mirror of
https://github.com/NVIDIA/cuda-samples.git
synced 2026-10-11 23:38:25 +08:00
Add and Update samples for CUDA 10.0
This commit is contained in:
Mahesh Doijade
@@ -51,457 +51,17 @@
|
||||
// CUDA Runtime error messages
|
||||
#ifdef __DRIVER_TYPES_H__
|
||||
static const char *_cudaGetErrorEnum(cudaError_t error) {
|
||||
switch (error) {
|
||||
case cudaSuccess:
|
||||
return "cudaSuccess";
|
||||
|
||||
case cudaErrorMissingConfiguration:
|
||||
return "cudaErrorMissingConfiguration";
|
||||
|
||||
case cudaErrorMemoryAllocation:
|
||||
return "cudaErrorMemoryAllocation";
|
||||
|
||||
case cudaErrorInitializationError:
|
||||
return "cudaErrorInitializationError";
|
||||
|
||||
case cudaErrorLaunchFailure:
|
||||
return "cudaErrorLaunchFailure";
|
||||
|
||||
case cudaErrorPriorLaunchFailure:
|
||||
return "cudaErrorPriorLaunchFailure";
|
||||
|
||||
case cudaErrorLaunchTimeout:
|
||||
return "cudaErrorLaunchTimeout";
|
||||
|
||||
case cudaErrorLaunchOutOfResources:
|
||||
return "cudaErrorLaunchOutOfResources";
|
||||
|
||||
case cudaErrorInvalidDeviceFunction:
|
||||
return "cudaErrorInvalidDeviceFunction";
|
||||
|
||||
case cudaErrorInvalidConfiguration:
|
||||
return "cudaErrorInvalidConfiguration";
|
||||
|
||||
case cudaErrorInvalidDevice:
|
||||
return "cudaErrorInvalidDevice";
|
||||
|
||||
case cudaErrorInvalidValue:
|
||||
return "cudaErrorInvalidValue";
|
||||
|
||||
case cudaErrorInvalidPitchValue:
|
||||
return "cudaErrorInvalidPitchValue";
|
||||
|
||||
case cudaErrorInvalidSymbol:
|
||||
return "cudaErrorInvalidSymbol";
|
||||
|
||||
case cudaErrorMapBufferObjectFailed:
|
||||
return "cudaErrorMapBufferObjectFailed";
|
||||
|
||||
case cudaErrorUnmapBufferObjectFailed:
|
||||
return "cudaErrorUnmapBufferObjectFailed";
|
||||
|
||||
case cudaErrorInvalidHostPointer:
|
||||
return "cudaErrorInvalidHostPointer";
|
||||
|
||||
case cudaErrorInvalidDevicePointer:
|
||||
return "cudaErrorInvalidDevicePointer";
|
||||
|
||||
case cudaErrorInvalidTexture:
|
||||
return "cudaErrorInvalidTexture";
|
||||
|
||||
case cudaErrorInvalidTextureBinding:
|
||||
return "cudaErrorInvalidTextureBinding";
|
||||
|
||||
case cudaErrorInvalidChannelDescriptor:
|
||||
return "cudaErrorInvalidChannelDescriptor";
|
||||
|
||||
case cudaErrorInvalidMemcpyDirection:
|
||||
return "cudaErrorInvalidMemcpyDirection";
|
||||
|
||||
case cudaErrorAddressOfConstant:
|
||||
return "cudaErrorAddressOfConstant";
|
||||
|
||||
case cudaErrorTextureFetchFailed:
|
||||
return "cudaErrorTextureFetchFailed";
|
||||
|
||||
case cudaErrorTextureNotBound:
|
||||
return "cudaErrorTextureNotBound";
|
||||
|
||||
case cudaErrorSynchronizationError:
|
||||
return "cudaErrorSynchronizationError";
|
||||
|
||||
case cudaErrorInvalidFilterSetting:
|
||||
return "cudaErrorInvalidFilterSetting";
|
||||
|
||||
case cudaErrorInvalidNormSetting:
|
||||
return "cudaErrorInvalidNormSetting";
|
||||
|
||||
case cudaErrorMixedDeviceExecution:
|
||||
return "cudaErrorMixedDeviceExecution";
|
||||
|
||||
case cudaErrorCudartUnloading:
|
||||
return "cudaErrorCudartUnloading";
|
||||
|
||||
case cudaErrorUnknown:
|
||||
return "cudaErrorUnknown";
|
||||
|
||||
case cudaErrorNotYetImplemented:
|
||||
return "cudaErrorNotYetImplemented";
|
||||
|
||||
case cudaErrorMemoryValueTooLarge:
|
||||
return "cudaErrorMemoryValueTooLarge";
|
||||
|
||||
case cudaErrorInvalidResourceHandle:
|
||||
return "cudaErrorInvalidResourceHandle";
|
||||
|
||||
case cudaErrorNotReady:
|
||||
return "cudaErrorNotReady";
|
||||
|
||||
case cudaErrorInsufficientDriver:
|
||||
return "cudaErrorInsufficientDriver";
|
||||
|
||||
case cudaErrorSetOnActiveProcess:
|
||||
return "cudaErrorSetOnActiveProcess";
|
||||
|
||||
case cudaErrorInvalidSurface:
|
||||
return "cudaErrorInvalidSurface";
|
||||
|
||||
case cudaErrorNoDevice:
|
||||
return "cudaErrorNoDevice";
|
||||
|
||||
case cudaErrorECCUncorrectable:
|
||||
return "cudaErrorECCUncorrectable";
|
||||
|
||||
case cudaErrorSharedObjectSymbolNotFound:
|
||||
return "cudaErrorSharedObjectSymbolNotFound";
|
||||
|
||||
case cudaErrorSharedObjectInitFailed:
|
||||
return "cudaErrorSharedObjectInitFailed";
|
||||
|
||||
case cudaErrorUnsupportedLimit:
|
||||
return "cudaErrorUnsupportedLimit";
|
||||
|
||||
case cudaErrorDuplicateVariableName:
|
||||
return "cudaErrorDuplicateVariableName";
|
||||
|
||||
case cudaErrorDuplicateTextureName:
|
||||
return "cudaErrorDuplicateTextureName";
|
||||
|
||||
case cudaErrorDuplicateSurfaceName:
|
||||
return "cudaErrorDuplicateSurfaceName";
|
||||
|
||||
case cudaErrorDevicesUnavailable:
|
||||
return "cudaErrorDevicesUnavailable";
|
||||
|
||||
case cudaErrorInvalidKernelImage:
|
||||
return "cudaErrorInvalidKernelImage";
|
||||
|
||||
case cudaErrorNoKernelImageForDevice:
|
||||
return "cudaErrorNoKernelImageForDevice";
|
||||
|
||||
case cudaErrorIncompatibleDriverContext:
|
||||
return "cudaErrorIncompatibleDriverContext";
|
||||
|
||||
case cudaErrorPeerAccessAlreadyEnabled:
|
||||
return "cudaErrorPeerAccessAlreadyEnabled";
|
||||
|
||||
case cudaErrorPeerAccessNotEnabled:
|
||||
return "cudaErrorPeerAccessNotEnabled";
|
||||
|
||||
case cudaErrorDeviceAlreadyInUse:
|
||||
return "cudaErrorDeviceAlreadyInUse";
|
||||
|
||||
case cudaErrorProfilerDisabled:
|
||||
return "cudaErrorProfilerDisabled";
|
||||
|
||||
case cudaErrorProfilerNotInitialized:
|
||||
return "cudaErrorProfilerNotInitialized";
|
||||
|
||||
case cudaErrorProfilerAlreadyStarted:
|
||||
return "cudaErrorProfilerAlreadyStarted";
|
||||
|
||||
case cudaErrorProfilerAlreadyStopped:
|
||||
return "cudaErrorProfilerAlreadyStopped";
|
||||
|
||||
/* Since CUDA 4.0*/
|
||||
case cudaErrorAssert:
|
||||
return "cudaErrorAssert";
|
||||
|
||||
case cudaErrorTooManyPeers:
|
||||
return "cudaErrorTooManyPeers";
|
||||
|
||||
case cudaErrorHostMemoryAlreadyRegistered:
|
||||
return "cudaErrorHostMemoryAlreadyRegistered";
|
||||
|
||||
case cudaErrorHostMemoryNotRegistered:
|
||||
return "cudaErrorHostMemoryNotRegistered";
|
||||
|
||||
/* Since CUDA 5.0 */
|
||||
case cudaErrorOperatingSystem:
|
||||
return "cudaErrorOperatingSystem";
|
||||
|
||||
case cudaErrorPeerAccessUnsupported:
|
||||
return "cudaErrorPeerAccessUnsupported";
|
||||
|
||||
case cudaErrorLaunchMaxDepthExceeded:
|
||||
return "cudaErrorLaunchMaxDepthExceeded";
|
||||
|
||||
case cudaErrorLaunchFileScopedTex:
|
||||
return "cudaErrorLaunchFileScopedTex";
|
||||
|
||||
case cudaErrorLaunchFileScopedSurf:
|
||||
return "cudaErrorLaunchFileScopedSurf";
|
||||
|
||||
case cudaErrorSyncDepthExceeded:
|
||||
return "cudaErrorSyncDepthExceeded";
|
||||
|
||||
case cudaErrorLaunchPendingCountExceeded:
|
||||
return "cudaErrorLaunchPendingCountExceeded";
|
||||
|
||||
case cudaErrorNotPermitted:
|
||||
return "cudaErrorNotPermitted";
|
||||
|
||||
case cudaErrorNotSupported:
|
||||
return "cudaErrorNotSupported";
|
||||
|
||||
/* Since CUDA 6.0 */
|
||||
case cudaErrorHardwareStackError:
|
||||
return "cudaErrorHardwareStackError";
|
||||
|
||||
case cudaErrorIllegalInstruction:
|
||||
return "cudaErrorIllegalInstruction";
|
||||
|
||||
case cudaErrorMisalignedAddress:
|
||||
return "cudaErrorMisalignedAddress";
|
||||
|
||||
case cudaErrorInvalidAddressSpace:
|
||||
return "cudaErrorInvalidAddressSpace";
|
||||
|
||||
case cudaErrorInvalidPc:
|
||||
return "cudaErrorInvalidPc";
|
||||
|
||||
case cudaErrorIllegalAddress:
|
||||
return "cudaErrorIllegalAddress";
|
||||
|
||||
/* Since CUDA 6.5*/
|
||||
case cudaErrorInvalidPtx:
|
||||
return "cudaErrorInvalidPtx";
|
||||
|
||||
case cudaErrorInvalidGraphicsContext:
|
||||
return "cudaErrorInvalidGraphicsContext";
|
||||
|
||||
case cudaErrorStartupFailure:
|
||||
return "cudaErrorStartupFailure";
|
||||
|
||||
case cudaErrorApiFailureBase:
|
||||
return "cudaErrorApiFailureBase";
|
||||
|
||||
/* Since CUDA 8.0*/
|
||||
case cudaErrorNvlinkUncorrectable:
|
||||
return "cudaErrorNvlinkUncorrectable";
|
||||
|
||||
/* Since CUDA 8.5*/
|
||||
case cudaErrorJitCompilerNotFound:
|
||||
return "cudaErrorJitCompilerNotFound";
|
||||
|
||||
/* Since CUDA 9.0*/
|
||||
case cudaErrorCooperativeLaunchTooLarge:
|
||||
return "cudaErrorCooperativeLaunchTooLarge";
|
||||
}
|
||||
|
||||
return "<unknown>";
|
||||
return cudaGetErrorName(error);
|
||||
}
|
||||
#endif
|
||||
|
||||
#ifdef __cuda_cuda_h__
|
||||
