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

fix(permute.h): incorrect comment in Tensor5DPermute20314 (#637)

Browse Source 
* fix(permute.h): incorrect comment in `Tensor5DPermute20314`

* typo in usage in example 39
This commit is contained in:
Wenzhuo Liu 2022-09-22 21:21:13 +08:00 committed by GitHub
parent 97bff52e8c
commit 7a458f00a6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
examples/39_gemm_permute/gemm_permute.cu
View File

@ -224,7 +224,7 @@ struct Options {
<< " permute([0, 2, 1, 3]) to be in shape of [B/D1, M, D1, N].\n\n" << " permute([0, 2, 1, 3]) to be in shape of [B/D1, M, D1, N].\n\n"
<< " 2) This example also profiles the performance of a normal GEMM kernel with output as permuted 5D Tensor." << " 2) This example also profiles the performance of a normal GEMM kernel with output as permuted 5D Tensor."
<< " The GEMM matrix output in shape of [M, N] is reshaped as [M/T1, T1, T2, T3, N/T2/T3] and then permuted" << " The GEMM matrix output in shape of [M, N] is reshaped as [M/T1, T1, T2, T3, N/T2/T3] and then permuted"
<< " with permute([2, 0, 3, 1, 4]) to be in shape of [T2, M/T1, T3, T1, N//T2/T3].\n\n" << " with permute([2, 0, 3, 1, 4]) to be in shape of [T2, M/T1, T3, T1, N/T2/T3].\n\n"
<< " Note: D1, T1, T2, T3 are compile-time constants defined in gemm_permute.cu\n\n" << " Note: D1, T1, T2, T3 are compile-time constants defined in gemm_permute.cu\n\n"
<< "Options:\n\n" << "Options:\n\n"
<< " --help If specified, displays this usage statement.\n\n" << " --help If specified, displays this usage statement.\n\n"

2
include/cutlass/layout/permute.h
View File

@ -254,7 +254,7 @@ public:
}; };
/// Permute layout function for 5-D permuted tensors with output matrix (dimension as [M, N]) reshaped /// Permute layout function for 5-D permuted tensors with output matrix (dimension as [M, N]) reshaped
/// as [M/T1, T1, T2, T3, N/T3]. Then perform permute([2, 0, 3, 1, 4]) on the corresponding output tensor. /// as [M/T1, T1, T2, T3, N/T2/T3]. Then perform permute([2, 0, 3, 1, 4]) on the corresponding output tensor.
template <int T1, int T2, int T3> template <int T1, int T2, int T3>
class Tensor5DPermute20314 { class Tensor5DPermute20314 {
public: public: