUDA_SUCCESS, @@ -14119,7 +14924,10 @@ CUresult CUDAAPI cuCtxDisablePeerAccess(CUcontext peerContext); * ::cuDeviceCanAccessPeer, * ::cudaDeviceGetP2PAttribute */ -CUresult CUDAAPI cuDeviceGetP2PAttribute(int* value, CUdevice_P2PAttribute attrib, CUdevice srcDevice, CUdevice dstDevice); +CUresult CUDAAPI cuDeviceGetP2PAttribute(int *value, + CUdevice_P2PAttribute attrib, + CUdevice srcDevice, + CUdevice dstDevice); #endif /* __CUDA_API_VERSION >= 8000 */ @@ -14168,12 +14976,13 @@ CUresult CUDAAPI cuDeviceGetP2PAttribute(int* value, CUdevice_P2PAttribute attri CUresult CUDAAPI cuGraphicsUnregisterResource(CUgraphicsResource resource); /** - * \brief Get an array through which to access a subresource of a mapped graphics resource. + * \brief Get an array through which to access a subresource of a mapped + * graphics resource. * * Returns in \p *pArray an array through which the subresource of the mapped * graphics resource \p resource which corresponds to array index \p arrayIndex - * and mipmap level \p mipLevel may be accessed. The value set in \p *pArray may - * change every time that \p resource is mapped. + * and mipmap level \p mipLevel may be accessed. The value set in \p *pArray + * may change every time that \p resource is mapped. * * If \p resource is not a texture then it cannot be accessed via an array and * ::CUDA_ERROR_NOT_MAPPED_AS_ARRAY is returned. @@ -14183,8 +14992,8 @@ CUresult CUDAAPI cuGraphicsUnregisterResource(CUgraphicsResource resource); * ::CUDA_ERROR_INVALID_VALUE is returned. * If \p resource is not mapped then ::CUDA_ERROR_NOT_MAPPED is returned. * - * \param pArray - Returned array through which a subresource of \p resource may be accessed - * \param resource - Mapped resource to access + * \param pArray - Returned array through which a subresource of \p + * resource may be accessed \param resource - Mapped resource to access * \param arrayIndex - Array index for array textures or cubemap face * index as defined by ::CUarray_cubemap_face for * cubemap textures for the subresource to access @@ -14205,23 +15014,27 @@ CUresult CUDAAPI cuGraphicsUnregisterResource(CUgraphicsResource resource); * ::cuGraphicsResourceGetMappedPointer, * ::cudaGraphicsSubResourceGetMappedArray */ -CUresult CUDAAPI cuGraphicsSubResourceGetMappedArray(CUarray *pArray, CUgraphicsResource resource, unsigned int arrayIndex, unsigned int mipLevel); +CUresult CUDAAPI cuGraphicsSubResourceGetMappedArray( + CUarray *pArray, CUgraphicsResource resource, unsigned int arrayIndex, + unsigned int mipLevel); #if __CUDA_API_VERSION >= 5000 /** - * \brief Get a mipmapped array through which to access a mapped graphics resource. + * \brief Get a mipmapped array through which to access a mapped graphics + * resource. * - * Returns in \p *pMipmappedArray a mipmapped array through which the mapped graphics - * resource \p resource. The value set in \p *pMipmappedArray may change every time - * that \p resource is mapped. + * Returns in \p *pMipmappedArray a mipmapped array through which the mapped + * graphics resource \p resource. The value set in \p *pMipmappedArray may + * change every time that \p resource is mapped. * - * If \p resource is not a texture then it cannot be accessed via a mipmapped array and + * If \p resource is not a texture then it cannot be accessed via a mipmapped + * array and * ::CUDA_ERROR_NOT_MAPPED_AS_ARRAY is returned. * If \p resource is not mapped then ::CUDA_ERROR_NOT_MAPPED is returned. * - * \param pMipmappedArray - Returned mipmapped array through which \p resource may be accessed - * \param resource - Mapped resource to access + * \param pMipmappedArray - Returned mipmapped array through which \p resource + * may be accessed \param resource - Mapped resource to access * * \return * ::CUDA_SUCCESS, @@ -14238,26 +15051,29 @@ CUresult CUDAAPI cuGraphicsSubResourceGetMappedArray(CUarray *pArray, CUgraphics * ::cuGraphicsResourceGetMappedPointer, * ::cudaGraphicsResourceGetMappedMipmappedArray */ -CUresult CUDAAPI cuGraphicsResourceGetMappedMipmappedArray(CUmipmappedArray *pMipmappedArray, CUgraphicsResource resource); +CUresult CUDAAPI cuGraphicsResourceGetMappedMipmappedArray( + CUmipmappedArray *pMipmappedArray, CUgraphicsResource resource); #endif /* __CUDA_API_VERSION >= 5000 */ #if __CUDA_API_VERSION >= 3020 /** - * \brief Get a device pointer through which to access a mapped graphics resource. + * \brief Get a device pointer through which to access a mapped graphics + * resource. * * Returns in \p *pDevPtr a pointer through which the mapped graphics resource * \p resource may be accessed. - * Returns in \p pSize the size of the memory in bytes which may be accessed from that pointer. - * The value set in \p pPointer may change every time that \p resource is mapped. + * Returns in \p pSize the size of the memory in bytes which may be accessed + * from that pointer. The value set in \p pPointer may change every time that \p + * resource is mapped. * * If \p resource is not a buffer then it cannot be accessed via a pointer and * ::CUDA_ERROR_NOT_MAPPED_AS_POINTER is returned. * If \p resource is not mapped then ::CUDA_ERROR_NOT_MAPPED is returned. * * - * \param pDevPtr - Returned pointer through which \p resource may be accessed - * \param pSize - Returned size of the buffer accessible starting at \p *pPointer - * \param resource - Mapped resource to access + * \param pDevPtr - Returned pointer through which \p resource may be + * accessed \param pSize - Returned size of the buffer accessible starting + * at \p *pPointer \param resource - Mapped resource to access * * \return * ::CUDA_SUCCESS, @@ -14275,7 +15091,8 @@ CUresult CUDAAPI cuGraphicsResourceGetMappedMipmappedArray(CUmipmappedArray *pMi * ::cuGraphicsSubResourceGetMappedArray, * ::cudaGraphicsResourceGetMappedPointer */ -CUresult CUDAAPI cuGraphicsResourceGetMappedPointer(CUdeviceptr *pDevPtr, size_t *pSize, CUgraphicsResource resource); +CUresult CUDAAPI cuGraphicsResourceGetMappedPointer( + CUdeviceptr *pDevPtr, size_t *pSize, CUgraphicsResource resource); #endif /* __CUDA_API_VERSION >= 3020 */ /** @@ -14289,7 +15106,8 @@ CUresult CUDAAPI cuGraphicsResourceGetMappedPointer(CUdeviceptr *pDevPtr, size_t * - ::CU_GRAPHICS_MAP_RESOURCE_FLAGS_NONE: Specifies no hints about how this * resource will be used. It is therefore assumed that this resource will be * read from and written to by CUDA kernels. This is the default value. - * - ::CU_GRAPHICS_MAP_RESOURCE_FLAGS_READONLY: Specifies that CUDA kernels which + * - ::CU_GRAPHICS_MAP_RESOURCE_FLAGS_READONLY: Specifies that CUDA kernels + which * access this resource will not write to this resource. * - ::CU_GRAPHICS_MAP_RESOURCE_FLAGS_WRITEDISCARD: Specifies that CUDA kernels * which access this resource will not read from this resource and will @@ -14298,7 +15116,8 @@ CUresult CUDAAPI cuGraphicsResourceGetMappedPointer(CUdeviceptr *pDevPtr, size_t * * If \p resource is presently mapped for access by CUDA then * ::CUDA_ERROR_ALREADY_MAPPED is returned. - * If \p flags is not one of the above values then ::CUDA_ERROR_INVALID_VALUE is returned. + * If \p flags is not one of the above values then ::CUDA_ERROR_INVALID_VALUE is + returned. * * \param resource - Registered resource to set flags for * \param flags - Parameters for resource mapping @@ -14317,7 +15136,8 @@ CUresult CUDAAPI cuGraphicsResourceGetMappedPointer(CUdeviceptr *pDevPtr, size_t * ::cuGraphicsMapResources, * ::cudaGraphicsResourceSetMapFlags */ -CUresult CUDAAPI cuGraphicsResourceSetMapFlags(CUgraphicsResource resource, unsigned int flags); +CUresult CUDAAPI cuGraphicsResourceSetMapFlags(CUgraphicsResource resource, + unsigned int flags); /** * \brief Map graphics resources for access by CUDA @@ -14330,11 +15150,12 @@ CUresult CUDAAPI cuGraphicsResourceSetMapFlags(CUgraphicsResource resource, unsi * application does so, the results are undefined. * * This function provides the synchronization guarantee that any graphics calls - * issued before ::cuGraphicsMapResources() will complete before any subsequent CUDA - * work issued in \p stream begins. + * issued before ::cuGraphicsMapResources() will complete before any subsequent + * CUDA work issued in \p stream begins. * - * If \p resources includes any duplicate entries then ::CUDA_ERROR_INVALID_HANDLE is returned. - * If any of \p resources are presently mapped for access by CUDA then ::CUDA_ERROR_ALREADY_MAPPED is returned. + * If \p resources includes any duplicate entries then + * ::CUDA_ERROR_INVALID_HANDLE is returned. If any of \p resources are presently + * mapped for access by CUDA then ::CUDA_ERROR_ALREADY_MAPPED is returned. * * \param count - Number of resources to map * \param resources - Resources to map for CUDA usage @@ -14357,7 +15178,9 @@ CUresult CUDAAPI cuGraphicsResourceSetMapFlags(CUgraphicsResource resource, unsi * ::cuGraphicsUnmapResources, * ::cudaGraphicsMapResources */ -CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, CUgraphicsResource *resources, CUstream hStream); +CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, + CUgraphicsResource *resources, + CUstream hStream); /** * \brief Unmap graphics resources. @@ -14367,13 +15190,14 @@ CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, CUgraphicsResource * * Once unmapped, the resources in \p resources may not be accessed by CUDA * until they are mapped again. * - * This function provides the synchronization guarantee that any CUDA work issued - * in \p stream before ::cuGraphicsUnmapResources() will complete before any - * subsequently issued graphics work begins. + * This function provides the synchronization guarantee that any CUDA work + * issued in \p stream before ::cuGraphicsUnmapResources() will complete before + * any subsequently issued graphics work begins. * * - * If \p resources includes any duplicate entries then ::CUDA_ERROR_INVALID_HANDLE is returned. - * If any of \p resources are not presently mapped for access by CUDA then ::CUDA_ERROR_NOT_MAPPED is returned. + * If \p resources includes any duplicate entries then + * ::CUDA_ERROR_INVALID_HANDLE is returned. If any of \p resources are not + * presently mapped for access by CUDA then ::CUDA_ERROR_NOT_MAPPED is returned. * * \param count - Number of resources to unmap * \param resources - Resources to unmap @@ -14394,264 +15218,318 @@ CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, CUgraphicsResource * * ::cuGraphicsMapResources, * ::cudaGraphicsUnmapResources */ -CUresult CUDAAPI cuGraphicsUnmapResources(unsigned int count, CUgraphicsResource *resources, CUstream hStream); +CUresult CUDAAPI cuGraphicsUnmapResources(unsigned int count, + CUgraphicsResource *resources, + CUstream hStream); /** @} */ /* END CUDA_GRAPHICS */ -CUresult CUDAAPI cuGetExportTable(const void **ppExportTable, const CUuuid *pExportTableId); - +CUresult CUDAAPI cuGetExportTable(const void **ppExportTable, + const CUuuid *pExportTableId); /** * CUDA API versioning support */ #if defined(__CUDA_API_VERSION_INTERNAL) - #undef cuMemHostRegister - #undef cuGraphicsResourceSetMapFlags - #undef cuLinkCreate - #undef cuLinkAddData - #undef cuLinkAddFile - #undef cuDeviceTotalMem - #undef cuCtxCreate - #undef cuModuleGetGlobal - #undef cuMemGetInfo - #undef cuMemAlloc - #undef cuMemAllocPitch - #undef cuMemFree - #undef cuMemGetAddressRange - #undef cuMemAllocHost - #undef cuMemHostGetDevicePointer - #undef cuMemcpyHtoD - #undef cuMemcpyDtoH - #undef cuMemcpyDtoD - #undef cuMemcpyDtoA - #undef cuMemcpyAtoD - #undef cuMemcpyHtoA - #undef cuMemcpyAtoH - #undef cuMemcpyAtoA - #undef cuMemcpyHtoAAsync - #undef cuMemcpyAtoHAsync - #undef cuMemcpy2D - #undef cuMemcpy2DUnaligned - #undef cuMemcpy3D - #undef cuMemcpyHtoDAsync - #undef cuMemcpyDtoHAsync - #undef cuMemcpyDtoDAsync - #undef cuMemcpy2DAsync - #undef cuMemcpy3DAsync - #undef cuMemsetD8 - #undef cuMemsetD16 - #undef cuMemsetD32 - #undef cuMemsetD2D8 - #undef cuMemsetD2D16 - #undef cuMemsetD2D32 - #undef cuArrayCreate - #undef cuArrayGetDescriptor - #undef cuArray3DCreate - #undef cuArray3DGetDescriptor - #undef cuTexRefSetAddress - #undef cuTexRefSetAddress2D - #undef cuTexRefGetAddress - #undef cuGraphicsResourceGetMappedPointer - #undef cuCtxDestroy - #undef cuCtxPopCurrent - #undef cuCtxPushCurrent - #undef cuStreamDestroy - #undef cuEventDestroy - #undef cuMemcpy - #undef cuMemcpyAsync - #undef cuMemcpyPeer - #undef cuMemcpyPeerAsync - #undef cuMemcpy3DPeer - #undef cuMemcpy3DPeerAsync - #undef cuMemsetD8Async - #undef cuMemsetD16Async - #undef cuMemsetD32Async - #undef cuMemsetD2D8Async - #undef cuMemsetD2D16Async - #undef cuMemsetD2D32Async - #undef cuStreamGetPriority - #undef cuStreamGetFlags - #undef cuStreamGetCtx - #undef cuStreamWaitEvent - #undef cuStreamAddCallback - #undef cuStreamAttachMemAsync - #undef cuStreamQuery - #undef cuStreamSynchronize - #undef cuEventRecord - #undef cuLaunchKernel - #undef cuLaunchHostFunc - #undef cuGraphicsMapResources - #undef cuGraphicsUnmapResources - #undef cuStreamWriteValue32 - #undef cuStreamWaitValue32 - #undef cuStreamWriteValue64 - #undef cuStreamWaitValue64 - #undef cuStreamBatchMemOp - #undef cuMemPrefetchAsync - #undef cuLaunchCooperativeKernel - #undef cuSignalExternalSemaphoresAsync - #undef cuWaitExternalSemaphoresAsync - #undef cuStreamBeginCapture - #undef cuStreamEndCapture - #undef cuStreamIsCapturing - #undef cuStreamGetCaptureInfo - #undef cuGraphLaunch +#undef cuMemHostRegister +#undef cuGraphicsResourceSetMapFlags +#undef cuLinkCreate +#undef cuLinkAddData +#undef cuLinkAddFile +#undef cuDeviceTotalMem +#undef cuCtxCreate +#undef cuModuleGetGlobal +#undef cuMemGetInfo +#undef cuMemAlloc +#undef cuMemAllocPitch +#undef cuMemFree +#undef cuMemGetAddressRange +#undef cuMemAllocHost +#undef cuMemHostGetDevicePointer +#undef cuMemcpyHtoD +#undef cuMemcpyDtoH +#undef cuMemcpyDtoD +#undef cuMemcpyDtoA +#undef cuMemcpyAtoD +#undef cuMemcpyHtoA +#undef cuMemcpyAtoH +#undef cuMemcpyAtoA +#undef cuMemcpyHtoAAsync +#undef cuMemcpyAtoHAsync +#undef cuMemcpy2D +#undef cuMemcpy2DUnaligned +#undef cuMemcpy3D +#undef cuMemcpyHtoDAsync +#undef cuMemcpyDtoHAsync +#undef cuMemcpyDtoDAsync +#undef cuMemcpy2DAsync +#undef cuMemcpy3DAsync +#undef cuMemsetD8 +#undef cuMemsetD16 +#undef cuMemsetD32 +#undef cuMemsetD2D8 +#undef cuMemsetD2D16 +#undef cuMemsetD2D32 +#undef cuArrayCreate +#undef cuArrayGetDescriptor +#undef cuArray3DCreate +#undef cuArray3DGetDescriptor +#undef cuTexRefSetAddress +#undef cuTexRefSetAddress2D +#undef cuTexRefGetAddress +#undef cuGraphicsResourceGetMappedPointer +#undef cuCtxDestroy +#undef cuCtxPopCurrent +#undef cuCtxPushCurrent +#undef cuStreamDestroy +#undef cuEventDestroy +#undef cuMemcpy +#undef cuMemcpyAsync +#undef cuMemcpyPeer +#undef cuMemcpyPeerAsync +#undef cuMemcpy3DPeer +#undef cuMemcpy3DPeerAsync +#undef cuMemsetD8Async +#undef cuMemsetD16Async +#undef cuMemsetD32Async +#undef cuMemsetD2D8Async +#undef cuMemsetD2D16Async +#undef cuMemsetD2D32Async +#undef cuStreamGetPriority +#undef cuStreamGetFlags +#undef cuStreamGetCtx +#undef cuStreamWaitEvent +#undef cuStreamAddCallback +#undef cuStreamAttachMemAsync +#undef cuStreamQuery +#undef cuStreamSynchronize +#undef cuEventRecord +#undef cuLaunchKernel +#undef cuLaunchHostFunc +#undef cuGraphicsMapResources +#undef cuGraphicsUnmapResources +#undef cuStreamWriteValue32 +#undef cuStreamWaitValue32 +#undef cuStreamWriteValue64 +#undef cuStreamWaitValue64 +#undef cuStreamBatchMemOp +#undef cuMemPrefetchAsync +#undef cuLaunchCooperativeKernel +#undef cuSignalExternalSemaphoresAsync +#undef cuWaitExternalSemaphoresAsync +#undef cuStreamBeginCapture +#undef cuStreamEndCapture +#undef cuStreamIsCapturing +#undef cuStreamGetCaptureInfo +#undef cuGraphLaunch #endif /* __CUDA_API_VERSION_INTERNAL */ -#if defined(__CUDA_API_VERSION_INTERNAL) || (__CUDA_API_VERSION >= 4000 && __CUDA_API_VERSION < 6050) -CUresult CUDAAPI cuMemHostRegister(void *p, size_t bytesize, unsigned int Flags); -#endif /* defined(__CUDA_API_VERSION_INTERNAL) || (__CUDA_API_VERSION >= 4000 && __CUDA_API_VERSION < 6050) */ +#if defined(__CUDA_API_VERSION_INTERNAL) || \ + (__CUDA_API_VERSION >= 4000 && __CUDA_API_VERSION < 6050) +CUresult CUDAAPI cuMemHostRegister(void *p, size_t bytesize, + unsigned int Flags); +#endif /* defined(__CUDA_API_VERSION_INTERNAL) || (__CUDA_API_VERSION >= 4000 \ + && __CUDA_API_VERSION < 6050) */ #if defined(__CUDA_API_VERSION_INTERNAL) || __CUDA_API_VERSION < 6050 -CUresult CUDAAPI cuGraphicsResourceSetMapFlags(CUgraphicsResource resource, unsigned int flags); +CUresult CUDAAPI cuGraphicsResourceSetMapFlags(CUgraphicsResource resource, + unsigned int flags); #endif /* defined(__CUDA_API_VERSION_INTERNAL) || __CUDA_API_VERSION < 6050 */ -#if defined(__CUDA_API_VERSION_INTERNAL) || (__CUDA_API_VERSION >= 5050 && __CUDA_API_VERSION < 6050) -CUresult CUDAAPI cuLinkCreate(unsigned int numOptions, CUjit_option *options, void **optionValues, CUlinkState *stateOut); -CUresult CUDAAPI cuLinkAddData(CUlinkState state, CUjitInputType type, void *data, size_t size, const char *name, - unsigned int numOptions, CUjit_option *options, void **optionValues); -CUresult CUDAAPI cuLinkAddFile(CUlinkState state, CUjitInputType type, const char *path, - unsigned int numOptions, CUjit_option *options, void **optionValues); -#endif /* __CUDA_API_VERSION_INTERNAL || (__CUDA_API_VERSION >= 5050 && __CUDA_API_VERSION < 6050) */ - -#if defined(__CUDA_API_VERSION_INTERNAL) || (__CUDA_API_VERSION >= 3020 && __CUDA_API_VERSION < 4010) -CUresult CUDAAPI cuTexRefSetAddress2D_v2(CUtexref hTexRef, const CUDA_ARRAY_DESCRIPTOR *desc, CUdeviceptr dptr, size_t Pitch); -#endif /* __CUDA_API_VERSION_INTERNAL || (__CUDA_API_VERSION >= 3020 && __CUDA_API_VERSION < 4010) */ +#if defined(__CUDA_API_VERSION_INTERNAL) || \ + (__CUDA_API_VERSION >= 5050 && __CUDA_API_VERSION < 6050) +CUresult CUDAAPI cuLinkCreate(unsigned int numOptions, CUjit_option *options, + void **optionValues, CUlinkState *stateOut); +CUresult CUDAAPI cuLinkAddData(CUlinkState state, CUjitInputType type, + void *data, size_t size, const char *name, + unsigned int numOptions, CUjit_option *options, + void **optionValues); +CUresult CUDAAPI cuLinkAddFile(CUlinkState state, CUjitInputType type, + const char *path, unsigned int numOptions, + CUjit_option *options, void **optionValues); +#endif /* __CUDA_API_VERSION_INTERNAL || (__CUDA_API_VERSION >= 5050 && \ + __CUDA_API_VERSION < 6050) */ + +#if defined(__CUDA_API_VERSION_INTERNAL) || \ + (__CUDA_API_VERSION >= 3020 && __CUDA_API_VERSION < 4010) +CUresult CUDAAPI cuTexRefSetAddress2D_v2(CUtexref hTexRef, + const CUDA_ARRAY_DESCRIPTOR *desc, + CUdeviceptr dptr, size_t Pitch); +#endif /* __CUDA_API_VERSION_INTERNAL || (__CUDA_API_VERSION >= 3020 && \ + __CUDA_API_VERSION < 4010) */ /** * CUDA API made obselete at API version 3020 */ #if defined(__CUDA_API_VERSION_INTERNAL) - #define CUdeviceptr CUdeviceptr_v1 - #define CUDA_MEMCPY2D_st CUDA_MEMCPY2D_v1_st - #define CUDA_MEMCPY2D CUDA_MEMCPY2D_v1 - #define CUDA_MEMCPY3D_st CUDA_MEMCPY3D_v1_st - #define CUDA_MEMCPY3D CUDA_MEMCPY3D_v1 - #define CUDA_ARRAY_DESCRIPTOR_st CUDA_ARRAY_DESCRIPTOR_v1_st - #define CUDA_ARRAY_DESCRIPTOR CUDA_ARRAY_DESCRIPTOR_v1 - #define CUDA_ARRAY3D_DESCRIPTOR_st CUDA_ARRAY3D_DESCRIPTOR_v1_st - #define CUDA_ARRAY3D_DESCRIPTOR CUDA_ARRAY3D_DESCRIPTOR_v1 +#define CUdeviceptr CUdeviceptr_v1 +#define CUDA_MEMCPY2D_st CUDA_MEMCPY2D_v1_st +#define CUDA_MEMCPY2D CUDA_MEMCPY2D_v1 +#define CUDA_MEMCPY3D_st CUDA_MEMCPY3D_v1_st +#define CUDA_MEMCPY3D CUDA_MEMCPY3D_v1 +#define CUDA_ARRAY_DESCRIPTOR_st CUDA_ARRAY_DESCRIPTOR_v1_st +#define CUDA_ARRAY_DESCRIPTOR CUDA_ARRAY_DESCRIPTOR_v1 +#define CUDA_ARRAY3D_DESCRIPTOR_st CUDA_ARRAY3D_DESCRIPTOR_v1_st +#define CUDA_ARRAY3D_DESCRIPTOR CUDA_ARRAY3D_DESCRIPTOR_v1 #endif /* CUDA_FORCE_LEGACY32_INTERNAL */ #if defined(__CUDA_API_VERSION_INTERNAL) || __CUDA_API_VERSION < 3020 typedef unsigned int CUdeviceptr; -typedef struct CUDA_MEMCPY2D_st -{ - unsigned int srcXInBytes; /**< Source X in bytes */ - unsigned int srcY; /**< Source Y */ - CUmemorytype srcMemoryType; /**< Source memory type (host, device, array) */ - const void *srcHost; /**< Source host pointer */ - CUdeviceptr srcDevice; /**< Source device pointer */ - CUarray srcArray; /**< Source array reference */ - unsigned int srcPitch; /**< Source pitch (ignored when src is array) */ - - unsigned int dstXInBytes; /**< Destination X in bytes */ - unsigned int dstY; /**< Destination Y */ - CUmemorytype dstMemoryType; /**< Destination memory type (host, device, array) */ - void *dstHost; /**< Destination host pointer */ - CUdeviceptr dstDevice; /**< Destination device pointer */ - CUarray dstArray; /**< Destination array reference */ - unsigned int dstPitch; /**< Destination pitch (ignored when dst is array) */ - - unsigned int WidthInBytes; /**< Width of 2D memory copy in bytes */ - unsigned int Height; /**< Height of 2D memory copy */ +typedef struct CUDA_MEMCPY2D_st { + unsigned int srcXInBytes; /**< Source X in bytes */ + unsigned int srcY; /**< Source Y */ + CUmemorytype srcMemoryType; /**< Source memory type (host, device, array) */ + const void *srcHost; /**< Source host pointer */ + CUdeviceptr srcDevice; /**< Source device pointer */ + CUarray srcArray; /**< Source array reference */ + unsigned int srcPitch; /**< Source pitch (ignored when src is array) */ + + unsigned int dstXInBytes; /**< Destination X in bytes */ + unsigned int dstY; /**< Destination Y */ + CUmemorytype + dstMemoryType; /**< Destination memory type (host, device, array) */ + void *dstHost; /**< Destination host pointer */ + CUdeviceptr dstDevice; /**< Destination device pointer */ + CUarray dstArray; /**< Destination array reference */ + unsigned int dstPitch; /**< Destination pitch (ignored when dst is array) */ + + unsigned int WidthInBytes; /**< Width of 2D memory copy in bytes */ + unsigned int Height; /**< Height of 2D memory copy */ } CUDA_MEMCPY2D; -typedef struct CUDA_MEMCPY3D_st -{ - unsigned int srcXInBytes; /**< Source X in bytes */ - unsigned int srcY; /**< Source Y */ - unsigned int srcZ; /**< Source Z */ - unsigned int srcLOD; /**< Source LOD */ - CUmemorytype srcMemoryType; /**< Source memory type (host, device, array) */ - const void *srcHost; /**< Source host pointer */ - CUdeviceptr srcDevice; /**< Source device pointer */ - CUarray srcArray; /**< Source array reference */ - void *reserved0; /**< Must be NULL */ - unsigned int srcPitch; /**< Source pitch (ignored when src is array) */ - unsigned int srcHeight; /**< Source height (ignored when src is array; may be 0 if Depth==1) */ - - unsigned int dstXInBytes; /**< Destination X in bytes */ - unsigned int dstY; /**< Destination Y */ - unsigned int dstZ; /**< Destination Z */ - unsigned int dstLOD; /**< Destination LOD */ - CUmemorytype dstMemoryType; /**< Destination memory type (host, device, array) */ - void *dstHost; /**< Destination host pointer */ - CUdeviceptr dstDevice; /**< Destination device pointer */ - CUarray dstArray; /**< Destination array reference */ - void *reserved1; /**< Must be NULL */ - unsigned int dstPitch; /**< Destination pitch (ignored when dst is array) */ - unsigned int dstHeight; /**< Destination height (ignored when dst is array; may be 0 if Depth==1) */ - - unsigned int WidthInBytes; /**< Width of 3D memory copy in bytes */ - unsigned int Height; /**< Height of 3D memory copy */ - unsigned int Depth; /**< Depth of 3D memory copy */ +typedef struct CUDA_MEMCPY3D_st { + unsigned int srcXInBytes; /**< Source X in bytes */ + unsigned int srcY; /**< Source Y */ + unsigned int srcZ; /**< Source Z */ + unsigned int srcLOD; /**< Source LOD */ + CUmemorytype srcMemoryType; /**< Source memory type (host, device, array) */ + const void *srcHost; /**< Source host pointer */ + CUdeviceptr srcDevice; /**< Source device pointer */ + CUarray srcArray; /**< Source array reference */ + void *reserved0; /**< Must be NULL */ + unsigned int srcPitch; /**< Source pitch (ignored when src is array) */ + unsigned int srcHeight; /**< Source height (ignored when src is array; may be + 0 if Depth==1) */ + + unsigned int dstXInBytes; /**< Destination X in bytes */ + unsigned int dstY; /**< Destination Y */ + unsigned int dstZ; /**< Destination Z */ + unsigned int dstLOD; /**< Destination LOD */ + CUmemorytype + dstMemoryType; /**< Destination memory type (host, device, array) */ + void *dstHost; /**< Destination host pointer */ + CUdeviceptr dstDevice; /**< Destination device pointer */ + CUarray dstArray; /**< Destination array reference */ + void *reserved1; /**< Must be NULL */ + unsigned int dstPitch; /**< Destination pitch (ignored when dst is array) */ + unsigned int dstHeight; /**< Destination height (ignored when dst is array; + may be 0 if Depth==1) */ + + unsigned int WidthInBytes; /**< Width of 3D memory copy in bytes */ + unsigned int Height; /**< Height of 3D memory copy */ + unsigned int Depth; /**< Depth of 3D memory copy */ } CUDA_MEMCPY3D; -typedef struct CUDA_ARRAY_DESCRIPTOR_st -{ - unsigned int Width; /**< Width of array */ - unsigned int Height; /**< Height of array */ +typedef struct CUDA_ARRAY_DESCRIPTOR_st { + unsigned int Width; /**< Width of array */ + unsigned int Height; /**< Height of array */ - CUarray_format Format; /**< Array format */ - unsigned int NumChannels; /**< Channels per array element */ + CUarray_format Format; /**< Array format */ + unsigned int NumChannels; /**< Channels per array element */ } CUDA_ARRAY_DESCRIPTOR; -typedef struct CUDA_ARRAY3D_DESCRIPTOR_st -{ - unsigned int Width; /**< Width of 3D array */ - unsigned int Height; /**< Height of 3D array */ - unsigned int Depth; /**< Depth of 3D array */ +typedef struct CUDA_ARRAY3D_DESCRIPTOR_st { + unsigned int Width; /**< Width of 3D array */ + unsigned int Height; /**< Height of 3D array */ + unsigned int Depth; /**< Depth of 3D array */ - CUarray_format Format; /**< Array format */ - unsigned int NumChannels; /**< Channels per array element */ - unsigned int Flags; /**< Flags */ + CUarray_format Format; /**< Array format */ + unsigned int NumChannels; /**< Channels per array element */ + unsigned int Flags; /**< Flags */ } CUDA_ARRAY3D_DESCRIPTOR; CUresult CUDAAPI cuDeviceTotalMem(unsigned int *bytes, CUdevice dev); CUresult CUDAAPI cuCtxCreate(CUcontext *pctx, unsigned int flags, CUdevice dev); -CUresult CUDAAPI cuModuleGetGlobal(CUdeviceptr *dptr, unsigned int *bytes, CUmodule hmod, const char *name); +CUresult CUDAAPI cuModuleGetGlobal(CUdeviceptr *dptr, unsigned int *bytes, + CUmodule hmod, const char *name); CUresult CUDAAPI cuMemGetInfo(unsigned int *free, unsigned int *total); CUresult CUDAAPI cuMemAlloc(CUdeviceptr *dptr, unsigned int bytesize); -CUresult CUDAAPI cuMemAllocPitch(CUdeviceptr *dptr, unsigned int *pPitch, unsigned int WidthInBytes, unsigned int Height, unsigned int ElementSizeBytes); +CUresult CUDAAPI cuMemAllocPitch(CUdeviceptr *dptr, unsigned int *pPitch, + unsigned int WidthInBytes, unsigned int Height, + unsigned int ElementSizeBytes); CUresult CUDAAPI cuMemFree(CUdeviceptr dptr); -CUresult CUDAAPI cuMemGetAddressRange(CUdeviceptr *pbase, unsigned int *psize, CUdeviceptr dptr); +CUresult CUDAAPI cuMemGetAddressRange(CUdeviceptr *pbase, unsigned int *psize, + CUdeviceptr dptr); CUresult CUDAAPI cuMemAllocHost(void **pp, unsigned int bytesize); -CUresult CUDAAPI cuMemHostGetDevicePointer(CUdeviceptr *pdptr, void *p, unsigned int Flags); -CUresult CUDAAPI cuMemcpyHtoD(CUdeviceptr dstDevice, const void *srcHost, unsigned int ByteCount); -CUresult CUDAAPI cuMemcpyDtoH(void *dstHost, CUdeviceptr srcDevice, unsigned int ByteCount); -CUresult CUDAAPI cuMemcpyDtoD(CUdeviceptr dstDevice, CUdeviceptr srcDevice, unsigned int ByteCount); -CUresult CUDAAPI cuMemcpyDtoA(CUarray dstArray, unsigned int dstOffset, CUdeviceptr srcDevice, unsigned int ByteCount); -CUresult CUDAAPI cuMemcpyAtoD(CUdeviceptr dstDevice, CUarray srcArray, unsigned int srcOffset, unsigned int ByteCount); -CUresult CUDAAPI cuMemcpyHtoA(CUarray dstArray, unsigned int dstOffset, const void *srcHost, unsigned int ByteCount); -CUresult CUDAAPI cuMemcpyAtoH(void *dstHost, CUarray srcArray, unsigned int srcOffset, unsigned int ByteCount); -CUresult CUDAAPI cuMemcpyAtoA(CUarray dstArray, unsigned int dstOffset, CUarray srcArray, unsigned int srcOffset, unsigned int ByteCount); -CUresult CUDAAPI cuMemcpyHtoAAsync(CUarray dstArray, unsigned int dstOffset, const void *srcHost, unsigned int ByteCount, CUstream hStream); -CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, unsigned int srcOffset, unsigned int ByteCount, CUstream hStream); +CUresult CUDAAPI cuMemHostGetDevicePointer(CUdeviceptr *pdptr, void *p, + unsigned int Flags); +CUresult CUDAAPI cuMemcpyHtoD(CUdeviceptr dstDevice, const void *srcHost, + unsigned int ByteCount); +CUresult CUDAAPI cuMemcpyDtoH(void *dstHost, CUdeviceptr srcDevice, + unsigned int ByteCount); +CUresult CUDAAPI cuMemcpyDtoD(CUdeviceptr dstDevice, CUdeviceptr srcDevice, + unsigned int ByteCount); +CUresult CUDAAPI cuMemcpyDtoA(CUarray dstArray, unsigned int dstOffset, + CUdeviceptr srcDevice, unsigned int ByteCount); +CUresult CUDAAPI cuMemcpyAtoD(CUdeviceptr dstDevice, CUarray srcArray, + unsigned int srcOffset, unsigned int ByteCount); +CUresult CUDAAPI cuMemcpyHtoA(CUarray dstArray, unsigned int dstOffset, + const void *srcHost, unsigned int ByteCount); +CUresult CUDAAPI cuMemcpyAtoH(void *dstHost, CUarray srcArray, + unsigned int srcOffset, unsigned int ByteCount); +CUresult CUDAAPI cuMemcpyAtoA(CUarray dstArray, unsigned int dstOffset, + CUarray srcArray, unsigned int srcOffset, + unsigned int ByteCount); +CUresult CUDAAPI cuMemcpyHtoAAsync(CUarray dstArray, unsigned int dstOffset, + const void *srcHost, unsigned int ByteCount, + CUstream hStream); +CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, + unsigned int srcOffset, + unsigned int ByteCount, CUstream hStream); CUresult CUDAAPI cuMemcpy2D(const CUDA_MEMCPY2D *pCopy); CUresult CUDAAPI cuMemcpy2DUnaligned(const CUDA_MEMCPY2D *pCopy); CUresult CUDAAPI cuMemcpy3D(const CUDA_MEMCPY3D *pCopy); -CUresult CUDAAPI cuMemcpyHtoDAsync(CUdeviceptr dstDevice, const void *srcHost, unsigned int ByteCount, CUstream hStream); -CUresult CUDAAPI cuMemcpyDtoHAsync(void *dstHost, CUdeviceptr srcDevice, unsigned int ByteCount, CUstream hStream); -CUresult CUDAAPI cuMemcpyDtoDAsync(CUdeviceptr dstDevice, CUdeviceptr srcDevice, unsigned int ByteCount, CUstream hStream); +CUresult CUDAAPI cuMemcpyHtoDAsync(CUdeviceptr dstDevice, const void *srcHost, + unsigned int ByteCount, CUstream hStream); +CUresult CUDAAPI cuMemcpyDtoHAsync(void *dstHost, CUdeviceptr srcDevice, + unsigned int ByteCount, CUstream hStream); +CUresult CUDAAPI cuMemcpyDtoDAsync(CUdeviceptr dstDevice, CUdeviceptr srcDevice, + unsigned int ByteCount, CUstream hStream); CUresult CUDAAPI cuMemcpy2DAsync(const CUDA_MEMCPY2D *pCopy, CUstream hStream); CUresult CUDAAPI cuMemcpy3DAsync(const CUDA_MEMCPY3D *pCopy, CUstream hStream); -CUresult CUDAAPI cuMemsetD8(CUdeviceptr dstDevice, unsigned char uc, unsigned int N); -CUresult CUDAAPI cuMemsetD16(CUdeviceptr dstDevice, unsigned short us, unsigned int N); -CUresult CUDAAPI cuMemsetD32(CUdeviceptr dstDevice, unsigned int ui, unsigned int N); -CUresult CUDAAPI cuMemsetD2D8(CUdeviceptr dstDevice, unsigned int dstPitch, unsigned char uc, unsigned int Width, unsigned int Height); -CUresult CUDAAPI cuMemsetD2D16(CUdeviceptr dstDevice, unsigned int dstPitch, unsigned short us, unsigned int Width, unsigned int Height); -CUresult CUDAAPI cuMemsetD2D32(CUdeviceptr dstDevice, unsigned int dstPitch, unsigned int ui, unsigned int Width, unsigned int Height); -CUresult CUDAAPI cuArrayCreate(CUarray *pHandle, const CUDA_ARRAY_DESCRIPTOR *pAllocateArray); -CUresult CUDAAPI cuArrayGetDescriptor(CUDA_ARRAY_DESCRIPTOR *pArrayDescriptor, CUarray hArray); -CUresult CUDAAPI cuArray3DCreate(CUarray *pHandle, const CUDA_ARRAY3D_DESCRIPTOR *pAllocateArray); -CUresult CUDAAPI cuArray3DGetDescriptor(CUDA_ARRAY3D_DESCRIPTOR *pArrayDescriptor, CUarray hArray); -CUresult CUDAAPI cuTexRefSetAddress(unsigned int *ByteOffset, CUtexref hTexRef, CUdeviceptr dptr, unsigned int bytes); -CUresult CUDAAPI cuTexRefSetAddress2D(CUtexref hTexRef, const CUDA_ARRAY_DESCRIPTOR *desc, CUdeviceptr dptr, unsigned int Pitch); +CUresult CUDAAPI cuMemsetD8(CUdeviceptr dstDevice, unsigned char uc, + unsigned int N); +CUresult CUDAAPI cuMemsetD16(CUdeviceptr dstDevice, unsigned short us, + unsigned int N); +CUresult CUDAAPI cuMemsetD32(CUdeviceptr dstDevice, unsigned int ui, + unsigned int N); +CUresult CUDAAPI cuMemsetD2D8(CUdeviceptr dstDevice, unsigned int dstPitch, + unsigned char uc, unsigned int Width, + unsigned int Height); +CUresult CUDAAPI cuMemsetD2D16(CUdeviceptr dstDevice, unsigned int dstPitch, + unsigned short us, unsigned int Width, + unsigned int Height); +CUresult CUDAAPI cuMemsetD2D32(CUdeviceptr dstDevice, unsigned int dstPitch, + unsigned int ui, unsigned int Width, + unsigned int Height); +CUresult CUDAAPI cuArrayCreate(CUarray *pHandle, + const CUDA_ARRAY_DESCRIPTOR *pAllocateArray); +CUresult CUDAAPI cuArrayGetDescriptor(CUDA_ARRAY_DESCRIPTOR *pArrayDescriptor, + CUarray hArray); +CUresult CUDAAPI cuArray3DCreate(CUarray *pHandle, + const CUDA_ARRAY3D_DESCRIPTOR *pAllocateArray); +CUresult CUDAAPI cuArray3DGetDescriptor( + CUDA_ARRAY3D_DESCRIPTOR *pArrayDescriptor, CUarray hArray); +CUresult CUDAAPI cuTexRefSetAddress(unsigned int *ByteOffset, CUtexref hTexRef, + CUdeviceptr dptr, unsigned int bytes); +CUresult CUDAAPI cuTexRefSetAddress2D(CUtexref hTexRef, + const CUDA_ARRAY_DESCRIPTOR *desc, + CUdeviceptr dptr, unsigned int Pitch); CUresult CUDAAPI cuTexRefGetAddress(CUdeviceptr *pdptr, CUtexref hTexRef); -CUresult CUDAAPI cuGraphicsResourceGetMappedPointer(CUdeviceptr *pDevPtr, unsigned int *pSize, CUgraphicsResource resource); +CUresult CUDAAPI cuGraphicsResourceGetMappedPointer( + CUdeviceptr *pDevPtr, unsigned int *pSize, CUgraphicsResource resource); #endif /* __CUDA_API_VERSION_INTERNAL || __CUDA_API_VERSION < 3020 */ #if defined(__CUDA_API_VERSION_INTERNAL) || __CUDA_API_VERSION < 4000 CUresult CUDAAPI cuCtxDestroy(CUcontext ctx); @@ -14661,85 +15539,162 @@ CUresult CUDAAPI cuStreamDestroy(CUstream hStream); CUresult CUDAAPI cuEventDestroy(CUevent hEvent); #endif /* __CUDA_API_VERSION_INTERNAL || __CUDA_API_VERSION < 4000 */ #if defined(__CUDA_API_VERSION_INTERNAL) - #undef CUdeviceptr - #undef CUDA_MEMCPY2D_st - #undef CUDA_MEMCPY2D - #undef CUDA_MEMCPY3D_st - #undef CUDA_MEMCPY3D - #undef CUDA_ARRAY_DESCRIPTOR_st - #undef CUDA_ARRAY_DESCRIPTOR - #undef CUDA_ARRAY3D_DESCRIPTOR_st - #undef CUDA_ARRAY3D_DESCRIPTOR +#undef CUdeviceptr +#undef CUDA_MEMCPY2D_st +#undef CUDA_MEMCPY2D +#undef CUDA_MEMCPY3D_st +#undef CUDA_MEMCPY3D +#undef CUDA_ARRAY_DESCRIPTOR_st +#undef CUDA_ARRAY_DESCRIPTOR +#undef CUDA_ARRAY3D_DESCRIPTOR_st +#undef CUDA_ARRAY3D_DESCRIPTOR #endif /* __CUDA_API_VERSION_INTERNAL */ #if defined(__CUDA_API_VERSION_INTERNAL) - CUresult CUDAAPI cuMemcpyHtoD_v2(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount); - CUresult CUDAAPI cuMemcpyDtoH_v2(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount); - CUresult CUDAAPI cuMemcpyDtoD_v2(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount); - CUresult CUDAAPI cuMemcpyDtoA_v2(CUarray dstArray, size_t dstOffset, CUdeviceptr srcDevice, size_t ByteCount); - CUresult CUDAAPI cuMemcpyAtoD_v2(CUdeviceptr dstDevice, CUarray srcArray, size_t srcOffset, size_t ByteCount); - CUresult CUDAAPI cuMemcpyHtoA_v2(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount); - CUresult CUDAAPI cuMemcpyAtoH_v2(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount); - CUresult CUDAAPI cuMemcpyAtoA_v2(CUarray dstArray, size_t dstOffset, CUarray srcArray, size_t srcOffset, size_t ByteCount); - CUresult CUDAAPI cuMemcpyHtoAAsync_v2(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount, CUstream hStream); - CUresult CUDAAPI cuMemcpyAtoHAsync_v2(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount, CUstream hStream); - CUresult CUDAAPI cuMemcpy2D_v2(const CUDA_MEMCPY2D *pCopy); - CUresult CUDAAPI cuMemcpy2DUnaligned_v2(const CUDA_MEMCPY2D *pCopy); - CUresult CUDAAPI cuMemcpy3D_v2(const CUDA_MEMCPY3D *pCopy); - CUresult CUDAAPI cuMemcpyHtoDAsync_v2(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount, CUstream hStream); - CUresult CUDAAPI cuMemcpyDtoHAsync_v2(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream); - CUresult CUDAAPI cuMemcpyDtoDAsync_v2(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream); - CUresult CUDAAPI cuMemcpy2DAsync_v2(const CUDA_MEMCPY2D *pCopy, CUstream hStream); - CUresult CUDAAPI cuMemcpy3DAsync_v2(const CUDA_MEMCPY3D *pCopy, CUstream hStream); - CUresult CUDAAPI cuMemsetD8_v2(CUdeviceptr dstDevice, unsigned char uc, size_t N); - CUresult CUDAAPI cuMemsetD16_v2(CUdeviceptr dstDevice, unsigned short us, size_t N); - CUresult CUDAAPI cuMemsetD32_v2(CUdeviceptr dstDevice, unsigned int ui, size_t N); - CUresult CUDAAPI cuMemsetD2D8_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height); - CUresult CUDAAPI cuMemsetD2D16_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height); - CUresult CUDAAPI cuMemsetD2D32_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height); - CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount); - CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount, CUstream hStream); - CUresult CUDAAPI cuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount); - CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount, CUstream hStream); - CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy); - CUresult CUDAAPI cuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER *pCopy, CUstream hStream); - - CUresult CUDAAPI cuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, size_t N, CUstream hStream); - CUresult CUDAAPI cuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, size_t N, CUstream hStream); - CUresult CUDAAPI cuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, size_t N, CUstream hStream); - CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height, CUstream hStream); - CUresult CUDAAPI cuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height, CUstream hStream); - CUresult CUDAAPI cuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height, CUstream hStream); - - CUresult CUDAAPI cuStreamGetPriority(CUstream hStream, int *priority); - CUresult CUDAAPI cuStreamGetFlags(CUstream hStream, unsigned int *flags); - CUresult CUDAAPI cuStreamGetCtx(CUstream hStream, CUcontext *pctx); - CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, unsigned int Flags); - CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, CUstreamCallback callback, void *userData, unsigned int flags); - CUresult CUDAAPI cuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, size_t length, unsigned int flags); - CUresult CUDAAPI cuStreamQuery(CUstream hStream); - CUresult CUDAAPI cuStreamSynchronize(CUstream hStream); - CUresult CUDAAPI cuEventRecord(CUevent hEvent, CUstream hStream); - CUresult CUDAAPI cuLaunchKernel(CUfunction f, unsigned int gridDimX, unsigned int gridDimY, unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, void **kernelParams, void **extra); - CUresult CUDAAPI cuLaunchHostFunc(CUstream hStream, CUhostFn fn, void *userData); - CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, CUgraphicsResource *resources, CUstream hStream); - CUresult CUDAAPI cuGraphicsUnmapResources(unsigned int count, CUgraphicsResource *resources, CUstream hStream); - CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags); - CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags); - CUresult CUDAAPI cuStreamWriteValue64(CUstream stream, CUdeviceptr addr, cuuint64_t value, unsigned int flags); - CUresult CUDAAPI cuStreamWaitValue64(CUstream stream, CUdeviceptr addr, cuuint64_t value, unsigned int flags); - CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, CUstreamBatchMemOpParams *paramArray, unsigned int flags); - CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice dstDevice, CUstream hStream); - CUresult CUDAAPI cuLaunchCooperativeKernel(CUfunction f, unsigned int gridDimX, unsigned int gridDimY, unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, void **kernelParams); - CUresult CUDAAPI cuSignalExternalSemaphoresAsync(const CUexternalSemaphore *extSemArray, const CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS *paramsArray, unsigned int numExtSems, CUstream stream); - CUresult CUDAAPI cuWaitExternalSemaphoresAsync(const CUexternalSemaphore *extSemArray, const CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS *paramsArray, unsigned int numExtSems, CUstream stream); - CUresult CUDAAPI cuStreamBeginCapture(CUstream hStream); - CUresult CUDAAPI cuStreamBeginCapture_ptsz(CUstream hStream); - CUresult CUDAAPI cuStreamBeginCapture_v2(CUstream hStream, CUstreamCaptureMode mode); - CUresult CUDAAPI cuStreamEndCapture(CUstream hStream, CUgraph *phGraph); - CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, CUstreamCaptureStatus *captureStatus); - CUresult CUDAAPI cuStreamGetCaptureInfo(CUstream hStream, CUstreamCaptureStatus *captureStatus, cuuint64_t *id); - CUresult CUDAAPI cuGraphLaunch(CUgraphExec hGraph, CUstream hStream); +CUresult CUDAAPI cuMemcpyHtoD_v2(CUdeviceptr dstDevice, const void *srcHost, + size_t ByteCount); +CUresult CUDAAPI cuMemcpyDtoH_v2(void *dstHost, CUdeviceptr srcDevice, + size_t ByteCount); +CUresult CUDAAPI cuMemcpyDtoD_v2(CUdeviceptr dstDevice, CUdeviceptr srcDevice, + size_t ByteCount); +CUresult CUDAAPI cuMemcpyDtoA_v2(CUarray dstArray, size_t dstOffset, + CUdeviceptr srcDevice, size_t ByteCount); +CUresult CUDAAPI cuMemcpyAtoD_v2(CUdeviceptr dstDevice, CUarray srcArray, + size_t srcOffset, size_t ByteCount); +CUresult CUDAAPI cuMemcpyHtoA_v2(CUarray dstArray, size_t dstOffset, + const void *srcHost, size_t ByteCount); +CUresult CUDAAPI cuMemcpyAtoH_v2(void *dstHost, CUarray srcArray, + size_t srcOffset, size_t ByteCount); +CUresult CUDAAPI cuMemcpyAtoA_v2(CUarray dstArray, size_t dstOffset, + CUarray srcArray, size_t srcOffset, + size_t ByteCount); +CUresult CUDAAPI cuMemcpyHtoAAsync_v2(CUarray dstArray, size_t dstOffset, + const void *srcHost, size_t ByteCount, + CUstream hStream); +CUresult CUDAAPI cuMemcpyAtoHAsync_v2(void *dstHost, CUarray srcArray, + size_t srcOffset, size_t ByteCount, + CUstream hStream); +CUresult CUDAAPI cuMemcpy2D_v2(const CUDA_MEMCPY2D *pCopy); +CUresult CUDAAPI cuMemcpy2DUnaligned_v2(const CUDA_MEMCPY2D *pCopy); +CUresult CUDAAPI cuMemcpy3D_v2(const CUDA_MEMCPY3D *pCopy); +CUresult CUDAAPI cuMemcpyHtoDAsync_v2(CUdeviceptr dstDevice, + const void *srcHost, size_t ByteCount, + CUstream hStream); +CUresult CUDAAPI cuMemcpyDtoHAsync_v2(void *dstHost, CUdeviceptr srcDevice, + size_t ByteCount, CUstream hStream); +CUresult CUDAAPI cuMemcpyDtoDAsync_v2(CUdeviceptr dstDevice, + CUdeviceptr srcDevice, size_t ByteCount, + CUstream hStream); +CUresult CUDAAPI cuMemcpy2DAsync_v2(const CUDA_MEMCPY2D *pCopy, + CUstream hStream); +CUresult CUDAAPI cuMemcpy3DAsync_v2(const CUDA_MEMCPY3D *pCopy, + CUstream hStream); +CUresult CUDAAPI cuMemsetD8_v2(CUdeviceptr dstDevice, unsigned char uc, + size_t N); +CUresult CUDAAPI cuMemsetD16_v2(CUdeviceptr dstDevice, unsigned short us, + size_t N); +CUresult CUDAAPI cuMemsetD32_v2(CUdeviceptr dstDevice, unsigned int ui, + size_t N); +CUresult CUDAAPI cuMemsetD2D8_v2(CUdeviceptr dstDevice, size_t dstPitch, + unsigned char uc, size_t Width, size_t Height); +CUresult CUDAAPI cuMemsetD2D16_v2(CUdeviceptr dstDevice, size_t dstPitch, + unsigned short us, size_t Width, + size_t Height); +CUresult CUDAAPI cuMemsetD2D32_v2(CUdeviceptr dstDevice, size_t dstPitch, + unsigned int ui, size_t Width, size_t Height); +CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount); +CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, + size_t ByteCount, CUstream hStream); +CUresult CUDAAPI cuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, + CUdeviceptr srcDevice, CUcontext srcContext, + size_t ByteCount); +CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, + CUdeviceptr srcDevice, CUcontext srcContext, + size_t ByteCount, CUstream hStream); +CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy); +CUresult CUDAAPI cuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER *pCopy, + CUstream hStream); + +CUresult CUDAAPI cuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, + size_t N, CUstream hStream); +CUresult CUDAAPI cuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, + size_t N, CUstream hStream); +CUresult CUDAAPI cuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, + size_t N, CUstream hStream); +CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned char uc, size_t Width, + size_t Height, CUstream hStream); +CUresult CUDAAPI cuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned short us, size_t Width, + size_t Height, CUstream hStream); +CUresult CUDAAPI cuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned int ui, size_t Width, + size_t Height, CUstream hStream); + +CUresult CUDAAPI cuStreamGetPriority(CUstream hStream, int *priority); +CUresult CUDAAPI cuStreamGetFlags(CUstream hStream, unsigned int *flags); +CUresult CUDAAPI cuStreamGetCtx(CUstream hStream, CUcontext *pctx); +CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, + unsigned int Flags); +CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, + CUstreamCallback callback, void *userData, + unsigned int flags); +CUresult CUDAAPI cuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, + size_t length, unsigned int flags); +CUresult CUDAAPI cuStreamQuery(CUstream hStream); +CUresult CUDAAPI cuStreamSynchronize(CUstream hStream); +CUresult CUDAAPI cuEventRecord(CUevent hEvent, CUstream hStream); +CUresult CUDAAPI cuLaunchKernel(CUfunction f, unsigned int gridDimX, + unsigned int gridDimY, unsigned int gridDimZ, + unsigned int blockDimX, unsigned int blockDimY, + unsigned int blockDimZ, + unsigned int sharedMemBytes, CUstream hStream, + void **kernelParams, void **extra); +CUresult CUDAAPI cuLaunchHostFunc(CUstream hStream, CUhostFn fn, + void *userData); +CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, + CUgraphicsResource *resources, + CUstream hStream); +CUresult CUDAAPI cuGraphicsUnmapResources(unsigned int count, + CUgraphicsResource *resources, + CUstream hStream); +CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, + cuuint32_t value, unsigned int flags); +CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, + cuuint32_t value, unsigned int flags); +CUresult CUDAAPI cuStreamWriteValue64(CUstream stream, CUdeviceptr addr, + cuuint64_t value, unsigned int flags); +CUresult CUDAAPI cuStreamWaitValue64(CUstream stream, CUdeviceptr addr, + cuuint64_t value, unsigned int flags); +CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, + CUstreamBatchMemOpParams *paramArray, + unsigned int flags); +CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, + CUdevice dstDevice, CUstream hStream); +CUresult CUDAAPI cuLaunchCooperativeKernel( + CUfunction f, unsigned int gridDimX, unsigned int gridDimY, + unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, + unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, + void **kernelParams); +CUresult CUDAAPI cuSignalExternalSemaphoresAsync( + const CUexternalSemaphore *extSemArray, + const CUDA_EXTERNAL_SEMAPHORE_SIGNAL_PARAMS *paramsArray, + unsigned int numExtSems, CUstream stream); +CUresult CUDAAPI cuWaitExternalSemaphoresAsync( + const CUexternalSemaphore *extSemArray, + const CUDA_EXTERNAL_SEMAPHORE_WAIT_PARAMS *paramsArray, + unsigned int numExtSems, CUstream stream); +CUresult CUDAAPI cuStreamBeginCapture(CUstream hStream); +CUresult CUDAAPI cuStreamBeginCapture_ptsz(CUstream hStream); +CUresult CUDAAPI cuStreamBeginCapture_v2(CUstream hStream, + CUstreamCaptureMode mode); +CUresult CUDAAPI cuStreamEndCapture(CUstream hStream, CUgraph *phGraph); +CUresult CUDAAPI cuStreamIsCapturing(CUstream hStream, + CUstreamCaptureStatus *captureStatus); +CUresult CUDAAPI cuStreamGetCaptureInfo(CUstream hStream, + CUstreamCaptureStatus *captureStatus, + cuuint64_t *id); +CUresult CUDAAPI cuGraphLaunch(CUgraphExec hGraph, CUstream hStream); #endif #ifdef __cplusplus diff --git a/libcuda/cuda_api_object.h b/libcuda/cuda_api_object.h index 51416f2..d292e22 100644 --- a/libcuda/cuda_api_object.h +++ b/libcuda/cuda_api_object.h @@ -8,9 +8,9 @@ #include "builtin_types.h" -#include "../src/gpgpu-sim/gpu-sim.h" -#include "../src/cuda-sim/ptx_ir.h" #include "../src/abstract_hardware_model.h" +#include "../src/cuda-sim/ptx_ir.h" +#include "../src/gpgpu-sim/gpu-sim.h" #include "cuobjdump.h" typedef std::list gpgpu_ptx_sim_arg_list_t; @@ -20,194 +20,198 @@ typedef unsigned long GLuint; #endif struct glbmap_entry { - GLuint m_bufferObj; - void *m_devPtr; - size_t m_size; - struct glbmap_entry *m_next; + GLuint m_bufferObj; + void *m_devPtr; + size_t m_size; + struct glbmap_entry *m_next; }; typedef struct glbmap_entry glbmap_entry_t; struct _cuda_device_id { - _cuda_device_id(gpgpu_sim* gpu) {m_id = 0; m_next = NULL; m_gpgpu=gpu;} - struct _cuda_device_id *next() { return m_next; } - unsigned num_shader() const { return m_gpgpu->get_config().num_shader(); } - int num_devices() const { - if( m_next == NULL ) return 1; - else return 1 + m_next->num_devices(); - } - struct _cuda_device_id *get_device( unsigned n ) - { - assert( n < (unsigned)num_devices() ); - struct _cuda_device_id *p=this; - for(unsigned i=0; im_next; - return p; - } - const struct cudaDeviceProp *get_prop() const - { - return m_gpgpu->get_prop(); - } - unsigned get_id() const { return m_id; } - - gpgpu_sim *get_gpgpu() { return m_gpgpu; } -private: - unsigned m_id; - class gpgpu_sim *m_gpgpu; - struct _cuda_device_id *m_next; + _cuda_device_id(gpgpu_sim *gpu) { + m_id = 0; + m_next = NULL; + m_gpgpu = gpu; + } + struct _cuda_device_id *next() { + return m_next; + } + unsigned num_shader() const { return m_gpgpu->get_config().num_shader(); } + int num_devices() const { + if (m_next == NULL) + return 1; + else + return 1 + m_next->num_devices(); + } + struct _cuda_device_id *get_device(unsigned n) { + assert(n < (unsigned)num_devices()); + struct _cuda_device_id *p = this; + for (unsigned i = 0; i < n; i++) p = p->m_next; + return p; + } + const struct cudaDeviceProp *get_prop() const { return m_gpgpu->get_prop(); } + unsigned get_id() const { return m_id; } + + gpgpu_sim *get_gpgpu() { return m_gpgpu; } + + private: + unsigned m_id; + class gpgpu_sim *m_gpgpu; + struct _cuda_device_id *m_next; }; struct CUctx_st { - CUctx_st( _cuda_device_id *gpu ) - { - m_gpu = gpu; - m_binary_info.cmem = 0; - m_binary_info.gmem = 0; - no_of_ptx=0; - } - - _cuda_device_id *get_device() { return m_gpu; } - - void add_binary( symbol_table *symtab, unsigned fat_cubin_handle ) - { - m_code[fat_cubin_handle] = symtab; - m_last_fat_cubin_handle = fat_cubin_handle; - } - - void add_ptxinfo( const char *deviceFun, const struct gpgpu_ptx_sim_info &info ) - { - symbol *s = m_code[m_last_fat_cubin_handle]->lookup(deviceFun); - assert( s != NULL ); - function_info *f = s->get_pc(); - assert( f != NULL ); - f->set_kernel_info(info); - } - - void add_ptxinfo( const struct gpgpu_ptx_sim_info &info ) - { - m_binary_info = info; - } - - void register_function( unsigned fat_cubin_handle, const char *hostFun, const char *deviceFun ) - { - if( m_code.find(fat_cubin_handle) != m_code.end() ) { - symbol *s = m_code[fat_cubin_handle]->lookup(deviceFun); - if(s != NULL) { - function_info *f = s->get_pc(); - assert( f != NULL ); - m_kernel_lookup[hostFun] = f; - } - else { - printf("Warning: cannot find deviceFun %s\n", deviceFun); - m_kernel_lookup[hostFun] = NULL; - } - // assert( s != NULL ); - // function_info *f = s->get_pc(); - // assert( f != NULL ); - // m_kernel_lookup[hostFun] = f; - } else { - m_kernel_lookup[hostFun] = NULL; - } - } - - void register_hostFun_function( const char*hostFun, function_info* f){ + CUctx_st(_cuda_device_id *gpu) { + m_gpu = gpu; + m_binary_info.cmem = 0; + m_binary_info.gmem = 0; + no_of_ptx = 0; + } + + _cuda_device_id *get_device() { return m_gpu; } + + void add_binary(symbol_table *symtab, unsigned fat_cubin_handle) { + m_code[fat_cubin_handle] = symtab; + m_last_fat_cubin_handle = fat_cubin_handle; + } + + void add_ptxinfo(const char *deviceFun, + const struct gpgpu_ptx_sim_info &info) { + symbol *s = m_code[m_last_fat_cubin_handle]->lookup(deviceFun); + assert(s != NULL); + function_info *f = s->get_pc(); + assert(f != NULL); + f->set_kernel_info(info); + } + + void add_ptxinfo(const struct gpgpu_ptx_sim_info &info) { + m_binary_info = info; + } + + void register_function(unsigned fat_cubin_handle, const char *hostFun, + const char *deviceFun) { + if (m_code.find(fat_cubin_handle) != m_code.end()) { + symbol *s = m_code[fat_cubin_handle]->lookup(deviceFun); + if (s != NULL) { + function_info *f = s->get_pc(); + assert(f != NULL); m_kernel_lookup[hostFun] = f; + } else { + printf("Warning: cannot find deviceFun %s\n", deviceFun); + m_kernel_lookup[hostFun] = NULL; + } + // assert( s != NULL ); + // function_info *f = s->get_pc(); + // assert( f != NULL ); + // m_kernel_lookup[hostFun] = f; + } else { + m_kernel_lookup[hostFun] = NULL; } - - function_info *get_kernel(const char *hostFun) - { - std::map::iterator i=m_kernel_lookup.find(hostFun); - assert( i != m_kernel_lookup.end() ); - return i->second; - } - - int no_of_ptx; - -private: - _cuda_device_id *m_gpu; // selected gpu - std::map m_code; // fat binary handle => global symbol table - unsigned m_last_fat_cubin_handle; - std::map m_kernel_lookup; // unique id (CUDA app function address) => kernel entry point - struct gpgpu_ptx_sim_info m_binary_info; - + } + + void register_hostFun_function(const char *hostFun, function_info *f) { + m_kernel_lookup[hostFun] = f; + } + + function_info *get_kernel(const char *hostFun) { + std::map::iterator i = + m_kernel_lookup.find(hostFun); + assert(i != m_kernel_lookup.end()); + return i->second; + } + + int no_of_ptx; + + private: + _cuda_device_id *m_gpu; // selected gpu + std::map + m_code; // fat binary handle => global symbol table + unsigned m_last_fat_cubin_handle; + std::map + m_kernel_lookup; // unique id (CUDA app function address) => kernel entry + // point + struct gpgpu_ptx_sim_info m_binary_info; }; class kernel_config { -public: - kernel_config( dim3 GridDim, dim3 BlockDim, size_t sharedMem, struct CUstream_st *stream ) - { - m_GridDim=GridDim; - m_BlockDim=BlockDim; - m_sharedMem=sharedMem; - m_stream = stream; - } - kernel_config() - { - m_GridDim=dim3(-1,-1,-1); - m_BlockDim=dim3(-1,-1,-1); - m_sharedMem=0; - m_stream =NULL; - } - void set_arg( const void *arg, size_t size, size_t offset ) - { - m_args.push_front( gpgpu_ptx_sim_arg(arg,size,offset) ); - } - dim3 grid_dim() const { return m_GridDim; } - dim3 block_dim() const { return m_BlockDim; } - void set_grid_dim(dim3 *d) { m_GridDim = *d; } - void set_block_dim(dim3 *d) { m_BlockDim = *d; } - gpgpu_ptx_sim_arg_list_t get_args() { return m_args; } - struct CUstream_st *get_stream() { return m_stream; } - -private: - dim3 m_GridDim; - dim3 m_BlockDim; - size_t m_sharedMem; - struct CUstream_st *m_stream; - gpgpu_ptx_sim_arg_list_t m_args; + public: + kernel_config(dim3 GridDim, dim3 BlockDim, size_t sharedMem, + struct CUstream_st *stream) { + m_GridDim = GridDim; + m_BlockDim = BlockDim; + m_sharedMem = sharedMem; + m_stream = stream; + } + kernel_config() { + m_GridDim = dim3(-1, -1, -1); + m_BlockDim = dim3(-1, -1, -1); + m_sharedMem = 0; + m_stream = NULL; + } + void set_arg(const void *arg, size_t size, size_t offset) { + m_args.push_front(gpgpu_ptx_sim_arg(arg, size, offset)); + } + dim3 grid_dim() const { return m_GridDim; } + dim3 block_dim() const { return m_BlockDim; } + void set_grid_dim(dim3 *d) { m_GridDim = *d; } + void set_block_dim(dim3 *d) { m_BlockDim = *d; } + gpgpu_ptx_sim_arg_list_t get_args() { return m_args; } + struct CUstream_st *get_stream() { + return m_stream; + } + + private: + dim3 m_GridDim; + dim3 m_BlockDim; + size_t m_sharedMem; + struct CUstream_st *m_stream; + gpgpu_ptx_sim_arg_list_t m_args; }; class cuda_runtime_api { - public: - cuda_runtime_api( gpgpu_context* ctx ) { - g_glbmap = NULL; - g_active_device = 0; //active gpu that runs the code - gpgpu_ctx = ctx; - } - // global list - std::list cuobjdumpSectionList; - std::list libSectionList; - std::list g_cuda_launch_stack; - std::mapfatbin_registered; - std::map fatbinmap; - std::map name_symtab; - std::map g_mallocPtr_Size; - //maps sm version number to set of filenames - std::map > version_filename; - std::map pinned_memory; //support for pinned memories added - std::map pinned_memory_size; - glbmap_entry_t* g_glbmap; - int g_active_device; //active gpu that runs the code - // backward pointer - class gpgpu_context* gpgpu_ctx; - // member function list - void cuobjdumpInit(); - void extract_code_using_cuobjdump(); - void extract_ptx_files_using_cuobjdump(CUctx_st *context); - std::list pruneSectionList(CUctx_st *context); - std::list mergeMatchingSections(std::string identifier); - std::list mergeSections(); - cuobjdumpELFSection* findELFSection(const std::string identifier); - cuobjdumpPTXSection* findPTXSection(const std::string identifier); - cuobjdumpPTXSection* findPTXSectionInList(std::list §ionlist, const std::string identifier); - void cuobjdumpRegisterFatBinary(unsigned int handle, const char* filename, CUctx_st *context); - kernel_info_t *gpgpu_cuda_ptx_sim_init_grid( const char *kernel_key, - gpgpu_ptx_sim_arg_list_t args, - struct dim3 gridDim, - struct dim3 blockDim, - struct CUctx_st* context ); - int load_static_globals( symbol_table *symtab, unsigned min_gaddr, unsigned max_gaddr, gpgpu_t *gpu ); - int load_constants( symbol_table *symtab, addr_t min_gaddr, gpgpu_t *gpu ); - + public: + cuda_runtime_api(gpgpu_context *ctx) { + g_glbmap = NULL; + g_active_device = 0; // active gpu that runs the code + gpgpu_ctx = ctx; + } + // global list + std::list cuobjdumpSectionList; + std::list libSectionList; + std::list g_cuda_launch_stack; + std::map fatbin_registered; + std::map fatbinmap; + std::map name_symtab; + std::map g_mallocPtr_Size; + // maps sm version number to set of filenames + std::map > version_filename; + std::map pinned_memory; // support for pinned memories added + std::map pinned_memory_size; + glbmap_entry_t *g_glbmap; + int g_active_device; // active gpu that runs the code + // backward pointer + class gpgpu_context *gpgpu_ctx; + // member function list + void cuobjdumpInit(); + void extract_code_using_cuobjdump(); + void extract_ptx_files_using_cuobjdump(CUctx_st *context); + std::list pruneSectionList(CUctx_st *context); + std::list mergeMatchingSections(std::string identifier); + std::list mergeSections(); + cuobjdumpELFSection *findELFSection(const std::string identifier); + cuobjdumpPTXSection *findPTXSection(const std::string identifier); + cuobjdumpPTXSection *findPTXSectionInList( + std::list §ionlist, const std::string identifier); + void cuobjdumpRegisterFatBinary(unsigned int handle, const char *filename, + CUctx_st *context); + kernel_info_t *gpgpu_cuda_ptx_sim_init_grid(const char *kernel_key, + gpgpu_ptx_sim_arg_list_t args, + struct dim3 gridDim, + struct dim3 blockDim, + struct CUctx_st *context); + int load_static_globals(symbol_table *symtab, unsigned min_gaddr, + unsigned max_gaddr, gpgpu_t *gpu); + int load_constants(symbol_table *symtab, addr_t min_gaddr, gpgpu_t *gpu); }; #endif /* __cuda_api_object_h__ */ diff --git a/libcuda/cuda_runtime_api.cc b/libcuda/cuda_runtime_api.cc index 273194e..cc01f12 100644 --- a/libcuda/cuda_runtime_api.cc +++ b/libcuda/cuda_runtime_api.cc @@ -2,16 +2,16 @@ // Changes Copyright 2009, Tor M. Aamodt, Ali Bakhoda and George L. Yuan // University of British Columbia -/* +/* * cuda_runtime_api.cc * - * Copyright © 2009 by Tor M. Aamodt, Wilson W. L. Fung, Ali Bakhoda, - * George L. Yuan and the University of British Columbia, Vancouver, + * Copyright © 2009 by Tor M. Aamodt, Wilson W. L. Fung, Ali Bakhoda, + * George L. Yuan and the University of British Columbia, Vancouver, * BC V6T 1Z4, All Rights Reserved. - * + * * THIS IS A LEGAL DOCUMENT BY DOWNLOADING GPGPU-SIM, YOU ARE AGREEING TO THESE * TERMS AND CONDITIONS. - * + * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE @@ -23,99 +23,100 @@ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. - * + * * NOTE: The files libcuda/cuda_runtime_api.c and src/cuda-sim/cuda-math.h * are derived from the CUDA Toolset available from http://www.nvidia.com/cuda - * (property of NVIDIA). The files benchmarks/BlackScholes/ and - * benchmarks/template/ are derived from the CUDA SDK available from - * http://www.nvidia.com/cuda (also property of NVIDIA). The files from - * src/intersim/ are derived from Booksim (a simulator provided with the - * textbook "Principles and Practices of Interconnection Networks" available - * from http://cva.stanford.edu/books/ppin/). As such, those files are bound by - * the corresponding legal terms and conditions set forth separately (original - * copyright notices are left in files from these sources and where we have - * modified a file our copyright notice appears before the original copyright - * notice). - * - * Using this version of GPGPU-Sim requires a complete installation of CUDA - * which is distributed seperately by NVIDIA under separate terms and + * (property of NVIDIA). The files benchmarks/BlackScholes/ and + * benchmarks/template/ are derived from the CUDA SDK available from + * http://www.nvidia.com/cuda (also property of NVIDIA). The files from + * src/intersim/ are derived from Booksim (a simulator provided with the + * textbook "Principles and Practices of Interconnection Networks" available + * from http://cva.stanford.edu/books/ppin/). As such, those files are bound by + * the corresponding legal terms and conditions set forth separately (original + * copyright notices are left in files from these sources and where we have + * modified a file our copyright notice appears before the original copyright + * notice). + * + * Using this version of GPGPU-Sim requires a complete installation of CUDA + * which is distributed seperately by NVIDIA under separate terms and * conditions. To use this version of GPGPU-Sim with OpenCL requires a * recent version of NVIDIA's drivers which support OpenCL. - * + * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are met: - * + * * 1. Redistributions of source code must retain the above copyright notice, * this list of conditions and the following disclaimer. - * + * * 2. Redistributions in binary form must reproduce the above copyright notice, * this list of conditions and the following disclaimer in the documentation * and/or other materials provided with the distribution. - * + * * 3. Neither the name of the University of British Columbia nor the names of * its contributors may be used to endorse or promote products derived from * this software without specific prior written permission. - * - * 4. This version of GPGPU-SIM is distributed freely for non-commercial use only. - * + * + * 4. This version of GPGPU-SIM is distributed freely for non-commercial use + * only. + * * 5. No nonprofit user may place any restrictions on the use of this software, * including as modified by the user, by any other authorized user. - * - * 6. GPGPU-SIM was developed primarily by Tor M. Aamodt, Wilson W. L. Fung, - * Ali Bakhoda, George L. Yuan, at the University of British Columbia, + * + * 6. GPGPU-SIM was developed primarily by Tor M. Aamodt, Wilson W. L. Fung, + * Ali Bakhoda, George L. Yuan, at the University of British Columbia, * Vancouver, BC V6T 1Z4 */ /* * Copyright 1993-2007 NVIDIA Corporation. All rights reserved. * - * NOTICE TO USER: + * NOTICE TO USER: * - * This source code is subject to NVIDIA ownership rights under U.S. and - * international Copyright laws. Users and possessors of this source code - * are hereby granted a nonexclusive, royalty-free license to use this code + * This source code is subject to NVIDIA ownership rights under U.S. and + * international Copyright laws. Users and possessors of this source code + * are hereby granted a nonexclusive, royalty-free license to use this code * in individual and commercial software. * - * NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE - * CODE FOR ANY PURPOSE. IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR - * IMPLIED WARRANTY OF ANY KIND. NVIDIA DISCLAIMS ALL WARRANTIES WITH - * REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF + * NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOURCE + * CODE FOR ANY PURPOSE. IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR + * IMPLIED WARRANTY OF ANY KIND. NVIDIA DISCLAIMS ALL WARRANTIES WITH + * REGARD TO THIS SOURCE CODE, INCLUDING ALL IMPLIED WARRANTIES OF * MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. - * IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, - * OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS - * OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE - * OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE - * OR PERFORMANCE OF THIS SOURCE CODE. + * IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, + * OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS + * OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE + * OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE + * OR PERFORMANCE OF THIS SOURCE CODE. * - * U.S. Government End Users. This source code is a "commercial item" as - * that term is defined at 48 C.F.R. 2.101 (OCT 1995), consisting of - * "commercial computer software" and "commercial computer software - * documentation" as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) - * and is provided to the U.S. Government only as a commercial end item. - * Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through - * 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the - * source code with only those rights set forth herein. + * U.S. Government End Users. This source code is a "commercial item" as + * that term is defined at 48 C.F.R. 2.101 (OCT 1995), consisting of + * "commercial computer software" and "commercial computer software + * documentation" as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) + * and is provided to the U.S. Government only as a commercial end item. + * Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through + * 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the + * source code with only those rights set forth herein. * - * Any use of this source code in individual and commercial software must + * Any use of this source code in individual and commercial software must * include, in the user documentation and internal comments to the code, * the above Disclaimer and U.S. Government End Users Notice. */ -#include +#include +#include #include +#include #include -#include #include -#include +#include #include -#include #include #include -#include +#include #ifdef OPENGL_SUPPORT #define GL_GLEXT_PROTOTYPES #ifdef __APPLE__ -#include // Apple's version of GLUT is here +#include // Apple's version of GLUT is here #else #include #endif @@ -150,23 +151,20 @@ #include #endif - /*DEVICE_BUILTIN*/ -struct cudaArray -{ - void *devPtr; - int devPtr32; - struct cudaChannelFormatDesc desc; - int width; - int height; - int size; //in bytes - unsigned dimensions; +struct cudaArray { + void *devPtr; + int devPtr32; + struct cudaChannelFormatDesc desc; + int width; + int height; + int size; // in bytes + unsigned dimensions; }; #if !defined(__dv) #if defined(__cplusplus) -#define __dv(v) \ - = v +#define __dv(v) = v #else /* __cplusplus */ #define __dv(v) #endif /* __cplusplus */ @@ -174,1962 +172,2115 @@ struct cudaArray cudaError_t g_last_cudaError = cudaSuccess; -void register_ptx_function( const char *name, function_info *impl ) -{ - // no longer need this +void register_ptx_function(const char *name, function_info *impl) { + // no longer need this } #if defined __APPLE__ -# define __my_func__ __PRETTY_FUNCTION__ +#define __my_func__ __PRETTY_FUNCTION__ +#else +#if defined __cplusplus ? __GNUC_PREREQ(2, 6) : __GNUC_PREREQ(2, 4) +#define __my_func__ __PRETTY_FUNCTION__ #else -# if defined __cplusplus ? __GNUC_PREREQ (2, 6) : __GNUC_PREREQ (2, 4) -# define __my_func__ __PRETTY_FUNCTION__ -# else -# if defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L -# define __my_func__ __func__ -# else -# define __my_func__ ((__const char *) 0) -# endif -# endif +#if defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L +#define __my_func__ __func__ +#else +#define __my_func__ ((__const char *)0) +#endif +#endif #endif -struct _cuda_device_id *gpgpu_context::GPGPUSim_Init() -{ - _cuda_device_id *the_device = the_gpgpusim->the_cude_device; - if( !the_device ) { - gpgpu_sim *the_gpu = gpgpu_ptx_sim_init_perf(); - - cudaDeviceProp *prop = (cudaDeviceProp *) calloc(sizeof(cudaDeviceProp),1); - snprintf(prop->name,256,"GPGPU-Sim_v%s", g_gpgpusim_version_string ); - prop->major = the_gpu->compute_capability_major(); - prop->minor = the_gpu->compute_capability_minor(); - prop->totalGlobalMem = 0x80000000 /* 2 GB */; - prop->memPitch = 0; - if(prop->major >= 2) { - prop->maxThreadsPerBlock = 1024; - prop->maxThreadsDim[0] = 1024; - prop->maxThreadsDim[1] = 1024; - } - else - { - prop->maxThreadsPerBlock = 512; - prop->maxThreadsDim[0] = 512; - prop->maxThreadsDim[1] = 512; - } - - prop->maxThreadsDim[2] = 64; - prop->maxGridSize[0] = 0x40000000; - prop->maxGridSize[1] = 0x40000000; - prop->maxGridSize[2] = 0x40000000; - prop->totalConstMem = 0x40000000; - prop->textureAlignment = 0; -// * TODO: Update the .config and xml files of all GPU config files with new value of sharedMemPerBlock and regsPerBlock - prop->sharedMemPerBlock = the_gpu->shared_mem_per_block(); +struct _cuda_device_id *gpgpu_context::GPGPUSim_Init() { + _cuda_device_id *the_device = the_gpgpusim->the_cude_device; + if (!the_device) { + gpgpu_sim *the_gpu = gpgpu_ptx_sim_init_perf(); + + cudaDeviceProp *prop = (cudaDeviceProp *)calloc(sizeof(cudaDeviceProp), 1); + snprintf(prop->name, 256, "GPGPU-Sim_v%s", g_gpgpusim_version_string); + prop->major = the_gpu->compute_capability_major(); + prop->minor = the_gpu->compute_capability_minor(); + prop->totalGlobalMem = 0x80000000 /* 2 GB */; + prop->memPitch = 0; + if (prop->major >= 2) { + prop->maxThreadsPerBlock = 1024; + prop->maxThreadsDim[0] = 1024; + prop->maxThreadsDim[1] = 1024; + } else { + prop->maxThreadsPerBlock = 512; + prop->maxThreadsDim[0] = 512; + prop->maxThreadsDim[1] = 512; + } + + prop->maxThreadsDim[2] = 64; + prop->maxGridSize[0] = 0x40000000; + prop->maxGridSize[1] = 0x40000000; + prop->maxGridSize[2] = 0x40000000; + prop->totalConstMem = 0x40000000; + prop->textureAlignment = 0; + // * TODO: Update the .config and xml files of all GPU config files + // with new value of sharedMemPerBlock and regsPerBlock + prop->sharedMemPerBlock = the_gpu->shared_mem_per_block(); #if (CUDART_VERSION > 5050) - prop->regsPerMultiprocessor = the_gpu->num_registers_per_core(); - prop->sharedMemPerMultiprocessor = the_gpu->shared_mem_size(); -#endif - prop->sharedMemPerBlock = the_gpu->shared_mem_per_block(); - prop->regsPerBlock = the_gpu->num_registers_per_block(); - prop->warpSize = the_gpu->wrp_size(); - prop->clockRate = the_gpu->shader_clock(); + prop->regsPerMultiprocessor = the_gpu->num_registers_per_core(); + prop->sharedMemPerMultiprocessor = the_gpu->shared_mem_size(); +#endif + prop->sharedMemPerBlock = the_gpu->shared_mem_per_block(); + prop->regsPerBlock = the_gpu->num_registers_per_block(); + prop->warpSize = the_gpu->wrp_size(); + prop->clockRate = the_gpu->shader_clock(); #if (CUDART_VERSION >= 2010) - prop->multiProcessorCount = the_gpu->get_config().num_shader(); + prop->multiProcessorCount = the_gpu->get_config().num_shader(); #endif #if (CUDART_VERSION >= 4000) - prop->maxThreadsPerMultiProcessor = the_gpu->threads_per_core(); + prop->maxThreadsPerMultiProcessor = the_gpu->threads_per_core(); #endif - the_gpu->set_prop(prop); - the_gpgpusim->the_cude_device = new _cuda_device_id(the_gpu); - the_device = the_gpgpusim->the_cude_device; - } - start_sim_thread(1); - return the_device; -} - -CUctx_st* GPGPUSim_Context(gpgpu_context * ctx) -{ - //static CUctx_st *the_context = NULL; - CUctx_st *the_context = ctx->the_gpgpusim->the_context; - if( the_context == NULL ) { - _cuda_device_id *the_gpu = ctx->GPGPUSim_Init(); - ctx->the_gpgpusim->the_context = new CUctx_st(the_gpu); - the_context = ctx->the_gpgpusim->the_context; - } - return the_context; -} - -gpgpu_context* GPGPU_Context() -{ - static gpgpu_context *gpgpu_ctx = NULL; - if( gpgpu_ctx == NULL ) { - gpgpu_ctx = new gpgpu_context(); - } - return gpgpu_ctx; + the_gpu->set_prop(prop); + the_gpgpusim->the_cude_device = new _cuda_device_id(the_gpu); + the_device = the_gpgpusim->the_cude_device; + } + start_sim_thread(1); + return the_device; +} + +CUctx_st *GPGPUSim_Context(gpgpu_context *ctx) { + // static CUctx_st *the_context = NULL; + CUctx_st *the_context = ctx->the_gpgpusim->the_context; + if (the_context == NULL) { + _cuda_device_id *the_gpu = ctx->GPGPUSim_Init(); + ctx->the_gpgpusim->the_context = new CUctx_st(the_gpu); + the_context = ctx->the_gpgpusim->the_context; + } + return the_context; +} + +gpgpu_context *GPGPU_Context() { + static gpgpu_context *gpgpu_ctx = NULL; + if (gpgpu_ctx == NULL) { + gpgpu_ctx = new gpgpu_context(); + } + return gpgpu_ctx; +} + +void ptxinfo_data::ptxinfo_addinfo() { + CUctx_st *context = GPGPUSim_Context(gpgpu_ctx); + if (!get_ptxinfo_kname()) { + /* This info is not per kernel (since CUDA 5.0 some info (e.g. gmem, and + * cmem) is added at the beginning for the whole binary ) */ + print_ptxinfo(); + context->add_ptxinfo(get_ptxinfo()); + clear_ptxinfo(); + return; + } + if (!strcmp("__cuda_dummy_entry__", get_ptxinfo_kname())) { + // this string produced by ptxas for empty ptx files (e.g., bandwidth test) + clear_ptxinfo(); + return; + } + print_ptxinfo(); + context->add_ptxinfo(get_ptxinfo_kname(), get_ptxinfo()); + clear_ptxinfo(); } - void ptxinfo_data::ptxinfo_addinfo() -{ - CUctx_st *context = GPGPUSim_Context(gpgpu_ctx); - if(!get_ptxinfo_kname()){ - /* This info is not per kernel (since CUDA 5.0 some info (e.g. gmem, and cmem) is added at the beginning for the whole binary ) */ - print_ptxinfo(); - context->add_ptxinfo(get_ptxinfo()); - clear_ptxinfo(); - return; - } - if( !strcmp("__cuda_dummy_entry__",get_ptxinfo_kname()) ) { - // this string produced by ptxas for empty ptx files (e.g., bandwidth test) - clear_ptxinfo(); - return; - } - print_ptxinfo(); - context->add_ptxinfo( get_ptxinfo_kname(), get_ptxinfo() ); - clear_ptxinfo(); -} - -void cuda_not_implemented( const char* func, unsigned line ) -{ - fflush(stdout); - fflush(stderr); - printf("\n\nGPGPU-Sim PTX: Execution error: CUDA API function \"%s()\" has not been implemented yet.\n" - " [$GPGPUSIM_ROOT/libcuda/%s around line %u]\n\n\n", - func,__FILE__, line ); - fflush(stdout); - abort(); +void cuda_not_implemented(const char *func, unsigned line) { + fflush(stdout); + fflush(stderr); + printf( + "\n\nGPGPU-Sim PTX: Execution error: CUDA API function \"%s()\" has not " + "been implemented yet.\n" + " [$GPGPUSIM_ROOT/libcuda/%s around line %u]\n\n\n", + func, __FILE__, line); + fflush(stdout); + abort(); } -void announce_call( const char* func ) -{ - printf("\n\nGPGPU-Sim PTX: CUDA API function \"%s\" has been called.\n", func); - fflush(stdout); +void announce_call(const char *func) { + printf("\n\nGPGPU-Sim PTX: CUDA API function \"%s\" has been called.\n", + func); + fflush(stdout); } -#define gpgpusim_ptx_error(msg, ...) gpgpusim_ptx_error_impl(__func__, __FILE__,__LINE__, msg, ##__VA_ARGS__) -#define gpgpusim_ptx_assert(cond,msg, ...) gpgpusim_ptx_assert_impl((cond),__func__, __FILE__,__LINE__, msg, ##__VA_ARGS__) +#define gpgpusim_ptx_error(msg, ...) \ + gpgpusim_ptx_error_impl(__func__, __FILE__, __LINE__, msg, ##__VA_ARGS__) +#define gpgpusim_ptx_assert(cond, msg, ...) \ + gpgpusim_ptx_assert_impl((cond), __func__, __FILE__, __LINE__, msg, \ + ##__VA_ARGS__) -void gpgpusim_ptx_error_impl( const char *func, const char *file, unsigned line, const char *msg, ... ) -{ - va_list ap; - char buf[1024]; - va_start(ap,msg); - vsnprintf(buf,1024,msg,ap); - va_end(ap); +void gpgpusim_ptx_error_impl(const char *func, const char *file, unsigned line, + const char *msg, ...) { + va_list ap; + char buf[1024]; + va_start(ap, msg); + vsnprintf(buf, 1024, msg, ap); + va_end(ap); - printf("GPGPU-Sim CUDA API: %s\n", buf); - printf(" [%s:%u : %s]\n", file, line, func ); - abort(); + printf("GPGPU-Sim CUDA API: %s\n", buf); + printf(" [%s:%u : %s]\n", file, line, func); + abort(); } -void gpgpusim_ptx_assert_impl( int test_value, const char *func, const char *file, unsigned line, const char *msg, ... ) -{ - va_list ap; - char buf[1024]; - va_start(ap,msg); - vsnprintf(buf,1024,msg,ap); - va_end(ap); +void gpgpusim_ptx_assert_impl(int test_value, const char *func, + const char *file, unsigned line, const char *msg, + ...) { + va_list ap; + char buf[1024]; + va_start(ap, msg); + vsnprintf(buf, 1024, msg, ap); + va_end(ap); - if ( test_value == 0 ) - gpgpusim_ptx_error_impl(func, file, line, msg); + if (test_value == 0) gpgpusim_ptx_error_impl(func, file, line, msg); } - -typedef std::map event_tracker_t; +typedef std::map event_tracker_t; int CUevent_st::m_next_event_uid; event_tracker_t g_timer_events; -extern int cuobjdump_lex_init(yyscan_t* scanner); -extern void cuobjdump_set_in (FILE * _in_str ,yyscan_t yyscanner ); -extern int cuobjdump_parse(yyscan_t scanner, struct cuobjdump_parser* parser, std::list &cuobjdumpSectionList); +extern int cuobjdump_lex_init(yyscan_t *scanner); +extern void cuobjdump_set_in(FILE *_in_str, yyscan_t yyscanner); +extern int cuobjdump_parse(yyscan_t scanner, struct cuobjdump_parser *parser, + std::list &cuobjdumpSectionList); extern int cuobjdump_lex_destroy(yyscan_t scanner); -enum cuobjdumpSectionType { - PTXSECTION=0, - ELFSECTION -}; - +enum cuobjdumpSectionType { PTXSECTION = 0, ELFSECTION }; // sectiontype: 0 for ptx, 1 for elf -void addCuobjdumpSection(int sectiontype, std::list &cuobjdumpSectionList){ - if (sectiontype) - cuobjdumpSectionList.push_front(new cuobjdumpELFSection()); - else - cuobjdumpSectionList.push_front(new cuobjdumpPTXSection()); - printf("## Adding new section %s\n", sectiontype?"ELF":"PTX"); -} - -void setCuobjdumparch(const char* arch, std::list &cuobjdumpSectionList){ - unsigned archnum; - sscanf(arch, "sm_%u", &archnum); - assert (archnum && "cannot have sm_0"); - printf("Adding arch: %s\n", arch); - cuobjdumpSectionList.front()->setArch(archnum); -} - -void setCuobjdumpidentifier(const char* identifier, std::list &cuobjdumpSectionList){ - printf("Adding identifier: %s\n", identifier); - cuobjdumpSectionList.front()->setIdentifier(identifier); -} - -void setCuobjdumpptxfilename(const char* filename, std::list &cuobjdumpSectionList){ - printf("Adding ptx filename: %s\n", filename); - cuobjdumpSection* x = cuobjdumpSectionList.front(); - if (dynamic_cast(x) == NULL){ - assert (0 && "You shouldn't be trying to add a ptxfilename to an elf section"); - } - (dynamic_cast(x))->setPTXfilename(filename); -} - -void setCuobjdumpelffilename(const char* filename, std::list &cuobjdumpSectionList){ - if (dynamic_cast(cuobjdumpSectionList.front()) == NULL){ - assert (0 && "You shouldn't be trying to add a elffilename to an ptx section"); - } - (dynamic_cast(cuobjdumpSectionList.front()))->setELFfilename(filename); -} - -void setCuobjdumpsassfilename(const char* filename, std::list &cuobjdumpSectionList){ - if (dynamic_cast(cuobjdumpSectionList.front()) == NULL){ - assert (0 && "You shouldn't be trying to add a sassfilename to an ptx section"); - } - (dynamic_cast(cuobjdumpSectionList.front()))->setSASSfilename(filename); -} - -//! Return the executable file of the process containing the PTX/SASS code +void addCuobjdumpSection(int sectiontype, + std::list &cuobjdumpSectionList) { + if (sectiontype) + cuobjdumpSectionList.push_front(new cuobjdumpELFSection()); + else + cuobjdumpSectionList.push_front(new cuobjdumpPTXSection()); + printf("## Adding new section %s\n", sectiontype ? "ELF" : "PTX"); +} + +void setCuobjdumparch(const char *arch, + std::list &cuobjdumpSectionList) { + unsigned archnum; + sscanf(arch, "sm_%u", &archnum); + assert(archnum && "cannot have sm_0"); + printf("Adding arch: %s\n", arch); + cuobjdumpSectionList.front()->setArch(archnum); +} + +void setCuobjdumpidentifier( + const char *identifier, + std::list &cuobjdumpSectionList) { + printf("Adding identifier: %s\n", identifier); + cuobjdumpSectionList.front()->setIdentifier(identifier); +} + +void setCuobjdumpptxfilename( + const char *filename, std::list &cuobjdumpSectionList) { + printf("Adding ptx filename: %s\n", filename); + cuobjdumpSection *x = cuobjdumpSectionList.front(); + if (dynamic_cast(x) == NULL) { + assert(0 && + "You shouldn't be trying to add a ptxfilename to an elf section"); + } + (dynamic_cast(x))->setPTXfilename(filename); +} + +void setCuobjdumpelffilename( + const char *filename, std::list &cuobjdumpSectionList) { + if (dynamic_cast(cuobjdumpSectionList.front()) == + NULL) { + assert(0 && + "You shouldn't be trying to add a elffilename to an ptx section"); + } + (dynamic_cast(cuobjdumpSectionList.front())) + ->setELFfilename(filename); +} + +void setCuobjdumpsassfilename( + const char *filename, std::list &cuobjdumpSectionList) { + if (dynamic_cast(cuobjdumpSectionList.front()) == + NULL) { + assert(0 && + "You shouldn't be trying to add a sassfilename to an ptx section"); + } + (dynamic_cast(cuobjdumpSectionList.front())) + ->setSASSfilename(filename); +} + +//! Return the executable file of the process containing the PTX/SASS code //! //! This Function returns the executable file ran by the process. This //! executable is supposed to contain the PTX/SASS code. It provides workaround -//! for processes running on valgrind by dereferencing /proc//exe within the -//! GPGPU-Sim process before calling cuobjdump to extract PTX/SASS. This is +//! for processes running on valgrind by dereferencing /proc//exe within +//! the GPGPU-Sim process before calling cuobjdump to extract PTX/SASS. This is //! needed because valgrind uses x86 emulation to detect memory leak. Other //! processes (e.g. cuobjdump) reading /proc//exe will see the emulator -//! executable instead of the application binary. -//! -std::string get_app_binary(){ - char self_exe_path[1025]; +//! executable instead of the application binary. +//! +std::string get_app_binary() { + char self_exe_path[1025]; #ifdef __APPLE__ - uint32_t size = sizeof(self_exe_path); - if( _NSGetExecutablePath(self_exe_path,&size) != 0 ) { - printf("GPGPU-Sim ** ERROR: _NSGetExecutablePath input buffer too small\n"); - exit(1); - } + uint32_t size = sizeof(self_exe_path); + if (_NSGetExecutablePath(self_exe_path, &size) != 0) { + printf("GPGPU-Sim ** ERROR: _NSGetExecutablePath input buffer too small\n"); + exit(1); + } #else - std::stringstream exec_link; - exec_link << "/proc/self/exe"; + std::stringstream exec_link; + exec_link << "/proc/self/exe"; - ssize_t path_length = readlink(exec_link.str().c_str(), self_exe_path, 1024); - assert(path_length != -1); - self_exe_path[path_length] = '\0'; + ssize_t path_length = readlink(exec_link.str().c_str(), self_exe_path, 1024); + assert(path_length != -1); + self_exe_path[path_length] = '\0'; #endif - printf("self exe links to: %s\n", self_exe_path); - return self_exe_path; + printf("self exe links to: %s\n", self_exe_path); + return self_exe_path; } -//above func gives abs path whereas this give just the name of application. -char* get_app_binary_name(std::string abs_path){ - char *self_exe_path; +// above func gives abs path whereas this give just the name of application. +char *get_app_binary_name(std::string abs_path) { + char *self_exe_path; #ifdef __APPLE__ - //TODO: get apple device and check the result. - printf("WARNING: not tested for Apple-mac devices \n"); - abort(); + // TODO: get apple device and check the result. + printf("WARNING: not tested for Apple-mac devices \n"); + abort(); #else - char* buf = strdup(abs_path.c_str()); - char *token = strtok(buf, "/"); - while(token !=NULL){ - self_exe_path = token; - token = strtok(NULL,"/"); - } + char *buf = strdup(abs_path.c_str()); + char *token = strtok(buf, "/"); + while (token != NULL) { + self_exe_path = token; + token = strtok(NULL, "/"); + } #endif - self_exe_path = strtok(self_exe_path, "."); - printf("self exe links to: %s\n", self_exe_path); - return self_exe_path; + self_exe_path = strtok(self_exe_path, "."); + printf("self exe links to: %s\n", self_exe_path); + return self_exe_path; } static int get_app_cuda_version() { - int app_cuda_version = 0; - char fname[1024]; - snprintf(fname,1024,"_app_cuda_version_XXXXXX"); - int fd=mkstemp(fname); - close(fd); - std::string app_cuda_version_command = "ldd " + get_app_binary() + " | grep libcudart.so | sed 's/.*libcudart.so.\\(.*\\) =>.*/\\1/' > " + fname; - system(app_cuda_version_command.c_str()); - FILE * cmd = fopen(fname, "r"); - char buf[256]; - while (fgets(buf, sizeof(buf), cmd) != 0) { - std::cout << buf; - app_cuda_version = atoi(buf); - } - fclose(cmd); - if ( app_cuda_version == 0 ) { - printf( "Error - Cannot detect the app's CUDA version.\n" ); - exit(1); - } - return app_cuda_version; + int app_cuda_version = 0; + char fname[1024]; + snprintf(fname, 1024, "_app_cuda_version_XXXXXX"); + int fd = mkstemp(fname); + close(fd); + std::string app_cuda_version_command = + "ldd " + get_app_binary() + + " | grep libcudart.so | sed 's/.*libcudart.so.\\(.*\\) =>.*/\\1/' > " + + fname; + system(app_cuda_version_command.c_str()); + FILE *cmd = fopen(fname, "r"); + char buf[256]; + while (fgets(buf, sizeof(buf), cmd) != 0) { + std::cout << buf; + app_cuda_version = atoi(buf); + } + fclose(cmd); + if (app_cuda_version == 0) { + printf("Error - Cannot detect the app's CUDA version.\n"); + exit(1); + } + return app_cuda_version; } //! Keep track of the association between filename and cubin handle -void cuda_runtime_api::cuobjdumpRegisterFatBinary(unsigned int handle, const char* filename, CUctx_st *context){ - fatbinmap[handle] = filename; +void cuda_runtime_api::cuobjdumpRegisterFatBinary(unsigned int handle, + const char *filename, + CUctx_st *context) { + fatbinmap[handle] = filename; } - - /******************************************************************************* - * Add internal cuda runtime API call to accept gpgpu_context * + * Add internal cuda runtime API call to accept gpgpu_context * *******************************************************************************/ -cudaError_t cudaSetDeviceInternal(int device, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //set the active device to run cuda - if ( device <= ctx->GPGPUSim_Init()->num_devices() ) { - ctx->api->g_active_device = device; - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorInvalidDevice; - } -} - -cudaError_t cudaGetDeviceInternal(int *device, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - *device = ctx->api->g_active_device; - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaDeviceGetLimitInternal( size_t* pValue, cudaLimit limit, gpgpu_context* gpgpu_ctx = NULL ) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - _cuda_device_id *dev = ctx->GPGPUSim_Init(); - const struct cudaDeviceProp *prop = dev->get_prop(); - const gpgpu_sim_config& config=dev->get_gpgpu()->get_config(); - switch(limit) { - case 0: // cudaLimitStackSize - *pValue=config.stack_limit(); - break; - case 2: // cudaLimitMallocHeapSize - *pValue=config.heap_limit(); - break; +cudaError_t cudaSetDeviceInternal(int device, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // set the active device to run cuda + if (device <= ctx->GPGPUSim_Init()->num_devices()) { + ctx->api->g_active_device = device; + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorInvalidDevice; + } +} + +cudaError_t cudaGetDeviceInternal(int *device, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + *device = ctx->api->g_active_device; + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaDeviceGetLimitInternal( + size_t *pValue, cudaLimit limit, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + _cuda_device_id *dev = ctx->GPGPUSim_Init(); + const struct cudaDeviceProp *prop = dev->get_prop(); + const gpgpu_sim_config &config = dev->get_gpgpu()->get_config(); + switch (limit) { + case 0: // cudaLimitStackSize + *pValue = config.stack_limit(); + break; + case 2: // cudaLimitMallocHeapSize + *pValue = config.heap_limit(); + break; #if (CUDART_VERSION > 5050) - case 3: // cudaLimitDevRuntimeSyncDepth - if(prop->major > 2){ - *pValue=config.sync_depth_limit(); - break; - } - else{ - printf("ERROR:Limit %d is not supported on this architecture \n", limit); - abort(); - } - case 4: // cudaLimitDevRuntimePendingLaunchCount - if(prop->major > 2){ - *pValue=config.pending_launch_count_limit(); - break; - } - else{ - printf("ERROR:Limit %d is not supported on this architecture \n",limit); - abort(); - } + case 3: // cudaLimitDevRuntimeSyncDepth + if (prop->major > 2) { + *pValue = config.sync_depth_limit(); + break; + } else { + printf("ERROR:Limit %d is not supported on this architecture \n", + limit); + abort(); + } + case 4: // cudaLimitDevRuntimePendingLaunchCount + if (prop->major > 2) { + *pValue = config.pending_launch_count_limit(); + break; + } else { + printf("ERROR:Limit %d is not supported on this architecture \n", + limit); + abort(); + } #endif - default: - printf("ERROR:Limit %d unimplemented \n",limit); - abort(); - } - return g_last_cudaError = cudaSuccess; - -} - - -void** cudaRegisterFatBinaryInternal( void *fatCubin, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } + default: + printf("ERROR:Limit %d unimplemented \n", limit); + abort(); + } + return g_last_cudaError = cudaSuccess; +} + +void **cudaRegisterFatBinaryInternal(void *fatCubin, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } #if (CUDART_VERSION < 2010) - printf("GPGPU-Sim PTX: ERROR ** this version of GPGPU-Sim requires CUDA 2.1 or higher\n"); - exit(1); + printf( + "GPGPU-Sim PTX: ERROR ** this version of GPGPU-Sim requires CUDA 2.1 or " + "higher\n"); + exit(1); #endif - CUctx_st *context = GPGPUSim_Context(ctx); - static unsigned next_fat_bin_handle = 1; - if(context->get_device()->get_gpgpu()->get_config().use_cuobjdump()) { - // The following workaround has only been verified on 64-bit systems. - if (sizeof(void*) == 4) - printf("GPGPU-Sim PTX: FatBin file name extraction has not been tested on 32-bit system.