#ifdef CUDA_DRIVER_API
|
||||
// CUDA Driver API errors
|
||||
static const char *_cudaGetErrorEnum(CUresult error) {
|
||||
switch (error) {
|
||||
case CUDA_SUCCESS:
|
||||
return "CUDA_SUCCESS";
|
||||
|
||||
case CUDA_ERROR_INVALID_VALUE:
|
||||
return "CUDA_ERROR_INVALID_VALUE";
|
||||
|
||||
case CUDA_ERROR_OUT_OF_MEMORY:
|
||||
return "CUDA_ERROR_OUT_OF_MEMORY";
|
||||
|
||||
case CUDA_ERROR_NOT_INITIALIZED:
|
||||
return "CUDA_ERROR_NOT_INITIALIZED";
|
||||
|
||||
case CUDA_ERROR_DEINITIALIZED:
|
||||
return "CUDA_ERROR_DEINITIALIZED";
|
||||
|
||||
case CUDA_ERROR_PROFILER_DISABLED:
|
||||
return "CUDA_ERROR_PROFILER_DISABLED";
|
||||
|
||||
case CUDA_ERROR_PROFILER_NOT_INITIALIZED:
|
||||
return "CUDA_ERROR_PROFILER_NOT_INITIALIZED";
|
||||
|
||||
case CUDA_ERROR_PROFILER_ALREADY_STARTED:
|
||||
return "CUDA_ERROR_PROFILER_ALREADY_STARTED";
|
||||
|
||||
case CUDA_ERROR_PROFILER_ALREADY_STOPPED:
|
||||
return "CUDA_ERROR_PROFILER_ALREADY_STOPPED";
|
||||
|
||||
case CUDA_ERROR_NO_DEVICE:
|
||||
return "CUDA_ERROR_NO_DEVICE";
|
||||
|
||||
case CUDA_ERROR_INVALID_DEVICE:
|
||||
return "CUDA_ERROR_INVALID_DEVICE";
|
||||
|
||||
case CUDA_ERROR_INVALID_IMAGE:
|
||||
return "CUDA_ERROR_INVALID_IMAGE";
|
||||
|
||||
case CUDA_ERROR_INVALID_CONTEXT:
|
||||
return "CUDA_ERROR_INVALID_CONTEXT";
|
||||
|
||||
case CUDA_ERROR_CONTEXT_ALREADY_CURRENT:
|
||||
return "CUDA_ERROR_CONTEXT_ALREADY_CURRENT";
|
||||
|
||||
case CUDA_ERROR_MAP_FAILED:
|
||||
return "CUDA_ERROR_MAP_FAILED";
|
||||
|
||||
case CUDA_ERROR_UNMAP_FAILED:
|
||||
return "CUDA_ERROR_UNMAP_FAILED";
|
||||
|
||||
case CUDA_ERROR_ARRAY_IS_MAPPED:
|
||||
return "CUDA_ERROR_ARRAY_IS_MAPPED";
|
||||
|
||||
case CUDA_ERROR_ALREADY_MAPPED:
|
||||
return "CUDA_ERROR_ALREADY_MAPPED";
|
||||
|
||||
case CUDA_ERROR_NO_BINARY_FOR_GPU:
|
||||
return "CUDA_ERROR_NO_BINARY_FOR_GPU";
|
||||
|
||||
case CUDA_ERROR_ALREADY_ACQUIRED:
|
||||
return "CUDA_ERROR_ALREADY_ACQUIRED";
|
||||
|
||||
case CUDA_ERROR_NOT_MAPPED:
|
||||
return "CUDA_ERROR_NOT_MAPPED";
|
||||
|
||||
case CUDA_ERROR_NOT_MAPPED_AS_ARRAY:
|
||||
return "CUDA_ERROR_NOT_MAPPED_AS_ARRAY";
|
||||
|
||||
case CUDA_ERROR_NOT_MAPPED_AS_POINTER:
|
||||
return "CUDA_ERROR_NOT_MAPPED_AS_POINTER";
|
||||
|
||||
case CUDA_ERROR_ECC_UNCORRECTABLE:
|
||||
return "CUDA_ERROR_ECC_UNCORRECTABLE";
|
||||
|
||||
case CUDA_ERROR_UNSUPPORTED_LIMIT:
|
||||
return "CUDA_ERROR_UNSUPPORTED_LIMIT";
|
||||
|
||||
case CUDA_ERROR_CONTEXT_ALREADY_IN_USE:
|
||||
return "CUDA_ERROR_CONTEXT_ALREADY_IN_USE";
|
||||
|
||||
case CUDA_ERROR_PEER_ACCESS_UNSUPPORTED:
|
||||
return "CUDA_ERROR_PEER_ACCESS_UNSUPPORTED";
|
||||
|
||||
case CUDA_ERROR_INVALID_PTX:
|
||||
return "CUDA_ERROR_INVALID_PTX";
|
||||
|
||||
case CUDA_ERROR_INVALID_GRAPHICS_CONTEXT:
|
||||
return "CUDA_ERROR_INVALID_GRAPHICS_CONTEXT";
|
||||
|
||||
case CUDA_ERROR_NVLINK_UNCORRECTABLE:
|
||||
return "CUDA_ERROR_NVLINK_UNCORRECTABLE";
|
||||
|
||||
case CUDA_ERROR_JIT_COMPILER_NOT_FOUND:
|
||||
return "CUDA_ERROR_JIT_COMPILER_NOT_FOUND";
|
||||
|
||||
case CUDA_ERROR_INVALID_SOURCE:
|
||||
return "CUDA_ERROR_INVALID_SOURCE";
|
||||
|
||||
case CUDA_ERROR_FILE_NOT_FOUND:
|
||||
return "CUDA_ERROR_FILE_NOT_FOUND";
|
||||
|
||||
case CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND:
|
||||
return "CUDA_ERROR_SHARED_OBJECT_SYMBOL_NOT_FOUND";
|
||||
|
||||
case CUDA_ERROR_SHARED_OBJECT_INIT_FAILED:
|
||||
return "CUDA_ERROR_SHARED_OBJECT_INIT_FAILED";
|
||||
|
||||
case CUDA_ERROR_OPERATING_SYSTEM:
|
||||
return "CUDA_ERROR_OPERATING_SYSTEM";
|
||||
|
||||
case CUDA_ERROR_INVALID_HANDLE:
|
||||
return "CUDA_ERROR_INVALID_HANDLE";
|
||||
|
||||
case CUDA_ERROR_NOT_FOUND:
|
||||
return "CUDA_ERROR_NOT_FOUND";
|
||||
|
||||
case CUDA_ERROR_NOT_READY:
|
||||
return "CUDA_ERROR_NOT_READY";
|
||||
|
||||
case CUDA_ERROR_ILLEGAL_ADDRESS:
|
||||
return "CUDA_ERROR_ILLEGAL_ADDRESS";
|
||||
|
||||
case CUDA_ERROR_LAUNCH_FAILED:
|
||||
return "CUDA_ERROR_LAUNCH_FAILED";
|
||||
|
||||
case CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES:
|
||||
return "CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES";
|
||||
|
||||
case CUDA_ERROR_LAUNCH_TIMEOUT:
|
||||
return "CUDA_ERROR_LAUNCH_TIMEOUT";
|
||||
|
||||
case CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING:
|
||||
return "CUDA_ERROR_LAUNCH_INCOMPATIBLE_TEXTURING";
|
||||
|
||||
case CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED:
|
||||
return "CUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED";
|
||||
|
||||
case CUDA_ERROR_PEER_ACCESS_NOT_ENABLED:
|
||||
return "CUDA_ERROR_PEER_ACCESS_NOT_ENABLED";
|
||||
|
||||
case CUDA_ERROR_PRIMARY_CONTEXT_ACTIVE:
|
||||
return "CUDA_ERROR_PRIMARY_CONTEXT_ACTIVE";
|
||||
|
||||
case CUDA_ERROR_CONTEXT_IS_DESTROYED:
|
||||
return "CUDA_ERROR_CONTEXT_IS_DESTROYED";
|
||||
|
||||
case CUDA_ERROR_ASSERT:
|
||||
return "CUDA_ERROR_ASSERT";
|
||||
|
||||
case CUDA_ERROR_TOO_MANY_PEERS:
|
||||
return "CUDA_ERROR_TOO_MANY_PEERS";
|
||||
|
||||
case CUDA_ERROR_HOST_MEMORY_ALREADY_REGISTERED:
|
||||
return "CUDA_ERROR_HOST_MEMORY_ALREADY_REGISTERED";
|
||||
|
||||
case CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED:
|
||||
return "CUDA_ERROR_HOST_MEMORY_NOT_REGISTERED";
|
||||
|
||||
case CUDA_ERROR_HARDWARE_STACK_ERROR:
|
||||
return "CUDA_ERROR_HARDWARE_STACK_ERROR";
|
||||
|
||||
case CUDA_ERROR_ILLEGAL_INSTRUCTION:
|
||||
return "CUDA_ERROR_ILLEGAL_INSTRUCTION";
|
||||
|
||||
case CUDA_ERROR_MISALIGNED_ADDRESS:
|
||||
return "CUDA_ERROR_MISALIGNED_ADDRESS";
|
||||
|
||||
