ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-29 11:16:59 +00:00
Code Issues Packages Projects Releases Wiki Activity

rm hard code inst cnt

Browse Source
This commit is contained in:
coderfeli
2025-02-21 09:00:16 +00:00
parent 751ddf8648
commit 97ffbdb040
1 changed files with 3 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
include/ck/tensor_operation/gpu/block/blockwise_gemm_pipeline_xdlops_b_preshuffle_v1.hpp
View File

@@ -141,6 +141,7 @@ struct BlockwiseGemmXdlops_pipeline_bpreshuffle_v1<BlockGemmPipelineScheduler::I
using Base::AMmaKStride;
using Base::BMmaKStride;
using Base::MWaves;
static constexpr index_t PrefetchStages = 2;
static constexpr index_t PrefillStages = 1;
@@ -182,9 +183,9 @@ struct BlockwiseGemmXdlops_pipeline_bpreshuffle_v1<BlockGemmPipelineScheduler::I
__device__ static constexpr auto HotLoopScheduler()
{
constexpr auto num_ds_read_inst_a = HotLoopInstList::A_LDS_Read_Inst_Num * 2;
constexpr auto num_ds_read_inst_a = HotLoopInstList::A_LDS_Read_Inst_Num;
constexpr auto num_buffer_load_inst_a = HotLoopInstList::A_Buffer_Load_Inst_Num;
constexpr auto num_buffer_load_inst_b = HotLoopInstList::B_Buffer_Load_Inst_Num * 2;
constexpr auto num_buffer_load_inst_b = HotLoopInstList::B_Buffer_Load_Inst_Num * MWaves;
// B global
static_for<0, num_buffer_load_inst_b, 1>{}([&](auto i) {