\n"); - - // This code will get the CUDA version the app was compiled with. - // We need this to determine how to handle the parsing of the binary. - // Making this a runtime variable based on the app, enables GPGPU-Sim compiled - // with a newer version of CUDA to run apps compiled with older versions of - // CUDA. This is especially useful for PTXPLUS execution. - //Skip cuda version check for pytorch application - std::string app_binary_path = get_app_binary(); - int pos = app_binary_path.find("python"); - if (pos==std::string::npos){ - // Not pytorch app : checking cuda version - int app_cuda_version = get_app_cuda_version(); - assert( app_cuda_version == CUDART_VERSION / 1000 && "The app must be compiled with same major version as the simulator." ); - } - - //int app_cuda_version = get_app_cuda_version(); - //assert( app_cuda_version == CUDART_VERSION / 1000 && "The app must be compiled with same major version as the simulator." ); - const char* filename; + CUctx_st *context = GPGPUSim_Context(ctx); + static unsigned next_fat_bin_handle = 1; + if (context->get_device()->get_gpgpu()->get_config().use_cuobjdump()) { + // The following workaround has only been verified on 64-bit systems. + if (sizeof(void *) == 4) + printf( + "GPGPU-Sim PTX: FatBin file name extraction has not been tested on " + "32-bit system.\n"); + + // This code will get the CUDA version the app was compiled with. + // We need this to determine how to handle the parsing of the binary. + // Making this a runtime variable based on the app, enables GPGPU-Sim + // compiled with a newer version of CUDA to run apps compiled with older + // versions of CUDA. This is especially useful for PTXPLUS execution. + // Skip cuda version check for pytorch application + std::string app_binary_path = get_app_binary(); + int pos = app_binary_path.find("python"); + if (pos == std::string::npos) { + // Not pytorch app : checking cuda version + int app_cuda_version = get_app_cuda_version(); + assert( + app_cuda_version == CUDART_VERSION / 1000 && + "The app must be compiled with same major version as the simulator."); + } + + // int app_cuda_version = get_app_cuda_version(); + // assert( app_cuda_version == CUDART_VERSION / 1000 && "The app must be + // compiled with same major version as the simulator." ); + const char *filename; #if CUDART_VERSION < 6000 - // FatBin handle from the .fatbin.c file (one of the intermediate files generated by NVCC) - typedef struct {int m; int v; const unsigned long long* d; char* f;} __fatDeviceText __attribute__ ((aligned (8))); - __fatDeviceText * fatDeviceText = (__fatDeviceText *) fatCubin; - - // Extract the source code file name that generate the given FatBin. - // - Obtains the pointer to the actual fatbin structure from the FatBin handle (fatCubin). - // - An integer inside the fatbin structure contains the relative offset to the source code file name. - // - This offset differs among different CUDA and GCC versions. - char * pfatbin = (char*) fatDeviceText->d; - int offset = *((int*)(pfatbin+48)); - filename = (pfatbin+16+offset); + // FatBin handle from the .fatbin.c file (one of the intermediate files + // generated by NVCC) + typedef struct { + int m; + int v; + const unsigned long long *d; + char *f; + } __fatDeviceText __attribute__((aligned(8))); + __fatDeviceText *fatDeviceText = (__fatDeviceText *)fatCubin; + + // Extract the source code file name that generate the given FatBin. + // - Obtains the pointer to the actual fatbin structure from the FatBin + // handle (fatCubin). + // - An integer inside the fatbin structure contains the relative offset to + // the source code file name. + // - This offset differs among different CUDA and GCC versions. + char *pfatbin = (char *)fatDeviceText->d; + int offset = *((int *)(pfatbin + 48)); + filename = (pfatbin + 16 + offset); #else - filename = "default"; + filename = "default"; #endif - // The extracted file name is associated with a fat_cubin_handle passed - // into cudaLaunch(). Inside cudaLaunch(), the associated file name is - // used to find the PTX/SASS section from cuobjdump, which contains the - // PTX/SASS code for the launched kernel function. - // This allows us to work around the fact that cuobjdump only outputs the - // file name associated with each section. - unsigned long long fat_cubin_handle = next_fat_bin_handle; - next_fat_bin_handle++; - printf("GPGPU-Sim PTX: __cudaRegisterFatBinary, fat_cubin_handle = %llu, filename=%s\n", fat_cubin_handle, filename); - /*! - * This function extracts all data from all files in first call - * then for next calls, only returns the appropriate number - */ - assert(fat_cubin_handle >= 1); - if (fat_cubin_handle==1) ctx->api->cuobjdumpInit(); - ctx->api->cuobjdumpRegisterFatBinary(fat_cubin_handle, filename, context); - - return (void**)fat_cubin_handle; - } + // The extracted file name is associated with a fat_cubin_handle passed + // into cudaLaunch(). Inside cudaLaunch(), the associated file name is + // used to find the PTX/SASS section from cuobjdump, which contains the + // PTX/SASS code for the launched kernel function. + // This allows us to work around the fact that cuobjdump only outputs the + // file name associated with each section. + unsigned long long fat_cubin_handle = next_fat_bin_handle; + next_fat_bin_handle++; + printf( + "GPGPU-Sim PTX: __cudaRegisterFatBinary, fat_cubin_handle = %llu, " + "filename=%s\n", + fat_cubin_handle, filename); + /*! + * This function extracts all data from all files in first call + * then for next calls, only returns the appropriate number + */ + assert(fat_cubin_handle >= 1); + if (fat_cubin_handle == 1) ctx->api->cuobjdumpInit(); + ctx->api->cuobjdumpRegisterFatBinary(fat_cubin_handle, filename, context); + + return (void **)fat_cubin_handle; + } #if (CUDART_VERSION < 8000) - else { - static unsigned source_num=1; - unsigned long long fat_cubin_handle = next_fat_bin_handle++; - __cudaFatCudaBinary *info = (__cudaFatCudaBinary *)fatCubin; - assert( info->version >= 3 ); - unsigned num_ptx_versions=0; - unsigned max_capability=0; - unsigned selected_capability=0; - bool found=false; - unsigned forced_max_capability = context->get_device()->get_gpgpu()->get_config().get_forced_max_capability(); - if (!info->ptx){ - printf("ERROR: Cannot find ptx code in cubin file\n" - "\tIf you are using CUDA 4.0 or higher, please enable -gpgpu_ptx_use_cuobjdump or downgrade to CUDA 3.1\n"); - exit(1); - } - while( info->ptx[num_ptx_versions].gpuProfileName != NULL ) { - unsigned capability=0; - sscanf(info->ptx[num_ptx_versions].gpuProfileName,"compute_%u",&capability); - printf("GPGPU-Sim PTX: __cudaRegisterFatBinary found PTX versions for '%s', ", info->ident); - printf("capability = %s\n", info->ptx[num_ptx_versions].gpuProfileName ); - if( forced_max_capability ) { - if( capability > max_capability && capability <= forced_max_capability ) { - found = true; - max_capability=capability; - selected_capability = num_ptx_versions; - } - } else { - if( capability > max_capability ) { - found = true; - max_capability=capability; - selected_capability = num_ptx_versions; - } - } - num_ptx_versions++; - } - if( found ) { - printf("GPGPU-Sim PTX: Loading PTX for %s, capability = %s\n", - info->ident, info->ptx[selected_capability].gpuProfileName ); - symbol_table *symtab; - const char *ptx = info->ptx[selected_capability].ptx; - if(context->get_device()->get_gpgpu()->get_config().convert_to_ptxplus() ) { - printf("GPGPU-Sim PTX: ERROR ** PTXPlus is only supported through cuobjdump\n" - "\tEither enable cuobjdump or disable PTXPlus in your configuration file\n"); - exit(1); - } else { - symtab=ctx->gpgpu_ptx_sim_load_ptx_from_string(ptx,source_num); - context->add_binary(symtab,fat_cubin_handle); - ctx->gpgpu_ptxinfo_load_from_string( ptx, source_num, max_capability, context->no_of_ptx ); - } - source_num++; - ctx->api->load_static_globals(symtab,STATIC_ALLOC_LIMIT,0xFFFFFFFF,context->get_device()->get_gpgpu()); - ctx->api->load_constants(symtab,STATIC_ALLOC_LIMIT,context->get_device()->get_gpgpu()); - } else { - printf("GPGPU-Sim PTX: warning -- did not find an appropriate PTX in cubin\n"); - } - return (void**)fat_cubin_handle; - } -#else - else { - printf("ERROR ** __cudaRegisterFatBinary() needs to be updated\n"); - abort(); + else { + static unsigned source_num = 1; + unsigned long long fat_cubin_handle = next_fat_bin_handle++; + __cudaFatCudaBinary *info = (__cudaFatCudaBinary *)fatCubin; + assert(info->version >= 3); + unsigned num_ptx_versions = 0; + unsigned max_capability = 0; + unsigned selected_capability = 0; + bool found = false; + unsigned forced_max_capability = context->get_device() + ->get_gpgpu() + ->get_config() + .get_forced_max_capability(); + if (!info->ptx) { + printf( + "ERROR: Cannot find ptx code in cubin file\n" + "\tIf you are using CUDA 4.0 or higher, please enable " + "-gpgpu_ptx_use_cuobjdump or downgrade to CUDA 3.1\n"); + exit(1); + } + while (info->ptx[num_ptx_versions].gpuProfileName != NULL) { + unsigned capability = 0; + sscanf(info->ptx[num_ptx_versions].gpuProfileName, "compute_%u", + &capability); + printf( + "GPGPU-Sim PTX: __cudaRegisterFatBinary found PTX versions for " + "'%s', ", + info->ident); + printf("capability = %s\n", info->ptx[num_ptx_versions].gpuProfileName); + if (forced_max_capability) { + if (capability > max_capability && + capability <= forced_max_capability) { + found = true; + max_capability = capability; + selected_capability = num_ptx_versions; } + } else { + if (capability > max_capability) { + found = true; + max_capability = capability; + selected_capability = num_ptx_versions; + } + } + num_ptx_versions++; + } + if (found) { + printf("GPGPU-Sim PTX: Loading PTX for %s, capability = %s\n", + info->ident, info->ptx[selected_capability].gpuProfileName); + symbol_table *symtab; + const char *ptx = info->ptx[selected_capability].ptx; + if (context->get_device() + ->get_gpgpu() + ->get_config() + .convert_to_ptxplus()) { + printf( + "GPGPU-Sim PTX: ERROR ** PTXPlus is only supported through " + "cuobjdump\n" + "\tEither enable cuobjdump or disable PTXPlus in your " + "configuration file\n"); + exit(1); + } else { + symtab = ctx->gpgpu_ptx_sim_load_ptx_from_string(ptx, source_num); + context->add_binary(symtab, fat_cubin_handle); + ctx->gpgpu_ptxinfo_load_from_string(ptx, source_num, max_capability, + context->no_of_ptx); + } + source_num++; + ctx->api->load_static_globals(symtab, STATIC_ALLOC_LIMIT, 0xFFFFFFFF, + context->get_device()->get_gpgpu()); + ctx->api->load_constants(symtab, STATIC_ALLOC_LIMIT, + context->get_device()->get_gpgpu()); + } else { + printf( + "GPGPU-Sim PTX: warning -- did not find an appropriate PTX in " + "cubin\n"); + } + return (void **)fat_cubin_handle; + } +#else + else { + printf("ERROR ** __cudaRegisterFatBinary() needs to be updated\n"); + abort(); + } #endif } -void cudaRegisterFunctionInternal( - void **fatCubinHandle, - const char *hostFun, - char *deviceFun, - const char *deviceName, - int thread_limit, - uint3 *tid, - uint3 *bid, - dim3 *bDim, - dim3 *gDim, - gpgpu_context *gpgpu_ctx = NULL -) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - unsigned fat_cubin_handle = (unsigned)(unsigned long long)fatCubinHandle; - printf("GPGPU-Sim PTX: __cudaRegisterFunction %s : hostFun 0x%p, fat_cubin_handle = %u\n", - deviceFun, hostFun, fat_cubin_handle); - if(context->get_device()->get_gpgpu()->get_config().use_cuobjdump()) - ctx->cuobjdumpParseBinary(fat_cubin_handle); - context->register_function( fat_cubin_handle, hostFun, deviceFun ); +void cudaRegisterFunctionInternal(void **fatCubinHandle, const char *hostFun, + char *deviceFun, const char *deviceName, + int thread_limit, uint3 *tid, uint3 *bid, + dim3 *bDim, dim3 *gDim, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + unsigned fat_cubin_handle = (unsigned)(unsigned long long)fatCubinHandle; + printf( + "GPGPU-Sim PTX: __cudaRegisterFunction %s : hostFun 0x%p, " + "fat_cubin_handle = %u\n", + deviceFun, hostFun, fat_cubin_handle); + if (context->get_device()->get_gpgpu()->get_config().use_cuobjdump()) + ctx->cuobjdumpParseBinary(fat_cubin_handle); + context->register_function(fat_cubin_handle, hostFun, deviceFun); } void cudaRegisterVarInternal( - void **fatCubinHandle, - char *hostVar, //pointer to...something - char *deviceAddress, //name of variable - const char *deviceName, //name of variable (same as above) - int ext, - int size, - int constant, - int global, - gpgpu_context *gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("GPGPU-Sim PTX: __cudaRegisterVar: hostVar = %p; deviceAddress = %s; deviceName = %s\n", hostVar, deviceAddress, deviceName); - printf("GPGPU-Sim PTX: __cudaRegisterVar: Registering const memory space of %d bytes\n", size); - if(GPGPUSim_Context(ctx)->get_device()->get_gpgpu()->get_config().use_cuobjdump()) - ctx->cuobjdumpParseBinary((unsigned)(unsigned long long)fatCubinHandle); - fflush(stdout); - if ( constant && !global && !ext ) { - ctx->func_sim->gpgpu_ptx_sim_register_const_variable(hostVar,deviceName,size); - } else if ( !constant && !global && !ext ) { - ctx->func_sim->gpgpu_ptx_sim_register_global_variable(hostVar,deviceName,size); - } else cuda_not_implemented(__my_func__,__LINE__); -} - -cudaError_t cudaConfigureCallInternal(dim3 gridDim, dim3 blockDim, size_t sharedMem, cudaStream_t stream, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; + void **fatCubinHandle, + char *hostVar, // pointer to...something + char *deviceAddress, // name of variable + const char *deviceName, // name of variable (same as above) + int ext, int size, int constant, int global, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf( + "GPGPU-Sim PTX: __cudaRegisterVar: hostVar = %p; deviceAddress = %s; " + "deviceName = %s\n", + hostVar, deviceAddress, deviceName); + printf( + "GPGPU-Sim PTX: __cudaRegisterVar: Registering const memory space of %d " + "bytes\n", + size); + if (GPGPUSim_Context(ctx) + ->get_device() + ->get_gpgpu() + ->get_config() + .use_cuobjdump()) + ctx->cuobjdumpParseBinary((unsigned)(unsigned long long)fatCubinHandle); + fflush(stdout); + if (constant && !global && !ext) { + ctx->func_sim->gpgpu_ptx_sim_register_const_variable(hostVar, deviceName, + size); + } else if (!constant && !global && !ext) { + ctx->func_sim->gpgpu_ptx_sim_register_global_variable(hostVar, deviceName, + size); + } else + cuda_not_implemented(__my_func__, __LINE__); +} + +cudaError_t cudaConfigureCallInternal(dim3 gridDim, dim3 blockDim, + size_t sharedMem, cudaStream_t stream, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + struct CUstream_st *s = (struct CUstream_st *)stream; + ctx->api->g_cuda_launch_stack.push_back( + kernel_config(gridDim, blockDim, sharedMem, s)); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI +cudaGetDeviceCountInternal(int *count, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + _cuda_device_id *dev = ctx->GPGPUSim_Init(); + *count = dev->num_devices(); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaGetDevicePropertiesInternal( + struct cudaDeviceProp *prop, int device, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + _cuda_device_id *dev = ctx->GPGPUSim_Init(); + if (device <= dev->num_devices()) { + *prop = *dev->get_prop(); + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorInvalidDevice; + } +} + +__host__ cudaError_t CUDARTAPI +cudaChooseDeviceInternal(int *device, const struct cudaDeviceProp *prop, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + _cuda_device_id *dev = ctx->GPGPUSim_Init(); + *device = dev->get_id(); + return g_last_cudaError = cudaSuccess; +} + +cudaError_t cudaSetupArgumentInternal(const void *arg, size_t size, + size_t offset, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + gpgpusim_ptx_assert(!ctx->api->g_cuda_launch_stack.empty(), + "empty launch stack"); + kernel_config &config = ctx->api->g_cuda_launch_stack.back(); + config.set_arg(arg, size, offset); + printf( + "GPGPU-Sim PTX: Setting up arguments for %zu bytes starting at " + "0x%llx..\n", + size, (unsigned long long)arg); + + return g_last_cudaError = cudaSuccess; +} + +cudaError_t cudaLaunchInternal(const char *hostFun, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + char *mode = getenv("PTX_SIM_MODE_FUNC"); + if (mode) sscanf(mode, "%u", &(ctx->func_sim->g_ptx_sim_mode)); + gpgpusim_ptx_assert(!ctx->api->g_cuda_launch_stack.empty(), + "empty launch stack"); + kernel_config config = ctx->api->g_cuda_launch_stack.back(); + { + dim3 gridDim = config.grid_dim(); + dim3 blockDim = config.block_dim(); + if (gridDim.x * gridDim.y * gridDim.z == 0 || + blockDim.x * blockDim.y * blockDim.z == 0) { + // can't launch + printf("can't launch a empty kernel\n"); + ctx->api->g_cuda_launch_stack.pop_back(); + return g_last_cudaError = cudaErrorInvalidConfiguration; + } + } + struct CUstream_st *stream = config.get_stream(); + + printf("\nGPGPU-Sim PTX: cudaLaunch for 0x%p (mode=%s) on stream %u\n", + hostFun, + (ctx->func_sim->g_ptx_sim_mode) ? "functional simulation" + : "performance simulation", + stream ? stream->get_uid() : 0); + kernel_info_t *grid = ctx->api->gpgpu_cuda_ptx_sim_init_grid( + hostFun, config.get_args(), config.grid_dim(), config.block_dim(), + context); + // do dynamic PDOM analysis for performance simulation scenario + std::string kname = grid->name(); + function_info *kernel_func_info = grid->entry(); + if (kernel_func_info->is_pdom_set()) { + printf("GPGPU-Sim PTX: PDOM analysis already done for %s \n", + kname.c_str()); + } else { + printf("GPGPU-Sim PTX: finding reconvergence points for \'%s\'...\n", + kname.c_str()); + kernel_func_info->do_pdom(); + kernel_func_info->set_pdom(); + } + dim3 gridDim = config.grid_dim(); + dim3 blockDim = config.block_dim(); + + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + checkpoint *g_checkpoint; + g_checkpoint = new checkpoint(); + class memory_space *global_mem; + global_mem = gpu->get_global_memory(); + + if (gpu->resume_option == 1 && (grid->get_uid() == gpu->resume_kernel)) { + char f1name[2048]; + snprintf(f1name, 2048, "checkpoint_files/global_mem_%d.txt", + grid->get_uid()); + + g_checkpoint->load_global_mem(global_mem, f1name); + for (int i = 0; i < gpu->resume_CTA; i++) grid->increment_cta_id(); + } + if (gpu->resume_option == 1 && (grid->get_uid() < gpu->resume_kernel)) { + char f1name[2048]; + snprintf(f1name, 2048, "checkpoint_files/global_mem_%d.txt", + grid->get_uid()); + + g_checkpoint->load_global_mem(global_mem, f1name); + printf("Skipping kernel %d as resuming from kernel %d\n", grid->get_uid(), + gpu->resume_kernel); + ctx->api->g_cuda_launch_stack.pop_back(); + return g_last_cudaError = cudaSuccess; + } + if (gpu->checkpoint_option == 1 && + (grid->get_uid() > gpu->checkpoint_kernel)) { + printf("Skipping kernel %d as checkpoint from kernel %d\n", grid->get_uid(), + gpu->checkpoint_kernel); + ctx->api->g_cuda_launch_stack.pop_back(); + return g_last_cudaError = cudaSuccess; + } + printf( + "GPGPU-Sim PTX: pushing kernel \'%s\' to stream %u, gridDim= (%u,%u,%u) " + "blockDim = (%u,%u,%u) \n", + kname.c_str(), stream ? stream->get_uid() : 0, gridDim.x, gridDim.y, + gridDim.z, blockDim.x, blockDim.y, blockDim.z); + stream_operation op(grid, ctx->func_sim->g_ptx_sim_mode, stream); + ctx->the_gpgpusim->g_stream_manager->push(op); + ctx->api->g_cuda_launch_stack.pop_back(); + return g_last_cudaError = cudaSuccess; +} + +cudaError_t cudaMallocInternal(void **devPtr, size_t size, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + *devPtr = context->get_device()->get_gpgpu()->gpu_malloc(size); + if (g_debug_execution >= 3) { + printf("GPGPU-Sim PTX: cudaMallocing %zu bytes starting at 0x%llx..\n", + size, (unsigned long long)*devPtr); + ctx->api->g_mallocPtr_Size[(unsigned long long)*devPtr] = size; + } + if (*devPtr) { + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorMemoryAllocation; + } +} + +cudaError_t cudaMallocHostInternal(void **ptr, size_t size, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + *ptr = malloc(size); + if (*ptr) { + // track pinned memory size allocated in the host so that same amount of + // memory is also allocated in GPU. + ctx->api->pinned_memory_size[*ptr] = size; + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorMemoryAllocation; + } +} + +__host__ cudaError_t CUDARTAPI +cudaMallocPitchInternal(void **devPtr, size_t *pitch, size_t width, + size_t height, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + unsigned malloc_width_inbytes = width; + printf("GPGPU-Sim PTX: cudaMallocPitch (width = %d)\n", malloc_width_inbytes); + CUctx_st *context = GPGPUSim_Context(ctx); + *devPtr = context->get_device()->get_gpgpu()->gpu_malloc( + malloc_width_inbytes * height); + pitch[0] = malloc_width_inbytes; + if (*devPtr) { + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorMemoryAllocation; + } +} + +cudaError_t cudaHostGetDevicePointerInternal(void **pDevice, void *pHost, + unsigned int flags, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // only cpu memory allocation happens in cudaHostAlloc. Linking with device + // pointer to pinned memory happens here. + // TODO: once kernel is executed, the contents in global pointer of GPU must + // be copied back to CPU host pointer! + flags = 0; + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + std::map::const_iterator i = + ctx->api->pinned_memory_size.find(pHost); + assert(i != ctx->api->pinned_memory_size.end()); + size_t size = i->second; + *pDevice = gpu->gpu_malloc(size); + if (g_debug_execution >= 3) { + printf("GPGPU-Sim PTX: cudaMallocing %zu bytes starting at 0x%llx..\n", + size, (unsigned long long)*pDevice); + ctx->api->g_mallocPtr_Size[(unsigned long long)*pDevice] = size; + } + if (*pDevice) { + ctx->api->pinned_memory[pHost] = pDevice; + // Copy contents in cpu to gpu + gpu->memcpy_to_gpu((size_t)*pDevice, pHost, size); + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorMemoryAllocation; + } +} + +__host__ cudaError_t CUDARTAPI cudaMallocArrayInternal( + struct cudaArray **array, const struct cudaChannelFormatDesc *desc, + size_t width, size_t height __dv(1), gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + unsigned size = + width * height * ((desc->x + desc->y + desc->z + desc->w) / 8); + CUctx_st *context = GPGPUSim_Context(ctx); + (*array) = (struct cudaArray *)malloc(sizeof(struct cudaArray)); + (*array)->desc = *desc; + (*array)->width = width; + (*array)->height = height; + (*array)->size = size; + (*array)->dimensions = 2; + ((*array)->devPtr32) = + (int)(long long)context->get_device()->get_gpgpu()->gpu_mallocarray(size); + printf("GPGPU-Sim PTX: cudaMallocArray: devPtr32 = %d\n", + ((*array)->devPtr32)); + ((*array)->devPtr) = (void *)(long long)((*array)->devPtr32); + if (((*array)->devPtr)) { + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorMemoryAllocation; + } +} + +__host__ cudaError_t CUDARTAPI +cudaMemcpyInternal(void *dst, const void *src, size_t count, + enum cudaMemcpyKind kind, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // CUctx_st *context = GPGPUSim_Context(); + // gpgpu_t *gpu = context->get_device()->get_gpgpu(); + if (g_debug_execution >= 3) + printf("GPGPU-Sim PTX: cudaMemcpy(): devPtr = %p\n", dst); + if (kind == cudaMemcpyHostToDevice) + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation(src, (size_t)dst, count, 0)); + else if (kind == cudaMemcpyDeviceToHost) + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation((size_t)src, dst, count, 0)); + else if (kind == cudaMemcpyDeviceToDevice) + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation((size_t)src, (size_t)dst, count, 0)); + else if (kind == cudaMemcpyDefault) { + if ((size_t)src >= GLOBAL_HEAP_START) { + if ((size_t)dst >= GLOBAL_HEAP_START) + ctx->the_gpgpusim->g_stream_manager->push(stream_operation( + (size_t)src, (size_t)dst, count, 0)); // device to device + else + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation((size_t)src, dst, count, 0)); // device to host } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); + if ((size_t)dst >= GLOBAL_HEAP_START) + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation(src, (size_t)dst, count, 0)); + else { + printf( + "GPGPU-Sim PTX: cudaMemcpy - ERROR : unsupported transfer: host to " + "host\n"); + abort(); + } } - struct CUstream_st *s = (struct CUstream_st *)stream; - ctx->api->g_cuda_launch_stack.push_back( kernel_config(gridDim,blockDim,sharedMem,s) ); - return g_last_cudaError = cudaSuccess; + } else { + printf("GPGPU-Sim PTX: cudaMemcpy - ERROR : unsupported cudaMemcpyKind\n"); + abort(); + } + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpyToArrayInternal( + struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, + size_t count, enum cudaMemcpyKind kind, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + size_t size = count; + printf("GPGPU-Sim PTX: cudaMemcpyToArray\n"); + if (kind == cudaMemcpyHostToDevice) + gpu->memcpy_to_gpu((size_t)(dst->devPtr), src, size); + else if (kind == cudaMemcpyDeviceToHost) + gpu->memcpy_from_gpu(dst->devPtr, (size_t)src, size); + else if (kind == cudaMemcpyDeviceToDevice) + gpu->memcpy_gpu_to_gpu((size_t)(dst->devPtr), (size_t)src, size); + else { + printf( + "GPGPU-Sim PTX: cudaMemcpyToArray - ERROR : unsupported " + "cudaMemcpyKind\n"); + abort(); + } + dst->devPtr32 = (unsigned)(size_t)(dst->devPtr); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpy2DInternal( + void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, + size_t height, enum cudaMemcpyKind kind, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + size_t size = spitch * height; + gpgpusim_ptx_assert((dpitch == spitch), + "different src and dst pitch not supported yet"); + if (kind == cudaMemcpyHostToDevice) + gpu->memcpy_to_gpu((size_t)dst, src, size); + else if (kind == cudaMemcpyDeviceToHost) + gpu->memcpy_from_gpu(dst, (size_t)src, size); + else if (kind == cudaMemcpyDeviceToDevice) + gpu->memcpy_gpu_to_gpu((size_t)dst, (size_t)src, size); + else { + printf( + "GPGPU-Sim PTX: cudaMemcpy2D - ERROR : unsupported cudaMemcpyKind\n"); + abort(); + } + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpy2DToArrayInternal( + struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, + size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + size_t size = spitch * height; + size_t channel_size = dst->desc.w + dst->desc.x + dst->desc.y + dst->desc.z; + gpgpusim_ptx_assert( + ((channel_size % 8) == 0), + "none byte multiple destination channel size not supported (sz=%u)", + channel_size); + unsigned elem_size = channel_size / 8; + gpgpusim_ptx_assert((dst->dimensions == 2), + "copy to none 2D array not supported"); + gpgpusim_ptx_assert((wOffset == 0), "non-zero wOffset not yet supported"); + gpgpusim_ptx_assert((hOffset == 0), "non-zero hOffset not yet supported"); + gpgpusim_ptx_assert((dst->height == (int)height), + "partial copy not supported"); + gpgpusim_ptx_assert((elem_size * dst->width == width), + "partial copy not supported"); + gpgpusim_ptx_assert((spitch == width), "spitch != width not supported"); + if (kind == cudaMemcpyHostToDevice) + gpu->memcpy_to_gpu((size_t)(dst->devPtr), src, size); + else if (kind == cudaMemcpyDeviceToHost) + gpu->memcpy_from_gpu(dst->devPtr, (size_t)src, size); + else if (kind == cudaMemcpyDeviceToDevice) + gpu->memcpy_gpu_to_gpu((size_t)dst->devPtr, (size_t)src, size); + else { + printf( + "GPGPU-Sim PTX: cudaMemcpy2D - ERROR : unsupported cudaMemcpyKind\n"); + abort(); + } + dst->devPtr32 = (unsigned)(size_t)(dst->devPtr); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpyToSymbolInternal( + const char *symbol, const void *src, size_t count, size_t offset __dv(0), + enum cudaMemcpyKind kind __dv(cudaMemcpyHostToDevice), + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // CUctx_st *context = GPGPUSim_Context(); + assert(kind == cudaMemcpyHostToDevice); + printf("GPGPU-Sim PTX: cudaMemcpyToSymbol: symbol = %p\n", symbol); + // stream_operation( const char *symbol, const void *src, size_t count, size_t + // offset ) + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation(src, symbol, count, offset, 0)); + // gpgpu_ptx_sim_memcpy_symbol(symbol,src,count,offset,1,context->get_device()->get_gpgpu()); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpyFromSymbolInternal( + void *dst, const char *symbol, size_t count, size_t offset __dv(0), + enum cudaMemcpyKind kind __dv(cudaMemcpyDeviceToHost), + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // CUctx_st *context = GPGPUSim_Context(); + assert(kind == cudaMemcpyDeviceToHost); + printf("GPGPU-Sim PTX: cudaMemcpyFromSymbol: symbol = %p\n", symbol); + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation(symbol, dst, count, offset, 0)); + // gpgpu_ptx_sim_memcpy_symbol(symbol,dst,count,offset,0,context->get_device()->get_gpgpu()); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpyAsyncInternal( + void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, + cudaStream_t stream, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + struct CUstream_st *s = (struct CUstream_st *)stream; + switch (kind) { + case cudaMemcpyHostToDevice: + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation(src, (size_t)dst, count, s)); + break; + case cudaMemcpyDeviceToHost: + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation((size_t)src, dst, count, s)); + break; + case cudaMemcpyDeviceToDevice: + ctx->the_gpgpusim->g_stream_manager->push( + stream_operation((size_t)src, (size_t)dst, count, s)); + break; + default: + abort(); + } + return g_last_cudaError = cudaSuccess; } -__host__ cudaError_t CUDARTAPI cudaGetDeviceCountInternal(int *count, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - _cuda_device_id *dev = ctx->GPGPUSim_Init(); - *count = dev->num_devices(); +#if (CUDART_VERSION >= 8000) +cudaError_t CUDARTAPI +cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlagsInternal( + int *numBlocks, const char *hostFunc, int blockSize, size_t dynamicSMemSize, + unsigned int flags, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + printf( + "GPGPU-Sim PTX: cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags " + "%p\n", + hostFunc); + CUctx_st *context = GPGPUSim_Context(ctx); + function_info *entry = context->get_kernel(hostFunc); + printf( + "Calculate Maxium Active Block with function ptr=%p, blockSize=%d, " + "SMemSize=%d\n", + hostFunc, blockSize, dynamicSMemSize); + if (flags == cudaOccupancyDefault) { + // create kernel_info based on entry + dim3 gridDim(context->get_device()->get_gpgpu()->max_cta_per_core() * + context->get_device()->get_gpgpu()->get_config().num_shader()); + dim3 blockDim(blockSize); + kernel_info_t result(gridDim, blockDim, entry); + // if(entry == NULL){ + // *numBlocks = 1; + // return g_last_cudaError = cudaErrorUnknown; + //} + *numBlocks = context->get_device()->get_gpgpu()->get_max_cta(result); + printf("Maximum size is %d with gridDim %d and blockDim %d\n", *numBlocks, + gridDim.x, blockDim.x); return g_last_cudaError = cudaSuccess; + } else { + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; + } } -__host__ cudaError_t CUDARTAPI cudaGetDevicePropertiesInternal(struct cudaDeviceProp *prop, int device, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - _cuda_device_id *dev = ctx->GPGPUSim_Init(); - if (device <= dev->num_devices() ) { - *prop= *dev->get_prop(); - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorInvalidDevice; - } -} +#endif +__host__ cudaError_t CUDARTAPI cudaMemsetInternal( + void *mem, int c, size_t count, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + gpu->gpu_memset((size_t)mem, c, count); + return g_last_cudaError = cudaSuccess; +} + +// memset operation is done but i think its not async? +__host__ cudaError_t CUDARTAPI +cudaMemsetAsyncInternal(void *mem, int c, size_t count, cudaStream_t stream = 0, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("GPGPU-Sim PTX: WARNING: Asynchronous memset not supported (%s)\n", + __my_func__); + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + gpu->gpu_memset((size_t)mem, c, count); + return g_last_cudaError = cudaSuccess; +} + +cudaError_t cudaGLMapBufferObjectInternal(void **devPtr, GLuint bufferObj, + gpgpu_context *gpgpu_ctx = NULL) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } +#ifdef OPENGL_SUPPORT + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + GLint buffer_size = 0; + CUctx_st *context = GPGPUSim_Context(ctx); + + glbmap_entry_t *p = ctx->api->g_glbmap; + while (p && p->m_bufferObj != bufferObj) p = p->m_next; + if (p == NULL) { + glBindBuffer(GL_ARRAY_BUFFER, bufferObj); + glGetBufferParameteriv(GL_ARRAY_BUFFER, GL_BUFFER_SIZE, &buffer_size); + assert(buffer_size != 0); + *devPtr = context->get_device()->get_gpgpu()->gpu_malloc(buffer_size); + + // create entry and insert to front of list + glbmap_entry_t *n = (glbmap_entry_t *)calloc(1, sizeof(glbmap_entry_t)); + n->m_next = ctx->api->g_glbmap; + ctx->api->g_glbmap = n; + + // initialize entry + n->m_bufferObj = bufferObj; + n->m_devPtr = *devPtr; + n->m_size = buffer_size; + + p = n; + } else { + buffer_size = p->m_size; + *devPtr = p->m_devPtr; + } + + if (*devPtr) { + char *data = (char *)calloc(p->m_size, 1); + glGetBufferSubData(GL_ARRAY_BUFFER, 0, buffer_size, data); + memcpy_to_gpu((size_t)*devPtr, data, buffer_size); + free(data); + printf( + "GPGPU-Sim PTX: cudaGLMapBufferObject %zu bytes starting at 0x%llx..\n", + (size_t)buffer_size, (unsigned long long)*devPtr); + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorMemoryAllocation; + } -__host__ cudaError_t CUDARTAPI cudaChooseDeviceInternal(int *device, const struct cudaDeviceProp *prop, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - _cuda_device_id *dev = ctx->GPGPUSim_Init(); - *device = dev->get_id(); - return g_last_cudaError = cudaSuccess; + return g_last_cudaError = cudaSuccess; +#else + fflush(stdout); + fflush(stderr); + printf( + "GPGPU-Sim PTX: GPGPU-Sim support for OpenGL integration disabled -- " + "exiting\n"); + fflush(stdout); + exit(50); +#endif } -cudaError_t cudaSetupArgumentInternal(const void *arg, size_t size, size_t offset, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - gpgpusim_ptx_assert( !ctx->api->g_cuda_launch_stack.empty(), "empty launch stack" ); - kernel_config &config = ctx->api->g_cuda_launch_stack.back(); - config.set_arg(arg,size,offset); - printf("GPGPU-Sim PTX: Setting up arguments for %zu bytes starting at 0x%llx..\n",size, (unsigned long long) arg); - - return g_last_cudaError = cudaSuccess; +#if CUDART_VERSION >= 6050 +CUresult cuLinkAddFileInternal(CUlinkState state, CUjitInputType type, + const char *path, unsigned int numOptions, + CUjit_option *options, void **optionValues, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + static bool addedFile = false; + if (addedFile) { + printf( + "GPGPU-Sim PTX: ERROR: cuLinkAddFile does not support multiple " + "files\n"); + abort(); + } + + // blocking + assert(type == CU_JIT_INPUT_PTX); + CUctx_st *context = GPGPUSim_Context(ctx); + char *file = getenv("PTX_JIT_PATH"); + if (file == NULL) { + printf("GPGPU-Sim PTX: ERROR: PTX_JIT_PATH has not been set\n"); + abort(); + } + strcat(file, "/"); + strcat(file, path); + symbol_table *symtab = ctx->gpgpu_ptx_sim_load_ptx_from_filename(file); + std::string fname(path); + ctx->api->name_symtab[fname] = symtab; + context->add_binary(symtab, 1); + ctx->api->load_static_globals(symtab, STATIC_ALLOC_LIMIT, 0xFFFFFFFF, + context->get_device()->get_gpgpu()); + ctx->api->load_constants(symtab, STATIC_ALLOC_LIMIT, + context->get_device()->get_gpgpu()); + addedFile = true; + return CUDA_SUCCESS; } +#endif -cudaError_t cudaLaunchInternal( const char *hostFun, gpgpu_context* gpgpu_ctx = NULL ) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st* context = GPGPUSim_Context(ctx); - char *mode = getenv("PTX_SIM_MODE_FUNC"); - if( mode ) - sscanf(mode,"%u", &(ctx->func_sim->g_ptx_sim_mode)); - gpgpusim_ptx_assert( !ctx->api->g_cuda_launch_stack.empty(), "empty launch stack" ); - kernel_config config = ctx->api->g_cuda_launch_stack.back(); - { - dim3 gridDim = config.grid_dim(); - dim3 blockDim = config.block_dim(); - if (gridDim.x * gridDim.y * gridDim.z == 0 || blockDim.x * blockDim.y * blockDim.z == 0) - { - //can't launch - printf("can't launch a empty kernel\n"); - ctx->api->g_cuda_launch_stack.pop_back(); - return g_last_cudaError = cudaErrorInvalidConfiguration; - } - } - struct CUstream_st *stream = config.get_stream(); - - printf("\nGPGPU-Sim PTX: cudaLaunch for 0x%p (mode=%s) on stream %u\n", hostFun, - (ctx->func_sim->g_ptx_sim_mode)?"functional simulation":"performance simulation", stream?stream->get_uid():0 ); - kernel_info_t *grid = ctx->api->gpgpu_cuda_ptx_sim_init_grid(hostFun,config.get_args(),config.grid_dim(),config.block_dim(),context); - //do dynamic PDOM analysis for performance simulation scenario - std::string kname = grid->name(); - function_info *kernel_func_info = grid->entry(); - if (kernel_func_info->is_pdom_set()) { - printf("GPGPU-Sim PTX: PDOM analysis already done for %s \n", kname.c_str() ); - } else { - printf("GPGPU-Sim PTX: finding reconvergence points for \'%s\'...\n", kname.c_str() ); - kernel_func_info->do_pdom(); - kernel_func_info->set_pdom(); - } - dim3 gridDim = config.grid_dim(); - dim3 blockDim = config.block_dim(); - - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - checkpoint *g_checkpoint; - g_checkpoint = new checkpoint(); - class memory_space *global_mem; - global_mem = gpu->get_global_memory(); - - if(gpu->resume_option ==1 && (grid->get_uid()==gpu->resume_kernel)) - { - - char f1name[2048]; - snprintf(f1name,2048,"checkpoint_files/global_mem_%d.txt", grid->get_uid()); - - g_checkpoint->load_global_mem(global_mem, f1name); - for (int i=0;iresume_CTA;i++) - grid->increment_cta_id(); - } - if(gpu->resume_option==1 && (grid->get_uid()resume_kernel)) - { - char f1name[2048]; - snprintf(f1name,2048,"checkpoint_files/global_mem_%d.txt", grid->get_uid()); - - g_checkpoint->load_global_mem(global_mem, f1name); - printf("Skipping kernel %d as resuming from kernel %d\n",grid->get_uid(),gpu->resume_kernel ); - ctx->api->g_cuda_launch_stack.pop_back(); - return g_last_cudaError = cudaSuccess; - - } - if(gpu->checkpoint_option==1 && (grid->get_uid()>gpu->checkpoint_kernel)) - { - printf("Skipping kernel %d as checkpoint from kernel %d\n",grid->get_uid(),gpu->checkpoint_kernel ); - ctx->api->g_cuda_launch_stack.pop_back(); - return g_last_cudaError = cudaSuccess; - - } - printf("GPGPU-Sim PTX: pushing kernel \'%s\' to stream %u, gridDim= (%u,%u,%u) blockDim = (%u,%u,%u) \n", - kname.c_str(), stream?stream->get_uid():0, gridDim.x,gridDim.y,gridDim.z,blockDim.x,blockDim.y,blockDim.z ); - stream_operation op(grid,ctx->func_sim->g_ptx_sim_mode,stream); - ctx->the_gpgpusim->g_stream_manager->push(op); - ctx->api->g_cuda_launch_stack.pop_back(); - return g_last_cudaError = cudaSuccess; -} - -cudaError_t cudaMallocInternal(void **devPtr, size_t size, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st* context = GPGPUSim_Context(ctx); - *devPtr = context->get_device()->get_gpgpu()->gpu_malloc(size); - if(g_debug_execution >= 3){ - printf("GPGPU-Sim PTX: cudaMallocing %zu bytes starting at 0x%llx..\n",size, (unsigned long long) *devPtr); - ctx->api->g_mallocPtr_Size[(unsigned long long)*devPtr] = size; - } - if ( *devPtr ) { - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorMemoryAllocation; - } -} +#if (CUDART_VERSION >= 2010) -cudaError_t cudaMallocHostInternal(void **ptr, size_t size, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - *ptr = malloc(size); - if ( *ptr ) { - //track pinned memory size allocated in the host so that same amount of memory is also allocated in GPU. - ctx->api->pinned_memory_size[*ptr]=size; - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorMemoryAllocation; - } +cudaError_t cudaHostAllocInternal(void **pHost, size_t bytes, + unsigned int flags, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + *pHost = malloc(bytes); + // need to track the size allocated so that cudaHostGetDevicePointer() can + // function properly. + // TODO: vary this function behavior based on flags value (following nvidia + // documentation) + ctx->api->pinned_memory_size[*pHost] = bytes; + if (*pHost) + return g_last_cudaError = cudaSuccess; + else + return g_last_cudaError = cudaErrorMemoryAllocation; } -__host__ cudaError_t CUDARTAPI cudaMallocPitchInternal(void **devPtr, size_t *pitch, size_t width, size_t height, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - unsigned malloc_width_inbytes = width; - printf("GPGPU-Sim PTX: cudaMallocPitch (width = %d)\n", malloc_width_inbytes); - CUctx_st* context = GPGPUSim_Context(ctx); - *devPtr = context->get_device()->get_gpgpu()->gpu_malloc(malloc_width_inbytes*height); - pitch[0] = malloc_width_inbytes; - if ( *devPtr ) { - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorMemoryAllocation; - } -} - -cudaError_t cudaHostGetDevicePointerInternal(void **pDevice, void *pHost, unsigned int flags, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //only cpu memory allocation happens in cudaHostAlloc. Linking with device pointer to pinned memory happens here. - //TODO: once kernel is executed, the contents in global pointer of GPU must be copied back to CPU host pointer! - flags=0; - CUctx_st* context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - std::map::const_iterator i = ctx->api->pinned_memory_size.find(pHost); - assert(i != ctx->api->pinned_memory_size.end()); - size_t size = i->second; - *pDevice = gpu->gpu_malloc(size); - if(g_debug_execution >= 3){ - printf("GPGPU-Sim PTX: cudaMallocing %zu bytes starting at 0x%llx..\n",size, (unsigned long long) *pDevice); - ctx->api->g_mallocPtr_Size[(unsigned long long)*pDevice] = size; - } - if ( *pDevice ) { - ctx->api->pinned_memory[pHost]=pDevice; - //Copy contents in cpu to gpu - gpu->memcpy_to_gpu((size_t)*pDevice,pHost,size); - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorMemoryAllocation; - } -} +#endif -__host__ cudaError_t CUDARTAPI cudaMallocArrayInternal(struct cudaArray **array, const struct cudaChannelFormatDesc *desc, size_t width, size_t height __dv(1), gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - unsigned size = width * height * ((desc->x + desc->y + desc->z + desc->w)/8); - CUctx_st* context = GPGPUSim_Context(ctx); - (*array) = (struct cudaArray*) malloc(sizeof(struct cudaArray)); - (*array)->desc = *desc; - (*array)->width = width; - (*array)->height = height; - (*array)->size = size; - (*array)->dimensions = 2; - ((*array)->devPtr32)= (int) (long long)context->get_device()->get_gpgpu()->gpu_mallocarray(size); - printf("GPGPU-Sim PTX: cudaMallocArray: devPtr32 = %d\n", ((*array)->devPtr32)); - ((*array)->devPtr) = (void*) (long long) ((*array)->devPtr32); - if ( ((*array)->devPtr) ) { - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorMemoryAllocation; - } -} - -__host__ cudaError_t CUDARTAPI cudaMemcpyInternal(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //CUctx_st *context = GPGPUSim_Context(); - //gpgpu_t *gpu = context->get_device()->get_gpgpu(); - if(g_debug_execution >= 3) - printf("GPGPU-Sim PTX: cudaMemcpy(): devPtr = %p\n", dst); - if( kind == cudaMemcpyHostToDevice ) - ctx->the_gpgpusim->g_stream_manager->push( stream_operation(src,(size_t)dst,count,0) ); - else if( kind == cudaMemcpyDeviceToHost ) - ctx->the_gpgpusim->g_stream_manager->push( stream_operation((size_t)src,dst,count,0) ); - else if( kind == cudaMemcpyDeviceToDevice ) - ctx->the_gpgpusim->g_stream_manager->push( stream_operation((size_t)src,(size_t)dst,count,0) ); - else if ( kind == cudaMemcpyDefault ) { - if ((size_t)src >= GLOBAL_HEAP_START) { - if ((size_t)dst >= GLOBAL_HEAP_START) - ctx->the_gpgpusim->g_stream_manager->push( stream_operation((size_t)src,(size_t)dst,count,0) ); // device to device - else - ctx->the_gpgpusim->g_stream_manager->push( stream_operation((size_t)src,dst,count,0) ); // device to host - } - else { - if ((size_t)dst >= GLOBAL_HEAP_START) - ctx->the_gpgpusim->g_stream_manager->push( stream_operation(src,(size_t)dst,count,0) ); - else { - printf("GPGPU-Sim PTX: cudaMemcpy - ERROR : unsupported transfer: host to host\n"); - abort(); - } - } - } - else { - printf("GPGPU-Sim PTX: cudaMemcpy - ERROR : unsupported cudaMemcpyKind\n"); - abort(); - } - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaMemcpyToArrayInternal(struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, size_t count, enum cudaMemcpyKind kind, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - size_t size = count; - printf("GPGPU-Sim PTX: cudaMemcpyToArray\n"); - if( kind == cudaMemcpyHostToDevice ) - gpu->memcpy_to_gpu( (size_t)(dst->devPtr), src, size); - else if( kind == cudaMemcpyDeviceToHost ) - gpu->memcpy_from_gpu( dst->devPtr, (size_t)src, size); - else if( kind == cudaMemcpyDeviceToDevice ) - gpu->memcpy_gpu_to_gpu( (size_t)(dst->devPtr), (size_t)src, size); - else { - printf("GPGPU-Sim PTX: cudaMemcpyToArray - ERROR : unsupported cudaMemcpyKind\n"); - abort(); - } - dst->devPtr32 = (unsigned) (size_t)(dst->devPtr); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaMemcpy2DInternal(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - size_t size = spitch*height; - gpgpusim_ptx_assert( (dpitch==spitch), "different src and dst pitch not supported yet" ); - if( kind == cudaMemcpyHostToDevice ) - gpu->memcpy_to_gpu( (size_t)dst, src, size ); - else if( kind == cudaMemcpyDeviceToHost ) - gpu->memcpy_from_gpu( dst, (size_t)src, size ); - else if( kind == cudaMemcpyDeviceToDevice ) - gpu->memcpy_gpu_to_gpu( (size_t)dst, (size_t)src, size); - else { - printf("GPGPU-Sim PTX: cudaMemcpy2D - ERROR : unsupported cudaMemcpyKind\n"); - abort(); - } - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaMemcpy2DToArrayInternal(struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - size_t size = spitch*height; - size_t channel_size = dst->desc.w+dst->desc.x+dst->desc.y+dst->desc.z; - gpgpusim_ptx_assert( ((channel_size%8) == 0), "none byte multiple destination channel size not supported (sz=%u)", channel_size ); - unsigned elem_size = channel_size/8; - gpgpusim_ptx_assert( (dst->dimensions==2), "copy to none 2D array not supported" ); - gpgpusim_ptx_assert( (wOffset==0), "non-zero wOffset not yet supported" ); - gpgpusim_ptx_assert( (hOffset==0), "non-zero hOffset not yet supported" ); - gpgpusim_ptx_assert( (dst->height == (int)height), "partial copy not supported" ); - gpgpusim_ptx_assert( (elem_size*dst->width == width), "partial copy not supported" ); - gpgpusim_ptx_assert( (spitch == width), "spitch != width not supported" ); - if( kind == cudaMemcpyHostToDevice ) - gpu->memcpy_to_gpu( (size_t)(dst->devPtr), src, size); - else if( kind == cudaMemcpyDeviceToHost ) - gpu->memcpy_from_gpu( dst->devPtr, (size_t)src, size); - else if( kind == cudaMemcpyDeviceToDevice ) - gpu->memcpy_gpu_to_gpu( (size_t)dst->devPtr, (size_t)src, size); - else { - printf("GPGPU-Sim PTX: cudaMemcpy2D - ERROR : unsupported cudaMemcpyKind\n"); - abort(); - } - dst->devPtr32 = (unsigned) (size_t)(dst->devPtr); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaMemcpyToSymbolInternal(const char *symbol, const void *src, size_t count, size_t offset __dv(0), enum cudaMemcpyKind kind __dv(cudaMemcpyHostToDevice), gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //CUctx_st *context = GPGPUSim_Context(); - assert(kind == cudaMemcpyHostToDevice); - printf("GPGPU-Sim PTX: cudaMemcpyToSymbol: symbol = %p\n", symbol); - //stream_operation( const char *symbol, const void *src, size_t count, size_t offset ) - ctx->the_gpgpusim->g_stream_manager->push( stream_operation(src,symbol,count,offset,0) ); - //gpgpu_ptx_sim_memcpy_symbol(symbol,src,count,offset,1,context->get_device()->get_gpgpu()); - return g_last_cudaError = cudaSuccess; -} +size_t getMaxThreadsPerBlock(struct cudaFuncAttributes *attr, + gpgpu_context *ctx) { + _cuda_device_id *dev = ctx->GPGPUSim_Init(); + struct cudaDeviceProp prop; + prop = *dev->get_prop(); -__host__ cudaError_t CUDARTAPI cudaMemcpyFromSymbolInternal(void *dst, const char *symbol, size_t count, size_t offset __dv(0), enum cudaMemcpyKind kind __dv(cudaMemcpyDeviceToHost), gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //CUctx_st *context = GPGPUSim_Context(); - assert(kind == cudaMemcpyDeviceToHost); - printf("GPGPU-Sim PTX: cudaMemcpyFromSymbol: symbol = %p\n", symbol); - ctx->the_gpgpusim->g_stream_manager->push( stream_operation(symbol,dst,count,offset,0) ); - //gpgpu_ptx_sim_memcpy_symbol(symbol,dst,count,offset,0,context->get_device()->get_gpgpu()); - return g_last_cudaError = cudaSuccess; -} + size_t max = prop.maxThreadsPerBlock; -__host__ cudaError_t CUDARTAPI cudaMemcpyAsyncInternal(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - struct CUstream_st *s = (struct CUstream_st *)stream; - switch( kind ) { - case cudaMemcpyHostToDevice: ctx->the_gpgpusim->g_stream_manager->push( stream_operation(src,(size_t)dst,count,s) ); break; - case cudaMemcpyDeviceToHost: ctx->the_gpgpusim->g_stream_manager->push( stream_operation((size_t)src,dst,count,s) ); break; - case cudaMemcpyDeviceToDevice: ctx->the_gpgpusim->g_stream_manager->push( stream_operation((size_t)src,(size_t)dst,count,s) ); break; - default: - abort(); - } - return g_last_cudaError = cudaSuccess; -} + if (attr->numRegs && (prop.regsPerBlock / attr->numRegs) < max) + max = prop.regsPerBlock / attr->numRegs; + if (attr->sharedSizeBytes && + (prop.sharedMemPerBlock / attr->sharedSizeBytes) < max) + max = prop.sharedMemPerBlock / attr->sharedSizeBytes; -#if (CUDART_VERSION >= 8000) -cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlagsInternal(int* numBlocks, const char *hostFunc, int blockSize, size_t dynamicSMemSize, unsigned int flags, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - printf("GPGPU-Sim PTX: cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags %p\n", hostFunc); - CUctx_st *context = GPGPUSim_Context(ctx); - function_info *entry = context->get_kernel(hostFunc); - printf("Calculate Maxium Active Block with function ptr=%p, blockSize=%d, SMemSize=%d\n", hostFunc, blockSize, dynamicSMemSize); - if (flags == cudaOccupancyDefault) { - //create kernel_info based on entry - dim3 gridDim(context->get_device()->get_gpgpu()->max_cta_per_core() - * context->get_device()->get_gpgpu()->get_config().num_shader()); - dim3 blockDim(blockSize); - kernel_info_t result(gridDim, blockDim, entry); - //if(entry == NULL){ - // *numBlocks = 1; - // return g_last_cudaError = cudaErrorUnknown; - //} - *numBlocks = context->get_device()->get_gpgpu()->get_max_cta(result); - printf("Maximum size is %d with gridDim %d and blockDim %d\n", *numBlocks, gridDim.x, blockDim.x); - return g_last_cudaError = cudaSuccess; - } else { - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; - } + return max; } +cudaError_t CUDARTAPI cudaFuncGetAttributesInternal( + struct cudaFuncAttributes *attr, const char *hostFun, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + function_info *entry = context->get_kernel(hostFun); + if (entry) { + const struct gpgpu_ptx_sim_info *kinfo = entry->get_kernel_info(); + attr->sharedSizeBytes = kinfo->smem; + attr->constSizeBytes = kinfo->cmem; + attr->localSizeBytes = kinfo->lmem; + attr->numRegs = kinfo->regs; + if (kinfo->maxthreads > 0) + attr->maxThreadsPerBlock = kinfo->maxthreads; + else + attr->maxThreadsPerBlock = getMaxThreadsPerBlock(attr, ctx); +#if CUDART_VERSION >= 3000 + attr->ptxVersion = kinfo->ptx_version; + attr->binaryVersion = kinfo->sm_target; #endif - -__host__ cudaError_t CUDARTAPI cudaMemsetInternal(void *mem, int c, size_t count, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - gpu->gpu_memset((size_t)mem, c, count); - return g_last_cudaError = cudaSuccess; + } + return g_last_cudaError = cudaSuccess; } -//memset operation is done but i think its not async? -__host__ cudaError_t CUDARTAPI cudaMemsetAsyncInternal(void *mem, int c, size_t count, cudaStream_t stream=0, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); +#if (CUDART_VERSION > 5000) +__host__ cudaError_t CUDARTAPI +cudaDeviceGetAttributeInternal(int *value, enum cudaDeviceAttr attr, int device, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + + const struct cudaDeviceProp *prop; + _cuda_device_id *dev = ctx->GPGPUSim_Init(); + + if (device <= dev->num_devices()) { + prop = dev->get_prop(); + switch (attr) { + case 1: + *value = prop->maxThreadsPerBlock; + break; + case 2: + *value = prop->maxThreadsDim[0]; + break; + case 3: + *value = prop->maxThreadsDim[1]; + break; + case 4: + *value = prop->maxThreadsDim[2]; + break; + case 5: + *value = prop->maxGridSize[0]; + break; + case 6: + *value = prop->maxGridSize[1]; + break; + case 7: + *value = prop->maxGridSize[2]; + break; + case 8: + *value = prop->sharedMemPerBlock; + break; + case 9: + *value = prop->totalConstMem; + break; + case 10: + *value = prop->warpSize; + break; + case 11: + *value = 16; // dummy value + break; + case 12: + *value = prop->regsPerBlock; + break; + case 13: + *value = 1480000; // for 1080ti + break; + case 14: + *value = prop->textureAlignment; + break; + case 15: + *value = 0; + break; + case 16: + *value = prop->multiProcessorCount; + break; + case 17: + case 18: + case 19: + *value = 0; + break; + case 21: + case 22: + case 23: + case 24: + case 25: + case 26: + case 27: + case 28: + case 42: + case 45: + case 46: + case 47: + case 48: + case 49: + case 52: + case 53: + case 55: + case 56: + case 57: + case 58: + case 59: + case 60: + case 61: + case 62: + case 63: + case 64: + case 66: + case 67: + case 69: + case 70: + case 71: + case 73: + case 74: + case 77: + *value = 1000; // dummy value + break; + case 29: + case 43: + case 54: + case 65: + case 68: + case 72: + *value = 10; // dummy value + break; + case 30: + case 51: + *value = 128; // dummy value + break; + case 31: + *value = 1; + break; + case 32: + *value = 0; + break; + case 33: + case 50: + *value = 0; // dummy value + break; + case 34: + *value = 0; + break; + case 35: + *value = 0; + break; + case 36: + *value = 1250000; // CK value for 1080ti + break; + case 37: + *value = 352; // value for 1080ti + break; + case 38: + *value = 3000000; // value for 1080ti + break; + case 39: + *value = dev->get_gpgpu()->threads_per_core(); + break; + case 40: + *value = 0; + break; + case 41: + *value = 0; + break; + case 75: // cudaDevAttrComputeCapabilityMajor + *value = prop->major; + break; + case 76: // cudaDevAttrComputeCapabilityMinor + *value = prop->minor; + break; + case 78: + *value = 0; // TODO: as of now, we dont support stream priorities. + break; + case 79: + *value = 0; + break; + case 80: + *value = 0; + break; +#if (CUDART_VERSION > 5050) + case 81: + *value = prop->sharedMemPerMultiprocessor; + break; + case 82: + *value = prop->regsPerMultiprocessor; + break; +#endif + case 83: + case 84: + case 85: + case 86: + *value = 0; + break; + case 87: + *value = 4; // dummy value + break; + case 88: + case 89: + *value = 0; + break; + default: + printf("ERROR: Attribute number %d unimplemented \n", attr); + abort(); } - printf("GPGPU-Sim PTX: WARNING: Asynchronous memset not supported (%s)\n", __my_func__); - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - gpu->gpu_memset((size_t)mem, c, count); - return g_last_cudaError = cudaSuccess; + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorInvalidDevice; + } } +#endif -cudaError_t cudaGLMapBufferObjectInternal(void** devPtr, GLuint bufferObj, gpgpu_context* gpgpu_ctx = NULL) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } -#ifdef OPENGL_SUPPORT - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - GLint buffer_size=0; - CUctx_st* context = GPGPUSim_Context(ctx); - - glbmap_entry_t *p = ctx->api->g_glbmap; - while ( p && p->m_bufferObj != bufferObj ) - p = p->m_next; - if ( p == NULL ) { - glBindBuffer(GL_ARRAY_BUFFER,bufferObj); - glGetBufferParameteriv(GL_ARRAY_BUFFER,GL_BUFFER_SIZE,&buffer_size); - assert( buffer_size != 0 ); - *devPtr = context->get_device()->get_gpgpu()->gpu_malloc(buffer_size); - - // create entry and insert to front of list - glbmap_entry_t *n = (glbmap_entry_t *) calloc(1,sizeof(glbmap_entry_t)); - n->m_next = ctx->api->g_glbmap; - ctx->api->g_glbmap = n; - - // initialize entry - n->m_bufferObj = bufferObj; - n->m_devPtr = *devPtr; - n->m_size = buffer_size; - - p = n; - } else { - buffer_size = p->m_size; - *devPtr = p->m_devPtr; - } - - if ( *devPtr ) { - char *data = (char *) calloc(p->m_size,1); - glGetBufferSubData(GL_ARRAY_BUFFER,0,buffer_size,data); - memcpy_to_gpu( (size_t) *devPtr, data, buffer_size ); - free(data); - printf("GPGPU-Sim PTX: cudaGLMapBufferObject %zu bytes starting at 0x%llx..\n", (size_t)buffer_size, - (unsigned long long) *devPtr); - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorMemoryAllocation; - } - - return g_last_cudaError = cudaSuccess; +__host__ cudaError_t CUDARTAPI cudaBindTextureInternal( + size_t *offset, const struct textureReference *texref, const void *devPtr, + const struct cudaChannelFormatDesc *desc, size_t size __dv(UINT_MAX), + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + printf( + "GPGPU-Sim PTX: in cudaBindTexture: sizeof(struct textureReference) = " + "%zu\n", + sizeof(struct textureReference)); + struct cudaArray *array; + array = (struct cudaArray *)malloc(sizeof(struct cudaArray)); + array->desc = *desc; + array->size = size; + array->width = size; + array->height = 1; + array->dimensions = 1; + array->devPtr = (void *)devPtr; + array->devPtr32 = (int)(long long)devPtr; + offset = 0; + printf("GPGPU-Sim PTX: size = %zu\n", size); + printf("GPGPU-Sim PTX: texref = %p, array = %p\n", texref, array); + printf("GPGPU-Sim PTX: devPtr32 = %x\n", array->devPtr32); + printf("GPGPU-Sim PTX: Name corresponding to textureReference: %s\n", + gpu->gpgpu_ptx_sim_findNamefromTexture(texref)); + printf("GPGPU-Sim PTX: ChannelFormatDesc: x=%d, y=%d, z=%d, w=%d\n", + desc->x, desc->y, desc->z, desc->w); + printf("GPGPU-Sim PTX: Texture Normalized? = %d\n", texref->normalized); + gpu->gpgpu_ptx_sim_bindTextureToArray(texref, array); + devPtr = (void *)(long long)array->devPtr32; + printf("GPGPU-Sim PTX: devPtr = %p\n", devPtr); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaBindTextureToArrayInternal( + const struct textureReference *texref, const struct cudaArray *array, + const struct cudaChannelFormatDesc *desc, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + printf("GPGPU-Sim PTX: in cudaBindTextureToArray: %p %p\n", texref, array); + printf("GPGPU-Sim PTX: devPtr32 = %x\n", array->devPtr32); + printf("GPGPU-Sim PTX: Name corresponding to textureReference: %s\n", + gpu->gpgpu_ptx_sim_findNamefromTexture(texref)); + printf("GPGPU-Sim PTX: Texture Normalized? = %d\n", texref->normalized); + gpu->gpgpu_ptx_sim_bindTextureToArray(texref, array); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaUnbindTextureInternal( + const struct textureReference *texref, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + printf( + "GPGPU-Sim PTX: in cudaUnbindTexture: sizeof(struct textureReference) = " + "%zu\n", + sizeof(struct textureReference)); + printf("GPGPU-Sim PTX: Name corresponding to textureReference: %s\n", + gpu->gpgpu_ptx_sim_findNamefromTexture(texref)); + + gpu->gpgpu_ptx_sim_unbindTexture(texref); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaLaunchKernelInternal( + const char *hostFun, dim3 gridDim, dim3 blockDim, const void **args, + size_t sharedMem, cudaStream_t stream, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + function_info *entry = context->get_kernel(hostFun); +#if CUDART_VERSION < 10000 + cudaConfigureCallInternal(gridDim, blockDim, sharedMem, stream, ctx); +#endif + for (unsigned i = 0; i < entry->num_args(); i++) { + std::pair p = entry->get_param_config(i); + cudaSetupArgumentInternal(args[i], p.first, p.second); + } + + cudaLaunchInternal(hostFun); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaStreamCreateInternal( + cudaStream_t *stream, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("GPGPU-Sim PTX: cudaStreamCreate\n"); +#if (CUDART_VERSION >= 3000) + *stream = new struct CUstream_st(); + ctx->the_gpgpusim->g_stream_manager->add_stream(*stream); #else - fflush(stdout); - fflush(stderr); - printf("GPGPU-Sim PTX: GPGPU-Sim support for OpenGL integration disabled -- exiting\n"); - fflush(stdout); - exit(50); + *stream = 0; + printf( + "GPGPU-Sim PTX: WARNING: Asynchronous kernel execution not supported " + "(%s)\n", + __my_func__); #endif -} - -#if CUDART_VERSION >= 6050 -CUresult -cuLinkAddFileInternal(CUlinkState state, CUjitInputType type, const char *path, - unsigned int numOptions, CUjit_option *options, void **optionValues, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - static bool addedFile = false; - if (addedFile){ - printf("GPGPU-Sim PTX: ERROR: cuLinkAddFile does not support multiple files\n"); - abort(); - } - - //blocking - assert(type==CU_JIT_INPUT_PTX); - CUctx_st *context = GPGPUSim_Context(ctx); - char *file = getenv("PTX_JIT_PATH"); - if(file==NULL){ - printf("GPGPU-Sim PTX: ERROR: PTX_JIT_PATH has not been set\n"); - abort(); - } - strcat(file,"/"); - strcat(file,path); - symbol_table *symtab = ctx->gpgpu_ptx_sim_load_ptx_from_filename( file ); - std::string fname(path); - ctx->api->name_symtab[fname] = symtab; - context->add_binary(symtab, 1); - ctx->api->load_static_globals(symtab,STATIC_ALLOC_LIMIT,0xFFFFFFFF,context->get_device()->get_gpgpu()); - ctx->api->load_constants(symtab,STATIC_ALLOC_LIMIT,context->get_device()->get_gpgpu()); - addedFile = true; - return CUDA_SUCCESS; -} -#endif - -#if (CUDART_VERSION >= 2010) - -cudaError_t cudaHostAllocInternal(void **pHost, size_t bytes, unsigned int flags, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - *pHost = malloc(bytes); - //need to track the size allocated so that cudaHostGetDevicePointer() can function properly. - //TODO: vary this function behavior based on flags value (following nvidia documentation) - ctx->api->pinned_memory_size[*pHost]=bytes; - if( *pHost ) - return g_last_cudaError = cudaSuccess; - else - return g_last_cudaError = cudaErrorMemoryAllocation; -} - -#endif - -size_t getMaxThreadsPerBlock(struct cudaFuncAttributes *attr, gpgpu_context *ctx) { - _cuda_device_id *dev = ctx->GPGPUSim_Init(); - struct cudaDeviceProp prop; - - prop = *dev->get_prop(); - - size_t max = prop.maxThreadsPerBlock; - - if (attr->numRegs && (prop.regsPerBlock / attr->numRegs) < max) - max = prop.regsPerBlock / attr->numRegs; - - if (attr->sharedSizeBytes && (prop.sharedMemPerBlock / attr->sharedSizeBytes) < max) - max = prop.sharedMemPerBlock / attr->sharedSizeBytes; - - return max; -} - -cudaError_t CUDARTAPI cudaFuncGetAttributesInternal(struct cudaFuncAttributes *attr, const char *hostFun, gpgpu_context* gpgpu_ctx = NULL ) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - function_info *entry = context->get_kernel(hostFun); - if( entry ) { - const struct gpgpu_ptx_sim_info *kinfo = entry->get_kernel_info(); - attr->sharedSizeBytes = kinfo->smem; - attr->constSizeBytes = kinfo->cmem; - attr->localSizeBytes = kinfo->lmem; - attr->numRegs = kinfo->regs; - if(kinfo->maxthreads > 0) - attr->maxThreadsPerBlock = kinfo->maxthreads; - else - attr->maxThreadsPerBlock = getMaxThreadsPerBlock(attr, ctx); -#if CUDART_VERSION >= 3000 - attr->ptxVersion = kinfo->ptx_version; - attr->binaryVersion = kinfo->sm_target; -#endif - } - return g_last_cudaError = cudaSuccess; -} - -#if (CUDART_VERSION > 5000) -__host__ cudaError_t CUDARTAPI cudaDeviceGetAttributeInternal(int *value, enum cudaDeviceAttr attr, int device, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - - const struct cudaDeviceProp *prop; - _cuda_device_id *dev = ctx->GPGPUSim_Init(); - - if (device <= dev->num_devices() ) { - prop = dev->get_prop(); - switch (attr) { - case 1: - *value= prop->maxThreadsPerBlock; - break; - case 2: - *value= prop->maxThreadsDim[0]; - break; - case 3: - *value= prop->maxThreadsDim[1]; - break; - case 4: - *value= prop->maxThreadsDim[2]; - break; - case 5: - *value= prop->maxGridSize[0]; - break; - case 6: - *value= prop->maxGridSize[1]; - break; - case 7: - *value= prop->maxGridSize[2]; - break; - case 8: - *value= prop->sharedMemPerBlock; - break; - case 9: - *value= prop->totalConstMem; - break; - case 10: - *value= prop->warpSize; - break; - case 11: - *value= 16;//dummy value - break; - case 12: - *value= prop->regsPerBlock; - break; - case 13: - *value= 1480000;//for 1080ti - break; - case 14: - *value= prop->textureAlignment ; - break; - case 15: - *value = 0; - break; - case 16: - *value= prop->multiProcessorCount ; - break; - case 17: - case 18: - case 19: - *value = 0; - break; - case 21: - case 22: - case 23: - case 24: - case 25: - case 26: - case 27: - case 28: - case 42: - case 45: - case 46: - case 47: - case 48: - case 49: - case 52: - case 53: - case 55: - case 56: - case 57: - case 58: - case 59: - case 60: - case 61: - case 62: - case 63: - case 64: - case 66: - case 67: - case 69: - case 70: - case 71: - case 73: - case 74: - case 77: - *value = 1000;//dummy value - break; - case 29: - case 43: - case 54: - case 65: - case 68: - case 72: - *value = 10;//dummy value - break; - case 30: - case 51: - *value = 128;//dummy value - break; - case 31: - *value = 1; - break; - case 32: - *value = 0; - break; - case 33: - case 50: - *value = 0;//dummy value - break; - case 34: - *value= 0; - break; - case 35: - *value = 0; - break; - case 36: - *value = 1250000;//CK value for 1080ti - break; - case 37: - *value = 352;//value for 1080ti - break; - case 38: - *value = 3000000;//value for 1080ti - break; - case 39: - *value= dev->get_gpgpu()->threads_per_core(); - break; - case 40: - *value= 0; - break; - case 41: - *value= 0; - break; - case 75://cudaDevAttrComputeCapabilityMajor - *value= prop->major ; - break; - case 76://cudaDevAttrComputeCapabilityMinor - *value= prop->minor ; - break; - case 78: - *value= 0 ; //TODO: as of now, we dont support stream priorities. - break; - case 79: - *value= 0; - break; - case 80: - *value= 0; - break; - #if (CUDART_VERSION > 5050) - case 81: - *value= prop->sharedMemPerMultiprocessor; - break; - case 82: - *value= prop->regsPerMultiprocessor; - break; - #endif - case 83: - case 84: - case 85: - case 86: - *value= 0; - break; - case 87: - *value= 4;//dummy value - break; - case 88: - case 89: - *value= 0; - break; - default: - printf("ERROR: Attribute number %d unimplemented \n",attr); - abort(); - } - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorInvalidDevice; - } -} -#endif - -__host__ cudaError_t CUDARTAPI cudaBindTextureInternal(size_t *offset, - const struct textureReference *texref, - const void *devPtr, - const struct cudaChannelFormatDesc *desc, - size_t size __dv(UINT_MAX), - gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - printf("GPGPU-Sim PTX: in cudaBindTexture: sizeof(struct textureReference) = %zu\n", sizeof(struct textureReference)); - struct cudaArray *array; - array = (struct cudaArray*) malloc(sizeof(struct cudaArray)); - array->desc = *desc; - array->size = size; - array->width = size; - array->height = 1; - array->dimensions = 1; - array->devPtr = (void*)devPtr; - array->devPtr32 = (int)(long long)devPtr; - offset = 0; - printf("GPGPU-Sim PTX: size = %zu\n", size); - printf("GPGPU-Sim PTX: texref = %p, array = %p\n", texref, array); - printf("GPGPU-Sim PTX: devPtr32 = %x\n", array->devPtr32); - printf("GPGPU-Sim PTX: Name corresponding to textureReference: %s\n", gpu->gpgpu_ptx_sim_findNamefromTexture(texref)); - printf("GPGPU-Sim PTX: ChannelFormatDesc: x=%d, y=%d, z=%d, w=%d\n", desc->x, desc->y, desc->z, desc->w); - printf("GPGPU-Sim PTX: Texture Normalized? = %d\n", texref->normalized); - gpu->gpgpu_ptx_sim_bindTextureToArray(texref, array); - devPtr = (void*)(long long)array->devPtr32; - printf("GPGPU-Sim PTX: devPtr = %p\n", devPtr); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaBindTextureToArrayInternal(const struct textureReference *texref, const struct cudaArray *array, const struct cudaChannelFormatDesc *desc, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - printf("GPGPU-Sim PTX: in cudaBindTextureToArray: %p %p\n", texref, array); - printf("GPGPU-Sim PTX: devPtr32 = %x\n", array->devPtr32); - printf("GPGPU-Sim PTX: Name corresponding to textureReference: %s\n", gpu->gpgpu_ptx_sim_findNamefromTexture(texref)); - printf("GPGPU-Sim PTX: Texture Normalized? = %d\n", texref->normalized); - gpu->gpgpu_ptx_sim_bindTextureToArray(texref, array); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaUnbindTextureInternal(const struct textureReference *texref, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - printf("GPGPU-Sim PTX: in cudaUnbindTexture: sizeof(struct textureReference) = %zu\n", sizeof(struct textureReference)); - printf("GPGPU-Sim PTX: Name corresponding to textureReference: %s\n", gpu->gpgpu_ptx_sim_findNamefromTexture(texref)); - - gpu->gpgpu_ptx_sim_unbindTexture(texref); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaLaunchKernelInternal( const char* hostFun, dim3 gridDim, dim3 blockDim, const void** args, size_t sharedMem, cudaStream_t stream, gpgpu_context* gpgpu_ctx = NULL ) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - function_info *entry = context->get_kernel(hostFun); -#if CUDART_VERSION < 10000 - cudaConfigureCallInternal(gridDim, blockDim, sharedMem, stream, ctx); -#endif - for(unsigned i = 0; i < entry->num_args(); i++){ - std::pair p = entry->get_param_config(i); - cudaSetupArgumentInternal(args[i], p.first, p.second); - } - - cudaLaunchInternal(hostFun); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaStreamCreateInternal(cudaStream_t *stream, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("GPGPU-Sim PTX: cudaStreamCreate\n"); -#if (CUDART_VERSION >= 3000) - *stream = new struct CUstream_st(); - ctx->the_gpgpusim->g_stream_manager->add_stream(*stream); -#else - *stream = 0; - printf("GPGPU-Sim PTX: WARNING: Asynchronous kernel execution not supported (%s)\n", __my_func__); -#endif - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaStreamDestroyInternal(cudaStream_t stream, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaStreamDestroyInternal( + cudaStream_t stream, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } #if (CUDART_VERSION >= 3000) - //per-stream synchronization required for application using external libraries without explicit synchronization in the code to - //avoid the stream_manager from spinning forever to destroy non-empty streams without making any forward progress. - stream->synchronize(); - ctx->the_gpgpusim->g_stream_manager->destroy_stream(stream); + // per-stream synchronization required for application using external + // libraries without explicit synchronization in the code to avoid the + // stream_manager from spinning forever to destroy non-empty streams without + // making any forward progress. + stream->synchronize(); + ctx->the_gpgpusim->g_stream_manager->destroy_stream(stream); #endif - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaStreamSynchronizeInternal(cudaStream_t stream, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaStreamSynchronizeInternal( + cudaStream_t stream, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } #if (CUDART_VERSION >= 3000) - if( stream == NULL ) - ctx->synchronize(); - return g_last_cudaError = cudaSuccess; - stream->synchronize(); + if (stream == NULL) ctx->synchronize(); + return g_last_cudaError = cudaSuccess; + stream->synchronize(); #else - printf("GPGPU-Sim PTX: WARNING: Asynchronous kernel execution not supported (%s)\n", __my_func__); + printf( + "GPGPU-Sim PTX: WARNING: Asynchronous kernel execution not supported " + "(%s)\n", + __my_func__); #endif - return g_last_cudaError = cudaSuccess; + return g_last_cudaError = cudaSuccess; } void __cudaRegisterTextureInternal( - void **fatCubinHandle, - const struct textureReference *hostVar, - const void **deviceAddress, - const char *deviceName, - int dim, - int norm, - int ext, - gpgpu_context* gpgpu_ctx = NULL -) //passes in a newly created textureReference -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - std::string devStr (deviceName); - #if (CUDART_VERSION > 4020) - if (devStr.size() > 2 && devStr.data()[0] == ':' && devStr.data()[1] == ':') - devStr = devStr.replace(0, 2, ""); - #endif - CUctx_st *context = GPGPUSim_Context(ctx); - gpgpu_t *gpu = context->get_device()->get_gpgpu(); - printf("GPGPU-Sim PTX: in __cudaRegisterTexture:\n"); - gpu->gpgpu_ptx_sim_bindNameToTexture(devStr.data(), hostVar, dim, norm, ext); - printf("GPGPU-Sim PTX: int dim = %d\n", dim); - printf("GPGPU-Sim PTX: int norm = %d\n", norm); - printf("GPGPU-Sim PTX: int ext = %d\n", ext); - printf("GPGPU-Sim PTX: Execution warning: Not finished implementing \"%s\"\n", __my_func__ ); -} - -cudaError_t cudaGLUnmapBufferObjectInternal(GLuint bufferObj, gpgpu_context* gpgpu_ctx = NULL) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } + void **fatCubinHandle, const struct textureReference *hostVar, + const void **deviceAddress, const char *deviceName, int dim, int norm, + int ext, + gpgpu_context *gpgpu_ctx = + NULL) // passes in a newly created textureReference +{ + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + std::string devStr(deviceName); +#if (CUDART_VERSION > 4020) + if (devStr.size() > 2 && devStr.data()[0] == ':' && devStr.data()[1] == ':') + devStr = devStr.replace(0, 2, ""); +#endif + CUctx_st *context = GPGPUSim_Context(ctx); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + printf("GPGPU-Sim PTX: in __cudaRegisterTexture:\n"); + gpu->gpgpu_ptx_sim_bindNameToTexture(devStr.data(), hostVar, dim, norm, ext); + printf("GPGPU-Sim PTX: int dim = %d\n", dim); + printf("GPGPU-Sim PTX: int norm = %d\n", norm); + printf("GPGPU-Sim PTX: int ext = %d\n", ext); + printf( + "GPGPU-Sim PTX: Execution warning: Not finished implementing \"%s\"\n", + __my_func__); +} + +cudaError_t cudaGLUnmapBufferObjectInternal(GLuint bufferObj, + gpgpu_context *gpgpu_ctx = NULL) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } #ifdef OPENGL_SUPPORT - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - CUctx_st* ctx = GPGPUSim_Context(ctx); - glbmap_entry_t *p = ctx->api->g_glbmap; - while ( p && p->m_bufferObj != bufferObj ) - p = p->m_next; - if ( p == NULL ) - return g_last_cudaError = cudaErrorUnknown; - - char *data = (char *) calloc(p->m_size,1); - memcpy_from_gpu( data,(size_t)p->m_devPtr,p->m_size ); - glBufferSubData(GL_ARRAY_BUFFER,0,p->m_size,data); - free(data); - - return g_last_cudaError = cudaSuccess; + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + CUctx_st *ctx = GPGPUSim_Context(ctx); + glbmap_entry_t *p = ctx->api->g_glbmap; + while (p && p->m_bufferObj != bufferObj) p = p->m_next; + if (p == NULL) return g_last_cudaError = cudaErrorUnknown; + + char *data = (char *)calloc(p->m_size, 1); + memcpy_from_gpu(data, (size_t)p->m_devPtr, p->m_size); + glBufferSubData(GL_ARRAY_BUFFER, 0, p->m_size, data); + free(data); + + return g_last_cudaError = cudaSuccess; #else - fflush(stdout); - fflush(stderr); - printf("GPGPU-Sim PTX: support for OpenGL integration disabled -- exiting\n"); - fflush(stdout); - exit(50); + fflush(stdout); + fflush(stderr); + printf("GPGPU-Sim PTX: support for OpenGL integration disabled -- exiting\n"); + fflush(stdout); + exit(50); #endif } #if CUDART_VERSION >= 3000 -__host__ cudaError_t CUDARTAPI cudaFuncSetCacheConfigInternal(const char *func, enum cudaFuncCache cacheConfig, gpgpu_context* gpgpu_ctx = NULL ) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUctx_st *context = GPGPUSim_Context(ctx); - context->get_device()->get_gpgpu()->set_cache_config(context->get_kernel(func)->get_name(), (FuncCache)cacheConfig); - return g_last_cudaError = cudaSuccess; +__host__ cudaError_t CUDARTAPI +cudaFuncSetCacheConfigInternal(const char *func, enum cudaFuncCache cacheConfig, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUctx_st *context = GPGPUSim_Context(ctx); + context->get_device()->get_gpgpu()->set_cache_config( + context->get_kernel(func)->get_name(), (FuncCache)cacheConfig); + return g_last_cudaError = cudaSuccess; } #endif #if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuLaunchKernelInternal(CUfunction f, - unsigned int gridDimX, - unsigned int gridDimY, - unsigned int gridDimZ, - unsigned int blockDimX, - unsigned int blockDimY, - unsigned int blockDimZ, - unsigned int sharedMemBytes, - CUstream hStream, - void **kernelParams, - void **extra, - gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - if (extra!=NULL){ - printf("GPGPU-Sim CUDA DRIVER API: ERROR: Currently do not support void** extra.\n"); - abort(); - } - const char *hostFun = (const char*) f; - CUctx_st *context = GPGPUSim_Context(ctx); - function_info *entry = context->get_kernel(hostFun); - cudaConfigureCallInternal(dim3(gridDimX, gridDimY, gridDimZ), dim3(blockDimX, blockDimY, blockDimZ), sharedMemBytes, (cudaStream_t) hStream, ctx); - for(unsigned i = 0; i < entry->num_args(); i++){ - std::pair p = entry->get_param_config(i); - cudaSetupArgumentInternal(kernelParams[i], p.first, p.second, ctx); - } - cudaLaunchInternal(hostFun, ctx); - return CUDA_SUCCESS; +CUresult CUDAAPI cuLaunchKernelInternal( + CUfunction f, unsigned int gridDimX, unsigned int gridDimY, + unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, + unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, + void **kernelParams, void **extra, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + if (extra != NULL) { + printf( + "GPGPU-Sim CUDA DRIVER API: ERROR: Currently do not support void** " + "extra.\n"); + abort(); + } + const char *hostFun = (const char *)f; + CUctx_st *context = GPGPUSim_Context(ctx); + function_info *entry = context->get_kernel(hostFun); + cudaConfigureCallInternal(dim3(gridDimX, gridDimY, gridDimZ), + dim3(blockDimX, blockDimY, blockDimZ), + sharedMemBytes, (cudaStream_t)hStream, ctx); + for (unsigned i = 0; i < entry->num_args(); i++) { + std::pair p = entry->get_param_config(i); + cudaSetupArgumentInternal(kernelParams[i], p.first, p.second, ctx); + } + cudaLaunchInternal(hostFun, ctx); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 4000 */ -CUevent_st *get_event(cudaEvent_t event) -{ - unsigned event_uid; +CUevent_st *get_event(cudaEvent_t event) { + unsigned event_uid; #if CUDART_VERSION >= 3000 - event_uid = event->get_uid(); + event_uid = event->get_uid(); #else - event_uid = event; + event_uid = event; #endif - event_tracker_t::iterator e = g_timer_events.find(event_uid); - if( e == g_timer_events.end() ) - return NULL; - return e->second; -} - -__host__ cudaError_t CUDARTAPI cudaEventRecordInternal(cudaEvent_t event, cudaStream_t stream, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUevent_st *e = get_event(event); - if( !e ) return g_last_cudaError = cudaErrorUnknown; - struct CUstream_st *s = (struct CUstream_st *)stream; - stream_operation op(e,s); - ctx->the_gpgpusim->g_stream_manager->push(op); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaStreamWaitEventInternal(cudaStream_t stream, cudaEvent_t event, unsigned int flags, gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //reference: https://www.cs.cmu.edu/afs/cs/academic/class/15668-s11/www/cuda-doc/html/group__CUDART__STREAM_gfe68d207dc965685d92d3f03d77b0876.html - CUevent_st *e = get_event(event); - if( !e ){ - printf("GPGPU-Sim API: Warning: cudaEventRecord has not been called on event before calling cudaStreamWaitEvent.\nNothing to be done.\n"); - return g_last_cudaError = cudaSuccess; - } - if (!stream){ - ctx->the_gpgpusim->g_stream_manager->pushCudaStreamWaitEventToAllStreams(e, flags); - } else { - struct CUstream_st *s = (struct CUstream_st *)stream; - stream_operation op(s,e,flags); - ctx->the_gpgpusim->g_stream_manager->push(op); - } - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaThreadExitInternal(gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - ctx->exit_simulation(); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaThreadSynchronizeInternal(gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //Called on host side - ctx->synchronize(); - return g_last_cudaError = cudaSuccess; -} - -cudaError_t CUDARTAPI cudaDeviceSynchronizeInternal(gpgpu_context* gpgpu_ctx = NULL) -{ - gpgpu_context *ctx; - if (gpgpu_ctx){ - ctx = gpgpu_ctx; - } else { - ctx = GPGPU_Context(); - } - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //Blocks until the device has completed all preceding requested tasks - ctx->synchronize(); - return g_last_cudaError = cudaSuccess; + event_tracker_t::iterator e = g_timer_events.find(event_uid); + if (e == g_timer_events.end()) return NULL; + return e->second; +} + +__host__ cudaError_t CUDARTAPI cudaEventRecordInternal( + cudaEvent_t event, cudaStream_t stream, gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUevent_st *e = get_event(event); + if (!e) return g_last_cudaError = cudaErrorUnknown; + struct CUstream_st *s = (struct CUstream_st *)stream; + stream_operation op(e, s); + ctx->the_gpgpusim->g_stream_manager->push(op); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaStreamWaitEventInternal( + cudaStream_t stream, cudaEvent_t event, unsigned int flags, + gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // reference: + // https://www.cs.cmu.edu/afs/cs/academic/class/15668-s11/www/cuda-doc/html/group__CUDART__STREAM_gfe68d207dc965685d92d3f03d77b0876.html + CUevent_st *e = get_event(event); + if (!e) { + printf( + "GPGPU-Sim API: Warning: cudaEventRecord has not been called on event " + "before calling cudaStreamWaitEvent.\nNothing to be done.\n"); + return g_last_cudaError = cudaSuccess; + } + if (!stream) { + ctx->the_gpgpusim->g_stream_manager->pushCudaStreamWaitEventToAllStreams( + e, flags); + } else { + struct CUstream_st *s = (struct CUstream_st *)stream; + stream_operation op(s, e, flags); + ctx->the_gpgpusim->g_stream_manager->push(op); + } + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI +cudaThreadExitInternal(gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + ctx->exit_simulation(); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI +cudaThreadSynchronizeInternal(gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // Called on host side + ctx->synchronize(); + return g_last_cudaError = cudaSuccess; +} + +cudaError_t CUDARTAPI +cudaDeviceSynchronizeInternal(gpgpu_context *gpgpu_ctx = NULL) { + gpgpu_context *ctx; + if (gpgpu_ctx) { + ctx = gpgpu_ctx; + } else { + ctx = GPGPU_Context(); + } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // Blocks until the device has completed all preceding requested tasks + ctx->synchronize(); + return g_last_cudaError = cudaSuccess; } /******************************************************************************* @@ -2145,142 +2296,148 @@ extern "C" { * * * * *******************************************************************************/ -cudaError_t cudaPeekAtLastError(void) -{ - return g_last_cudaError; -} +cudaError_t cudaPeekAtLastError(void) { return g_last_cudaError; } -__host__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) -{ - return cudaMallocInternal(devPtr, size); +__host__ cudaError_t CUDARTAPI cudaMalloc(void **devPtr, size_t size) { + return cudaMallocInternal(devPtr, size); } -__host__ cudaError_t CUDARTAPI cudaMallocHost(void **ptr, size_t size) -{ - return cudaMallocHostInternal(ptr, size); +__host__ cudaError_t CUDARTAPI cudaMallocHost(void **ptr, size_t size) { + return cudaMallocHostInternal(ptr, size); } -__host__ cudaError_t CUDARTAPI cudaMallocPitch(void **devPtr, size_t *pitch, size_t width, size_t height) -{ - return cudaMallocPitchInternal(devPtr, pitch, width, height); +__host__ cudaError_t CUDARTAPI cudaMallocPitch(void **devPtr, size_t *pitch, + size_t width, size_t height) { + return cudaMallocPitchInternal(devPtr, pitch, width, height); } -__host__ cudaError_t CUDARTAPI cudaMallocArray(struct cudaArray **array, const struct cudaChannelFormatDesc *desc, size_t width, size_t height __dv(1)) -{ - return cudaMallocArrayInternal(array, desc, width, height); +__host__ cudaError_t CUDARTAPI cudaMallocArray( + struct cudaArray **array, const struct cudaChannelFormatDesc *desc, + size_t width, size_t height __dv(1)) { + return cudaMallocArrayInternal(array, desc, width, height); } -__host__ cudaError_t CUDARTAPI cudaFree(void *devPtr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - // TODO... manage g_global_mem space? - return g_last_cudaError = cudaSuccess; +__host__ cudaError_t CUDARTAPI cudaFree(void *devPtr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // TODO... manage g_global_mem space? + return g_last_cudaError = cudaSuccess; } -__host__ cudaError_t CUDARTAPI cudaFreeHost(void *ptr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - free (ptr); // this will crash the system if called twice - return g_last_cudaError = cudaSuccess; +__host__ cudaError_t CUDARTAPI cudaFreeHost(void *ptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + free(ptr); // this will crash the system if called twice + return g_last_cudaError = cudaSuccess; } -__host__ cudaError_t CUDARTAPI cudaFreeArray(struct cudaArray *array) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - // TODO... manage g_global_mem space? - return g_last_cudaError = cudaSuccess; +__host__ cudaError_t CUDARTAPI cudaFreeArray(struct cudaArray *array) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // TODO... manage g_global_mem space? + return g_last_cudaError = cudaSuccess; }; - /******************************************************************************* * * * * * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaMemcpy(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind) -{ - return cudaMemcpyInternal(dst, src, count, kind); +__host__ cudaError_t CUDARTAPI cudaMemcpy(void *dst, const void *src, + size_t count, + enum cudaMemcpyKind kind) { + return cudaMemcpyInternal(dst, src, count, kind); } -__host__ cudaError_t CUDARTAPI cudaMemcpyToArray(struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, size_t count, enum cudaMemcpyKind kind) -{ - return cudaMemcpyToArrayInternal(dst, wOffset, hOffset, src, count, kind); +__host__ cudaError_t CUDARTAPI cudaMemcpyToArray(struct cudaArray *dst, + size_t wOffset, size_t hOffset, + const void *src, size_t count, + enum cudaMemcpyKind kind) { + return cudaMemcpyToArrayInternal(dst, wOffset, hOffset, src, count, kind); } - -__host__ cudaError_t CUDARTAPI cudaMemcpyFromArray(void *dst, const struct cudaArray *src, size_t wOffset, size_t hOffset, size_t count, enum cudaMemcpyKind kind) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaMemcpyFromArray(void *dst, + const struct cudaArray *src, + size_t wOffset, + size_t hOffset, size_t count, + enum cudaMemcpyKind kind) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } - -__host__ cudaError_t CUDARTAPI cudaMemcpyArrayToArray(struct cudaArray *dst, size_t wOffsetDst, size_t hOffsetDst, const struct cudaArray *src, size_t wOffsetSrc, size_t hOffsetSrc, size_t count, enum cudaMemcpyKind kind __dv(cudaMemcpyDeviceToDevice)) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaMemcpyArrayToArray( + struct cudaArray *dst, size_t wOffsetDst, size_t hOffsetDst, + const struct cudaArray *src, size_t wOffsetSrc, size_t hOffsetSrc, + size_t count, enum cudaMemcpyKind kind __dv(cudaMemcpyDeviceToDevice)) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } - -__host__ cudaError_t CUDARTAPI cudaMemcpy2D(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind) -{ - return cudaMemcpy2DInternal(dst, dpitch, src, spitch, width, height, kind); +__host__ cudaError_t CUDARTAPI cudaMemcpy2D(void *dst, size_t dpitch, + const void *src, size_t spitch, + size_t width, size_t height, + enum cudaMemcpyKind kind) { + return cudaMemcpy2DInternal(dst, dpitch, src, spitch, width, height, kind); } -__host__ cudaError_t CUDARTAPI cudaMemcpy2DToArray(struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind) -{ - return cudaMemcpy2DToArrayInternal(dst, wOffset, hOffset, src, spitch, width, height, kind); +__host__ cudaError_t CUDARTAPI cudaMemcpy2DToArray( + struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, + size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind) { + return cudaMemcpy2DToArrayInternal(dst, wOffset, hOffset, src, spitch, width, + height, kind); } -__host__ cudaError_t CUDARTAPI cudaMemcpy2DFromArray(void *dst, size_t dpitch, const struct cudaArray *src, size_t wOffset, size_t hOffset, size_t width, size_t height, enum