From e38c2e31033dc6880d92eff4977c40f2eb6cff4a Mon Sep 17 00:00:00 2001 From: Chenggang Zhao Date: Fri, 22 Aug 2025 17:32:04 +0800 Subject: [PATCH] Remove comments --- deep_gemm/include/deep_gemm/impls/sm100_fp8_gemm_1d2d.cuh | 2 -- 1 file changed, 2 deletions(-) diff --git a/deep_gemm/include/deep_gemm/impls/sm100_fp8_gemm_1d2d.cuh b/deep_gemm/include/deep_gemm/impls/sm100_fp8_gemm_1d2d.cuh index 455e600..e04db3c 100644 --- a/deep_gemm/include/deep_gemm/impls/sm100_fp8_gemm_1d2d.cuh +++ b/deep_gemm/include/deep_gemm/impls/sm100_fp8_gemm_1d2d.cuh @@ -513,13 +513,11 @@ sm100_fp8_gemm_1d2d_impl(float* sfb, int* grouped_layout, } // Flush all stages in the pipeline to make TMA stores visible to the next kernel - // TODO: do we actually need this? if (epilogue_thread_idx_in_warpgroup == 0) cute::tma_store_wait<0>(); // Deallocate tensor memory by warp 1 // NOTES: warp 0 is waiting TMA store - // TODO: do we need 2 SM allocation? if (epilogue_warp_idx == 1) Allocator().free(0, kNumTmemCols); }