case CUDA_ERROR_INVALID_ADDRESS_SPACE:
|
||||
return "CUDA_ERROR_INVALID_ADDRESS_SPACE";
|
||||
|
||||
case CUDA_ERROR_INVALID_PC:
|
||||
return "CUDA_ERROR_INVALID_PC";
|
||||
|
||||
case CUDA_ERROR_COOPERATIVE_LAUNCH_TOO_LARGE:
|
||||
return "CUDA_ERROR_COOPERATIVE_LAUNCH_TOO_LARGE";
|
||||
|
||||
case CUDA_ERROR_NOT_PERMITTED:
|
||||
return "CUDA_ERROR_NOT_PERMITTED";
|
||||
|
||||
case CUDA_ERROR_NOT_SUPPORTED:
|
||||
return "CUDA_ERROR_NOT_SUPPORTED";
|
||||
|
||||
case CUDA_ERROR_UNKNOWN:
|
||||
return "CUDA_ERROR_UNKNOWN";
|
||||
}
|
||||
|
||||
return "<unknown>";
|
||||
static char unknown[] = "<unknown>";
|
||||
const char *ret = NULL;
|
||||
cuGetErrorName(error, &ret);
|
||||
return ret ? ret : unknown;
|
||||
}
|
||||
#endif
|
||||
|
||||
@@ -1067,18 +627,19 @@ inline int _ConvertSMVer2Cores(int major, int minor) {
|
||||
} sSMtoCores;
|
||||
|
||||
sSMtoCores nGpuArchCoresPerSM[] = {
|
||||
{0x30, 192}, // Kepler Generation (SM 3.0) GK10x class
|
||||
{0x32, 192}, // Kepler Generation (SM 3.2) GK10x class
|
||||
{0x35, 192}, // Kepler Generation (SM 3.5) GK11x class
|
||||
{0x37, 192}, // Kepler Generation (SM 3.7) GK21x class
|
||||
{0x50, 128}, // Maxwell Generation (SM 5.0) GM10x class
|
||||
{0x52, 128}, // Maxwell Generation (SM 5.2) GM20x class
|
||||
{0x53, 128}, // Maxwell Generation (SM 5.3) GM20x class
|
||||
{0x60, 64}, // Pascal Generation (SM 6.0) GP100 class
|
||||
{0x61, 128}, // Pascal Generation (SM 6.1) GP10x class
|
||||
{0x62, 128}, // Pascal Generation (SM 6.2) GP10x class
|
||||
{0x70, 64}, // Volta Generation (SM 7.0) GV100 class
|
||||
{0x72, 64}, // Volta Generation (SM 7.2) GV11b class
|
||||
{0x30, 192},
|
||||
{0x32, 192},
|
||||
{0x35, 192},
|
||||
{0x37, 192},
|
||||
{0x50, 128},
|
||||
{0x52, 128},
|
||||
{0x53, 128},
|
||||
{0x60, 64},
|
||||
{0x61, 128},
|
||||
{0x62, 128},
|
||||
{0x70, 64},
|
||||
{0x72, 64},
|
||||
{0x75, 64},
|
||||
{-1, -1}};
|
||||
|
||||
int index = 0;
|
||||
@@ -1155,7 +716,7 @@ inline int gpuDeviceInit(int devID) {
|
||||
inline int gpuGetMaxGflopsDeviceId() {
|
||||
int current_device = 0, sm_per_multiproc = 0;
|
||||
int max_perf_device = 0;
|
||||
int device_count = 0, best_SM_arch = 0;
|
||||
int device_count = 0;
|
||||
int devices_prohibited = 0;
|
||||
|
||||
uint64_t max_compute_perf = 0;
|
||||
@@ -1169,30 +730,6 @@ inline int gpuGetMaxGflopsDeviceId() {
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
// Find the best major SM Architecture GPU device
|
||||
while (current_device < device_count) {
|
||||
cudaGetDeviceProperties(&deviceProp, current_device);
|
||||
|
||||
// If this GPU is not running on Compute Mode prohibited,
|
||||
// then we can add it to the list
|
||||
if (deviceProp.computeMode != cudaComputeModeProhibited) {
|
||||
if (deviceProp.major > 0 && deviceProp.major < 9999) {
|
||||
best_SM_arch = MAX(best_SM_arch, deviceProp.major);
|
||||
}
|
||||
} else {
|
||||
devices_prohibited++;
|
||||
}
|
||||
|
||||
current_device++;
|
||||
}
|
||||
|
||||
if (devices_prohibited == device_count) {
|
||||
fprintf(stderr,
|
||||
"gpuGetMaxGflopsDeviceId() CUDA error:"
|
||||
" all devices have compute mode prohibited.\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
// Find the best CUDA capable GPU device
|
||||
current_device = 0;
|
||||
|
||||
@@ -1213,23 +750,23 @@ inline int gpuGetMaxGflopsDeviceId() {
|
||||
sm_per_multiproc * deviceProp.clockRate;
|
||||
|
||||
if (compute_perf > max_compute_perf) {
|
||||
// If we find GPU with SM major > 2, search only these
|
||||
if (best_SM_arch > 2) {
|
||||
// If our device==dest_SM_arch, choose this, or else pass
|
||||
if (deviceProp.major == best_SM_arch) {
|
||||
max_compute_perf = compute_perf;
|
||||
max_perf_device = current_device;
|
||||
}
|
||||
} else {
|
||||
max_compute_perf = compute_perf;
|
||||
max_perf_device = current_device;
|
||||
}
|
||||
max_compute_perf = compute_perf;
|
||||
max_perf_device = current_device;
|
||||
}
|
||||
} else {
|
||||
devices_prohibited++;
|
||||
}
|
||||
|
||||
++current_device;
|
||||
}
|
||||
|
||||
if (devices_prohibited == device_count) {
|
||||
fprintf(stderr,
|
||||
"gpuGetMaxGflopsDeviceId() CUDA error:"
|
||||
" all devices have compute mode prohibited.\n");
|
||||
exit(EXIT_FAILURE);
|
||||
}
|
||||
|
||||
return max_perf_device;
|
||||
}
|
||||
|
||||
|
||||
@@ -122,18 +122,19 @@ inline int _ConvertSMVer2CoresDRV(int major, int minor) {
|
||||
} sSMtoCores;
|
||||
|
||||
sSMtoCores nGpuArchCoresPerSM[] = {
|
||||
{0x30, 192}, // Kepler Generation (SM 3.0) GK10x class
|
||||
{0x32, 192}, // Kepler Generation (SM 3.2) GK10x class
|
||||
{0x35, 192}, // Kepler Generation (SM 3.5) GK11x class
|
||||
{0x37, 192}, // Kepler Generation (SM 3.7) GK21x class
|
||||
{0x50, 128}, // Maxwell Generation (SM 5.0) GM10x class
|
||||
{0x52, 128}, // Maxwell Generation (SM 5.2) GM20x class
|
||||
{0x53, 128}, // Maxwell Generation (SM 5.3) GM20x class
|
||||
{0x60, 64}, // Pascal Generation (SM 6.0) GP100 class
|
||||
{0x61, 128}, // Pascal Generation (SM 6.1) GP10x class
|
||||
{0x62, 128}, // Pascal Generation (SM 6.2) GP10x class
|
||||
{0x70, 64}, // Volta Generation (SM 7.0) GV100 class
|
||||
{0x72, 64}, // Volta Generation (SM 7.2) GV11b class
|
||||
{0x30, 192},
|
||||
{0x32, 192},
|
||||
{0x35, 192},
|
||||
{0x37, 192},
|
||||
{0x50, 128},
|
||||
{0x52, 128},
|
||||
{0x53, 128},
|
||||
{0x60, 64},
|
||||
{0x61, 128},
|
||||
{0x62, 128},
|
||||
{0x70, 64},
|
||||
{0x72, 64},
|
||||
{0x75, 64},
|
||||
{-1, -1}};
|
||||
|
||||
int index = 0;
|
||||
|
||||
Reference in New Issue
Block a user