cudaMemcpyKind kind) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaMemcpy2DFromArray( + void *dst, size_t dpitch, const struct cudaArray *src, size_t wOffset, + size_t hOffset, size_t width, size_t height, enum cudaMemcpyKind kind) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t CUDARTAPI cudaMemcpy2DArrayToArray(struct cudaArray *dst, size_t wOffsetDst, size_t hOffsetDst, const struct cudaArray *src, size_t wOffsetSrc, size_t hOffsetSrc, size_t width, size_t height, enum cudaMemcpyKind kind __dv(cudaMemcpyDeviceToDevice)) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaMemcpy2DArrayToArray( + struct cudaArray *dst, size_t wOffsetDst, size_t hOffsetDst, + const struct cudaArray *src, size_t wOffsetSrc, size_t hOffsetSrc, + size_t width, size_t height, + enum cudaMemcpyKind kind __dv(cudaMemcpyDeviceToDevice)) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t CUDARTAPI cudaMemcpyToSymbol(const char *symbol, const void *src, size_t count, size_t offset __dv(0), enum cudaMemcpyKind kind __dv(cudaMemcpyHostToDevice)) -{ - return cudaMemcpyToSymbolInternal(symbol, src, count, offset, kind); +__host__ cudaError_t CUDARTAPI cudaMemcpyToSymbol( + const char *symbol, const void *src, size_t count, size_t offset __dv(0), + enum cudaMemcpyKind kind __dv(cudaMemcpyHostToDevice)) { + return cudaMemcpyToSymbolInternal(symbol, src, count, offset, kind); } - -__host__ cudaError_t CUDARTAPI cudaMemcpyFromSymbol(void *dst, const char *symbol, size_t count, size_t offset __dv(0), enum cudaMemcpyKind kind __dv(cudaMemcpyDeviceToHost)) -{ - return cudaMemcpyFromSymbolInternal(dst, symbol, count, offset, kind); +__host__ cudaError_t CUDARTAPI cudaMemcpyFromSymbol( + void *dst, const char *symbol, size_t count, size_t offset __dv(0), + enum cudaMemcpyKind kind __dv(cudaMemcpyDeviceToHost)) { + return cudaMemcpyFromSymbolInternal(dst, symbol, count, offset, kind); } -__host__ cudaError_t CUDARTAPI cudaMemGetInfo (size_t *free, size_t *total){ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //placeholder; should interact with cudaMalloc and cudaFree? - *free = 10000000000; - *total = 10000000000; +__host__ cudaError_t CUDARTAPI cudaMemGetInfo(size_t *free, size_t *total) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // placeholder; should interact with cudaMalloc and cudaFree? + *free = 10000000000; + *total = 10000000000; - return g_last_cudaError = cudaSuccess; + return g_last_cudaError = cudaSuccess; } /******************************************************************************* @@ -2289,391 +2446,377 @@ __host__ cudaError_t CUDARTAPI cudaMemGetInfo (size_t *free, size_t *total){ * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) -{ - return cudaMemcpyAsyncInternal(dst, src, count, kind, stream); -} - - -__host__ cudaError_t CUDARTAPI cudaMemcpyToArrayAsync(struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; -} - - -__host__ cudaError_t CUDARTAPI cudaMemcpyFromArrayAsync(void *dst, const struct cudaArray *src, size_t wOffset, size_t hOffset, size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; -} - - -__host__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; -} - - -__host__ cudaError_t CUDARTAPI cudaMemcpy2DToArrayAsync(struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; -} - - -__host__ cudaError_t CUDARTAPI cudaMemcpy2DFromArrayAsync(void *dst, size_t dpitch, const struct cudaArray *src, size_t wOffset, size_t hOffset, size_t width, size_t height, enum cudaMemcpyKind kind, cudaStream_t stream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaMemcpyAsync(void *dst, const void *src, + size_t count, + enum cudaMemcpyKind kind, + cudaStream_t stream) { + return cudaMemcpyAsyncInternal(dst, src, count, kind, stream); +} + +__host__ cudaError_t CUDARTAPI cudaMemcpyToArrayAsync( + struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, + size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpyFromArrayAsync( + void *dst, const struct cudaArray *src, size_t wOffset, size_t hOffset, + size_t count, enum cudaMemcpyKind kind, cudaStream_t stream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpy2DAsync(void *dst, size_t dpitch, + const void *src, size_t spitch, + size_t width, size_t height, + enum cudaMemcpyKind kind, + cudaStream_t stream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpy2DToArrayAsync( + struct cudaArray *dst, size_t wOffset, size_t hOffset, const void *src, + size_t spitch, size_t width, size_t height, enum cudaMemcpyKind kind, + cudaStream_t stream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; +} + +__host__ cudaError_t CUDARTAPI cudaMemcpy2DFromArrayAsync( + void *dst, size_t dpitch, const struct cudaArray *src, size_t wOffset, + size_t hOffset, size_t width, size_t height, enum cudaMemcpyKind kind, + cudaStream_t stream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } #if (CUDART_VERSION >= 8000) -cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int* numBlocks, const char *hostFunc, int blockSize, size_t dynamicSMemSize, unsigned int flags) -{ - return cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlagsInternal(numBlocks, hostFunc, blockSize, dynamicSMemSize, flags); +cudaError_t CUDARTAPI cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlags( + int *numBlocks, const char *hostFunc, int blockSize, size_t dynamicSMemSize, + unsigned int flags) { + return cudaOccupancyMaxActiveBlocksPerMultiprocessorWithFlagsInternal( + numBlocks, hostFunc, blockSize, dynamicSMemSize, flags); } #endif - - /******************************************************************************* * * * * * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaMemset(void *mem, int c, size_t count) -{ - return cudaMemsetInternal(mem, c, count); +__host__ cudaError_t CUDARTAPI cudaMemset(void *mem, int c, size_t count) { + return cudaMemsetInternal(mem, c, count); } -//memset operation is done but i think its not async? -__host__ cudaError_t CUDARTAPI cudaMemsetAsync(void *mem, int c, size_t count, cudaStream_t stream=0) -{ - return cudaMemsetAsyncInternal(mem, c, count, stream=0); +// memset operation is done but i think its not async? +__host__ cudaError_t CUDARTAPI cudaMemsetAsync(void *mem, int c, size_t count, + cudaStream_t stream = 0) { + return cudaMemsetAsyncInternal(mem, c, count, stream = 0); } -__host__ cudaError_t CUDARTAPI cudaMemset2D(void *mem, size_t pitch, int c, size_t width, size_t height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaMemset2D(void *mem, size_t pitch, int c, + size_t width, size_t height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } - - /******************************************************************************* * * * * * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaGetSymbolAddress(void **devPtr, const char *symbol) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaGetSymbolAddress(void **devPtr, + const char *symbol) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } - -__host__ cudaError_t CUDARTAPI cudaGetSymbolSize(size_t *size, const char *symbol) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaGetSymbolSize(size_t *size, + const char *symbol) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } - - /******************************************************************************* * * * * * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) -{ - return cudaGetDeviceCountInternal(count); +__host__ cudaError_t CUDARTAPI cudaGetDeviceCount(int *count) { + return cudaGetDeviceCountInternal(count); } -__host__ cudaError_t CUDARTAPI cudaGetDeviceProperties(struct cudaDeviceProp *prop, int device) -{ - return cudaGetDevicePropertiesInternal(prop, device); +__host__ cudaError_t CUDARTAPI +cudaGetDeviceProperties(struct cudaDeviceProp *prop, int device) { + return cudaGetDevicePropertiesInternal(prop, device); } #if (CUDART_VERSION > 5000) -__host__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDeviceAttr attr, int device) -{ - return cudaDeviceGetAttributeInternal(value, attr, device); +__host__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, + enum cudaDeviceAttr attr, + int device) { + return cudaDeviceGetAttributeInternal(value, attr, device); } #endif -__host__ cudaError_t CUDARTAPI cudaChooseDevice(int *device, const struct cudaDeviceProp *prop) -{ - return cudaChooseDeviceInternal(device, prop); +__host__ cudaError_t CUDARTAPI +cudaChooseDevice(int *device, const struct cudaDeviceProp *prop) { + return cudaChooseDeviceInternal(device, prop); } -__host__ cudaError_t CUDARTAPI cudaSetDevice(int device) -{ - return cudaSetDeviceInternal(device); +__host__ cudaError_t CUDARTAPI cudaSetDevice(int device) { + return cudaSetDeviceInternal(device); } -__host__ cudaError_t CUDARTAPI cudaGetDevice(int *device) -{ - return cudaGetDeviceInternal(device); +__host__ cudaError_t CUDARTAPI cudaGetDevice(int *device) { + return cudaGetDeviceInternal(device); } -__host__ cudaError_t CUDARTAPI cudaDeviceGetLimit( size_t* pValue, cudaLimit limit ) -{ - return cudaDeviceGetLimitInternal( pValue, limit ); +__host__ cudaError_t CUDARTAPI cudaDeviceGetLimit(size_t *pValue, + cudaLimit limit) { + return cudaDeviceGetLimitInternal(pValue, limit); } -__host__ cudaError_t CUDARTAPI cudaStreamGetPriority ( cudaStream_t hStream, int* priority ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaSuccess; - +__host__ cudaError_t CUDARTAPI cudaStreamGetPriority(cudaStream_t hStream, + int *priority) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaSuccess; } -__host__ cudaError_t CUDARTAPI cudaDeviceGetPCIBusId ( - char *pciBusId, - int len, - int device -) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaDeviceGetPCIBusId(char *pciBusId, int len, + int device) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t CUDARTAPI cudaIpcGetMemHandle( cudaIpcMemHandle_t* handle, void* devPtr ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaIpcGetMemHandle(cudaIpcMemHandle_t *handle, + void *devPtr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t cudaIpcOpenMemHandle( - void **devPtr, - cudaIpcMemHandle_t handle, - unsigned int flags -) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t cudaIpcOpenMemHandle(void **devPtr, + cudaIpcMemHandle_t handle, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t CUDARTAPI cudaDestroyTextureObject(cudaTextureObject_t texObject) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI +cudaDestroyTextureObject(cudaTextureObject_t texObject) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } - /******************************************************************************* * * * * * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaBindTexture(size_t *offset, - const struct textureReference *texref, - const void *devPtr, - const struct cudaChannelFormatDesc *desc, - size_t size __dv(UINT_MAX)) -{ - return cudaBindTextureInternal(offset, texref, devPtr, desc, size __dv(UINT_MAX)); +__host__ cudaError_t CUDARTAPI cudaBindTexture( + size_t *offset, const struct textureReference *texref, const void *devPtr, + const struct cudaChannelFormatDesc *desc, size_t size __dv(UINT_MAX)) { + return cudaBindTextureInternal(offset, texref, devPtr, desc, + size __dv(UINT_MAX)); } +__host__ cudaError_t CUDARTAPI cudaBindTextureToArray( + const struct textureReference *texref, const struct cudaArray *array, + const struct cudaChannelFormatDesc *desc) { + return cudaBindTextureToArrayInternal(texref, array, desc); +} -__host__ cudaError_t CUDARTAPI cudaBindTextureToArray(const struct textureReference *texref, const struct cudaArray *array, const struct cudaChannelFormatDesc *desc) -{ - return cudaBindTextureToArrayInternal(texref, array, desc); +__host__ cudaError_t CUDARTAPI +cudaUnbindTexture(const struct textureReference *texref) { + return cudaUnbindTextureInternal(texref); } -__host__ cudaError_t CUDARTAPI cudaUnbindTexture(const struct textureReference *texref) -{ - return cudaUnbindTextureInternal(texref); +__host__ cudaError_t CUDARTAPI cudaGetTextureAlignmentOffset( + size_t *offset, const struct textureReference *texref) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t CUDARTAPI cudaGetTextureAlignmentOffset(size_t *offset, const struct textureReference *texref) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaGetTextureReference( + const struct textureReference **texref, const char *symbol) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t CUDARTAPI cudaGetTextureReference(const struct textureReference **texref, const char *symbol) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaGetChannelDesc( + struct cudaChannelFormatDesc *desc, const struct cudaArray *array) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + *desc = array->desc; + return g_last_cudaError = cudaSuccess; } -__host__ cudaError_t CUDARTAPI cudaGetChannelDesc(struct cudaChannelFormatDesc *desc, const struct cudaArray *array) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - *desc = array->desc; - return g_last_cudaError = cudaSuccess; +__host__ struct cudaChannelFormatDesc CUDARTAPI cudaCreateChannelDesc( + int x, int y, int z, int w, enum cudaChannelFormatKind f) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + struct cudaChannelFormatDesc dummy; + dummy.x = x; + dummy.y = y; + dummy.z = z; + dummy.w = w; + dummy.f = f; + return dummy; } +__host__ cudaError_t CUDARTAPI cudaGetLastError(void) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return g_last_cudaError; +} -__host__ struct cudaChannelFormatDesc CUDARTAPI cudaCreateChannelDesc(int x, int y, int z, int w, enum cudaChannelFormatKind f) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - struct cudaChannelFormatDesc dummy; - dummy.x = x; - dummy.y = y; - dummy.z = z; - dummy.w = w; - dummy.f = f; - return dummy; +__host__ const char *cudaGetErrorName(cudaError_t error) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return NULL; } -__host__ cudaError_t CUDARTAPI cudaGetLastError(void) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return g_last_cudaError; +__host__ const char *CUDARTAPI cudaGetErrorString(cudaError_t error) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + if (g_last_cudaError == cudaSuccess) return "no error"; + char buf[1024]; + snprintf(buf, 1024, "<>", + g_last_cudaError); + return strdup(buf); } -__host__ const char *cudaGetErrorName(cudaError_t error) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return NULL; -} - -__host__ const char* CUDARTAPI cudaGetErrorString(cudaError_t error) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - if( g_last_cudaError == cudaSuccess ) - return "no error"; - char buf[1024]; - snprintf(buf,1024,"<>", g_last_cudaError); - return strdup(buf); -} - -__host__ cudaError_t CUDARTAPI cudaSetupArgument(const void *arg, size_t size, size_t offset) -{ - return cudaSetupArgumentInternal(arg, size, offset); +__host__ cudaError_t CUDARTAPI cudaSetupArgument(const void *arg, size_t size, + size_t offset) { + return cudaSetupArgumentInternal(arg, size, offset); } - -__host__ cudaError_t CUDARTAPI cudaLaunch( const char *hostFun ) -{ - return cudaLaunchInternal( hostFun ); +__host__ cudaError_t CUDARTAPI cudaLaunch(const char *hostFun) { + return cudaLaunchInternal(hostFun); } -__host__ cudaError_t CUDARTAPI cudaLaunchKernel( const char* hostFun, dim3 gridDim, dim3 blockDim, const void** args, size_t sharedMem, cudaStream_t stream ) -{ - return cudaLaunchKernelInternal(hostFun, gridDim, blockDim, args, sharedMem, stream); +__host__ cudaError_t CUDARTAPI cudaLaunchKernel(const char *hostFun, + dim3 gridDim, dim3 blockDim, + const void **args, + size_t sharedMem, + cudaStream_t stream) { + return cudaLaunchKernelInternal(hostFun, gridDim, blockDim, args, sharedMem, + stream); } - /******************************************************************************* * * * * * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaStreamCreate(cudaStream_t *stream) -{ - return cudaStreamCreateInternal(stream); +__host__ cudaError_t CUDARTAPI cudaStreamCreate(cudaStream_t *stream) { + return cudaStreamCreateInternal(stream); } -//TODO: introduce priorities -__host__ cudaError_t CUDARTAPI cudaStreamCreateWithPriority(cudaStream_t *stream, unsigned int flags, int priority) { - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return cudaStreamCreate(stream); +// TODO: introduce priorities +__host__ cudaError_t CUDARTAPI cudaStreamCreateWithPriority( + cudaStream_t *stream, unsigned int flags, int priority) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return cudaStreamCreate(stream); } -__host__ cudaError_t CUDARTAPI cudaDeviceGetStreamPriorityRange(int* leastPriority, int* greatestPriority) { - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return cudaSuccess; +__host__ cudaError_t CUDARTAPI +cudaDeviceGetStreamPriorityRange(int *leastPriority, int *greatestPriority) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return cudaSuccess; } -__host__ __device__ cudaError_t CUDARTAPI cudaStreamCreateWithFlags(cudaStream_t *stream, unsigned int flags) { - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return cudaStreamCreate(stream); +__host__ __device__ cudaError_t CUDARTAPI +cudaStreamCreateWithFlags(cudaStream_t *stream, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return cudaStreamCreate(stream); } -__host__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) -{ - return cudaStreamDestroyInternal(stream); +__host__ cudaError_t CUDARTAPI cudaStreamDestroy(cudaStream_t stream) { + return cudaStreamDestroyInternal(stream); } -__host__ cudaError_t CUDARTAPI cudaStreamSynchronize(cudaStream_t stream) -{ - return cudaStreamSynchronizeInternal(stream); +__host__ cudaError_t CUDARTAPI cudaStreamSynchronize(cudaStream_t stream) { + return cudaStreamSynchronizeInternal(stream); } -__host__ cudaError_t CUDARTAPI cudaStreamQuery(cudaStream_t stream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } +__host__ cudaError_t CUDARTAPI cudaStreamQuery(cudaStream_t stream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } #if (CUDART_VERSION >= 3000) - if( stream == NULL ) - return g_last_cudaError = cudaErrorInvalidResourceHandle; - return g_last_cudaError = stream->empty()?cudaSuccess:cudaErrorNotReady; + if (stream == NULL) return g_last_cudaError = cudaErrorInvalidResourceHandle; + return g_last_cudaError = stream->empty() ? cudaSuccess : cudaErrorNotReady; #else - printf("GPGPU-Sim PTX: WARNING: Asynchronous kernel execution not supported (%s)\n", __my_func__); - return g_last_cudaError = cudaSuccess; // it is always success because all cuda calls are synchronous + printf( + "GPGPU-Sim PTX: WARNING: Asynchronous kernel execution not supported " + "(%s)\n", + __my_func__); + return g_last_cudaError = cudaSuccess; // it is always success because all + // cuda calls are synchronous #endif } @@ -2683,119 +2826,108 @@ __host__ cudaError_t CUDARTAPI cudaStreamQuery(cudaStream_t stream) * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaEventCreate(cudaEvent_t *event) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUevent_st *e = new CUevent_st(false); - g_timer_events[e->get_uid()] = e; +__host__ cudaError_t CUDARTAPI cudaEventCreate(cudaEvent_t *event) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUevent_st *e = new CUevent_st(false); + g_timer_events[e->get_uid()] = e; #if CUDART_VERSION >= 3000 - *event = e; + *event = e; #else - *event = e->get_uid(); + *event = e->get_uid(); #endif - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, cudaStream_t stream) -{ - return cudaEventRecordInternal(event, stream); -} - -__host__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, cudaEvent_t event, unsigned int flags) -{ - return cudaStreamWaitEventInternal(stream, event, flags); -} - -__host__ cudaError_t CUDARTAPI cudaEventQuery(cudaEvent_t event) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUevent_st *e = get_event(event); - if( e == NULL ) { - return g_last_cudaError = cudaErrorInvalidValue; - } else if( e->done() ) { - return g_last_cudaError = cudaSuccess; - } else { - return g_last_cudaError = cudaErrorNotReady; - } + return g_last_cudaError = cudaSuccess; } -__host__ cudaError_t CUDARTAPI cudaEventSynchronize(cudaEvent_t event) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("GPGPU-Sim API: cudaEventSynchronize ** waiting for event\n"); - fflush(stdout); - CUevent_st *e = (CUevent_st*) event; - while( !e->done() ) - ; - printf("GPGPU-Sim API: cudaEventSynchronize ** event detected\n"); - fflush(stdout); - return g_last_cudaError = cudaSuccess; +__host__ cudaError_t CUDARTAPI cudaEventRecord(cudaEvent_t event, + cudaStream_t stream) { + return cudaEventRecordInternal(event, stream); } -__host__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - CUevent_st *e = get_event(event); - unsigned event_uid = e->get_uid(); - event_tracker_t::iterator pe = g_timer_events.find(event_uid); - if( pe == g_timer_events.end() ) - return g_last_cudaError = cudaErrorInvalidValue; - g_timer_events.erase(pe); - return g_last_cudaError = cudaSuccess; +__host__ cudaError_t CUDARTAPI cudaStreamWaitEvent(cudaStream_t stream, + cudaEvent_t event, + unsigned int flags) { + return cudaStreamWaitEventInternal(stream, event, flags); } - -__host__ cudaError_t CUDARTAPI cudaEventElapsedTime(float *ms, cudaEvent_t start, cudaEvent_t end) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - time_t elapsed_time; - CUevent_st *s = get_event(start); - CUevent_st *e = get_event(end); - if( s==NULL || e==NULL ) - return g_last_cudaError = cudaErrorUnknown; - elapsed_time = e->clock() - s->clock(); - *ms = 1000*elapsed_time; - return g_last_cudaError = cudaSuccess; +__host__ cudaError_t CUDARTAPI cudaEventQuery(cudaEvent_t event) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUevent_st *e = get_event(event); + if (e == NULL) { + return g_last_cudaError = cudaErrorInvalidValue; + } else if (e->done()) { + return g_last_cudaError = cudaSuccess; + } else { + return g_last_cudaError = cudaErrorNotReady; + } +} + +__host__ cudaError_t CUDARTAPI cudaEventSynchronize(cudaEvent_t event) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("GPGPU-Sim API: cudaEventSynchronize ** waiting for event\n"); + fflush(stdout); + CUevent_st *e = (CUevent_st *)event; + while (!e->done()) + ; + printf("GPGPU-Sim API: cudaEventSynchronize ** event detected\n"); + fflush(stdout); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaEventDestroy(cudaEvent_t event) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + CUevent_st *e = get_event(event); + unsigned event_uid = e->get_uid(); + event_tracker_t::iterator pe = g_timer_events.find(event_uid); + if (pe == g_timer_events.end()) + return g_last_cudaError = cudaErrorInvalidValue; + g_timer_events.erase(pe); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t CUDARTAPI cudaEventElapsedTime(float *ms, + cudaEvent_t start, + cudaEvent_t end) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + time_t elapsed_time; + CUevent_st *s = get_event(start); + CUevent_st *e = get_event(end); + if (s == NULL || e == NULL) return g_last_cudaError = cudaErrorUnknown; + elapsed_time = e->clock() - s->clock(); + *ms = 1000 * elapsed_time; + return g_last_cudaError = cudaSuccess; } - - /******************************************************************************* * * * * * * *******************************************************************************/ -__host__ cudaError_t CUDARTAPI cudaThreadExit(void) -{ - return cudaThreadExitInternal(); +__host__ cudaError_t CUDARTAPI cudaThreadExit(void) { + return cudaThreadExitInternal(); } -__host__ cudaError_t CUDARTAPI cudaThreadSynchronize(void) -{ - return cudaThreadSynchronizeInternal(); +__host__ cudaError_t CUDARTAPI cudaThreadSynchronize(void) { + return cudaThreadSynchronizeInternal(); } -int CUDARTAPI __cudaSynchronizeThreads(void**, void*) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return cudaThreadExit(); +int CUDARTAPI __cudaSynchronizeThreads(void **, void *) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return cudaThreadExit(); } - - /******************************************************************************* * * * * @@ -2804,39 +2936,40 @@ int CUDARTAPI __cudaSynchronizeThreads(void**, void*) #if (CUDART_VERSION >= 3010) int dummy0() { - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } -return 0; } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return 0; +} int dummy1() { - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } -return 2 << 20; } + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return 2 << 20; +} typedef int (*ExportedFunction)(); static ExportedFunction exportTable[3] = {&dummy0, &dummy0, &dummy0}; -__host__ cudaError_t CUDARTAPI cudaGetExportTable(const void **ppExportTable, const cudaUUID_t *pExportTableId) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("cudaGetExportTable: UUID = "); - for (int s = 0; s < 16; s++) { - printf("%#2x ", (unsigned char) (pExportTableId->bytes[s])); - } - *ppExportTable = &exportTable; +__host__ cudaError_t CUDARTAPI cudaGetExportTable( + const void **ppExportTable, const cudaUUID_t *pExportTableId) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("cudaGetExportTable: UUID = "); + for (int s = 0; s < 16; s++) { + printf("%#2x ", (unsigned char)(pExportTableId->bytes[s])); + } + *ppExportTable = &exportTable; - printf("\n"); - return g_last_cudaError = cudaSuccess; + printf("\n"); + return g_last_cudaError = cudaSuccess; } #endif - /******************************************************************************* * * * * @@ -2845,875 +2978,862 @@ __host__ cudaError_t CUDARTAPI cudaGetExportTable(const void **ppExportTable, co //#include "../../cuobjdump_to_ptxplus/cuobjdump_parser.h" -//extracts all ptx files from binary and dumps into prog_name.unique_no.sm_<>.ptx files -void cuda_runtime_api::extract_ptx_files_using_cuobjdump(CUctx_st *context){ - char command[1000]; - char *pytorch_bin = getenv("PYTORCH_BIN"); - std::string app_binary = get_app_binary(); - +// extracts all ptx files from binary and dumps into +// prog_name.unique_no.sm_<>.ptx files +void cuda_runtime_api::extract_ptx_files_using_cuobjdump(CUctx_st *context) { + char command[1000]; + char *pytorch_bin = getenv("PYTORCH_BIN"); + std::string app_binary = get_app_binary(); + + char ptx_list_file_name[1024]; + snprintf(ptx_list_file_name, 1024, "_cuobjdump_list_ptx_XXXXXX"); + int fd2 = mkstemp(ptx_list_file_name); + close(fd2); + + if (pytorch_bin != NULL && strlen(pytorch_bin) != 0) { + app_binary = std::string(pytorch_bin); + } + + // only want file names + snprintf(command, 1000, + "$CUDA_INSTALL_PATH/bin/cuobjdump -lptx %s | cut -d \":\" -f 2 | " + "awk '{$1=$1}1' > %s", + app_binary.c_str(), ptx_list_file_name); + if (system(command) != 0) { + printf("WARNING: Failed to execute cuobjdump to get list of ptx files \n"); + exit(0); + } + if (!gpgpu_ctx->device_runtime->g_cdp_enabled) { + // based on the list above, dump ptx files individually. Format of dumped + // ptx file is prog_name.unique_no.sm_<>.ptx - char ptx_list_file_name[1024]; - snprintf(ptx_list_file_name,1024,"_cuobjdump_list_ptx_XXXXXX"); - int fd2=mkstemp(ptx_list_file_name); - close(fd2); - - if (pytorch_bin!=NULL && strlen(pytorch_bin)!=0){ - app_binary = std::string(pytorch_bin); - } - - //only want file names - snprintf(command,1000,"$CUDA_INSTALL_PATH/bin/cuobjdump -lptx %s | cut -d \":\" -f 2 | awk '{$1=$1}1' > %s", app_binary.c_str(), ptx_list_file_name); - if( system(command) != 0 ) { - printf("WARNING: Failed to execute cuobjdump to get list of ptx files \n"); + std::ifstream infile(ptx_list_file_name); + std::string line; + while (std::getline(infile, line)) { + // int pos = line.find(std::string(get_app_binary_name(app_binary))); + const char *ptx_file = line.c_str(); + printf("Extracting specific PTX file named %s \n", ptx_file); + snprintf(command, 1000, "$CUDA_INSTALL_PATH/bin/cuobjdump -xptx %s %s", + ptx_file, app_binary.c_str()); + if (system(command) != 0) { + printf("ERROR: command: %s failed \n", command); exit(0); - } - if(!gpgpu_ctx->device_runtime->g_cdp_enabled) { - //based on the list above, dump ptx files individually. Format of dumped ptx file is prog_name.unique_no.sm_<>.ptx - - std::ifstream infile(ptx_list_file_name); - std::string line; - while (std::getline(infile, line)) - { - //int pos = line.find(std::string(get_app_binary_name(app_binary))); - const char *ptx_file = line.c_str(); - printf("Extracting specific PTX file named %s \n",ptx_file); - snprintf(command,1000,"$CUDA_INSTALL_PATH/bin/cuobjdump -xptx %s %s", ptx_file, app_binary.c_str()); - if (system(command)!=0) { - printf("ERROR: command: %s failed \n",command); - exit(0); - } - context->no_of_ptx++; - } + } + context->no_of_ptx++; } + } - if(!context->no_of_ptx){ - printf("WARNING: Number of ptx in the executable file are 0. One of the reasons might be\n"); - printf("\t1. CDP is enabled\n"); - printf("\t2. When using PyTorch, PYTORCH_BIN is not set correctly\n"); - } + if (!context->no_of_ptx) { + printf( + "WARNING: Number of ptx in the executable file are 0. One of the " + "reasons might be\n"); + printf("\t1. CDP is enabled\n"); + printf("\t2. When using PyTorch, PYTORCH_BIN is not set correctly\n"); + } - std::ifstream infile(ptx_list_file_name); - std::string line; - while (std::getline(infile, line)) - { - //int pos = line.find(std::string(get_app_binary_name(app_binary))); - int pos1 = line.find("sm_"); - int pos2 = line.find_last_of("."); - if (pos1==std::string::npos&&pos2==std::string::npos){ - printf("ERROR: PTX list is not in correct format"); - exit(0); - } - std::string vstr = line.substr(pos1+3,pos2-pos1-3); - int version = atoi(vstr.c_str()); - if (version_filename.find(version)==version_filename.end()){ - version_filename[version] = std::set(); - } - version_filename[version].insert(line); + std::ifstream infile(ptx_list_file_name); + std::string line; + while (std::getline(infile, line)) { + // int pos = line.find(std::string(get_app_binary_name(app_binary))); + int pos1 = line.find("sm_"); + int pos2 = line.find_last_of("."); + if (pos1 == std::string::npos && pos2 == std::string::npos) { + printf("ERROR: PTX list is not in correct format"); + exit(0); } - + std::string vstr = line.substr(pos1 + 3, pos2 - pos1 - 3); + int version = atoi(vstr.c_str()); + if (version_filename.find(version) == version_filename.end()) { + version_filename[version] = std::set(); + } + version_filename[version].insert(line); + } } - //! Call cuobjdump to extract everything (-elf -sass -ptx) /*! * This Function extract the whole PTX (for all the files) using cuobjdump - * to _cuobjdump_complete_output_XXXXXX then runs a parser to chop it up with each binary in - * its own file - * It is also responsible for extracting the libraries linked to the binary if the option is - * enabled + * to _cuobjdump_complete_output_XXXXXX then runs a parser to chop it up + *with each binary in its own file It is also responsible for extracting the + *libraries linked to the binary if the option is enabled * */ -void cuda_runtime_api::extract_code_using_cuobjdump(){ - CUctx_st *context = GPGPUSim_Context(gpgpu_ctx); - - //prevent the dumping by cuobjdump everytime we execute the code! - const char *override_cuobjdump = getenv("CUOBJDUMP_SIM_FILE"); - char command[1000]; - std::string app_binary = get_app_binary(); - //Running cuobjdump using dynamic link to current process - snprintf(command,1000,"md5sum %s ", app_binary.c_str()); - printf("Running md5sum using \"%s\"\n", command); - if(system(command)){ - std::cout << "Failed to execute: " << command << std::endl; +void cuda_runtime_api::extract_code_using_cuobjdump() { + CUctx_st *context = GPGPUSim_Context(gpgpu_ctx); + + // prevent the dumping by cuobjdump everytime we execute the code! + const char *override_cuobjdump = getenv("CUOBJDUMP_SIM_FILE"); + char command[1000]; + std::string app_binary = get_app_binary(); + // Running cuobjdump using dynamic link to current process + snprintf(command, 1000, "md5sum %s ", app_binary.c_str()); + printf("Running md5sum using \"%s\"\n", command); + if (system(command)) { + std::cout << "Failed to execute: " << command << std::endl; + exit(1); + } + // Running cuobjdump using dynamic link to current process + // Needs the option '-all' to extract PTX from CDP-enabled binary + + // dump ptx for all individial ptx files into sepearte files which is later + // used by ptxas. + int result = 0; +#if (CUDART_VERSION >= 6000) + extract_ptx_files_using_cuobjdump(context); + return; +#endif + // TODO: redundant to dump twice. how can it be prevented? + // dump only for specific arch + char fname[1024]; + if ((override_cuobjdump == NULL) || (strlen(override_cuobjdump) == 0)) { + snprintf(fname, 1024, "_cuobjdump_complete_output_XXXXXX"); + int fd = mkstemp(fname); + close(fd); + if (!gpgpu_ctx->device_runtime->g_cdp_enabled) + snprintf(command, 1000, + "$CUDA_INSTALL_PATH/bin/cuobjdump -ptx -elf -sass %s > %s", + app_binary.c_str(), fname); + else + snprintf(command, 1000, + "$CUDA_INSTALL_PATH/bin/cuobjdump -ptx -elf -sass -all %s > %s", + app_binary.c_str(), fname); + bool parse_output = true; + result = system(command); + if (result) { + if (context->get_device() + ->get_gpgpu() + ->get_config() + .experimental_lib_support() && + (result == 65280)) { + // Some CUDA application may exclusively use kernels provided by CUDA + // libraries (e.g. CUBLAS). Skipping cuobjdump extraction from the + // executable for this case. + // 65280 is the return code from cuobjdump denoting the specific error + // (tested on CUDA 4.0/4.1/4.2) + printf("WARNING: Failed to execute: %s\n", command); + printf(" Executable binary does not contain any GPU kernel.\n"); + parse_output = false; + } else { + printf("ERROR: Failed to execute: %s\n", command); exit(1); + } } - // Running cuobjdump using dynamic link to current process - // Needs the option '-all' to extract PTX from CDP-enabled binary - //dump ptx for all individial ptx files into sepearte files which is later used by ptxas. - int result=0; -#if (CUDART_VERSION >= 6000) - extract_ptx_files_using_cuobjdump(context); - return; -#endif - //TODO: redundant to dump twice. how can it be prevented? - //dump only for specific arch - char fname[1024]; - if ((override_cuobjdump == NULL) || (strlen(override_cuobjdump)==0)) { - snprintf(fname,1024,"_cuobjdump_complete_output_XXXXXX"); - int fd=mkstemp(fname); - close(fd); - if(!gpgpu_ctx->device_runtime->g_cdp_enabled) - snprintf(command,1000,"$CUDA_INSTALL_PATH/bin/cuobjdump -ptx -elf -sass %s > %s", app_binary.c_str(), fname); - else - snprintf(command,1000,"$CUDA_INSTALL_PATH/bin/cuobjdump -ptx -elf -sass -all %s > %s", app_binary.c_str(), fname); - bool parse_output = true; - result = system(command); - if(result) { - if (context->get_device()->get_gpgpu()->get_config().experimental_lib_support() && (result == 65280)) { - // Some CUDA application may exclusively use kernels provided by CUDA - // libraries (e.g. CUBLAS). Skipping cuobjdump extraction from the - // executable for this case. - // 65280 is the return code from cuobjdump denoting the specific error (tested on CUDA 4.0/4.1/4.2) - printf("WARNING: Failed to execute: %s\n", command); - printf(" Executable binary does not contain any GPU kernel.\n"); - parse_output = false; - } else { - printf("ERROR: Failed to execute: %s\n", command); - exit(1); - } - } + if (parse_output) { + printf("Parsing file %s\n", fname); + FILE *cuobjdump_in; + cuobjdump_in = fopen(fname, "r"); + + struct cuobjdump_parser parser; + parser.elfserial = 1; + parser.ptxserial = 1; + cuobjdump_lex_init(&(parser.scanner)); + cuobjdump_set_in(cuobjdump_in, (parser.scanner)); + cuobjdump_parse(parser.scanner, &parser, cuobjdumpSectionList); + cuobjdump_lex_destroy(parser.scanner); + fclose(cuobjdump_in); + printf("Done parsing!!!\n"); + } else { + printf("Parsing skipped for %s\n", fname); + } + + if (context->get_device() + ->get_gpgpu() + ->get_config() + .experimental_lib_support()) { + // Experimental library support + // Currently only for cufft + + std::stringstream cmd; + cmd << "ldd " << app_binary + << " | grep $CUDA_INSTALL_PATH | awk \'{print $3}\' > _tempfile_.txt"; + int result = system(cmd.str().c_str()); + if (result) { + std::cout << "Failed to execute: " << cmd.str() << std::endl; + exit(1); + } + std::ifstream libsf; + libsf.open("_tempfile_.txt"); + if (!libsf.is_open()) { + std::cout << "Failed to open: _tempfile_.txt" << std::endl; + exit(1); + } - if (parse_output) { - printf("Parsing file %s\n", fname); - FILE *cuobjdump_in; - cuobjdump_in = fopen(fname, "r"); - - struct cuobjdump_parser parser; - parser.elfserial = 1; - parser.ptxserial = 1; - cuobjdump_lex_init(&(parser.scanner)); - cuobjdump_set_in(cuobjdump_in, (parser.scanner)); - cuobjdump_parse(parser.scanner, &parser, cuobjdumpSectionList); - cuobjdump_lex_destroy(parser.scanner); - fclose(cuobjdump_in); - printf("Done parsing!!!\n"); - } else { - printf("Parsing skipped for %s\n", fname); + // Save the original section list + std::list tmpsl = cuobjdumpSectionList; + cuobjdumpSectionList.clear(); + + std::string line; + std::getline(libsf, line); + std::cout << "DOING: " << line << std::endl; + int cnt = 1; + while (libsf.good()) { + std::stringstream libcodfn; + libcodfn << "_cuobjdump_complete_lib_" << cnt << "_"; + cmd.str(""); // resetting + cmd << "$CUDA_INSTALL_PATH/bin/cuobjdump -ptx -elf -sass "; + cmd << line; + cmd << " > "; + cmd << libcodfn.str(); + std::cout << "Running cuobjdump on " << line << std::endl; + std::cout << "Using command: " << cmd.str() << std::endl; + result = system(cmd.str().c_str()); + if (result) { + printf("ERROR: Failed to execute: %s\n", command); + exit(1); } + std::cout << "Done" << std::endl; + + std::cout << "Trying to parse " << libcodfn.str() << std::endl; + FILE *cuobjdump_in; + cuobjdump_in = fopen(libcodfn.str().c_str(), "r"); + struct cuobjdump_parser parser; + parser.elfserial = 1; + parser.ptxserial = 1; + cuobjdump_lex_init(&(parser.scanner)); + cuobjdump_set_in(cuobjdump_in, (parser.scanner)); + cuobjdump_parse(parser.scanner, &parser, cuobjdumpSectionList); + cuobjdump_lex_destroy(parser.scanner); + fclose(cuobjdump_in); + std::getline(libsf, line); + } + libSectionList = cuobjdumpSectionList; - if (context->get_device()->get_gpgpu()->get_config().experimental_lib_support()){ - //Experimental library support - //Currently only for cufft - - std::stringstream cmd; - cmd << "ldd " << app_binary << " | grep $CUDA_INSTALL_PATH | awk \'{print $3}\' > _tempfile_.txt"; - int result = system(cmd.str().c_str()); - if(result){ - std::cout << "Failed to execute: " << cmd.str() << std::endl; - exit(1); - } - std::ifstream libsf; - libsf.open("_tempfile_.txt"); - if(!libsf.is_open()) { - std::cout << "Failed to open: _tempfile_.txt" << std::endl; - exit(1); - } - - //Save the original section list - std::list tmpsl = cuobjdumpSectionList; - cuobjdumpSectionList.clear(); - - std::string line; - std::getline(libsf, line); - std::cout << "DOING: " << line << std::endl; - int cnt=1; - while(libsf.good()){ - std::stringstream libcodfn; - libcodfn << "_cuobjdump_complete_lib_" << cnt << "_"; - cmd.str(""); //resetting - cmd << "$CUDA_INSTALL_PATH/bin/cuobjdump -ptx -elf -sass "; - cmd << line; - cmd << " > "; - cmd << libcodfn.str(); - std::cout << "Running cuobjdump on " << line << std::endl; - std::cout << "Using command: " << cmd.str() << std::endl; - result = system(cmd.str().c_str()); - if(result) {printf("ERROR: Failed to execute: %s\n", command); exit(1);} - std::cout << "Done" << std::endl; - - std::cout << "Trying to parse " << libcodfn.str() << std::endl; - FILE *cuobjdump_in; - cuobjdump_in = fopen(libcodfn.str().c_str(), "r"); - struct cuobjdump_parser parser; - parser.elfserial = 1; - parser.ptxserial = 1; - cuobjdump_lex_init(&(parser.scanner)); - cuobjdump_set_in(cuobjdump_in, (parser.scanner)); - cuobjdump_parse(parser.scanner, &parser, cuobjdumpSectionList); - cuobjdump_lex_destroy(parser.scanner); - fclose(cuobjdump_in); - std::getline(libsf, line); - } - libSectionList = cuobjdumpSectionList; - - //Restore the original section list - cuobjdumpSectionList = tmpsl; - } - } else { - printf("GPGPU-Sim PTX: overriding cuobjdump with '%s' (CUOBJDUMP_SIM_FILE is set)\n", override_cuobjdump); - snprintf(fname,1024, "%s",override_cuobjdump); + // Restore the original section list + cuobjdumpSectionList = tmpsl; } + } else { + printf( + "GPGPU-Sim PTX: overriding cuobjdump with '%s' (CUOBJDUMP_SIM_FILE is " + "set)\n", + override_cuobjdump); + snprintf(fname, 1024, "%s", override_cuobjdump); + } } //! Read file into char* -//TODO: convert this to C++ streams, will be way cleaner -char* readfile (const std::string filename){ - assert (filename != ""); - FILE* fp = fopen(filename.c_str(),"r"); - if (!fp) { - std::cout << "ERROR: Could not open file %s for reading\n" << filename << std::endl; - assert (0); - } - // finding size of the file - int filesize= 0; - fseek (fp , 0 , SEEK_END); - - filesize = ftell (fp); - fseek (fp, 0, SEEK_SET); - // allocate and copy the entire ptx - char* ret = (char*)malloc((filesize +1)* sizeof(char)); - fread(ret,1,filesize,fp); - ret[filesize]='\0'; - fclose(fp); - return ret; +// TODO: convert this to C++ streams, will be way cleaner +char *readfile(const std::string filename) { + assert(filename != ""); + FILE *fp = fopen(filename.c_str(), "r"); + if (!fp) { + std::cout << "ERROR: Could not open file %s for reading\n" + << filename << std::endl; + assert(0); + } + // finding size of the file + int filesize = 0; + fseek(fp, 0, SEEK_END); + + filesize = ftell(fp); + fseek(fp, 0, SEEK_SET); + // allocate and copy the entire ptx + char *ret = (char *)malloc((filesize + 1) * sizeof(char)); + fread(ret, 1, filesize, fp); + ret[filesize] = '\0'; + fclose(fp); + return ret; } //! Function that helps debugging -void printSectionList(std::list sl) { - std::list::iterator iter; - for ( iter = sl.begin(); - iter != sl.end(); - iter++ - ){ - (*iter)->print(); - } +void printSectionList(std::list sl) { + std::list::iterator iter; + for (iter = sl.begin(); iter != sl.end(); iter++) { + (*iter)->print(); + } } //! Remove unecessary sm versions from the section list -std::list cuda_runtime_api::pruneSectionList(CUctx_st *context) { - unsigned forced_max_capability = context->get_device()->get_gpgpu()->get_config().get_forced_max_capability(); - - //For ptxplus, force the max capability to 19 if it's higher or unspecified(0) - if (context->get_device()->get_gpgpu()->get_config().convert_to_ptxplus()){ - if ( (forced_max_capability == 0) || - (forced_max_capability >= 20)){ - printf("GPGPU-Sim: WARNING: Capability >= 20 are not supported in PTXPlus\n\tSetting forced_max_capability to 19\n"); - forced_max_capability = 19; - } - } - - std::list prunedList; - - //Find the highest capability (that is lower than the forced maximum) for each cubin file - //and set it in cuobjdumpSectionMap. Do this only for ptx sections - std::map cuobjdumpSectionMap; - int min_ptx_capability_found=0; - for ( std::list::iterator iter = cuobjdumpSectionList.begin(); - iter != cuobjdumpSectionList.end(); - iter++){ - unsigned capability = (*iter)->getArch(); - if(dynamic_cast(*iter) != NULL){ - if(capabilitygetIdentifier())==cuobjdumpSectionMap.end()) - || (cuobjdumpSectionMap[(*iter)->getIdentifier()] < capability)) - cuobjdumpSectionMap[(*iter)->getIdentifier()] = capability; - } - } - } - - //Throw away the sections with the lower capabilites and push those with the highest in - //the pruned list - for ( std::list::iterator iter = cuobjdumpSectionList.begin(); - iter != cuobjdumpSectionList.end(); - iter++){ - unsigned capability = (*iter)->getArch(); - if(capability == cuobjdumpSectionMap[(*iter)->getIdentifier()]){ - prunedList.push_back(*iter); - } else { - delete *iter; - } - } - if(prunedList.empty()){ - printf("Error: No PTX sections found with sm capability that is lower than current forced maximum capability \n minimum ptx capability found = %u, maximum forced ptx capability = %u \n User might want to change either the forced maximum capability from gpgpusim configuration or update the compilation to generate the required PTX version\n",min_ptx_capability_found,forced_max_capability); - abort(); - } - return prunedList; +std::list cuda_runtime_api::pruneSectionList( + CUctx_st *context) { + unsigned forced_max_capability = context->get_device() + ->get_gpgpu() + ->get_config() + .get_forced_max_capability(); + + // For ptxplus, force the max capability to 19 if it's higher or + // unspecified(0) + if (context->get_device()->get_gpgpu()->get_config().convert_to_ptxplus()) { + if ((forced_max_capability == 0) || (forced_max_capability >= 20)) { + printf( + "GPGPU-Sim: WARNING: Capability >= 20 are not supported in " + "PTXPlus\n\tSetting forced_max_capability to 19\n"); + forced_max_capability = 19; + } + } + + std::list