diff --git a/examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu b/examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu index 5d1987ec..2d4799f9 100644 --- a/examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu +++ b/examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu @@ -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. diff --git a/examples/cute/tutorial/blackwell/05_mma_tma_epi_sm100.cu b/examples/cute/tutorial/blackwell/05_mma_tma_epi_sm100.cu index ad0abb26..8a058c5c 100644 --- a/examples/cute/tutorial/blackwell/05_mma_tma_epi_sm100.cu +++ b/examples/cute/tutorial/blackwell/05_mma_tma_epi_sm100.cu @@ -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.