From 84a27b392659f8d939f910143d1d8ae09ab31c50 Mon Sep 17 00:00:00 2001 From: xiangjiaojun <37390337+JJXiangJiaoJun@users.noreply.github.com> Date: Thu, 31 Jul 2025 10:11:04 +0800 Subject: [PATCH] fix: examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu GridDim miscalculated (#2492) * fix: examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu Launch dimGrid error * feat: add cta tiler * Update examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu use cluster_layout_vmnk instead of cta_tiler Co-authored-by: Junkai-Wu * feat: remove cta_tiler --------- Co-authored-by: qinghongzeng Co-authored-by: Junkai-Wu --- examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) 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 1fdf1edd..e771287e 100644 --- a/examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu +++ b/examples/cute/tutorial/blackwell/04_mma_tma_2sm_sm100.cu @@ -573,8 +573,8 @@ void gemm_host_f16xf16_f32_f32_tnt(TypeA const* device_ptr_A, LayoutA layout_A, dim3 dimBlock(128); dim3 dimCluster(size<0>(cluster_shape), size<1>(cluster_shape), size<2>(cluster_shape)); - dim3 dimGrid(round_up(size(ceil_div(Gemm_M, bM)), dimCluster.x), - round_up(size(ceil_div(Gemm_N, bN)), dimCluster.y)); + dim3 dimGrid(size(ceil_div(Gemm_M, bM * size<1>(cluster_layout_vmnk))) * dimCluster.x, + size(ceil_div(Gemm_N, bN * size<2>(cluster_layout_vmnk))) * dimCluster.y); int smemBytes = sizeof(SMEMStorage); auto* kernel_ptr = &gemm_device