Move g_cuda_device_launch_param_map and g_cuda_device_launch_op

Signed-off-by: Mengchi Zhang <[email protected]>

3 files changed, 42 insertions, 45 deletions

diff --git a/src/cuda-sim/cuda_device_runtime.cc b/src/cuda-sim/cuda_device_runtime.cc
index 22304db..326a1c5 100644
--- a/src/cuda-sim/cuda_device_runtime.cc
+++ b/src/cuda-sim/cuda_device_runtime.cc
@@ -27,44 +27,7 @@ unsigned long long g_max_total_param_size = 0;
       std::cout.flush(); \
    }
 
-class device_launch_config_t {
 
-public:
-    device_launch_config_t() {}
-
-    device_launch_config_t(dim3 _grid_dim,
-        dim3 _block_dim,
-        unsigned int _shared_mem,
-        function_info * _entry):
-            grid_dim(_grid_dim),
-            block_dim(_block_dim),
-            shared_mem(_shared_mem),
-            entry(_entry) {}
-    
-    dim3 grid_dim;
-    dim3 block_dim;
-    unsigned int shared_mem;
-    function_info * entry;
-
-};
-
-class device_launch_operation_t {
-
-public:
-    device_launch_operation_t() {}
-    device_launch_operation_t(kernel_info_t *_grid,
-        CUstream_st * _stream) :
-            grid(_grid), stream(_stream) {}
-
-    kernel_info_t * grid; //a new child grid
-
-    CUstream_st * stream; 
-
-};
-
-
-std::map<void *, device_launch_config_t> g_cuda_device_launch_param_map;
-std::list<device_launch_operation_t> g_cuda_device_launch_op;
 //extern stream_manager *g_stream_manager();
 
 //Handling device runtime api:
@@ -142,7 +105,7 @@ void cuda_device_runtime::gpgpusim_cuda_getParameterBufferV2(const ptx_instructi
 
 //Handling device runtime api:
 //cudaError_t cudaLaunchDeviceV2(void *parameterBuffer, cudaStream_t stream)
-void gpgpusim_cuda_launchDeviceV2(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func) {
+void cuda_device_runtime::gpgpusim_cuda_launchDeviceV2(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func) {
     DEV_RUNTIME_REPORT("Calling cudaLaunchDeviceV2");
 
     unsigned n_return = target_func->has_return();
@@ -263,7 +226,7 @@ void gpgpusim_cuda_launchDeviceV2(const ptx_instruction * pI, ptx_thread_info *
 //Handling device runtime api:
 //cudaError_t cudaStreamCreateWithFlags ( cudaStream_t* pStream, unsigned int  flags)
 //flags can only be cudaStreamNonBlocking
-void gpgpusim_cuda_streamCreateWithFlags(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func) {
+void cuda_device_runtime::gpgpusim_cuda_streamCreateWithFlags(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func) {
     DEV_RUNTIME_REPORT("Calling cudaStreamCreateWithFlags");
 
     unsigned n_return = target_func->has_return();
diff --git a/src/cuda-sim/cuda_device_runtime.h b/src/cuda-sim/cuda_device_runtime.h
index dd908ae..b49d0eb 100644
--- a/src/cuda-sim/cuda_device_runtime.h
+++ b/src/cuda-sim/cuda_device_runtime.h
@@ -3,10 +3,40 @@
 #if (CUDART_VERSION >= 5000)
 #pragma once
 
-void gpgpusim_cuda_launchDeviceV2(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func);
-void gpgpusim_cuda_streamCreateWithFlags(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func);
-#endif
-#if (CUDART_VERSION >= 5000)
+class device_launch_config_t {
+
+public:
+    device_launch_config_t() {}
+
+    device_launch_config_t(dim3 _grid_dim,
+        dim3 _block_dim,
+        unsigned int _shared_mem,
+        function_info * _entry):
+            grid_dim(_grid_dim),
+            block_dim(_block_dim),
+            shared_mem(_shared_mem),
+            entry(_entry) {}
+
+    dim3 grid_dim;
+    dim3 block_dim;
+    unsigned int shared_mem;
+    function_info * entry;
+
+};
+
+class device_launch_operation_t {
+
+public:
+    device_launch_operation_t() {}
+    device_launch_operation_t(kernel_info_t *_grid,
+        CUstream_st * _stream) :
+            grid(_grid), stream(_stream) {}
+
+    kernel_info_t * grid; //a new child grid
+
+    CUstream_st * stream;
+
+};
 
 class gpgpu_context;
 
@@ -17,8 +47,12 @@ class cuda_device_runtime {
 	    gpgpu_ctx = ctx;
 	}
 	unsigned long long g_total_param_size;
+	std::map<void *, device_launch_config_t> g_cuda_device_launch_param_map;
+	std::list<device_launch_operation_t> g_cuda_device_launch_op;
 	// backward pointer
 	class gpgpu_context* gpgpu_ctx;
+	void gpgpusim_cuda_launchDeviceV2(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func);
+	void gpgpusim_cuda_streamCreateWithFlags(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func);
 	void gpgpusim_cuda_getParameterBufferV2(const ptx_instruction * pI, ptx_thread_info * thread, const function_info * target_func);
 	void launch_all_device_kernels();
 	void launch_one_device_kernel();
diff --git a/src/cuda-sim/instructions.cc b/src/cuda-sim/instructions.cc
index dfe5af6..565340c 100644
--- a/src/cuda-sim/instructions.cc
+++ b/src/cuda-sim/instructions.cc
@@ -2101,11 +2101,11 @@ void call_impl( const ptx_instruction *pI, ptx_thread_info *thread )
 	  return;
    }
    else if(fname == "cudaLaunchDeviceV2") {
-      gpgpusim_cuda_launchDeviceV2(pI, thread, target_func);
+      target_func->gpgpu_ctx->device_runtime->gpgpusim_cuda_launchDeviceV2(pI, thread, target_func);
 	  return;
    }
    else if(fname == "cudaStreamCreateWithFlags") {
-      gpgpusim_cuda_streamCreateWithFlags(pI, thread, target_func);
+      target_func->gpgpu_ctx->device_runtime->gpgpusim_cuda_streamCreateWithFlags(pI, thread, target_func);
 	  return;
    }
 #endif