leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 26 Packages Projects Releases Wiki Activity

CUTLASS 2.8 (#363)

Browse Source 
CUTLASS 2.8
This commit is contained in:
Manish Gupta
2021-11-19 13:26:35 -08:00
committed by GitHub
parent 6fc5008803
commit 808c25337a
127 changed files with 18555 additions and 1338 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/unit/gemm/device/testbed_complex.h
+30 -30
View File
@@ -102,33 +102,33 @@ struct TestbedComplex : public Testbed<Gemm> {
}
/// Returns true if the CUDA device is sufficient to execute the kernel.
bool sufficient() const {
//
// Determine SMEM requirements and waive if not satisfied
//
int smem_size = int(sizeof(typename Gemm::GemmKernel::SharedStorage));
cudaDeviceProp properties;
int device_idx;
cudaError_t result = cudaGetDevice(&device_idx);
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDevice() API call failed.");
}
result = cudaGetDeviceProperties(&properties, device_idx);
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerMultiprocessor < smem_size) {
return false;
}
return true;
}
bool sufficient() const {
//
// Determine SMEM requirements and waive if not satisfied
//
int smem_size = int(sizeof(typename Gemm::GemmKernel::SharedStorage));
cudaDeviceProp properties;
int device_idx;
cudaError_t result = cudaGetDevice(&device_idx);
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDevice() API call failed.");
}
result = cudaGetDeviceProperties(&properties, device_idx);
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerMultiprocessor < smem_size) {
return false;
}
return true;
}
/// Executes one test
bool run(
@@ -145,9 +145,9 @@ struct TestbedComplex : public Testbed<Gemm> {
return true;
}
//
// Initialize workspace
//
//
// Initialize workspace
//
this->initialize(problem_size);