leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 25 Packages Projects Releases Wiki Activity

Fix example bug (#2351)

Browse Source
This commit is contained in:
botbw
2025-07-30 22:12:33 -04:00
committed by GitHub
parent 26b7450023
commit da47886e34
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/cute/tutorial/sgemm_sm80.cu
+2 -1
View File
@@ -232,6 +232,7 @@ gemm_device(ProblemShape shape_MNK, CtaTiler cta_tiler,
// Size of the register pipeline // Size of the register pipeline
auto K_BLOCK_MAX = size<2>(tCrA); auto K_BLOCK_MAX = size<2>(tCrA);
CUTE_STATIC_ASSERT_V(K_BLOCK_MAX == size<2>(tXrA));
// PREFETCH register pipeline // PREFETCH register pipeline
if (K_BLOCK_MAX > 1) { if (K_BLOCK_MAX > 1) {
@@ -371,7 +372,7 @@ gemm_tn(int m, int n, int k,
Layout<Shape<_16,_8>,Stride<_8,_1>>{}, // Thr layout 16x8 k-major Layout<Shape<_16,_8>,Stride<_8,_1>>{}, // Thr layout 16x8 k-major
Layout<Shape< _1,_8>>{}); // Val layout 1x8 n-major Layout<Shape< _1,_8>>{}); // Val layout 1x8 n-major
TiledMMA mmaC = make_tiled_mma(SM80_16x8x8_F16F16F16F16_TN{}, TiledMMA mmaC = make_tiled_mma(SM80_16x8x16_F16F16F16F16_TN{},
Layout<Shape<_2,_2>>{}, // 2x2x1 MMA Atoms Layout<Shape<_2,_2>>{}, // 2x2x1 MMA Atoms
Tile<_32,_32,_16>{}); // 32x32x16 Tiled MMA for LDSM Tile<_32,_32,_16>{}); // 32x32x16 Tiled MMA for LDSM