NVIDIA/cutlass
1
0
Fork 0
mirror of https://github.com/NVIDIA/cutlass.git synced 2026-04-19 22:38:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix typo (#3012)

Browse Source
This commit is contained in:
Tianqi Zhang (张天启)
2026-02-27 16:25:35 +08:00
committed by GitHub
parent 518327d631
commit c651d660d2
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu
View File

@@ -461,7 +461,7 @@ void gemm_host_f16xf16_f32_f32_tnt(TypeA const* device_ptr_A, LayoutA layout_A,
// Shape_MNK: (_256,_256,_16) // MmaM, MmaN, MmaK (MmaK is constant for each instr.)
// LayoutA_TV: (_2,(_128,_16)):(_128,(_1,_256)) // TV -> MmaCoordinate mapping for A matrix
// LayoutB_TV: (_2,(_128,_16)):(_128,(_1,_256)) // TV -> MmaCoordinate mapping for B matrix
// LayoutC_TV: (_2,(_128,_256)):(_128,(_1,_256)) // TV -> MmaCoordinate mapping for B matrix
// LayoutC_TV: (_2,(_128,_256)):(_128,(_1,_256)) // TV -> MmaCoordinate mapping for C matrix
// Define MMA tiler sizes (static)
auto bM = tile_size<0>(tiled_mma); // MMA Tile M. We'll use 1 MMAs per MMA Tile M.

2
examples/cute/tutorial/blackwell/05_mma_tma_epi_sm100.cu
View File

@@ -515,7 +515,7 @@ void gemm_host_f16xf16_f32_f32_tnt(TypeA const* device_ptr_A, LayoutA layout_A,
// Shape_MNK: (_256,_256,_16) // MmaM, MmaN, MmaK (MmaK is constant for each instr.)
// LayoutA_TV: (_2,(_128,_16)):(_128,(_1,_256)) // TV -> MmaCoordinate mapping for A matrix
// LayoutB_TV: (_2,(_128,_16)):(_128,(_1,_256)) // TV -> MmaCoordinate mapping for B matrix
// LayoutC_TV: (_2,(_128,_256)):(_128,(_1,_256)) // TV -> MmaCoordinate mapping for B matrix
// LayoutC_TV: (_2,(_128,_256)):(_128,(_1,_256)) // TV -> MmaCoordinate mapping for C matrix
// Define MMA tiler sizes (static)
auto bM = tile_size<0>(tiled_mma); // MMA Tile M. We'll use 1 MMAs per MMA Tile M.