CUTLASS 3.0.0 (#786)

* CUTLASS 3.0.0
This commit is contained in:
Vijay Thakkar
2023-01-23 20:55:28 -05:00
committed by GitHub
parent 66d9cddc83
commit 277bd6e537
377 changed files with 76396 additions and 1186 deletions
@@ -186,6 +186,34 @@ public:
tensor_D_reference.sync_device();
}
bool sufficient() const {
//
// Determine SMEM requirements and waive if not satisfied
//
int smem_size = int(sizeof(typename Conv2d::UnderlyingKernel::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(
cutlass::conv::Conv2dProblemSize const &problem_size,
@@ -193,6 +221,14 @@ public:
ElementCompute alpha = ElementCompute(1),
ElementCompute beta = ElementCompute(0)) {
// Waive test if insufficient CUDA device
if (!sufficient()) {
if (CUTLASS_TEST_UNIT_ENABLE_WARNINGS) {
std::cerr << "Test waived due to insufficient CUDA device." << std::endl;
}
return true;
}
#if 0 //display conv2d problem size for debugging
std::cout << problem_size << std::endl
<< "alpha, beta: (" << float(alpha) << ", " << float(beta) << ")" << std::endl
@@ -328,6 +328,7 @@ TEST(
DepthwiseFpropProblemSizes_filter5x5()));
}
#if 0
////////////////////////////////////////////////////////////////////////////////
TEST(
SM60_Device_Depthwise_conv2d_Fprop_Direct_Conv_Optimized_f16nhwc_f16nhwc_f16nhwc_simt_f16,
@@ -424,3 +425,5 @@ TEST(
EXPECT_TRUE(test::conv::device::TestSpecificDepthwiseDirectConv2d<Direct2dConv>(
DepthwiseFpropProblemSizes_filter5x37()));
}
#endif