prunedList; + + // Find the highest capability (that is lower than the forced maximum) for + // each cubin file and set it in cuobjdumpSectionMap. Do this only for ptx + // sections + std::map cuobjdumpSectionMap; + int min_ptx_capability_found = 0; + for (std::list::iterator iter = + cuobjdumpSectionList.begin(); + iter != cuobjdumpSectionList.end(); iter++) { + unsigned capability = (*iter)->getArch(); + if (dynamic_cast(*iter) != NULL) { + if (capability < min_ptx_capability_found || + min_ptx_capability_found == 0) + min_ptx_capability_found = capability; + if (capability <= forced_max_capability || forced_max_capability == 0) { + if ((cuobjdumpSectionMap.find((*iter)->getIdentifier()) == + cuobjdumpSectionMap.end()) || + (cuobjdumpSectionMap[(*iter)->getIdentifier()] < capability)) + cuobjdumpSectionMap[(*iter)->getIdentifier()] = capability; + } + } + } + + // Throw away the sections with the lower capabilites and push those with the + // highest in the pruned list + for (std::list::iterator iter = + cuobjdumpSectionList.begin(); + iter != cuobjdumpSectionList.end(); iter++) { + unsigned capability = (*iter)->getArch(); + if (capability == cuobjdumpSectionMap[(*iter)->getIdentifier()]) { + prunedList.push_back(*iter); + } else { + delete *iter; + } + } + if (prunedList.empty()) { + printf( + "Error: No PTX sections found with sm capability that is lower than " + "current forced maximum capability \n minimum ptx capability found = " + "%u, maximum forced ptx capability = %u \n User might want to change " + "either the forced maximum capability from gpgpusim configuration or " + "update the compilation to generate the required PTX version\n", + min_ptx_capability_found, forced_max_capability); + abort(); + } + return prunedList; } //! Merge all PTX sections that have a specific identifier into one file -std::list cuda_runtime_api::mergeMatchingSections(std::string identifier){ - const char *ptxcode = ""; - std::list::iterator old_iter; - cuobjdumpPTXSection* old_ptxsection = NULL; - cuobjdumpPTXSection* ptxsection; - std::list mergedList; - - for ( std::list::iterator iter = cuobjdumpSectionList.begin(); - iter != cuobjdumpSectionList.end(); - iter++){ - if((ptxsection=dynamic_cast(*iter)) != NULL && - strcmp(ptxsection->getIdentifier().c_str(), identifier.c_str()) == 0){ - // Read and remove the last PTX section - if (old_ptxsection != NULL) { - ptxcode = readfile(old_ptxsection->getPTXfilename()); - // remove ptx file? - delete *old_iter; - } - - // Append all the PTX from the last PTX section into the current PTX section - // Add 50 to ptxcode to ignore the information regarding version/target/address_size - if (strlen(ptxcode) >= 50) { - FILE *ptxfile = fopen((ptxsection->getPTXfilename()).c_str(), "a"); - fprintf(ptxfile, "%s", ptxcode + 50); - fclose(ptxfile); - } - - old_iter = iter; - old_ptxsection = ptxsection; - } - // Store all non-PTX sections and PTX sections with non-matching identifiers - else { - mergedList.push_back(*iter); - } - } - - // Store the final PTX section - mergedList.push_back(*old_iter); - - return mergedList; +std::list cuda_runtime_api::mergeMatchingSections( + std::string identifier) { + const char *ptxcode = ""; + std::list::iterator old_iter; + cuobjdumpPTXSection *old_ptxsection = NULL; + cuobjdumpPTXSection *ptxsection; + std::list mergedList; + + for (std::list::iterator iter = + cuobjdumpSectionList.begin(); + iter != cuobjdumpSectionList.end(); iter++) { + if ((ptxsection = dynamic_cast(*iter)) != NULL && + strcmp(ptxsection->getIdentifier().c_str(), identifier.c_str()) == 0) { + // Read and remove the last PTX section + if (old_ptxsection != NULL) { + ptxcode = readfile(old_ptxsection->getPTXfilename()); + // remove ptx file? + delete *old_iter; + } + + // Append all the PTX from the last PTX section into the current PTX + // section Add 50 to ptxcode to ignore the information regarding + // version/target/address_size + if (strlen(ptxcode) >= 50) { + FILE *ptxfile = fopen((ptxsection->getPTXfilename()).c_str(), "a"); + fprintf(ptxfile, "%s", ptxcode + 50); + fclose(ptxfile); + } + + old_iter = iter; + old_ptxsection = ptxsection; + } + // Store all non-PTX sections and PTX sections with non-matching identifiers + else { + mergedList.push_back(*iter); + } + } + + // Store the final PTX section + mergedList.push_back(*old_iter); + + return mergedList; } //! Merge any PTX sections with matching identifiers -std::list cuda_runtime_api::mergeSections(){ - std::vector identifier; - cuobjdumpPTXSection* ptxsection; - - // Add all identifiers present in PTX sections to a vector - for ( std::list::iterator iter = cuobjdumpSectionList.begin(); - iter != cuobjdumpSectionList.end(); - iter++){ - if((ptxsection=dynamic_cast(*iter)) != NULL){ - std::string current_id = ptxsection->getIdentifier(); - - // If we haven't yet seen a given identifier, add it to the vector - if (std::find(identifier.begin(), identifier.end(), current_id) == identifier.end()) { - identifier.push_back(current_id); - } - } - } - - // Call mergeMatchingSections on all identifiers in the vector - for ( std::vector::iterator iter = identifier.begin(); - iter != identifier.end(); - iter++) { - cuobjdumpSectionList = mergeMatchingSections(*iter); - } - - return cuobjdumpSectionList; -} - - -//! Within the section list, find the ELF section corresponding to a given identifier -cuobjdumpELFSection* findELFSectionInList(std::list sectionlist, const std::string identifier){ - - std::list::iterator iter; - for ( iter = sectionlist.begin(); - iter != sectionlist.end(); - iter++ - ){ - cuobjdumpELFSection* elfsection; - if((elfsection=dynamic_cast(*iter)) != NULL){ - if(elfsection->getIdentifier() == identifier) - return elfsection; - } - } - return NULL; +std::list cuda_runtime_api::mergeSections() { + std::vector identifier; + cuobjdumpPTXSection *ptxsection; + + // Add all identifiers present in PTX sections to a vector + for (std::list::iterator iter = + cuobjdumpSectionList.begin(); + iter != cuobjdumpSectionList.end(); iter++) { + if ((ptxsection = dynamic_cast(*iter)) != NULL) { + std::string current_id = ptxsection->getIdentifier(); + + // If we haven't yet seen a given identifier, add it to the vector + if (std::find(identifier.begin(), identifier.end(), current_id) == + identifier.end()) { + identifier.push_back(current_id); + } + } + } + + // Call mergeMatchingSections on all identifiers in the vector + for (std::vector::iterator iter = identifier.begin(); + iter != identifier.end(); iter++) { + cuobjdumpSectionList = mergeMatchingSections(*iter); + } + + return cuobjdumpSectionList; +} + +//! Within the section list, find the ELF section corresponding to a given +//! identifier +cuobjdumpELFSection *findELFSectionInList( + std::list sectionlist, const std::string identifier) { + std::list::iterator iter; + for (iter = sectionlist.begin(); iter != sectionlist.end(); iter++) { + cuobjdumpELFSection *elfsection; + if ((elfsection = dynamic_cast(*iter)) != NULL) { + if (elfsection->getIdentifier() == identifier) return elfsection; + } + } + return NULL; } //! Find an ELF section in all the known lists -cuobjdumpELFSection* cuda_runtime_api::findELFSection(const std::string identifier){ - cuobjdumpELFSection* sec = findELFSectionInList(cuobjdumpSectionList, identifier); - if (sec!=NULL)return sec; - sec = findELFSectionInList(libSectionList, identifier); - if (sec!=NULL)return sec; - std::cout << "Could not find " << identifier << std::endl; - assert(0 && "Could not find the required ELF section"); - return NULL; -} - -//! Within the section list, find the PTX section corresponding to a given identifier -cuobjdumpPTXSection* cuda_runtime_api::findPTXSectionInList(std::list §ionlist, const std::string identifier){ - std::list::iterator iter; - for ( iter = sectionlist.begin(); - iter != sectionlist.end(); - iter++ - ){ - cuobjdumpPTXSection* ptxsection; - if((ptxsection=dynamic_cast(*iter)) != NULL){ - if(ptxsection->getIdentifier() == identifier) - return ptxsection; - else { - if(gpgpu_ctx->device_runtime->g_cdp_enabled) { - printf("Warning: __cudaRegisterFatBinary needs %s, but find PTX section with %s\n", - identifier.c_str(), ptxsection->getIdentifier().c_str()); - return ptxsection; - } - } - } - } - return NULL; +cuobjdumpELFSection *cuda_runtime_api::findELFSection( + const std::string identifier) { + cuobjdumpELFSection *sec = + findELFSectionInList(cuobjdumpSectionList, identifier); + if (sec != NULL) return sec; + sec = findELFSectionInList(libSectionList, identifier); + if (sec != NULL) return sec; + std::cout << "Could not find " << identifier << std::endl; + assert(0 && "Could not find the required ELF section"); + return NULL; +} + +//! Within the section list, find the PTX section corresponding to a given +//! identifier +cuobjdumpPTXSection *cuda_runtime_api::findPTXSectionInList( + std::list §ionlist, const std::string identifier) { + std::list::iterator iter; + for (iter = sectionlist.begin(); iter != sectionlist.end(); iter++) { + cuobjdumpPTXSection *ptxsection; + if ((ptxsection = dynamic_cast(*iter)) != NULL) { + if (ptxsection->getIdentifier() == identifier) + return ptxsection; + else { + if (gpgpu_ctx->device_runtime->g_cdp_enabled) { + printf( + "Warning: __cudaRegisterFatBinary needs %s, but find PTX section " + "with %s\n", + identifier.c_str(), ptxsection->getIdentifier().c_str()); + return ptxsection; + } + } + } + } + return NULL; } //! Find an PTX section in all the known lists -cuobjdumpPTXSection* cuda_runtime_api::findPTXSection(const std::string identifier){ - cuobjdumpPTXSection* sec = findPTXSectionInList(cuobjdumpSectionList, identifier); - if (sec!=NULL)return sec; - sec = findPTXSectionInList(libSectionList, identifier); - if (sec!=NULL)return sec; - std::cout << "Could not find " << identifier << std::endl; - assert(0 && "Could not find the required PTX section"); - return NULL; +cuobjdumpPTXSection *cuda_runtime_api::findPTXSection( + const std::string identifier) { + cuobjdumpPTXSection *sec = + findPTXSectionInList(cuobjdumpSectionList, identifier); + if (sec != NULL) return sec; + sec = findPTXSectionInList(libSectionList, identifier); + if (sec != NULL) return sec; + std::cout << "Could not find " << identifier << std::endl; + assert(0 && "Could not find the required PTX section"); + return NULL; } - - //! Extract the code using cuobjdump and remove unnecessary sections -void cuda_runtime_api::cuobjdumpInit(){ - CUctx_st *context = GPGPUSim_Context(gpgpu_ctx); - extract_code_using_cuobjdump(); //extract all the output of cuobjdump to _cuobjdump_*.* - const char* pre_load = getenv("CUOBJDUMP_SIM_FILE"); - if (pre_load ==NULL || strlen(pre_load)==0){ - cuobjdumpSectionList = pruneSectionList(context); - cuobjdumpSectionList = mergeSections(); - } +void cuda_runtime_api::cuobjdumpInit() { + CUctx_st *context = GPGPUSim_Context(gpgpu_ctx); + extract_code_using_cuobjdump(); // extract all the output of cuobjdump to + // _cuobjdump_*.* + const char *pre_load = getenv("CUOBJDUMP_SIM_FILE"); + if (pre_load == NULL || strlen(pre_load) == 0) { + cuobjdumpSectionList = pruneSectionList(context); + cuobjdumpSectionList = mergeSections(); + } } - //! Either submit PTX for simulation or convert SASS to PTXPlus and submit it -void gpgpu_context::cuobjdumpParseBinary(unsigned int handle){ - - CUctx_st *context = GPGPUSim_Context(this); - if(api->fatbin_registered[handle]) return; - api->fatbin_registered[handle] = true; - std::string fname = api->fatbinmap[handle]; - - if (api->name_symtab.find(fname) != api->name_symtab.end()) { - symbol_table *symtab = api->name_symtab[fname]; - context->add_binary(symtab, handle); - return; - } - symbol_table *symtab; +void gpgpu_context::cuobjdumpParseBinary(unsigned int handle) { + CUctx_st *context = GPGPUSim_Context(this); + if (api->fatbin_registered[handle]) return; + api->fatbin_registered[handle] = true; + std::string fname = api->fatbinmap[handle]; + + if (api->name_symtab.find(fname) != api->name_symtab.end()) { + symbol_table *symtab = api->name_symtab[fname]; + context->add_binary(symtab, handle); + return; + } + symbol_table *symtab; #if (CUDART_VERSION >= 6000) - //loops through all ptx files from smallest sm version to largest - std::map >::iterator itr_m; - for (itr_m = api->version_filename.begin(); itr_m!=api->version_filename.end(); itr_m++){ - std::set::iterator itr_s; - for (itr_s = itr_m->second.begin(); itr_s!=itr_m->second.end(); itr_s++){ - std::string ptx_filename = *itr_s; - printf("GPGPU-Sim PTX: Parsing %s\n",ptx_filename.c_str()); - symtab = gpgpu_ptx_sim_load_ptx_from_filename( ptx_filename.c_str() ); - } - } - api->name_symtab[fname] = symtab; - context->add_binary(symtab, handle); - api->load_static_globals(symtab,STATIC_ALLOC_LIMIT,0xFFFFFFFF,context->get_device()->get_gpgpu()); - api->load_constants(symtab,STATIC_ALLOC_LIMIT,context->get_device()->get_gpgpu()); - for (itr_m = api->version_filename.begin(); itr_m!=api->version_filename.end(); itr_m++){ - std::set::iterator itr_s; - for (itr_s = itr_m->second.begin(); itr_s!=itr_m->second.end(); itr_s++){ - std::string ptx_filename = *itr_s; - printf("GPGPU-Sim PTX: Loading PTXInfo from %s\n",ptx_filename.c_str()); - gpgpu_ptx_info_load_from_filename( ptx_filename.c_str(), itr_m->first ); - } - } - return; + // loops through all ptx files from smallest sm version to largest + std::map >::iterator itr_m; + for (itr_m = api->version_filename.begin(); + itr_m != api->version_filename.end(); itr_m++) { + std::set::iterator itr_s; + for (itr_s = itr_m->second.begin(); itr_s != itr_m->second.end(); itr_s++) { + std::string ptx_filename = *itr_s; + printf("GPGPU-Sim PTX: Parsing %s\n", ptx_filename.c_str()); + symtab = gpgpu_ptx_sim_load_ptx_from_filename(ptx_filename.c_str()); + } + } + api->name_symtab[fname] = symtab; + context->add_binary(symtab, handle); + api->load_static_globals(symtab, STATIC_ALLOC_LIMIT, 0xFFFFFFFF, + context->get_device()->get_gpgpu()); + api->load_constants(symtab, STATIC_ALLOC_LIMIT, + context->get_device()->get_gpgpu()); + for (itr_m = api->version_filename.begin(); + itr_m != api->version_filename.end(); itr_m++) { + std::set::iterator itr_s; + for (itr_s = itr_m->second.begin(); itr_s != itr_m->second.end(); itr_s++) { + std::string ptx_filename = *itr_s; + printf("GPGPU-Sim PTX: Loading PTXInfo from %s\n", ptx_filename.c_str()); + gpgpu_ptx_info_load_from_filename(ptx_filename.c_str(), itr_m->first); + } + } + return; #endif - unsigned max_capability = 0; - for ( std::list::iterator iter = api->cuobjdumpSectionList.begin(); - iter != api->cuobjdumpSectionList.end(); - iter++){ - unsigned capability = (*iter)->getArch(); - if (capability > max_capability) max_capability = capability; - } - if (max_capability > 20) printf("WARNING: No guarantee that PTX will be parsed for SM version %u\n", max_capability); - if (max_capability == 0) max_capability=context->get_device()->get_gpgpu()->get_config().get_forced_max_capability(); - - cuobjdumpPTXSection* ptx = NULL; - const char* pre_load = getenv("CUOBJDUMP_SIM_FILE"); - if(pre_load==NULL || strlen(pre_load)==0) - ptx = api->findPTXSection(fname); - char *ptxcode; - const char *override_ptx_name = getenv("PTX_SIM_KERNELFILE"); - if (override_ptx_name == NULL or getenv("PTX_SIM_USE_PTX_FILE") == NULL or strlen(getenv("PTX_SIM_USE_PTX_FILE"))==0) { - ptxcode = readfile(ptx->getPTXfilename()); - } else { - printf("GPGPU-Sim PTX: overriding embedded ptx with '%s' (PTX_SIM_USE_PTX_FILE is set)\n", override_ptx_name); - ptxcode = readfile(override_ptx_name); - } - if(context->get_device()->get_gpgpu()->get_config().convert_to_ptxplus() ) { - cuobjdumpELFSection* elfsection = api->findELFSection(ptx->getIdentifier()); - assert (elfsection!= NULL); - char *ptxplus_str = ptxinfo->gpgpu_ptx_sim_convert_ptx_and_sass_to_ptxplus( - ptx->getPTXfilename(), - elfsection->getELFfilename(), - elfsection->getSASSfilename()); - symtab=gpgpu_ptx_sim_load_ptx_from_string(ptxplus_str, handle); - printf("Adding %s with cubin handle %u\n", ptx->getPTXfilename().c_str(), handle); - context->add_binary(symtab, handle); - gpgpu_ptxinfo_load_from_string( ptxcode, handle, max_capability, context->no_of_ptx ); - delete[] ptxplus_str; - } else { - symtab=gpgpu_ptx_sim_load_ptx_from_string(ptxcode, handle); - //if CUOBJDUMP_SIM_FILE is not set, ptx is NULL. So comment below. - //printf("Adding %s with cubin handle %u\n", ptx->getPTXfilename().c_str(), handle); - context->add_binary(symtab, handle); - gpgpu_ptxinfo_load_from_string( ptxcode, handle, max_capability, context->no_of_ptx ); - } - api->load_static_globals(symtab,STATIC_ALLOC_LIMIT,0xFFFFFFFF,context->get_device()->get_gpgpu()); - api->load_constants(symtab,STATIC_ALLOC_LIMIT,context->get_device()->get_gpgpu()); - api->name_symtab[fname] = symtab; - - //TODO: Remove temporarily files as per configurations + unsigned max_capability = 0; + for (std::list::iterator iter = + api->cuobjdumpSectionList.begin(); + iter != api->cuobjdumpSectionList.end(); iter++) { + unsigned capability = (*iter)->getArch(); + if (capability > max_capability) max_capability = capability; + } + if (max_capability > 20) + printf("WARNING: No guarantee that PTX will be parsed for SM version %u\n", + max_capability); + if (max_capability == 0) + max_capability = context->get_device() + ->get_gpgpu() + ->get_config() + .get_forced_max_capability(); + + cuobjdumpPTXSection *ptx = NULL; + const char *pre_load = getenv("CUOBJDUMP_SIM_FILE"); + if (pre_load == NULL || strlen(pre_load) == 0) + ptx = api->findPTXSection(fname); + char *ptxcode; + const char *override_ptx_name = getenv("PTX_SIM_KERNELFILE"); + if (override_ptx_name == NULL or getenv("PTX_SIM_USE_PTX_FILE") == NULL or + strlen(getenv("PTX_SIM_USE_PTX_FILE")) == 0) { + ptxcode = readfile(ptx->getPTXfilename()); + } else { + printf( + "GPGPU-Sim PTX: overriding embedded ptx with '%s' " + "(PTX_SIM_USE_PTX_FILE is set)\n", + override_ptx_name); + ptxcode = readfile(override_ptx_name); + } + if (context->get_device()->get_gpgpu()->get_config().convert_to_ptxplus()) { + cuobjdumpELFSection *elfsection = api->findELFSection(ptx->getIdentifier()); + assert(elfsection != NULL); + char *ptxplus_str = ptxinfo->gpgpu_ptx_sim_convert_ptx_and_sass_to_ptxplus( + ptx->getPTXfilename(), elfsection->getELFfilename(), + elfsection->getSASSfilename()); + symtab = gpgpu_ptx_sim_load_ptx_from_string(ptxplus_str, handle); + printf("Adding %s with cubin handle %u\n", ptx->getPTXfilename().c_str(), + handle); + context->add_binary(symtab, handle); + gpgpu_ptxinfo_load_from_string(ptxcode, handle, max_capability, + context->no_of_ptx); + delete[] ptxplus_str; + } else { + symtab = gpgpu_ptx_sim_load_ptx_from_string(ptxcode, handle); + // if CUOBJDUMP_SIM_FILE is not set, ptx is NULL. So comment below. + // printf("Adding %s with cubin handle %u\n", ptx->getPTXfilename().c_str(), + // handle); + context->add_binary(symtab, handle); + gpgpu_ptxinfo_load_from_string(ptxcode, handle, max_capability, + context->no_of_ptx); + } + api->load_static_globals(symtab, STATIC_ALLOC_LIMIT, 0xFFFFFFFF, + context->get_device()->get_gpgpu()); + api->load_constants(symtab, STATIC_ALLOC_LIMIT, + context->get_device()->get_gpgpu()); + api->name_symtab[fname] = symtab; + + // TODO: Remove temporarily files as per configurations } } extern "C" { -void** CUDARTAPI __cudaRegisterFatBinary( void *fatCubin ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return cudaRegisterFatBinaryInternal(fatCubin); +void **CUDARTAPI __cudaRegisterFatBinary(void *fatCubin) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return cudaRegisterFatBinaryInternal(fatCubin); } -void CUDARTAPI __cudaRegisterFatBinaryEnd( void **fatCubinHandle ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } +void CUDARTAPI __cudaRegisterFatBinaryEnd(void **fatCubinHandle) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } } -unsigned CUDARTAPI __cudaPushCallConfiguration(dim3 gridDim, - dim3 blockDim, - size_t sharedMem = 0, - struct CUstream_st *stream = 0) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cudaConfigureCallInternal(gridDim, blockDim, sharedMem, stream); +unsigned CUDARTAPI __cudaPushCallConfiguration(dim3 gridDim, dim3 blockDim, + size_t sharedMem = 0, + struct CUstream_st *stream = 0) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cudaConfigureCallInternal(gridDim, blockDim, sharedMem, stream); } -cudaError_t CUDARTAPI __cudaPopCallConfiguration( - dim3 *gridDim, - dim3 *blockDim, - size_t *sharedMem, - void *stream -) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return g_last_cudaError = cudaSuccess; +cudaError_t CUDARTAPI __cudaPopCallConfiguration(dim3 *gridDim, dim3 *blockDim, + size_t *sharedMem, + void *stream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return g_last_cudaError = cudaSuccess; } - -void CUDARTAPI __cudaRegisterFunction( - void **fatCubinHandle, - const char *hostFun, - char *deviceFun, - const char *deviceName, - int thread_limit, - uint3 *tid, - uint3 *bid, - dim3 *bDim, - dim3 *gDim -) { - cudaRegisterFunctionInternal( - fatCubinHandle, - hostFun, - deviceFun, - deviceName, - thread_limit, - tid, - bid, - bDim, - gDim - ); - +void CUDARTAPI __cudaRegisterFunction(void **fatCubinHandle, + const char *hostFun, char *deviceFun, + const char *deviceName, int thread_limit, + uint3 *tid, uint3 *bid, dim3 *bDim, + dim3 *gDim) { + cudaRegisterFunctionInternal(fatCubinHandle, hostFun, deviceFun, deviceName, + thread_limit, tid, bid, bDim, gDim); } extern void __cudaRegisterVar( - void **fatCubinHandle, - char *hostVar, //pointer to...something - char *deviceAddress, //name of variable - const char *deviceName, //name of variable (same as above) - int ext, - int size, - int constant, - int global ) -{ - cudaRegisterVarInternal( - fatCubinHandle, - hostVar, - deviceAddress, - deviceName, - ext, - size, - constant, - global ); -} - -__host__ cudaError_t CUDARTAPI cudaConfigureCall(dim3 gridDim, dim3 blockDim, size_t sharedMem, cudaStream_t stream) -{ - return cudaConfigureCallInternal(gridDim, blockDim, sharedMem, stream); + void **fatCubinHandle, + char *hostVar, // pointer to...something + char *deviceAddress, // name of variable + const char *deviceName, // name of variable (same as above) + int ext, int size, int constant, int global) { + cudaRegisterVarInternal(fatCubinHandle, hostVar, deviceAddress, deviceName, + ext, size, constant, global); } -void __cudaUnregisterFatBinary(void **fatCubinHandle) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } +__host__ cudaError_t CUDARTAPI cudaConfigureCall(dim3 gridDim, dim3 blockDim, + size_t sharedMem, + cudaStream_t stream) { + return cudaConfigureCallInternal(gridDim, blockDim, sharedMem, stream); } -cudaError_t cudaDeviceReset ( void ) { - // Should reset the simulated GPU - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return g_last_cudaError = cudaSuccess; +void __cudaUnregisterFatBinary(void **fatCubinHandle) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } } -cudaError_t CUDARTAPI cudaDeviceSynchronize(void) -{ - return cudaDeviceSynchronizeInternal(); +cudaError_t cudaDeviceReset(void) { + // Should reset the simulated GPU + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return g_last_cudaError = cudaSuccess; } -void __cudaRegisterShared( - void **fatCubinHandle, - void **devicePtr -) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - // we don't do anything here - printf("GPGPU-Sim PTX: __cudaRegisterShared\n" ); +cudaError_t CUDARTAPI cudaDeviceSynchronize(void) { + return cudaDeviceSynchronizeInternal(); } -void CUDARTAPI __cudaRegisterSharedVar( - void **fatCubinHandle, - void **devicePtr, - size_t size, - size_t alignment, - int storage -) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - // we don't do anything here - printf("GPGPU-Sim PTX: __cudaRegisterSharedVar\n" ); +void __cudaRegisterShared(void **fatCubinHandle, void **devicePtr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // we don't do anything here + printf("GPGPU-Sim PTX: __cudaRegisterShared\n"); } -void __cudaRegisterTexture( - void **fatCubinHandle, - const struct textureReference *hostVar, - const void **deviceAddress, - const char *deviceName, - int dim, - int norm, - int ext -) //passes in a newly created textureReference -{ - __cudaRegisterTextureInternal(fatCubinHandle, hostVar, deviceAddress, deviceName, dim, norm, ext); +void CUDARTAPI __cudaRegisterSharedVar(void **fatCubinHandle, void **devicePtr, + size_t size, size_t alignment, + int storage) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // we don't do anything here + printf("GPGPU-Sim PTX: __cudaRegisterSharedVar\n"); } - -char __cudaInitModule( - void **fatCubinHandle -) +void __cudaRegisterTexture( + void **fatCubinHandle, const struct textureReference *hostVar, + const void **deviceAddress, const char *deviceName, int dim, int norm, + int ext) // passes in a newly created textureReference { - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; + __cudaRegisterTextureInternal(fatCubinHandle, hostVar, deviceAddress, + deviceName, dim, norm, ext); } +char __cudaInitModule(void **fatCubinHandle) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; +} -cudaError_t cudaGLRegisterBufferObject(GLuint bufferObj) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("GPGPU-Sim PTX: Execution warning: ignoring call to \"%s\"\n", __my_func__ ); - return g_last_cudaError = cudaSuccess; +cudaError_t cudaGLRegisterBufferObject(GLuint bufferObj) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("GPGPU-Sim PTX: Execution warning: ignoring call to \"%s\"\n", + __my_func__); + return g_last_cudaError = cudaSuccess; } -cudaError_t cudaGLMapBufferObject(void** devPtr, GLuint bufferObj) -{ - return cudaGLMapBufferObjectInternal(devPtr, bufferObj); +cudaError_t cudaGLMapBufferObject(void **devPtr, GLuint bufferObj) { + return cudaGLMapBufferObjectInternal(devPtr, bufferObj); } -cudaError_t cudaGLUnmapBufferObject(GLuint bufferObj) -{ - return cudaGLUnmapBufferObjectInternal(bufferObj); +cudaError_t cudaGLUnmapBufferObject(GLuint bufferObj) { + return cudaGLUnmapBufferObjectInternal(bufferObj); } -cudaError_t cudaGLUnregisterBufferObject(GLuint bufferObj) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("GPGPU-Sim PTX: Execution warning: ignoring call to \"%s\"\n", __my_func__ ); - return g_last_cudaError = cudaSuccess; +cudaError_t cudaGLUnregisterBufferObject(GLuint bufferObj) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("GPGPU-Sim PTX: Execution warning: ignoring call to \"%s\"\n", + __my_func__); + return g_last_cudaError = cudaSuccess; } #if (CUDART_VERSION >= 2010) -cudaError_t CUDARTAPI cudaHostAlloc(void **pHost, size_t bytes, unsigned int flags) -{ - return cudaHostAllocInternal(pHost, bytes, flags); +cudaError_t CUDARTAPI cudaHostAlloc(void **pHost, size_t bytes, + unsigned int flags) { + return cudaHostAllocInternal(pHost, bytes, flags); } -cudaError_t CUDARTAPI cudaHostGetDevicePointer(void **pDevice, void *pHost, unsigned int flags) -{ - return cudaHostGetDevicePointerInternal(pDevice, pHost, flags); +cudaError_t CUDARTAPI cudaHostGetDevicePointer(void **pDevice, void *pHost, + unsigned int flags) { + return cudaHostGetDevicePointerInternal(pDevice, pHost, flags); } -__host__ cudaError_t CUDARTAPI cudaPointerGetAttributes( - cudaPointerAttributes *attributes, - const void *ptr -) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI +cudaPointerGetAttributes(cudaPointerAttributes *attributes, const void *ptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t CUDARTAPI cudaDeviceCanAccessPeer( - int *canAccessPeer, - int device, - int peerDevice -) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaDeviceCanAccessPeer(int *canAccessPeer, + int device, + int peerDevice) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -__host__ cudaError_t CUDARTAPI cudaDeviceEnablePeerAccess( - int peerDevice, - unsigned int flags -) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +__host__ cudaError_t CUDARTAPI cudaDeviceEnablePeerAccess(int peerDevice, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -cudaError_t CUDARTAPI cudaSetValidDevices(int *device_arr, int len) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +cudaError_t CUDARTAPI cudaSetValidDevices(int *device_arr, int len) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -cudaError_t CUDARTAPI cudaSetDeviceFlags( int flags ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - // This flag is implicitly always on (unless you are using the driver API). It is safe for GPGPU-Sim to - // just ignore it. - if ( cudaDeviceMapHost == flags ) { - return g_last_cudaError = cudaSuccess; - } else { - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; - } +cudaError_t CUDARTAPI cudaSetDeviceFlags(int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // This flag is implicitly always on (unless you are using the driver API). It + // is safe for GPGPU-Sim to just ignore it. + if (cudaDeviceMapHost == flags) { + return g_last_cudaError = cudaSuccess; + } else { + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; + } } - -cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, const char *hostFun ) -{ - return cudaFuncGetAttributesInternal(attr, hostFun ); +cudaError_t CUDARTAPI cudaFuncGetAttributes(struct cudaFuncAttributes *attr, + const char *hostFun) { + return cudaFuncGetAttributesInternal(attr, hostFun); } -cudaError_t CUDARTAPI cudaDriverGetVersion(int *driverVersion) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - *driverVersion = CUDART_VERSION; - return g_last_cudaError = cudaSuccess; +cudaError_t CUDARTAPI cudaDriverGetVersion(int *driverVersion) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + *driverVersion = CUDART_VERSION; + return g_last_cudaError = cudaSuccess; } -cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - *runtimeVersion = CUDART_VERSION; - return g_last_cudaError = cudaSuccess; +cudaError_t CUDARTAPI cudaRuntimeGetVersion(int *runtimeVersion) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + *runtimeVersion = CUDART_VERSION; + return g_last_cudaError = cudaSuccess; } #if CUDART_VERSION >= 3000 -__host__ cudaError_t CUDARTAPI cudaFuncSetCacheConfig(const char *func, enum cudaFuncCache cacheConfig ) -{ - return cudaFuncSetCacheConfigInternal(func, cacheConfig); +__host__ cudaError_t CUDARTAPI +cudaFuncSetCacheConfig(const char *func, enum cudaFuncCache cacheConfig) { + return cudaFuncSetCacheConfigInternal(func, cacheConfig); } -//Jin: hack for cdp -__host__ cudaError_t CUDARTAPI cudaDeviceSetLimit(enum cudaLimit limit, size_t value) { - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - return g_last_cudaError = cudaSuccess; +// Jin: hack for cdp +__host__ cudaError_t CUDARTAPI cudaDeviceSetLimit(enum cudaLimit limit, + size_t value) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + return g_last_cudaError = cudaSuccess; } //#if CUDART_VERSION >= 9000 -//__host__ cudaError_t cudaFuncSetAttribute ( const void* func, enum cudaFuncAttribute attr, int value ) { +//__host__ cudaError_t cudaFuncSetAttribute ( const void* func, enum +// cudaFuncAttribute attr, int value ) { - //ignore this Attribute for now, and the default is that carveout = cudaSharedmemCarveoutDefault; // (-1) +// ignore this Attribute for now, and the default is that carveout = +// cudaSharedmemCarveoutDefault; // (-1) // return g_last_cudaError = cudaSuccess; //} - #endif #endif - #if CUDART_VERSION >= 9000 /** * \brief Set attributes for a given function * * This function sets the attributes of a function specified via \p entry. * The parameter \p entry must be a pointer to a function that executes - * on the device. The parameter specified by \p entry must be declared as a \p __global__ - * function. The enumeration defined by \p attr is set to the value defined by \p value - * If the specified function does not exist, then ::cudaErrorInvalidDeviceFunction is returned. - * If the specified attribute cannot be written, or if the value is incorrect, - * then ::cudaErrorInvalidValue is returned. + * on the device. The parameter specified by \p entry must be declared as a \p + * __global__ function. The enumeration defined by \p attr is set to the value + * defined by \p value If the specified function does not exist, then + * ::cudaErrorInvalidDeviceFunction is returned. If the specified attribute + * cannot be written, or if the value is incorrect, then ::cudaErrorInvalidValue + * is returned. * * Valid values for \p attr are: - * ::cuFuncAttrMaxDynamicSharedMem - Maximum size of dynamic shared memory per block - * ::cudaFuncAttributePreferredSharedMemoryCarveout - Preferred shared memory-L1 cache split ratio + * ::cuFuncAttrMaxDynamicSharedMem - Maximum size of dynamic shared memory per + * block + * ::cudaFuncAttributePreferredSharedMemoryCarveout - Preferred shared memory-L1 + * cache split ratio * * \param entry - Function to get attributes of * \param attr - Attribute to set @@ -3726,216 +3846,235 @@ __host__ cudaError_t CUDARTAPI cudaDeviceSetLimit(enum cudaLimit limit, size_t v * ::cudaErrorInvalidValue * \notefnerr * - * \ref ::cudaLaunchKernel(const T *func, dim3 gridDim, dim3 blockDim, void **args, size_t sharedMem, cudaStream_t stream) "cudaLaunchKernel (C++ API)", - * \ref ::cudaFuncSetCacheConfig(T*, enum cudaFuncCache) "cudaFuncSetCacheConfig (C++ API)", - * \ref ::cudaFuncGetAttributes(struct cudaFuncAttributes*, const void*) "cudaFuncGetAttributes (C API)", + * \ref ::cudaLaunchKernel(const T *func, dim3 gridDim, dim3 blockDim, void + * **args, size_t sharedMem, cudaStream_t stream) "cudaLaunchKernel (C++ API)", + * \ref ::cudaFuncSetCacheConfig(T*, enum cudaFuncCache) "cudaFuncSetCacheConfig + * (C++ API)", \ref ::cudaFuncGetAttributes(struct cudaFuncAttributes*, const + * void*) "cudaFuncGetAttributes (C API)", * ::cudaSetDoubleForDevice, * ::cudaSetDoubleForHost, * \ref ::cudaSetupArgument(T, size_t) "cudaSetupArgument (C++ API)" */ -cudaError_t CUDARTAPI cudaFuncSetAttribute(const void *func, enum cudaFuncAttribute attr, int value) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("GPGPU-Sim PTX: Execution warning: ignoring call to \"%s ( func=%p, attr=%d, value=%d )\"\n", - __my_func__, func, attr, value ); - return g_last_cudaError = cudaSuccess; +cudaError_t CUDARTAPI cudaFuncSetAttribute(const void *func, + enum cudaFuncAttribute attr, + int value) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf( + "GPGPU-Sim PTX: Execution warning: ignoring call to \"%s ( func=%p, " + "attr=%d, value=%d )\"\n", + __my_func__, func, attr, value); + return g_last_cudaError = cudaSuccess; } #endif -cudaError_t CUDARTAPI cudaGLSetGLDevice(int device) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("GPGPU-Sim PTX: Execution warning: ignoring call to \"%s\"\n", __my_func__ ); - return g_last_cudaError = cudaErrorUnknown; +cudaError_t CUDARTAPI cudaGLSetGLDevice(int device) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("GPGPU-Sim PTX: Execution warning: ignoring call to \"%s\"\n", + __my_func__); + return g_last_cudaError = cudaErrorUnknown; } -typedef void* HGPUNV; +typedef void *HGPUNV; -cudaError_t CUDARTAPI cudaWGLGetDevice(int *device, HGPUNV hGpu) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaErrorUnknown; +cudaError_t CUDARTAPI cudaWGLGetDevice(int *device, HGPUNV hGpu) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaErrorUnknown; } -void CUDARTAPI __cudaMutexOperation(int lock) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); +void CUDARTAPI __cudaMutexOperation(int lock) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); } -void CUDARTAPI __cudaTextureFetch(const void *tex, void *index, int integer, void *val) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); +void CUDARTAPI __cudaTextureFetch(const void *tex, void *index, int integer, + void *val) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); } - } namespace cuda_math { -void CUDARTAPI __cudaMutexOperation(int lock) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); +void CUDARTAPI __cudaMutexOperation(int lock) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); } -void CUDARTAPI __cudaTextureFetch(const void *tex, void *index, int integer, void *val) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); +void CUDARTAPI __cudaTextureFetch(const void *tex, void *index, int integer, + void *val) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); } -int CUDARTAPI __cudaSynchronizeThreads(void**, void*) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //TODO This function should syncronize if we support Asyn kernel calls - return g_last_cudaError = cudaSuccess; +int CUDARTAPI __cudaSynchronizeThreads(void **, void *) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // TODO This function should syncronize if we support Asyn kernel calls + return g_last_cudaError = cudaSuccess; } -} +} // namespace cuda_math //////// /// static functions -int cuda_runtime_api::load_static_globals( symbol_table *symtab, unsigned min_gaddr, unsigned max_gaddr, gpgpu_t *gpu ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf( "GPGPU-Sim PTX: loading globals with explicit initializers... \n" ); - fflush(stdout); - int ng_bytes=0; - symbol_table::iterator g=symtab->global_iterator_begin(); - - for ( ; g!=symtab->global_iterator_end(); g++) { - symbol *global = *g; - if ( global->has_initializer() ) { - printf( "GPGPU-Sim PTX: initializing '%s' ... ", global->name().c_str() ); - unsigned addr=global->get_address(); - const type_info *type = global->type(); - type_info_key ti=type->get_key(); - size_t size; - int t; - ti.type_decode(size,t); - int nbytes = size/8; - int offset=0; - std::list init_list = global->get_initializer(); - for ( std::list::iterator i=init_list.begin(); i!=init_list.end(); i++ ) { - operand_info op = *i; - ptx_reg_t value = op.get_literal_value(); - assert( (addr+offset+nbytes) < min_gaddr ); // min_gaddr is start of "heap" for cudaMalloc - gpu->get_global_memory()->write(addr+offset,nbytes,&value,NULL,NULL); // assuming little endian here - offset+=nbytes; - ng_bytes+=nbytes; - } - printf(" wrote %u bytes\n", offset ); - } - } - printf( "GPGPU-Sim PTX: finished loading globals (%u bytes total).\n", ng_bytes ); - fflush(stdout); - return ng_bytes; -} - -int cuda_runtime_api::load_constants( symbol_table *symtab, addr_t min_gaddr, gpgpu_t *gpu ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf( "GPGPU-Sim PTX: loading constants with explicit initializers... " ); - fflush(stdout); - int nc_bytes = 0; - symbol_table::iterator g=symtab->const_iterator_begin(); - - for ( ; g!=symtab->const_iterator_end(); g++) { - symbol *constant = *g; - if ( constant->is_const() && constant->has_initializer() ) { - - // get the constant element data size - int basic_type; - size_t num_bits; - constant->type()->get_key().type_decode(num_bits,basic_type); - - std::list init_list = constant->get_initializer(); - int nbytes_written = 0; - for ( std::list::iterator i=init_list.begin(); i!=init_list.end(); i++ ) { - operand_info op = *i; - ptx_reg_t value = op.get_literal_value(); - int nbytes = num_bits/8; - switch ( op.get_type() ) { - case int_t: assert(nbytes >= 1); break; - case float_op_t: assert(nbytes == 4); break; - case double_op_t: assert(nbytes >= 4); break; // account for double DEMOTING - default: - abort(); - } - unsigned addr=constant->get_address() + nbytes_written; - assert( addr+nbytes < min_gaddr ); - - gpu->get_global_memory()->write(addr,nbytes,&value,NULL,NULL); // assume little endian (so u8 is the first byte in u32) - nc_bytes+=nbytes; - nbytes_written += nbytes; - } - } - } - printf( " done.\n"); - fflush(stdout); - return nc_bytes; -} - -kernel_info_t * cuda_runtime_api::gpgpu_cuda_ptx_sim_init_grid( const char *hostFun, - gpgpu_ptx_sim_arg_list_t args, - struct dim3 gridDim, - struct dim3 blockDim, - CUctx_st* context ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - function_info *entry = context->get_kernel(hostFun); - gpgpu_t* gpu= context->get_device()->get_gpgpu(); - /* - Passing a snapshot of the GPU's current texture mapping to the kernel's info - as kernels should use texture bindings present at the time of their launch. - */ - kernel_info_t *result = new kernel_info_t(gridDim,blockDim,entry,gpu->getNameArrayMapping(),gpu->getNameInfoMapping()); - if( entry == NULL ) { - printf("GPGPU-Sim PTX: ERROR launching kernel -- no PTX implementation found for %p\n", hostFun); - abort(); - } - unsigned argcount=args.size(); - unsigned argn=1; - for( gpgpu_ptx_sim_arg_list_t::iterator a = args.begin(); a != args.end(); a++ ) { - entry->add_param_data(argcount-argn,&(*a)); - argn++; - } - - entry->finalize(result->get_param_memory()); - gpgpu_ctx->func_sim->g_ptx_kernel_count++; - fflush(stdout); - - if(g_debug_execution >= 4){ - entry->ptx_jit_config(g_mallocPtr_Size, result->get_param_memory(), (gpgpu_t *) context->get_device()->get_gpgpu(), gridDim, blockDim); +int cuda_runtime_api::load_static_globals(symbol_table *symtab, + unsigned min_gaddr, + unsigned max_gaddr, gpgpu_t *gpu) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("GPGPU-Sim PTX: loading globals with explicit initializers... \n"); + fflush(stdout); + int ng_bytes = 0; + symbol_table::iterator g = symtab->global_iterator_begin(); + + for (; g != symtab->global_iterator_end(); g++) { + symbol *global = *g; + if (global->has_initializer()) { + printf("GPGPU-Sim PTX: initializing '%s' ... ", + global->name().c_str()); + unsigned addr = global->get_address(); + const type_info *type = global->type(); + type_info_key ti = type->get_key(); + size_t size; + int t; + ti.type_decode(size, t); + int nbytes = size / 8; + int offset = 0; + std::list init_list = global->get_initializer(); + for (std::list::iterator i = init_list.begin(); + i != init_list.end(); i++) { + operand_info op = *i; + ptx_reg_t value = op.get_literal_value(); + assert((addr + offset + nbytes) < + min_gaddr); // min_gaddr is start of "heap" for cudaMalloc + gpu->get_global_memory()->write(addr + offset, nbytes, &value, NULL, + NULL); // assuming little endian here + offset += nbytes; + ng_bytes += nbytes; + } + printf(" wrote %u bytes\n", offset); + } + } + printf("GPGPU-Sim PTX: finished loading globals (%u bytes total).\n", + ng_bytes); + fflush(stdout); + return ng_bytes; +} + +int cuda_runtime_api::load_constants(symbol_table *symtab, addr_t min_gaddr, + gpgpu_t *gpu) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("GPGPU-Sim PTX: loading constants with explicit initializers... "); + fflush(stdout); + int nc_bytes = 0; + symbol_table::iterator g = symtab->const_iterator_begin(); + + for (; g != symtab->const_iterator_end(); g++) { + symbol *constant = *g; + if (constant->is_const() && constant->has_initializer()) { + // get the constant element data size + int basic_type; + size_t num_bits; + constant->type()->get_key().type_decode(num_bits, basic_type); + + std::list init_list = constant->get_initializer(); + int nbytes_written = 0; + for (std::list::iterator i = init_list.begin(); + i != init_list.end(); i++) { + operand_info op = *i; + ptx_reg_t value = op.get_literal_value(); + int nbytes = num_bits / 8; + switch (op.get_type()) { + case int_t: + assert(nbytes >= 1); + break; + case float_op_t: + assert(nbytes == 4); + break; + case double_op_t: + assert(nbytes >= 4); + break; // account for double DEMOTING + default: + abort(); + } + unsigned addr = constant->get_address() + nbytes_written; + assert(addr + nbytes < min_gaddr); + + gpu->get_global_memory()->write( + addr, nbytes, &value, NULL, + NULL); // assume little endian (so u8 is the first byte in u32) + nc_bytes += nbytes; + nbytes_written += nbytes; + } } - - return result; + } + printf(" done.