From f118b10cc49bf771755f2afdee04cde0359d745e Mon Sep 17 00:00:00 2001
From: Amruth <amruth@ece.ubc.ca>
Date: Wed, 11 Apr 2018 12:07:34 -0700
Subject: clean up

---
 libcuda/cuda_runtime_api.cc | 36 +++---------------------------------
 1 file changed, 3 insertions(+), 33 deletions(-)

(limited to 'libcuda/cuda_runtime_api.cc')

diff --git a/libcuda/cuda_runtime_api.cc b/libcuda/cuda_runtime_api.cc
index 184325c..abdc345 100644
--- a/libcuda/cuda_runtime_api.cc
+++ b/libcuda/cuda_runtime_api.cc
@@ -349,15 +349,7 @@ class _cuda_device_id *GPGPUSim_Init()
 		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.
-		 * Previously, this was thought as sharedMemPerMultiprocessor and is being used in many places.
-		 * Check whether all the instances of shared_mem_size(), gpgpu_shmem_size or sharedMemPerBlock are meant to use sharedMemPerBlock or sharedMemPerMultiprocessor.
-		*/
-		prop->sharedMemPerBlock = 49152;
-#if (CUDART_VERSION > 5000)
-		prop->sharedMemPerMultiprocessor = the_gpu->shared_mem_size();
-#endif
+		prop->sharedMemPerBlock = the_gpu->shared_mem_size();
 		prop->regsPerBlock = the_gpu->num_registers_per_core();
 		prop->warpSize = the_gpu->wrp_size();
 		prop->clockRate = the_gpu->shader_clock();
@@ -848,17 +840,8 @@ __host__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDevic
         if (device <= dev->num_devices() )  {
                 prop = dev->get_prop();
                 switch (attr) {
-                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];
+                        *value= prop->maxGridSize[1];
                         break;
                 case 6:
                         *value= prop->maxGridSize[1];
@@ -866,12 +849,6 @@ __host__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDevic
                 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;
@@ -884,14 +861,11 @@ __host__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDevic
                 case 16:
                         *value= prop->multiProcessorCount ;
                         break;
-                case 34:
-                        *value= 0;
-                        break;
                 case 39:
                         *value= dev->get_gpgpu()->threads_per_core();
                         break;
                 case 75:
-                        *value= 9 ;
+                        *value= 8 ;
                         break;
                 case 76:
                         *value= 3 ;
@@ -899,9 +873,6 @@ __host__ cudaError_t CUDARTAPI cudaDeviceGetAttribute(int *value, enum cudaDevic
                 case 78:
                         *value= 0 ; //TODO: as of now, we dont support stream priorities.
                         break;
-                case 81:
-                        *value= prop->sharedMemPerMultiprocessor;
-                        break;
 		default:
 			printf("ERROR: implement the attribute numbered %d \n",attr);
 			abort();
@@ -1911,7 +1882,6 @@ void cuobjdumpParseBinary(unsigned int handle){
 		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);
-	//max_capability=context->get_device()->get_gpgpu()->get_config().get_forced_max_capability();
 
 	cuobjdumpPTXSection* ptx = NULL;
 	const char* pre_load = getenv("CUOBJDUMP_SIM_FILE");
-- 
cgit v1.3