leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 29 Packages Projects Releases Wiki Activity

2.9 fixes for nvrtc (#480)

Browse Source 
* Use platform::is_same instead of std::is_same

* Don't hide cuComplex include from nvrtc

* Typo fixed

* Remove comment rename
This commit is contained in:
Stepan Tezyunichev
2022-04-29 09:06:52 -04:00
committed by GitHub
parent 21c1fa3849
commit 86ce09aed1
13 changed files with 36 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cutlass/conv/kernel/implicit_gemm_convolution.h
+1 -1
View File
@@ -121,7 +121,7 @@ struct ImplicitGemmConvolution {
// Conv2d row-major matrix C (KxRSC)
// Conv3d row-major matrix C (KxTRSC)
static int const kWgradCStrideIdx =
cutlass::platform::is_same<LayoutC, cutlass::layout::TensorNHWC>::value ? 2 : 3;
platform::is_same<LayoutC, cutlass::layout::TensorNHWC>::value ? 2 : 3;
/// This chooses the appropriate stride element of the C tensor.
static int const kTensorCStrideIdx =
include/cutlass/conv/kernel/implicit_gemm_convolution_fusion.h
+1 -1
View File
@@ -123,7 +123,7 @@ struct ImplicitGemmConvolutionFusion {
// Conv2d row-major matrix C (KxRSC)
// Conv3d row-major matrix C (KxTRSC)
static int const kWgradCStrideIdx =
cutlass::platform::is_same<LayoutC, cutlass::layout::TensorNHWC>::value ? 2 : 3;
platform::is_same<LayoutC, cutlass::layout::TensorNHWC>::value ? 2 : 3;
/// This chooses the appropriate stride element of the C tensor.
static int const kTensorCStrideIdx =
include/cutlass/conv/kernel/implicit_gemm_convolution_strided_dgrad.h
+5 -5
View File
@@ -121,20 +121,20 @@ struct ImplicitGemmConvolutionStridedDgrad {
// Conv2d row-major matrix C (KxRSC)
// Conv3d row-major matrix C (KxTRSC)
static int const kWgradCStrideIdx =
cutlass::platform::is_same<LayoutC, cutlass::layout::TensorNHWC>::value ? 2 : 3;
platform::is_same<LayoutC, cutlass::layout::TensorNHWC>::value ? 2 : 3;
/// This chooses the appropriate stride element of the C tensor.
static int const kTensorCStrideIdx =
(kConvolutionalOperator == conv::Operator::kWgrad ? kWgradCStrideIdx : 0);
// Strided dgrad uses a specialized threadblock swizzle for functionality and performance
static_assert((std::is_same<ThreadblockSwizzle,
static_assert((platform::is_same<ThreadblockSwizzle,
threadblock::StridedDgradHorizontalThreadblockSwizzle>::value) ||
(std::is_same<ThreadblockSwizzle,
(platform::is_same<ThreadblockSwizzle,
threadblock::StridedDgradIdentityThreadblockSwizzle<1>>::value) ||
(std::is_same<ThreadblockSwizzle,
(platform::is_same<ThreadblockSwizzle,
threadblock::StridedDgradIdentityThreadblockSwizzle<4>>::value) ||
(std::is_same<ThreadblockSwizzle,
(platform::is_same<ThreadblockSwizzle,
threadblock::StridedDgradIdentityThreadblockSwizzle<8>>::value),
"Needs ThreadblockSwizzle type specialized for strided dgrad");
include/cutlass/conv/kernel/implicit_gemm_convolution_with_fused_epilogue.h
+1 -1
View File
@@ -121,7 +121,7 @@ struct ImplicitGemmConvolutionWithFusedEpilogue {
// Conv2d row-major matrix C (KxRSC)
// Conv3d row-major matrix C (KxTRSC)
static int const kWgradCStrideIdx =
cutlass::platform::is_same<LayoutC, cutlass::layout::TensorNHWC>::value ? 2 : 3;
platform::is_same<LayoutC, cutlass::layout::TensorNHWC>::value ? 2 : 3;
/// This chooses the appropriate stride element of the C tensor.
static int const kTensorCStrideIdx =