\n"); + fflush(stdout); + return nc_bytes; +} + +kernel_info_t *cuda_runtime_api::gpgpu_cuda_ptx_sim_init_grid( + const char *hostFun, gpgpu_ptx_sim_arg_list_t args, struct dim3 gridDim, + struct dim3 blockDim, CUctx_st *context) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + function_info *entry = context->get_kernel(hostFun); + gpgpu_t *gpu = context->get_device()->get_gpgpu(); + /* + Passing a snapshot of the GPU's current texture mapping to the kernel's info + as kernels should use texture bindings present at the time of their launch. + */ + kernel_info_t *result = + new kernel_info_t(gridDim, blockDim, entry, gpu->getNameArrayMapping(), + gpu->getNameInfoMapping()); + if (entry == NULL) { + printf( + "GPGPU-Sim PTX: ERROR launching kernel -- no PTX implementation found " + "for %p\n", + hostFun); + abort(); + } + unsigned argcount = args.size(); + unsigned argn = 1; + for (gpgpu_ptx_sim_arg_list_t::iterator a = args.begin(); a != args.end(); + a++) { + entry->add_param_data(argcount - argn, &(*a)); + argn++; + } + + entry->finalize(result->get_param_memory()); + gpgpu_ctx->func_sim->g_ptx_kernel_count++; + fflush(stdout); + + if (g_debug_execution >= 4) { + entry->ptx_jit_config(g_mallocPtr_Size, result->get_param_memory(), + (gpgpu_t *)context->get_device()->get_gpgpu(), + gridDim, blockDim); + } + + return result; } /******************************************************************************* @@ -3945,2934 +4084,2912 @@ kernel_info_t * cuda_runtime_api::gpgpu_cuda_ptx_sim_init_grid( const char *host *******************************************************************************/ //***extra api for pytorch*** -CUresult CUDAAPI cuGetErrorString(CUresult error, const char **pStr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuGetErrorName(CUresult error, const char **pStr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuInit(unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuDriverGetVersion(int *driverVersion) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cudaError_t e = cudaDriverGetVersion(driverVersion); - assert(e == cudaSuccess); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuDeviceGet(CUdevice *device, int ordinal) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - int deviceI = -1; - cudaError_t e = cudaGetDevice(&deviceI); - assert(e == cudaSuccess); - assert(deviceI!=-1); - *device = deviceI; - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuDeviceGetCount(int *count) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cudaError_t e = cudaGetDeviceCount(count); - assert(e == cudaSuccess); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuDeviceGetName(char *name, int len, CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - assert(len>=10); - strcpy(name, "GPGPU-Sim"); - return CUDA_SUCCESS; +CUresult CUDAAPI cuGetErrorString(CUresult error, const char **pStr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuGetErrorName(CUresult error, const char **pStr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuInit(unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuDriverGetVersion(int *driverVersion) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cudaError_t e = cudaDriverGetVersion(driverVersion); + assert(e == cudaSuccess); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuDeviceGet(CUdevice *device, int ordinal) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + int deviceI = -1; + cudaError_t e = cudaGetDevice(&deviceI); + assert(e == cudaSuccess); + assert(deviceI != -1); + *device = deviceI; + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuDeviceGetCount(int *count) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cudaError_t e = cudaGetDeviceCount(count); + assert(e == cudaSuccess); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuDeviceGetName(char *name, int len, CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + assert(len >= 10); + strcpy(name, "GPGPU-Sim"); + return CUDA_SUCCESS; } #if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuDeviceTotalMem(size_t *bytes, CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - *bytes = 20000000000;//dummy value - return CUDA_SUCCESS; +CUresult CUDAAPI cuDeviceTotalMem(size_t *bytes, CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + *bytes = 20000000000; // dummy value + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 3020 */ #if (CUDART_VERSION > 5000) -CUresult CUDAAPI cuDeviceGetAttribute(int *pi, CUdevice_attribute attrib, CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cudaError_t e = cudaDeviceGetAttribute(pi, (cudaDeviceAttr)attrib, dev); - assert(e == cudaSuccess); +CUresult CUDAAPI cuDeviceGetAttribute(int *pi, CUdevice_attribute attrib, + CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cudaError_t e = cudaDeviceGetAttribute(pi, (cudaDeviceAttr)attrib, dev); + assert(e == cudaSuccess); - return CUDA_SUCCESS; + return CUDA_SUCCESS; } #endif -CUresult CUDAAPI cuDeviceGetProperties(CUdevprop *prop, CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDeviceGetProperties(CUdevprop *prop, CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuDeviceComputeCapability(int *major, int *minor, CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDeviceComputeCapability(int *major, int *minor, + CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #if CUDART_VERSION >= 7000 -CUresult CUDAAPI cuDevicePrimaryCtxRetain(CUcontext *pctx, CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDevicePrimaryCtxRetain(CUcontext *pctx, CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuDevicePrimaryCtxRelease(CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDevicePrimaryCtxRelease(CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuDevicePrimaryCtxSetFlags(CUdevice dev, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDevicePrimaryCtxSetFlags(CUdevice dev, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuDevicePrimaryCtxGetState(CUdevice dev, unsigned int *flags, int *active) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDevicePrimaryCtxGetState(CUdevice dev, unsigned int *flags, + int *active) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuDevicePrimaryCtxReset(CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDevicePrimaryCtxReset(CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 7000 */ #if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuCtxCreate(CUcontext *pctx, unsigned int flags, CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxCreate(CUcontext *pctx, unsigned int flags, + CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 3020 */ #if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuCtxDestroy(CUcontext ctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxDestroy(CUcontext ctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 4000 */ #if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuCtxPushCurrent(CUcontext ctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxPushCurrent(CUcontext ctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxPopCurrent(CUcontext *pctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxPopCurrent(CUcontext *pctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxSetCurrent(CUcontext ctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxSetCurrent(CUcontext ctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxGetCurrent(CUcontext *pctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxGetCurrent(CUcontext *pctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 4000 */ -CUresult CUDAAPI cuCtxGetDevice(CUdevice *device) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxGetDevice(CUdevice *device) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #if CUDART_VERSION >= 7000 -CUresult CUDAAPI cuCtxGetFlags(unsigned int *flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxGetFlags(unsigned int *flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 7000 */ -CUresult CUDAAPI cuCtxSynchronize(void) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxSynchronize(void) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxSetLimit(CUlimit limit, size_t value) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxSetLimit(CUlimit limit, size_t value) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxGetLimit(size_t *pvalue, CUlimit limit) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxGetLimit(size_t *pvalue, CUlimit limit) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxGetCacheConfig(CUfunc_cache *pconfig) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxGetCacheConfig(CUfunc_cache *pconfig) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxSetCacheConfig(CUfunc_cache config) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxSetCacheConfig(CUfunc_cache config) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #if CUDART_VERSION >= 4020 -CUresult CUDAAPI cuCtxGetSharedMemConfig(CUsharedconfig *pConfig) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxGetSharedMemConfig(CUsharedconfig *pConfig) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxSetSharedMemConfig(CUsharedconfig config) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxSetSharedMemConfig(CUsharedconfig config) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif -CUresult CUDAAPI cuCtxGetApiVersion(CUcontext ctx, unsigned int *version) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxGetApiVersion(CUcontext ctx, unsigned int *version) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxGetStreamPriorityRange(int *leastPriority, int *greatestPriority) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxGetStreamPriorityRange(int *leastPriority, + int *greatestPriority) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxAttach(CUcontext *pctx, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxAttach(CUcontext *pctx, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuCtxDetach(CUcontext ctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuCtxDetach(CUcontext ctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuModuleLoad(CUmodule *module, const char *fname) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleLoad(CUmodule *module, const char *fname) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuModuleLoadData(CUmodule *module, const void *image) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleLoadData(CUmodule *module, const void *image) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuModuleLoadDataEx(CUmodule *module, const void *image, unsigned int numOptions, CUjit_option *options, void **optionValues) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleLoadDataEx(CUmodule *module, const void *image, + unsigned int numOptions, + CUjit_option *options, + void **optionValues) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuModuleLoadFatBinary(CUmodule *module, const void *fatCubin) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleLoadFatBinary(CUmodule *module, const void *fatCubin) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuModuleUnload(CUmodule hmod) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleUnload(CUmodule hmod) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuModuleGetFunction(CUfunction *hfunc, CUmodule hmod, const char *name) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleGetFunction(CUfunction *hfunc, CUmodule hmod, + const char *name) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuModuleGetGlobal(CUdeviceptr *dptr, size_t *bytes, CUmodule hmod, const char *name) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleGetGlobal(CUdeviceptr *dptr, size_t *bytes, + CUmodule hmod, const char *name) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 3020 */ -CUresult CUDAAPI cuModuleGetTexRef(CUtexref *pTexRef, CUmodule hmod, const char *name) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleGetTexRef(CUtexref *pTexRef, CUmodule hmod, + const char *name) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuModuleGetSurfRef(CUsurfref *pSurfRef, CUmodule hmod, const char *name) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuModuleGetSurfRef(CUsurfref *pSurfRef, CUmodule hmod, + const char *name) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #if CUDART_VERSION >= 6050 -CUresult CUDAAPI -cuLinkCreate(unsigned int numOptions, CUjit_option *options, void **optionValues, CUlinkState *stateOut) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //currently do not support options or multiple CUlinkStates - return CUDA_SUCCESS; -} - -CUresult CUDAAPI -cuLinkAddData(CUlinkState state, CUjitInputType type, void *data, size_t size, const char *name, - unsigned int numOptions, CUjit_option *options, void **optionValues) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - assert(type==CU_JIT_INPUT_PTX); - cuda_not_implemented(__my_func__,__LINE__); - return CUDA_ERROR_UNKNOWN; -} - -CUresult CUDAAPI -cuLinkAddFile(CUlinkState state, CUjitInputType type, const char *path, - unsigned int numOptions, CUjit_option *options, void **optionValues) -{ - return cuLinkAddFileInternal(state, type, path, - numOptions, options, optionValues); +CUresult CUDAAPI cuLinkCreate(unsigned int numOptions, CUjit_option *options, + void **optionValues, CUlinkState *stateOut) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // currently do not support options or multiple CUlinkStates + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuLinkAddData(CUlinkState state, CUjitInputType type, + void *data, size_t size, const char *name, + unsigned int numOptions, CUjit_option *options, + void **optionValues) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + assert(type == CU_JIT_INPUT_PTX); + cuda_not_implemented(__my_func__, __LINE__); + return CUDA_ERROR_UNKNOWN; +} + +CUresult CUDAAPI cuLinkAddFile(CUlinkState state, CUjitInputType type, + const char *path, unsigned int numOptions, + CUjit_option *options, void **optionValues) { + return cuLinkAddFileInternal(state, type, path, numOptions, options, + optionValues); } #endif #if CUDART_VERSION >= 5050 -CUresult CUDAAPI -cuLinkComplete(CUlinkState state, void **cubinOut, size_t *sizeOut) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //all cuLink* function are implemented to block until completion so nothing to do here - return CUDA_SUCCESS; +CUresult CUDAAPI cuLinkComplete(CUlinkState state, void **cubinOut, + size_t *sizeOut) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // all cuLink* function are implemented to block until completion so nothing + // to do here + return CUDA_SUCCESS; } -CUresult CUDAAPI -cuLinkDestroy(CUlinkState state) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - //currently do not support options or multiple CUlinkStates - return CUDA_SUCCESS; +CUresult CUDAAPI cuLinkDestroy(CUlinkState state) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + // currently do not support options or multiple CUlinkStates + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 5050 */ #if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuMemGetInfo(size_t *free, size_t *total) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuMemAlloc(CUdeviceptr *dptr, size_t bytesize) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuMemAllocPitch(CUdeviceptr *dptr, size_t *pPitch, size_t WidthInBytes, size_t Height, unsigned int ElementSizeBytes) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuMemFree(CUdeviceptr dptr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuMemGetAddressRange(CUdeviceptr *pbase, size_t *psize, CUdeviceptr dptr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuMemAllocHost(void **pp, size_t bytesize) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemGetInfo(size_t *free, size_t *total) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemAlloc(CUdeviceptr *dptr, size_t bytesize) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemAllocPitch(CUdeviceptr *dptr, size_t *pPitch, + size_t WidthInBytes, size_t Height, + unsigned int ElementSizeBytes) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemFree(CUdeviceptr dptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemGetAddressRange(CUdeviceptr *pbase, size_t *psize, + CUdeviceptr dptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemAllocHost(void **pp, size_t bytesize) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 3020 */ -CUresult CUDAAPI cuMemFreeHost(void *p) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemFreeHost(void *p) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemHostAlloc(void **pp, size_t bytesize, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemHostAlloc(void **pp, size_t bytesize, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuMemHostGetDevicePointer(CUdeviceptr *pdptr, void *p, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemHostGetDevicePointer(CUdeviceptr *pdptr, void *p, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 3020 */ -CUresult CUDAAPI cuMemHostGetFlags(unsigned int *pFlags, void *p) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemHostGetFlags(unsigned int *pFlags, void *p) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #if CUDART_VERSION >= 6000 -CUresult CUDAAPI cuMemAllocManaged(CUdeviceptr *dptr, size_t bytesize, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemAllocManaged(CUdeviceptr *dptr, size_t bytesize, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 6000 */ #if CUDART_VERSION >= 4010 -CUresult CUDAAPI cuDeviceGetByPCIBusId(CUdevice *dev, const char *pciBusId) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDeviceGetByPCIBusId(CUdevice *dev, const char *pciBusId) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuDeviceGetPCIBusId(char *pciBusId, int len, CUdevice dev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDeviceGetPCIBusId(char *pciBusId, int len, CUdevice dev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuIpcGetEventHandle(CUipcEventHandle *pHandle, CUevent event) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuIpcGetEventHandle(CUipcEventHandle *pHandle, CUevent event) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuIpcOpenEventHandle(CUevent *phEvent, CUipcEventHandle handle) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuIpcOpenEventHandle(CUevent *phEvent, + CUipcEventHandle handle) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuIpcGetMemHandle(CUipcMemHandle *pHandle, CUdeviceptr dptr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuIpcGetMemHandle(CUipcMemHandle *pHandle, CUdeviceptr dptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuIpcOpenMemHandle(CUdeviceptr *pdptr, CUipcMemHandle handle, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuIpcOpenMemHandle(CUdeviceptr *pdptr, CUipcMemHandle handle, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuIpcCloseMemHandle(CUdeviceptr dptr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuIpcCloseMemHandle(CUdeviceptr dptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 4010 */ #if CUDART_VERSION >= 6050 -CUresult CUDAAPI cuMemHostRegister(void *p, size_t bytesize, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -__host__ cudaError_t cudaHostRegister(void* ptr, size_t size, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t cudaProfilerStart ( ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return g_last_cudaError = cudaSuccess; -} - -__host__ cudaError_t cudaProfilerStop ( ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return g_last_cudaError = cudaSuccess; +CUresult CUDAAPI cuMemHostRegister(void *p, size_t bytesize, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +__host__ cudaError_t cudaHostRegister(void *ptr, size_t size, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t cudaProfilerStart() { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return g_last_cudaError = cudaSuccess; +} + +__host__ cudaError_t cudaProfilerStop() { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return g_last_cudaError = cudaSuccess; } #endif #if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuMemHostUnregister(void *p) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemHostUnregister(void *p) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, + CUdeviceptr srcDevice, CUcontext srcContext, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 4000 */ #if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuMemcpyHtoD(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuMemcpyHtoD(CUdeviceptr dstDevice, const void *srcHost, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyDtoH(void *dstHost, CUdeviceptr srcDevice, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyDtoD(CUdeviceptr dstDevice, CUdeviceptr srcDevice, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyDtoA(CUarray dstArray, size_t dstOffset, + CUdeviceptr srcDevice, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyAtoD(CUdeviceptr dstDevice, CUarray srcArray, + size_t srcOffset, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyHtoA(CUarray dstArray, size_t dstOffset, + const void *srcHost, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyAtoH(void *dstHost, CUarray srcArray, size_t srcOffset, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyAtoA(CUarray dstArray, size_t dstOffset, + CUarray srcArray, size_t srcOffset, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpy2D(const CUDA_MEMCPY2D *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpy2DUnaligned(const CUDA_MEMCPY2D *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpy3D(const CUDA_MEMCPY3D *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 3020 */ -CUresult CUDAAPI cuMemcpyDtoH(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 4000 +CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, + size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, + CUdeviceptr srcDevice, CUcontext srcContext, + size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 4000 */ -CUresult CUDAAPI cuMemcpyDtoD(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 3020 +CUresult CUDAAPI cuMemcpyHtoDAsync(CUdeviceptr dstDevice, const void *srcHost, + size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyDtoHAsync(void *dstHost, CUdeviceptr srcDevice, + size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyDtoDAsync(CUdeviceptr dstDevice, CUdeviceptr srcDevice, + size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyHtoAAsync(CUarray dstArray, size_t dstOffset, + const void *srcHost, size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, + size_t srcOffset, size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpy2DAsync(const CUDA_MEMCPY2D *pCopy, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemcpy3DAsync(const CUDA_MEMCPY3D *pCopy, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 3020 */ -CUresult CUDAAPI cuMemcpyDtoA(CUarray dstArray, size_t dstOffset, CUdeviceptr srcDevice, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 4000 +CUresult CUDAAPI cuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER *pCopy, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 4000 */ -CUresult CUDAAPI cuMemcpyAtoD(CUdeviceptr dstDevice, CUarray srcArray, size_t srcOffset, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 3020 +CUresult CUDAAPI cuMemsetD8(CUdeviceptr dstDevice, unsigned char uc, size_t N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD16(CUdeviceptr dstDevice, unsigned short us, + size_t N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD32(CUdeviceptr dstDevice, unsigned int ui, size_t N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD2D8(CUdeviceptr dstDevice, size_t dstPitch, + unsigned char uc, size_t Width, size_t Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD2D16(CUdeviceptr dstDevice, size_t dstPitch, + unsigned short us, size_t Width, size_t Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD2D32(CUdeviceptr dstDevice, size_t dstPitch, + unsigned int ui, size_t Width, size_t Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, + size_t N, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, + size_t N, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, + size_t N, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned char uc, size_t Width, + size_t Height, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned short us, size_t Width, + size_t Height, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned int ui, size_t Width, + size_t Height, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuArrayCreate(CUarray *pHandle, + const CUDA_ARRAY_DESCRIPTOR *pAllocateArray) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuArrayGetDescriptor(CUDA_ARRAY_DESCRIPTOR *pArrayDescriptor, + CUarray hArray) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 3020 */ -CUresult CUDAAPI cuMemcpyHtoA(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuArrayDestroy(CUarray hArray) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpyAtoH(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 3020 +CUresult CUDAAPI cuArray3DCreate( + CUarray *pHandle, const CUDA_ARRAY3D_DESCRIPTOR *pAllocateArray) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuArray3DGetDescriptor( + CUDA_ARRAY3D_DESCRIPTOR *pArrayDescriptor, CUarray hArray) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 3020 */ -CUresult CUDAAPI cuMemcpyAtoA(CUarray dstArray, size_t dstOffset, CUarray srcArray, size_t srcOffset, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 5000 + +CUresult CUDAAPI +cuMipmappedArrayCreate(CUmipmappedArray *pHandle, + const CUDA_ARRAY3D_DESCRIPTOR *pMipmappedArrayDesc, + unsigned int numMipmapLevels) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMipmappedArrayGetLevel(CUarray *pLevelArray, + CUmipmappedArray hMipmappedArray, + unsigned int level) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpy2D(const CUDA_MEMCPY2D *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#endif /* CUDART_VERSION >= 5000 */ + +/** @} */ /* END CUDA_MEM */ + +#if CUDART_VERSION >= 4000 +CUresult CUDAAPI cuPointerGetAttribute(void *data, + CUpointer_attribute attribute, + CUdeviceptr ptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 4000 */ -CUresult CUDAAPI cuMemcpy2DUnaligned(const CUDA_MEMCPY2D *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 8000 +__host__ cudaError_t CUDARTAPI cudaCreateTextureObject( + cudaTextureObject_t *pTexObject, const cudaResourceDesc *pResDesc, + const cudaTextureDesc *pTexDesc, const cudaResourceViewDesc *pResViewDesc) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cuda_not_implemented(__my_func__, __LINE__); + return g_last_cudaError = cudaSuccess; +} + +CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, + CUdevice dstDevice, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemAdvise(CUdeviceptr devPtr, size_t count, + CUmem_advise advice, CUdevice device) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemRangeGetAttribute(void *data, size_t dataSize, + CUmem_range_attribute attribute, + CUdeviceptr devPtr, size_t count) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemRangeGetAttributes(void **data, size_t *dataSizes, + CUmem_range_attribute *attributes, + size_t numAttributes, + CUdeviceptr devPtr, size_t count) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 8000 */ -CUresult CUDAAPI cuMemcpy3D(const CUDA_MEMCPY3D *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 6000 +CUresult CUDAAPI cuPointerSetAttribute(const void *value, + CUpointer_attribute attribute, + CUdeviceptr ptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 3020 */ +#endif /* CUDART_VERSION >= 6000 */ -#if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 7000 +CUresult CUDAAPI cuPointerGetAttributes(unsigned int numAttributes, + CUpointer_attribute *attributes, + void **data, CUdeviceptr ptr) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 7000 */ -CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +/** @} */ /* END CUDA_UNIFIED */ + +CUresult CUDAAPI cuStreamCreate(CUstream *phStream, unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuStreamCreateWithPriority(CUstream *phStream, + unsigned int flags, int priority) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 4000 */ -#if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuMemcpyHtoDAsync(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuStreamGetPriority(CUstream hStream, int *priority) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpyDtoHAsync(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuStreamGetFlags(CUstream hStream, unsigned int *flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpyDtoDAsync(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpyHtoAAsync(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, + CUstreamCallback callback, void *userData, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpyAtoHAsync(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 6000 + +CUresult CUDAAPI cuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, + size_t length, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpy2DAsync(const CUDA_MEMCPY2D *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#endif /* CUDART_VERSION >= 6000 */ + +CUresult CUDAAPI cuStreamQuery(CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemcpy3DAsync(const CUDA_MEMCPY3D *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuStreamSynchronize(CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 3020 */ #if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuStreamDestroy(CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 4000 */ -#if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuMemsetD8(CUdeviceptr dstDevice, unsigned char uc, size_t N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +/** @} */ /* END CUDA_STREAM */ + +CUresult CUDAAPI cuEventCreate(CUevent *phEvent, unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemsetD16(CUdeviceptr dstDevice, unsigned short us, size_t N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuEventRecord(CUevent hEvent, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemsetD32(CUdeviceptr dstDevice, unsigned int ui, size_t N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuEventQuery(CUevent hEvent) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemsetD2D8(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuEventSynchronize(CUevent hEvent) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemsetD2D16(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 4000 +CUresult CUDAAPI cuEventDestroy(CUevent hEvent) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 4000 */ -CUresult CUDAAPI cuMemsetD2D32(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuEventElapsedTime(float *pMilliseconds, CUevent hStart, + CUevent hEnd) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, size_t N, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 8000 +CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, + cuuint32_t value, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, + cuuint32_t value, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, + CUstreamBatchMemOpParams *paramArray, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 8000 */ -CUresult CUDAAPI cuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, size_t N, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +/** @} */ /* END CUDA_EVENT */ + +CUresult CUDAAPI cuFuncGetAttribute(int *pi, CUfunction_attribute attrib, + CUfunction hfunc) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, size_t N, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuFuncSetCacheConfig(CUfunction hfunc, CUfunc_cache config) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 4020 +CUresult CUDAAPI cuFuncSetSharedMemConfig(CUfunction hfunc, + CUsharedconfig config) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif -CUresult CUDAAPI cuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 4000 +CUresult CUDAAPI cuLaunchKernel(CUfunction f, unsigned int gridDimX, + unsigned int gridDimY, unsigned int gridDimZ, + unsigned int blockDimX, unsigned int blockDimY, + unsigned int blockDimZ, + unsigned int sharedMemBytes, CUstream hStream, + void **kernelParams, void **extra) { + return cuLaunchKernelInternal(f, gridDimX, gridDimY, gridDimZ, blockDimX, + blockDimY, blockDimZ, sharedMemBytes, hStream, + kernelParams, extra); } +#endif /* CUDART_VERSION >= 4000 */ -CUresult CUDAAPI cuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +/** @} */ /* END CUDA_EXEC */ + +CUresult CUDAAPI cuFuncSetBlockShape(CUfunction hfunc, int x, int y, int z) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuArrayCreate(CUarray *pHandle, const CUDA_ARRAY_DESCRIPTOR *pAllocateArray) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuFuncSetSharedSize(CUfunction hfunc, unsigned int bytes) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuArrayGetDescriptor(CUDA_ARRAY_DESCRIPTOR *pArrayDescriptor, CUarray hArray) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuParamSetSize(CUfunction hfunc, unsigned int numbytes) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 3020 */ +CUresult CUDAAPI cuParamSeti(CUfunction hfunc, int offset, unsigned int value) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} -CUresult CUDAAPI cuArrayDestroy(CUarray hArray) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -#if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuArray3DCreate(CUarray *pHandle, const CUDA_ARRAY3D_DESCRIPTOR *pAllocateArray) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuArray3DGetDescriptor(CUDA_ARRAY3D_DESCRIPTOR *pArrayDescriptor, CUarray hArray) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif /* CUDART_VERSION >= 3020 */ - -#if CUDART_VERSION >= 5000 - -CUresult CUDAAPI cuMipmappedArrayCreate(CUmipmappedArray *pHandle, const CUDA_ARRAY3D_DESCRIPTOR *pMipmappedArrayDesc, unsigned int numMipmapLevels) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuMipmappedArrayGetLevel(CUarray *pLevelArray, CUmipmappedArray hMipmappedArray, unsigned int level) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuMipmappedArrayDestroy(CUmipmappedArray hMipmappedArray) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -#endif /* CUDART_VERSION >= 5000 */ - -/** @} */ /* END CUDA_MEM */ - - -#if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuPointerGetAttribute(void *data, CUpointer_attribute attribute, CUdeviceptr ptr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuParamSetf(CUfunction hfunc, int offset, float value) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 4000 */ - -#if CUDART_VERSION >= 8000 -__host__ cudaError_t CUDARTAPI cudaCreateTextureObject ( cudaTextureObject_t* pTexObject, const cudaResourceDesc* pResDesc, const cudaTextureDesc* pTexDesc, const cudaResourceViewDesc* pResViewDesc ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cuda_not_implemented(__my_func__,__LINE__); - return g_last_cudaError = cudaSuccess; +CUresult CUDAAPI cuParamSetv(CUfunction hfunc, int offset, void *ptr, + unsigned int numbytes) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice dstDevice, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuLaunch(CUfunction f) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemAdvise(CUdeviceptr devPtr, size_t count, CUmem_advise advice, CUdevice device) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuLaunchGrid(CUfunction f, int grid_width, int grid_height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemRangeGetAttribute(void *data, size_t dataSize, CUmem_range_attribute attribute, CUdeviceptr devPtr, size_t count) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuLaunchGridAsync(CUfunction f, int grid_width, + int grid_height, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuMemRangeGetAttributes(void **data, size_t *dataSizes, CUmem_range_attribute *attributes, size_t numAttributes, CUdeviceptr devPtr, size_t count) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuParamSetTexRef(CUfunction hfunc, int texunit, + CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 8000 */ +/** @} */ /* END CUDA_EXEC_DEPRECATED */ -#if CUDART_VERSION >= 6000 -CUresult CUDAAPI cuPointerSetAttribute(const void *value, CUpointer_attribute attribute, CUdeviceptr ptr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif /* CUDART_VERSION >= 6000 */ +#if CUDART_VERSION >= 6050 -#if CUDART_VERSION >= 7000 -CUresult CUDAAPI cuPointerGetAttributes(unsigned int numAttributes, CUpointer_attribute *attributes, void **data, CUdeviceptr ptr) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessor( + int *numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags( + int *numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuOccupancyMaxPotentialBlockSize( + int *minGridSize, int *blockSize, CUfunction func, + CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, + int blockSizeLimit) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuOccupancyMaxPotentialBlockSizeWithFlags( + int *minGridSize, int *blockSize, CUfunction func, + CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, + int blockSizeLimit, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 7000 */ -/** @} */ /* END CUDA_UNIFIED */ - - -CUresult CUDAAPI cuStreamCreate(CUstream *phStream, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} +/** @} */ /* END CUDA_OCCUPANCY */ +#endif /* CUDART_VERSION >= 6050 */ -CUresult CUDAAPI cuStreamCreateWithPriority(CUstream *phStream, unsigned int flags, int priority) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetArray(CUtexref hTexRef, CUarray hArray, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } - -CUresult CUDAAPI cuStreamGetPriority(CUstream hStream, int *priority) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetMipmappedArray(CUtexref hTexRef, + CUmipmappedArray hMipmappedArray, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuStreamGetFlags(CUstream hStream, unsigned int *flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 3020 +CUresult CUDAAPI cuTexRefSetAddress(size_t *ByteOffset, CUtexref hTexRef, + CUdeviceptr dptr, size_t bytes) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefSetAddress2D(CUtexref hTexRef, + const CUDA_ARRAY_DESCRIPTOR *desc, + CUdeviceptr dptr, size_t Pitch) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 3020 */ - -CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetFormat(CUtexref hTexRef, CUarray_format fmt, + int NumPackedComponents) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, CUstreamCallback callback, void *userData, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetAddressMode(CUtexref hTexRef, int dim, + CUaddress_mode am) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#if CUDART_VERSION >= 6000 - -CUresult CUDAAPI cuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, size_t length, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetFilterMode(CUtexref hTexRef, CUfilter_mode fm) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 6000 */ - -CUresult CUDAAPI cuStreamQuery(CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetMipmapFilterMode(CUtexref hTexRef, + CUfilter_mode fm) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuStreamSynchronize(CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetMipmapLevelBias(CUtexref hTexRef, float bias) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuStreamDestroy(CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetMipmapLevelClamp(CUtexref hTexRef, + float minMipmapLevelClamp, + float maxMipmapLevelClamp) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 4000 */ - -/** @} */ /* END CUDA_STREAM */ - - -CUresult CUDAAPI cuEventCreate(CUevent *phEvent, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetMaxAnisotropy(CUtexref hTexRef, + unsigned int maxAniso) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuEventRecord(CUevent hEvent, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetBorderColor(CUtexref hTexRef, float *pBorderColor) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuEventQuery(CUevent hEvent) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefSetFlags(CUtexref hTexRef, unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuEventSynchronize(CUevent hEvent) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 3020 +CUresult CUDAAPI cuTexRefGetAddress(CUdeviceptr *pdptr, CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION >= 3020 */ -#if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuEventDestroy(CUevent hEvent) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuTexRefGetArray(CUarray *phArray, CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetMipmappedArray(CUmipmappedArray *phMipmappedArray, + CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetAddressMode(CUaddress_mode *pam, CUtexref hTexRef, + int dim) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetFilterMode(CUfilter_mode *pfm, CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetFormat(CUarray_format *pFormat, int *pNumChannels, + CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetMipmapFilterMode(CUfilter_mode *pfm, + CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetMipmapLevelBias(float *pbias, CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetMipmapLevelClamp(float *pminMipmapLevelClamp, + float *pmaxMipmapLevelClamp, + CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetMaxAnisotropy(int *pmaxAniso, CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetBorderColor(float *pBorderColor, CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefGetFlags(unsigned int *pFlags, CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefCreate(CUtexref *pTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexRefDestroy(CUtexref hTexRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuSurfRefSetArray(CUsurfref hSurfRef, CUarray hArray, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 4000 */ -CUresult CUDAAPI cuEventElapsedTime(float *pMilliseconds, CUevent hStart, CUevent hEnd) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} +/** @} */ /* END CUDA_SURFREF */ -#if CUDART_VERSION >= 8000 -CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if CUDART_VERSION >= 5000 +CUresult CUDAAPI +cuTexObjectCreate(CUtexObject *pTexObject, const CUDA_RESOURCE_DESC *pResDesc, + const CUDA_TEXTURE_DESC *pTexDesc, + const CUDA_RESOURCE_VIEW_DESC *pResViewDesc) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexObjectDestroy(CUtexObject texObject) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, + CUtexObject texObject) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexObjectGetTextureDesc(CUDA_TEXTURE_DESC *pTexDesc, + CUtexObject texObject) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuTexObjectGetResourceViewDesc( + CUDA_RESOURCE_VIEW_DESC *pResViewDesc, CUtexObject texObject) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} +/** @} */ /* END CUDA_TEXOBJECT */ -CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, CUstreamBatchMemOpParams *paramArray, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuSurfObjectCreate(CUsurfObject *pSurfObject, + const CUDA_RESOURCE_DESC *pResDesc) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#endif /* CUDART_VERSION >= 8000 */ - -/** @} */ /* END CUDA_EVENT */ - -CUresult CUDAAPI cuFuncGetAttribute(int *pi, CUfunction_attribute attrib, CUfunction hfunc) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuSurfObjectDestroy(CUsurfObject surfObject) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuFuncSetCacheConfig(CUfunction hfunc, CUfunc_cache config) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuSurfObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, + CUsurfObject surfObject) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#if CUDART_VERSION >= 4020 -CUresult CUDAAPI cuFuncSetSharedMemConfig(CUfunction hfunc, CUsharedconfig config) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif +#endif /* CUDART_VERSION >= 5000 */ #if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuLaunchKernel(CUfunction f, - unsigned int gridDimX, - unsigned int gridDimY, - unsigned int gridDimZ, - unsigned int blockDimX, - unsigned int blockDimY, - unsigned int blockDimZ, - unsigned int sharedMemBytes, - CUstream hStream, - void **kernelParams, - void **extra) -{ - return cuLaunchKernelInternal(f, gridDimX, gridDimY, gridDimZ, blockDimX, blockDimY, blockDimZ, sharedMemBytes, hStream, kernelParams, extra); -} -#endif /* CUDART_VERSION >= 4000 */ - -/** @} */ /* END CUDA_EXEC */ - - -CUresult CUDAAPI cuFuncSetBlockShape(CUfunction hfunc, int x, int y, int z) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuFuncSetSharedSize(CUfunction hfunc, unsigned int bytes) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuParamSetSize(CUfunction hfunc, unsigned int numbytes) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuParamSeti(CUfunction hfunc, int offset, unsigned int value) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuParamSetf(CUfunction hfunc, int offset, float value) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuParamSetv(CUfunction hfunc, int offset, void *ptr, unsigned int numbytes) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuLaunch(CUfunction f) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuLaunchGrid(CUfunction f, int grid_width, int grid_height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuLaunchGridAsync(CUfunction f, int grid_width, int grid_height, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - -CUresult CUDAAPI cuParamSetTexRef(CUfunction hfunc, int texunit, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -/** @} */ /* END CUDA_EXEC_DEPRECATED */ - - -#if CUDART_VERSION >= 6050 - -CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessor(int *numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuDeviceCanAccessPeer(int *canAccessPeer, CUdevice dev, + CUdevice peerDev) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuDeviceGetP2PAttribute(int *value, + CUdevice_P2PAttribute attrib, + CUdevice srcDevice, + CUdevice dstDevice) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuCtxEnablePeerAccess(CUcontext peerContext, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuCtxDisablePeerAccess(CUcontext peerContext) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuOccupancyMaxActiveBlocksPerMultiprocessorWithFlags(int *numBlocks, CUfunction func, int blockSize, size_t dynamicSMemSize, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} +/** @} */ /* END CUDA_PEER_ACCESS */ +#endif /* CUDART_VERSION >= 4000 */ -CUresult CUDAAPI cuOccupancyMaxPotentialBlockSize(int *minGridSize, int *blockSize, CUfunction func, CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, int blockSizeLimit) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuGraphicsUnregisterResource(CUgraphicsResource resource) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuOccupancyMaxPotentialBlockSizeWithFlags(int *minGridSize, int *blockSize, CUfunction func, CUoccupancyB2DSize blockSizeToDynamicSMemSize, size_t dynamicSMemSize, int blockSizeLimit, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuGraphicsSubResourceGetMappedArray( + CUarray *pArray, CUgraphicsResource resource, unsigned int arrayIndex, + unsigned int mipLevel) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -/** @} */ /* END CUDA_OCCUPANCY */ -#endif /* CUDART_VERSION >= 6050 */ +#if CUDART_VERSION >= 5000 -CUresult CUDAAPI cuTexRefSetArray(CUtexref hTexRef, CUarray hArray, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuGraphicsResourceGetMappedMipmappedArray( + CUmipmappedArray *pMipmappedArray, CUgraphicsResource resource) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuTexRefSetMipmappedArray(CUtexref hTexRef, CUmipmappedArray hMipmappedArray, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} +#endif /* CUDART_VERSION >= 5000 */ #if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuTexRefSetAddress(size_t *ByteOffset, CUtexref hTexRef, CUdeviceptr dptr, size_t bytes) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefSetAddress2D(CUtexref hTexRef, const CUDA_ARRAY_DESCRIPTOR *desc, CUdeviceptr dptr, size_t Pitch) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuGraphicsResourceGetMappedPointer( + CUdeviceptr *pDevPtr, size_t *pSize, CUgraphicsResource resource) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } #endif /* CUDART_VERSION >= 3020 */ -CUresult CUDAAPI cuTexRefSetFormat(CUtexref hTexRef, CUarray_format fmt, int NumPackedComponents) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefSetAddressMode(CUtexref hTexRef, int dim, CUaddress_mode am) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuGraphicsResourceSetMapFlags(CUgraphicsResource resource, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuTexRefSetFilterMode(CUtexref hTexRef, CUfilter_mode fm) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, + CUgraphicsResource *resources, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuTexRefSetMipmapFilterMode(CUtexref hTexRef, CUfilter_mode fm) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult CUDAAPI cuGraphicsUnmapResources(unsigned int count, + CUgraphicsResource *resources, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -CUresult CUDAAPI cuTexRefSetMipmapLevelBias(CUtexref hTexRef, float bias) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} +/** @} */ /* END CUDA_GRAPHICS */ -CUresult CUDAAPI cuTexRefSetMipmapLevelClamp(CUtexref hTexRef, float minMipmapLevelClamp, float maxMipmapLevelClamp) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} +CUresult CUDAAPI cuGetExportTable(const void **ppExportTable, + const CUuuid *pExportTableId) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + cudaError_t e = cudaGetExportTable(ppExportTable, pExportTableId); + assert(e == cudaSuccess); + return CUDA_SUCCESS; +} + +#if defined(CUDART_VERSION_INTERNAL) || \ + (CUDART_VERSION >= 4000 && CUDART_VERSION < 6050) +CUresult CUDAAPI cuMemHostRegister(void *p, size_t bytesize, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +#endif /* defined(CUDART_VERSION_INTERNAL) || (CUDART_VERSION >= 4000 && \ + CUDART_VERSION < 6050) */ + +#if defined(CUDART_VERSION_INTERNAL) || \ + (CUDART_VERSION >= 5050 && CUDART_VERSION < 6050) +CUresult CUDAAPI cuLinkCreate(unsigned int numOptions, CUjit_option *options, + void **optionValues, CUlinkState *stateOut) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuLinkAddData(CUlinkState state, CUjitInputType type, + void *data, size_t size, const char *name, + unsigned int numOptions, CUjit_option *options, + void **optionValues) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuLinkAddFile(CUlinkState state, CUjitInputType type, + const char *path, unsigned int numOptions, + CUjit_option *options, void **optionValues) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +#endif /* CUDART_VERSION_INTERNAL || (CUDART_VERSION >= 5050 && CUDART_VERSION \ + < 6050) */ + +#if defined(CUDART_VERSION_INTERNAL) || \ + (CUDART_VERSION >= 3020 && CUDART_VERSION < 4010) +CUresult CUDAAPI cuTexRefSetAddress2D_v2(CUtexref hTexRef, + const CUDA_ARRAY_DESCRIPTOR *desc, + CUdeviceptr dptr, size_t Pitch) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +#endif /* CUDART_VERSION_INTERNAL || (CUDART_VERSION >= 3020 && CUDART_VERSION \ + < 4010) */ -CUresult CUDAAPI cuTexRefSetMaxAnisotropy(CUtexref hTexRef, unsigned int maxAniso) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if defined(CUDART_VERSION_INTERNAL) || CUDART_VERSION < 4000 +CUresult CUDAAPI cuCtxDestroy(CUcontext ctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuCtxPopCurrent(CUcontext *pctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuCtxPushCurrent(CUcontext ctx) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamDestroy(CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuEventDestroy(CUevent hEvent) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif /* CUDART_VERSION_INTERNAL || CUDART_VERSION < 4000 */ -CUresult CUDAAPI cuTexRefSetBorderColor(CUtexref hTexRef, float *pBorderColor) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +#if defined(CUDART_VERSION_INTERNAL) +CUresult CUDAAPI cuMemcpyHtoD_v2(CUdeviceptr dstDevice, const void *srcHost, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyDtoH_v2(void *dstHost, CUdeviceptr srcDevice, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyDtoD_v2(CUdeviceptr dstDevice, CUdeviceptr srcDevice, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyDtoA_v2(CUarray dstArray, size_t dstOffset, + CUdeviceptr srcDevice, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyAtoD_v2(CUdeviceptr dstDevice, CUarray srcArray, + size_t srcOffset, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyHtoA_v2(CUarray dstArray, size_t dstOffset, + const void *srcHost, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyAtoH_v2(void *dstHost, CUarray srcArray, + size_t srcOffset, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyAtoA_v2(CUarray dstArray, size_t dstOffset, + CUarray srcArray, size_t srcOffset, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyHtoAAsync_v2(CUarray dstArray, size_t dstOffset, + const void *srcHost, size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyAtoHAsync_v2(void *dstHost, CUarray srcArray, + size_t srcOffset, size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpy2D_v2(const CUDA_MEMCPY2D *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpy2DUnaligned_v2(const CUDA_MEMCPY2D *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpy3D_v2(const CUDA_MEMCPY3D *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyHtoDAsync_v2(CUdeviceptr dstDevice, + const void *srcHost, size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyDtoHAsync_v2(void *dstHost, CUdeviceptr srcDevice, + size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyDtoDAsync_v2(CUdeviceptr dstDevice, + CUdeviceptr srcDevice, size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpy2DAsync_v2(const CUDA_MEMCPY2D *pCopy, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpy3DAsync_v2(const CUDA_MEMCPY3D *pCopy, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD8_v2(CUdeviceptr dstDevice, unsigned char uc, + size_t N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD16_v2(CUdeviceptr dstDevice, unsigned short us, + size_t N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD32_v2(CUdeviceptr dstDevice, unsigned int ui, + size_t N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD2D8_v2(CUdeviceptr dstDevice, size_t dstPitch, + unsigned char uc, size_t Width, + size_t Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD2D16_v2(CUdeviceptr dstDevice, size_t dstPitch, + unsigned short us, size_t Width, + size_t Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD2D32_v2(CUdeviceptr dstDevice, size_t dstPitch, + unsigned int ui, size_t Width, + size_t Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, + size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, + CUdeviceptr srcDevice, CUcontext srcContext, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, + CUdeviceptr srcDevice, CUcontext srcContext, + size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER *pCopy, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, + size_t N, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, + size_t N, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, + size_t N, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned char uc, size_t Width, + size_t Height, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned short us, size_t Width, + size_t Height, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, + unsigned int ui, size_t Width, + size_t Height, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +CUresult CUDAAPI cuStreamGetPriority(CUstream hStream, int *priority) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamGetFlags(CUstream hStream, unsigned int *flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, + CUstreamCallback callback, void *userData, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, + size_t length, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamQuery(CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamSynchronize(CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuEventRecord(CUevent hEvent, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuLaunchKernel(CUfunction f, unsigned int gridDimX, + unsigned int gridDimY, unsigned int gridDimZ, + unsigned int blockDimX, unsigned int blockDimY, + unsigned int blockDimZ, + unsigned int sharedMemBytes, CUstream hStream, + void **kernelParams, void **extra) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, + CUgraphicsResource *resources, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuGraphicsUnmapResources(unsigned int count, + CUgraphicsResource *resources, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, + CUdevice dstDevice, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, + cuuint32_t value, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, + cuuint32_t value, unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, + CUstreamBatchMemOpParams *paramArray, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } +#endif -CUresult CUDAAPI cuTexRefSetFlags(CUtexref hTexRef, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +CUresult cuProfilerInitialize(const char *configFile, const char *outputFile, + CUoutput_mode outputMode) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult cuProfilerStart(void) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +CUresult cuProfilerStop(void) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } -#if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuTexRefGetAddress(CUdeviceptr *pdptr, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif /* CUDART_VERSION >= 3020 */ +//_ptds -CUresult CUDAAPI cuTexRefGetArray(CUarray *phArray, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetMipmappedArray(CUmipmappedArray *phMipmappedArray, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetAddressMode(CUaddress_mode *pam, CUtexref hTexRef, int dim) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetFilterMode(CUfilter_mode *pfm, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetFormat(CUarray_format *pFormat, int *pNumChannels, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetMipmapFilterMode(CUfilter_mode *pfm, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetMipmapLevelBias(float *pbias, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetMipmapLevelClamp(float *pminMipmapLevelClamp, float *pmaxMipmapLevelClamp, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetMaxAnisotropy(int *pmaxAniso, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetBorderColor(float *pBorderColor, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefGetFlags(unsigned int *pFlags, CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefCreate(CUtexref *pTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexRefDestroy(CUtexref hTexRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuSurfRefSetArray(CUsurfref hSurfRef, CUarray hArray, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuSurfRefGetArray(CUarray *phArray, CUsurfref hSurfRef) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -/** @} */ /* END CUDA_SURFREF */ - -#if CUDART_VERSION >= 5000 -CUresult CUDAAPI cuTexObjectCreate(CUtexObject *pTexObject, const CUDA_RESOURCE_DESC *pResDesc, const CUDA_TEXTURE_DESC *pTexDesc, const CUDA_RESOURCE_VIEW_DESC *pResViewDesc) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexObjectDestroy(CUtexObject texObject) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, CUtexObject texObject) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexObjectGetTextureDesc(CUDA_TEXTURE_DESC *pTexDesc, CUtexObject texObject) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuTexObjectGetResourceViewDesc(CUDA_RESOURCE_VIEW_DESC *pResViewDesc, CUtexObject texObject) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -/** @} */ /* END CUDA_TEXOBJECT */ - - -CUresult CUDAAPI cuSurfObjectCreate(CUsurfObject *pSurfObject, const CUDA_RESOURCE_DESC *pResDesc) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuSurfObjectDestroy(CUsurfObject surfObject) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuSurfObjectGetResourceDesc(CUDA_RESOURCE_DESC *pResDesc, CUsurfObject surfObject) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -#endif /* CUDART_VERSION >= 5000 */ - -#if CUDART_VERSION >= 4000 -CUresult CUDAAPI cuDeviceCanAccessPeer(int *canAccessPeer, CUdevice dev, CUdevice peerDev) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuDeviceGetP2PAttribute(int* value, CUdevice_P2PAttribute attrib, CUdevice srcDevice, CUdevice dstDevice) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuCtxEnablePeerAccess(CUcontext peerContext, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuCtxDisablePeerAccess(CUcontext peerContext) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -/** @} */ /* END CUDA_PEER_ACCESS */ -#endif /* CUDART_VERSION >= 4000 */ - - -CUresult CUDAAPI cuGraphicsUnregisterResource(CUgraphicsResource resource) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuGraphicsSubResourceGetMappedArray(CUarray *pArray, CUgraphicsResource resource, unsigned int arrayIndex, unsigned int mipLevel) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -#if CUDART_VERSION >= 5000 - -CUresult CUDAAPI cuGraphicsResourceGetMappedMipmappedArray(CUmipmappedArray *pMipmappedArray, CUgraphicsResource resource) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -#endif /* CUDART_VERSION >= 5000 */ - -#if CUDART_VERSION >= 3020 -CUresult CUDAAPI cuGraphicsResourceGetMappedPointer(CUdeviceptr *pDevPtr, size_t *pSize, CUgraphicsResource resource) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif /* CUDART_VERSION >= 3020 */ - -CUresult CUDAAPI cuGraphicsResourceSetMapFlags(CUgraphicsResource resource, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, CUgraphicsResource *resources, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -CUresult CUDAAPI cuGraphicsUnmapResources(unsigned int count, CUgraphicsResource *resources, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -/** @} */ /* END CUDA_GRAPHICS */ - -CUresult CUDAAPI cuGetExportTable(const void **ppExportTable, const CUuuid *pExportTableId) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - cudaError_t e = cudaGetExportTable(ppExportTable, pExportTableId); - assert(e == cudaSuccess); - return CUDA_SUCCESS; -} - - -#if defined(CUDART_VERSION_INTERNAL) || (CUDART_VERSION >= 4000 && CUDART_VERSION < 6050) -CUresult CUDAAPI cuMemHostRegister(void *p, size_t bytesize, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif /* defined(CUDART_VERSION_INTERNAL) || (CUDART_VERSION >= 4000 && CUDART_VERSION < 6050) */ - -#if defined(CUDART_VERSION_INTERNAL) || (CUDART_VERSION >= 5050 && CUDART_VERSION < 6050) -CUresult CUDAAPI cuLinkCreate(unsigned int numOptions, CUjit_option *options, void **optionValues, CUlinkState *stateOut) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -CUresult CUDAAPI cuLinkAddData(CUlinkState state, CUjitInputType type, void *data, size_t size, const char *name, - unsigned int numOptions, CUjit_option *options, void **optionValues) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -CUresult CUDAAPI cuLinkAddFile(CUlinkState state, CUjitInputType type, const char *path, - unsigned int numOptions, CUjit_option *options, void **optionValues) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif /* CUDART_VERSION_INTERNAL || (CUDART_VERSION >= 5050 && CUDART_VERSION < 6050) */ - -#if defined(CUDART_VERSION_INTERNAL) || (CUDART_VERSION >= 3020 && CUDART_VERSION < 4010) -CUresult CUDAAPI cuTexRefSetAddress2D_v2(CUtexref hTexRef, const CUDA_ARRAY_DESCRIPTOR *desc, CUdeviceptr dptr, size_t Pitch) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif /* CUDART_VERSION_INTERNAL || (CUDART_VERSION >= 3020 && CUDART_VERSION < 4010) */ - -#if defined(CUDART_VERSION_INTERNAL) || CUDART_VERSION < 4000 -CUresult CUDAAPI cuCtxDestroy(CUcontext ctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -CUresult CUDAAPI cuCtxPopCurrent(CUcontext *pctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -CUresult CUDAAPI cuCtxPushCurrent(CUcontext ctx) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -CUresult CUDAAPI cuStreamDestroy(CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -CUresult CUDAAPI cuEventDestroy(CUevent hEvent) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif /* CUDART_VERSION_INTERNAL || CUDART_VERSION < 4000 */ - -#if defined(CUDART_VERSION_INTERNAL) - CUresult CUDAAPI cuMemcpyHtoD_v2(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyDtoH_v2(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyDtoD_v2(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyDtoA_v2(CUarray dstArray, size_t dstOffset, CUdeviceptr srcDevice, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyAtoD_v2(CUdeviceptr dstDevice, CUarray srcArray, size_t srcOffset, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyHtoA_v2(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyAtoH_v2(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyAtoA_v2(CUarray dstArray, size_t dstOffset, CUarray srcArray, size_t srcOffset, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyHtoAAsync_v2(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyAtoHAsync_v2(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpy2D_v2(const CUDA_MEMCPY2D *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpy2DUnaligned_v2(const CUDA_MEMCPY2D *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpy3D_v2(const CUDA_MEMCPY3D *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyHtoDAsync_v2(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyDtoHAsync_v2(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyDtoDAsync_v2(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpy2DAsync_v2(const CUDA_MEMCPY2D *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpy3DAsync_v2(const CUDA_MEMCPY3D *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD8_v2(CUdeviceptr dstDevice, unsigned char uc, size_t N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD16_v2(CUdeviceptr dstDevice, unsigned short us, size_t N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD32_v2(CUdeviceptr dstDevice, unsigned int ui, size_t N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD2D8_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD2D16_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD2D32_v2(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpy(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyAsync(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyPeer(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpyPeerAsync(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpy3DPeer(const CUDA_MEMCPY3D_PEER *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemcpy3DPeerAsync(const CUDA_MEMCPY3D_PEER *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - CUresult CUDAAPI cuMemsetD8Async(CUdeviceptr dstDevice, unsigned char uc, size_t N, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD16Async(CUdeviceptr dstDevice, unsigned short us, size_t N, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD32Async(CUdeviceptr dstDevice, unsigned int ui, size_t N, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD2D8Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD2D16Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned short us, size_t Width, size_t Height, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemsetD2D32Async(CUdeviceptr dstDevice, size_t dstPitch, unsigned int ui, size_t Width, size_t Height, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - CUresult CUDAAPI cuStreamGetPriority(CUstream hStream, int *priority) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamGetFlags(CUstream hStream, unsigned int *flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamWaitEvent(CUstream hStream, CUevent hEvent, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamAddCallback(CUstream hStream, CUstreamCallback callback, void *userData, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamAttachMemAsync(CUstream hStream, CUdeviceptr dptr, size_t length, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamQuery(CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamSynchronize(CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuEventRecord(CUevent hEvent, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuLaunchKernel(CUfunction f, unsigned int gridDimX, unsigned int gridDimY, unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, void **kernelParams, void **extra) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuGraphicsMapResources(unsigned int count, CUgraphicsResource *resources, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuGraphicsUnmapResources(unsigned int count, CUgraphicsResource *resources, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuMemPrefetchAsync(CUdeviceptr devPtr, size_t count, CUdevice dstDevice, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamWriteValue32(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamWaitValue32(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - CUresult CUDAAPI cuStreamBatchMemOp(CUstream stream, unsigned int count, CUstreamBatchMemOpParams *paramArray, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -#endif - -CUresult cuProfilerInitialize ( const char* configFile, const char* outputFile, CUoutput_mode outputMode ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -CUresult cuProfilerStart ( void ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -CUresult cuProfilerStop ( void ) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -//_ptds - -extern "C" CUresult CUDAAPI cuMemcpy_ptds(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -extern "C" CUresult CUDAAPI cuMemcpyPeer_ptds(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - extern "C" CUresult CUDAAPI cuMemcpyHtoD_v2_ptds(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpyDtoH_v2_ptds(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpyDtoD_v2_ptds(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpy2DUnaligned_v2_ptds(const CUDA_MEMCPY2D *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpy3D_v2_ptds(const CUDA_MEMCPY3D *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpy3DPeer_ptds(const CUDA_MEMCPY3D_PEER *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuMemsetD8_v2_ptds(CUdeviceptr dstDevice, unsigned char uc, unsigned int N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuMemsetD16_v2_ptds(CUdeviceptr dstDevice, unsigned short us, unsigned int N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuMemsetD32_v2_ptds(CUdeviceptr dstDevice, unsigned int ui, unsigned int N) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuMemsetD2D8_v2_ptds(CUdeviceptr dstDevice, unsigned int dstPitch, unsigned char uc, unsigned int Width, unsigned int Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuMemsetD2D16_v2_ptds(CUdeviceptr dstDevice, unsigned int dstPitch, unsigned short us, unsigned int Width, unsigned int Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuMemsetD2D32_v2_ptds(CUdeviceptr dstDevice, unsigned int dstPitch, unsigned int ui, unsigned int Width, unsigned int Height) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +extern "C" CUresult CUDAAPI cuMemcpy_ptds(CUdeviceptr dst, CUdeviceptr src, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuMemcpyPeer_ptds(CUdeviceptr dstDevice, + CUcontext dstContext, + CUdeviceptr srcDevice, + CUcontext srcContext, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuMemcpyHtoD_v2_ptds(CUdeviceptr dstDevice, + const void *srcHost, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpyDtoH_v2_ptds(void *dstHost, + CUdeviceptr srcDevice, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpyDtoD_v2_ptds(CUdeviceptr dstDevice, + CUdeviceptr srcDevice, + size_t ByteCount) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI +cuMemcpy2DUnaligned_v2_ptds(const CUDA_MEMCPY2D *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpy3D_v2_ptds(const CUDA_MEMCPY3D *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI +cuMemcpy3DPeer_ptds(const CUDA_MEMCPY3D_PEER *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemsetD8_v2_ptds(CUdeviceptr dstDevice, + unsigned char uc, + unsigned int N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemsetD16_v2_ptds(CUdeviceptr dstDevice, + unsigned short us, + unsigned int N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemsetD32_v2_ptds(CUdeviceptr dstDevice, + unsigned int ui, + unsigned int N) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemsetD2D8_v2_ptds(CUdeviceptr dstDevice, + unsigned int dstPitch, + unsigned char uc, + unsigned int Width, + unsigned int Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemsetD2D16_v2_ptds(CUdeviceptr dstDevice, + unsigned int dstPitch, + unsigned short us, + unsigned int Width, + unsigned int Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemsetD2D32_v2_ptds(CUdeviceptr dstDevice, + unsigned int dstPitch, + unsigned int ui, + unsigned int Width, + unsigned int Height) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } //_ptsz -extern "C" CUresult CUDAAPI cuMemcpy3DPeer_ptsz(const CUDA_MEMCPY3D_PEER *pCopy) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -extern "C" CUresult CUDAAPI cuMemcpyAsync_ptsz(CUdeviceptr dst, CUdeviceptr src, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -extern "C" CUresult CUDAAPI cuMemcpyPeerAsync_ptsz(CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, CUcontext srcContext, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpyHtoAAsync_v2_ptsz(CUarray dstArray, size_t dstOffset, const void *srcHost, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpyAtoHAsync_v2_ptsz(void *dstHost, CUarray srcArray, size_t srcOffset, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpyHtoDAsync_v2_ptsz(CUdeviceptr dstDevice, const void *srcHost, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpyDtoHAsync_v2_ptsz(void *dstHost, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpyDtoDAsync_v2_ptsz(CUdeviceptr dstDevice, CUdeviceptr srcDevice, size_t ByteCount, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpy2DAsync_v2_ptsz(const CUDA_MEMCPY2D *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpy3DAsync_v2_ptsz(const CUDA_MEMCPY3D *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemcpy3DPeerAsync_ptsz(const CUDA_MEMCPY3D_PEER *pCopy, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - extern "C" CUresult CUDAAPI cuMemsetD8Async_ptsz(CUdeviceptr dstDevice, unsigned char uc, size_t N, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuMemsetD2D8Async_ptsz(CUdeviceptr dstDevice, size_t dstPitch, unsigned char uc, size_t Width, size_t Height, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuLaunchKernel_ptsz(CUfunction f, unsigned int gridDimX, unsigned int gridDimY, unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, void **kernelParams, void **extra) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuEventRecord_ptsz(CUevent hEvent, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuStreamWriteValue32_ptsz(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuStreamWaitValue32_ptsz(CUstream stream, CUdeviceptr addr, cuuint32_t value, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - extern "C" CUresult CUDAAPI cuStreamBatchMemOp_ptsz(CUstream stream, unsigned int count, CUstreamBatchMemOpParams *paramArray, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuStreamGetPriority_ptsz(CUstream hStream, int *priority) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuStreamGetFlags_ptsz(CUstream hStream, unsigned int *flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - -extern "C" CUresult CUDAAPI cuStreamWaitEvent_ptsz(CUstream hStream, CUevent hEvent, unsigned int Flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -extern "C" CUresult CUDAAPI cuStreamAddCallback_ptsz(CUstream hStream, CUstreamCallback callback, void *userData, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -extern "C" CUresult CUDAAPI cuStreamSynchronize_ptsz(CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - extern "C" CUresult CUDAAPI cuStreamQuery_ptsz(CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} -extern "C" CUresult CUDAAPI cuStreamAttachMemAsync_ptsz(CUstream hStream, CUdeviceptr dptr, size_t length, unsigned int flags) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - -extern "C" CUresult CUDAAPI cuGraphicsMapResources_ptsz(unsigned int count, CUgraphicsResource *resources, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - -extern "C" CUresult CUDAAPI cuGraphicsUnmapResources_ptsz(unsigned int count, CUgraphicsResource *resources, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; -} - - extern "C" CUresult CUDAAPI cuMemPrefetchAsync_ptsz(CUdeviceptr devPtr, size_t count, CUdevice dstDevice, CUstream hStream) -{ - if(g_debug_execution >= 3){ - announce_call(__my_func__); - } - printf("WARNING: this function has not been implemented yet."); - return CUDA_SUCCESS; +extern "C" CUresult CUDAAPI +cuMemcpy3DPeer_ptsz(const CUDA_MEMCPY3D_PEER *pCopy) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuMemcpyAsync_ptsz(CUdeviceptr dst, CUdeviceptr src, + size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuMemcpyPeerAsync_ptsz( + CUdeviceptr dstDevice, CUcontext dstContext, CUdeviceptr srcDevice, + CUcontext srcContext, size_t ByteCount, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpyHtoAAsync_v2_ptsz(CUarray dstArray, + size_t dstOffset, + const void *srcHost, + size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpyAtoHAsync_v2_ptsz(void *dstHost, + CUarray srcArray, + size_t srcOffset, + size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpyHtoDAsync_v2_ptsz(CUdeviceptr dstDevice, + const void *srcHost, + size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpyDtoHAsync_v2_ptsz(void *dstHost, + CUdeviceptr srcDevice, + size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpyDtoDAsync_v2_ptsz(CUdeviceptr dstDevice, + CUdeviceptr srcDevice, + size_t ByteCount, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpy2DAsync_v2_ptsz(const CUDA_MEMCPY2D *pCopy, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemcpy3DAsync_v2_ptsz(const CUDA_MEMCPY3D *pCopy, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI +cuMemcpy3DPeerAsync_ptsz(const CUDA_MEMCPY3D_PEER *pCopy, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuMemsetD8Async_ptsz(CUdeviceptr dstDevice, + unsigned char uc, size_t N, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuMemsetD2D8Async_ptsz(CUdeviceptr dstDevice, + size_t dstPitch, + unsigned char uc, + size_t Width, size_t Height, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuLaunchKernel_ptsz( + CUfunction f, unsigned int gridDimX, unsigned int gridDimY, + unsigned int gridDimZ, unsigned int blockDimX, unsigned int blockDimY, + unsigned int blockDimZ, unsigned int sharedMemBytes, CUstream hStream, + void **kernelParams, void **extra) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuEventRecord_ptsz(CUevent hEvent, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuStreamWriteValue32_ptsz(CUstream stream, + CUdeviceptr addr, + cuuint32_t value, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuStreamWaitValue32_ptsz(CUstream stream, + CUdeviceptr addr, + cuuint32_t value, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuStreamBatchMemOp_ptsz( + CUstream stream, unsigned int count, CUstreamBatchMemOpParams *paramArray, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuStreamGetPriority_ptsz(CUstream hStream, + int *priority) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuStreamGetFlags_ptsz(CUstream hStream, + unsigned int *flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuStreamWaitEvent_ptsz(CUstream hStream, + CUevent hEvent, + unsigned int Flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuStreamAddCallback_ptsz(CUstream hStream, + CUstreamCallback callback, + void *userData, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuStreamSynchronize_ptsz(CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuStreamQuery_ptsz(CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} +extern "C" CUresult CUDAAPI cuStreamAttachMemAsync_ptsz(CUstream hStream, + CUdeviceptr dptr, + size_t length, + unsigned int flags) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuGraphicsMapResources_ptsz( + unsigned int count, CUgraphicsResource *resources, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuGraphicsUnmapResources_ptsz( + unsigned int count, CUgraphicsResource *resources, CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; +} + +extern "C" CUresult CUDAAPI cuMemPrefetchAsync_ptsz(CUdeviceptr devPtr, + size_t count, + CUdevice dstDevice, + CUstream hStream) { + if (g_debug_execution >= 3) { + announce_call(__my_func__); + } + printf("WARNING: this function has not been implemented yet."); + return CUDA_SUCCESS; } diff --git a/libcuda/cuobjdump.h b/libcuda/cuobjdump.h index 6ab6778..38afa4c 100644 --- a/libcuda/cuobjdump.h +++ b/libcuda/cuobjdump.h @@ -1,80 +1,81 @@ #ifndef __cuobjdump_h__ #define __cuobjdump_h__ -#include -#include #include +#include +#include -typedef void * yyscan_t; +typedef void *yyscan_t; struct cuobjdump_parser { - yyscan_t scanner; - int elfserial; - int ptxserial; - FILE *ptxfile; - FILE *elffile; - FILE *sassfile; - char filename [1024]; + yyscan_t scanner; + int elfserial; + int ptxserial; + FILE *ptxfile; + FILE *elffile; + FILE *sassfile; + char filename[1024]; }; class cuobjdumpSection { -public: - //Constructor - cuobjdumpSection() { - arch = 0; - identifier = ""; - } - virtual ~cuobjdumpSection() {} - unsigned getArch() {return arch;} - void setArch(unsigned a) {arch = a;} - std::string getIdentifier() {return identifier;} - void setIdentifier(std::string i) {identifier = i;} - virtual void print(){std::cout << "cuobjdump Section: unknown type" << std::endl;} -private: - unsigned arch; - std::string identifier; + public: + // Constructor + cuobjdumpSection() { + arch = 0; + identifier = ""; + } + virtual ~cuobjdumpSection() {} + unsigned getArch() { return arch; } + void setArch(unsigned a) { arch = a; } + std::string getIdentifier() { return identifier; } + void setIdentifier(std::string i) { identifier = i; } + virtual void print() { + std::cout << "cuobjdump Section: unknown type" << std::endl; + } + + private: + unsigned arch; + std::string identifier; }; -class cuobjdumpELFSection : public cuobjdumpSection -{ -public: - cuobjdumpELFSection() {} - virtual ~cuobjdumpELFSection() { - elffilename = ""; - sassfilename = ""; - } - std::string getELFfilename() {return elffilename;} - void setELFfilename(std::string f) {elffilename = f;} - std::string getSASSfilename() {return sassfilename;} - void setSASSfilename(std::string f) {sassfilename = f;} - virtual void print() { - std::cout << "ELF Section:" << std::endl; - std::cout << "arch: sm_" << getArch() << std::endl; - std::cout << "identifier: " << getIdentifier() << std::endl; - std::cout << "elf filename: " << getELFfilename() << std::endl; - std::cout << "sass filename: " << getSASSfilename() << std::endl; - std::cout << std::endl; - } -private: - std::string elffilename; - std::string sassfilename; +class cuobjdumpELFSection : public cuobjdumpSection { + public: + cuobjdumpELFSection() {} + virtual ~cuobjdumpELFSection() { + elffilename = ""; + sassfilename = ""; + } + std::string getELFfilename() { return elffilename; } + void setELFfilename(std::string f) { elffilename = f; } + std::string getSASSfilename() { return sassfilename; } + void setSASSfilename(std::string f) { sassfilename = f; } + virtual void print() { + std::cout << "ELF Section:" << std::endl; + std::cout << "arch: sm_" << getArch() << std::endl; + std::cout << "identifier: " << getIdentifier() << std::endl; + std::cout << "elf filename: " << getELFfilename() << std::endl; + std::cout << "sass filename: " << getSASSfilename() << std::endl; + std::cout << std::endl; + } + + private: + std::string elffilename; + std::string sassfilename; }; -class cuobjdumpPTXSection : public cuobjdumpSection -{ -public: - cuobjdumpPTXSection(){ - ptxfilename = ""; - } - std::string getPTXfilename() {return ptxfilename;} - void setPTXfilename(std::string f) {ptxfilename = f;} - virtual void print() { - std::cout << "PTX Section:" << std::endl; - std::cout << "arch: sm_" << getArch() << std::endl; - std::cout << "identifier: " << getIdentifier() << std::endl; - std::cout << "ptx filename: " << getPTXfilename() << std::endl; - std::cout << std::endl; - } -private: - std::string ptxfilename; +class cuobjdumpPTXSection : public cuobjdumpSection { + public: + cuobjdumpPTXSection() { ptxfilename = ""; } + std::string getPTXfilename() { return ptxfilename; } + void setPTXfilename(std::string f) { ptxfilename = f; } + virtual void print() { + std::cout << "PTX Section:" << std::endl; + std::cout << "arch: sm_" << getArch() << std::endl; + std::cout << "identifier: " << getIdentifier() << std::endl; + std::cout << "ptx filename: " << getPTXfilename() << std::endl; + std::cout << std::endl; + } + + private: + std::string ptxfilename; }; #endif /* __cuobjdump_h__ */ diff --git a/libcuda/gpgpu_context.h b/libcuda/gpgpu_context.h index 61d7507..d0cd7c4 100644 --- a/libcuda/gpgpu_context.h +++ b/libcuda/gpgpu_context.h @@ -1,76 +1,83 @@ #ifndef __gpgpu_context_h__ #define __gpgpu_context_h__ -#include "cuda_api_object.h" -#include "../src/cuda-sim/ptx_loader.h" -#include "../src/cuda-sim/ptx_parser.h" -#include "../src/gpgpusim_entrypoint.h" #include "../src/cuda-sim/cuda-sim.h" #include "../src/cuda-sim/cuda_device_runtime.h" #include "../src/cuda-sim/ptx-stats.h" +#include "../src/cuda-sim/ptx_loader.h" +#include "../src/cuda-sim/ptx_parser.h" +#include "../src/gpgpusim_entrypoint.h" +#include "cuda_api_object.h" class gpgpu_context { - public: - gpgpu_context() { - g_global_allfiles_symbol_table = NULL; - sm_next_access_uid=0; - warp_inst_sm_next_uid=0; - operand_info_sm_next_uid = 1; - kernel_info_m_next_uid = 1; - g_num_ptx_inst_uid = 0; - g_ptx_cta_info_uid = 1; - symbol_sm_next_uid = 1; - function_info_sm_next_uid = 1; - debug_tensorcore = 0; - api = new cuda_runtime_api(this); - ptxinfo = new ptxinfo_data(this); - ptx_parser = new ptx_recognizer(this); - the_gpgpusim = new GPGPUsim_ctx(this); - func_sim = new cuda_sim(this); - device_runtime = new cuda_device_runtime(this); - stats = new ptx_stats(this); - } - // global list - symbol_table *g_global_allfiles_symbol_table; - const char *g_filename; - unsigned sm_next_access_uid; - unsigned warp_inst_sm_next_uid; - unsigned operand_info_sm_next_uid;//uid for operand_info - unsigned kernel_info_m_next_uid;//uid for kernel_info_t - unsigned g_num_ptx_inst_uid; //uid for ptx inst inside ptx_instruction - unsigned long long g_ptx_cta_info_uid; - unsigned symbol_sm_next_uid; //uid for symbol - unsigned function_info_sm_next_uid; - std::vector s_g_pc_to_insn; // a direct mapping from PC to instruction - bool debug_tensorcore; + public: + gpgpu_context() { + g_global_allfiles_symbol_table = NULL; + sm_next_access_uid = 0; + warp_inst_sm_next_uid = 0; + operand_info_sm_next_uid = 1; + kernel_info_m_next_uid = 1; + g_num_ptx_inst_uid = 0; + g_ptx_cta_info_uid = 1; + symbol_sm_next_uid = 1; + function_info_sm_next_uid = 1; + debug_tensorcore = 0; + api = new cuda_runtime_api(this); + ptxinfo = new ptxinfo_data(this); + ptx_parser = new ptx_recognizer(this); + the_gpgpusim = new GPGPUsim_ctx(this); + func_sim = new cuda_sim(this); + device_runtime = new cuda_device_runtime(this); + stats = new ptx_stats(this); + } + // global list + symbol_table *g_global_allfiles_symbol_table; + const char *g_filename; + unsigned sm_next_access_uid; + unsigned warp_inst_sm_next_uid; + unsigned operand_info_sm_next_uid; // uid for operand_info + unsigned kernel_info_m_next_uid; // uid for kernel_info_t + unsigned g_num_ptx_inst_uid; // uid for ptx inst inside ptx_instruction + unsigned long long g_ptx_cta_info_uid; + unsigned symbol_sm_next_uid; // uid for symbol + unsigned function_info_sm_next_uid; + std::vector + s_g_pc_to_insn; // a direct mapping from PC to instruction + bool debug_tensorcore; - // objects pointers for each file - cuda_runtime_api* api; - ptxinfo_data* ptxinfo; - ptx_recognizer* ptx_parser; - GPGPUsim_ctx* the_gpgpusim; - cuda_sim* func_sim; - cuda_device_runtime* device_runtime; - ptx_stats* stats; - // member function list - void synchronize(); - void exit_simulation(); - void print_simulation_time(); - int gpgpu_opencl_ptx_sim_main_perf( kernel_info_t *grid ); - void cuobjdumpParseBinary(unsigned int handle); - class symbol_table *gpgpu_ptx_sim_load_ptx_from_string( const char *p, unsigned source_num ); - class symbol_table *gpgpu_ptx_sim_load_ptx_from_filename( const char *filename ); - void gpgpu_ptx_info_load_from_filename( const char *filename, unsigned sm_version); - void gpgpu_ptxinfo_load_from_string( const char *p_for_info, unsigned source_num, unsigned sm_version=20, int no_of_ptx=0 ); - void print_ptx_file( const char *p, unsigned source_num, const char *filename ); - class symbol_table* init_parser(const char*); - class gpgpu_sim *gpgpu_ptx_sim_init_perf(); - void start_sim_thread(int api); - struct _cuda_device_id *GPGPUSim_Init(); - void ptx_reg_options(option_parser_t opp); - const ptx_instruction* pc_to_instruction(unsigned pc); - const warp_inst_t *ptx_fetch_inst( address_type pc ); - unsigned translate_pc_to_ptxlineno(unsigned pc); + // objects pointers for each file + cuda_runtime_api *api; + ptxinfo_data *ptxinfo; + ptx_recognizer *ptx_parser; + GPGPUsim_ctx *the_gpgpusim; + cuda_sim *func_sim; + cuda_device_runtime *device_runtime; + ptx_stats *stats; + // member function list + void synchronize(); + void exit_simulation(); + void print_simulation_time(); + int gpgpu_opencl_ptx_sim_main_perf(kernel_info_t *grid); + void cuobjdumpParseBinary(unsigned int handle); + class symbol_table *gpgpu_ptx_sim_load_ptx_from_string(const char *p, + unsigned source_num); + class symbol_table *gpgpu_ptx_sim_load_ptx_from_filename( + const char *filename); + void gpgpu_ptx_info_load_from_filename(const char *filename, + unsigned sm_version); + void gpgpu_ptxinfo_load_from_string(const char *p_for_info, + unsigned source_num, + unsigned sm_version = 20, + int no_of_ptx = 0); + void print_ptx_file(const char *p, unsigned source_num, const char *filename); + class symbol_table *init_parser(const char *); + class gpgpu_sim *gpgpu_ptx_sim_init_perf(); + void start_sim_thread(int api); + struct _cuda_device_id *GPGPUSim_Init(); + void ptx_reg_options(option_parser_t opp); + const ptx_instruction *pc_to_instruction(unsigned pc); + const warp_inst_t *ptx_fetch_inst(address_type pc); + unsigned translate_pc_to_ptxlineno(unsigned pc); }; -gpgpu_context* GPGPU_Context(); +gpgpu_context *GPGPU_Context(); #endif /* __gpgpu_context_h__ */ -- cgit v1.3