Skip to content

GitLab

Commit 70e7069c authored by rocking's avatar rocking
Browse files

Use sgpr for shuffleM_index

parent b58d5a7d
Hide whitespace changes
Inline Side-by-side
Showing with 2 additions and 4 deletions
include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_gemm_multiple_d_welford_first_half_xdl_cshuffle.hpp
View file @ 70e7069c
...@@ -902,9 +902,7 @@ struct GridwiseGemmMultipleDWelfordFirstHalf_xdl_cshuffle ...@@ -902,9 +902,7 @@ struct GridwiseGemmMultipleDWelfordFirstHalf_xdl_cshuffle
static_assert(num_access == sfc_der_global.GetNumOfAccess(), "wrong!"); static_assert(num_access == sfc_der_global.GetNumOfAccess(), "wrong!");
// TODO - SGPR int shuffleM_index = __builtin_amdgcn_readfirstlane(0);
int shuffleM_index = 0;
static_for<0, num_access, 1>{}([&](auto access_id) { static_for<0, num_access, 1>{}([&](auto access_id) {
// make sure it's safe to read from LDS // make sure it's safe to read from LDS
block_sync_lds(); block_sync_lds();
...@@ -984,7 +982,7 @@ struct GridwiseGemmMultipleDWelfordFirstHalf_xdl_cshuffle ...@@ -984,7 +982,7 @@ struct GridwiseGemmMultipleDWelfordFirstHalf_xdl_cshuffle
constexpr int shuffleMInc = constexpr int shuffleMInc =
de_global_step[I1] / de_global_step[I1] /
c_shuffle_block_desc_mblock_mperblock_nblock_nperblock.GetLength(I1); c_shuffle_block_desc_mblock_mperblock_nblock_nperblock.GetLength(I1);
shuffleM_index += shuffleMInc; shuffleM_index = __builtin_amdgcn_readfirstlane(shuffleM_index + shuffleMInc);
} }
}); // copy c, d, e + welford }); // copy c, d, e + welford
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment