ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-04 13:41:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK-Tile] Add the API to load SGPR (#2878)

Browse Source 
* Have a workable version for SGPR

* have a workable version for atomic add

* Revert "have a workable version for atomic add"

This reverts commit 792377a590c26cfff9c8f545d9a9e8484a7422eb.

* substitute with the new sgpr read api

* update the CHANGELOG

* have a workable version for atomic add

* Revert "have a workable version for atomic add"

This reverts commit 792377a590c26cfff9c8f545d9a9e8484a7422eb.

* change to static for logic

* have a workable version for atomic add

* Revert "have a workable version for atomic add"

This reverts commit 792377a590c26cfff9c8f545d9a9e8484a7422eb.
This commit is contained in:
Thomas Ning
2025-09-23 01:23:56 -07:00
committed by GitHub
parent b6e8994386
commit 2cbbf5dcb3
40 changed files with 273 additions and 167 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
include/ck_tile/ops/fused_moe/pipeline/fused_moegemm_pipeline_flatmm_uk.hpp
View File

@@ -184,17 +184,17 @@ struct FusedMoeGemmPipeline_FlatmmUk
index_t nr_1 = kargs.hidden_size / BlockShape::Warp_N1;
index_t kr_1 = shared_intermediate_size_1 / BlockShape::Warp_K1;
const IndexDataType expert_id = __builtin_amdgcn_readfirstlane(
const IndexDataType expert_id = amd_wave_read_first_lane(
reinterpret_cast<const IndexDataType*>(kargs.sorted_expert_ids_ptr)[sorted_tile_id]);
index_t expert_stride_0 = shared_intermediate_size_0 * kargs.hidden_size;
index_t expert_stride_1 = shared_intermediate_size_1 * kargs.hidden_size;
// nr*kr*w
index_t interm_idx_nr0 = __builtin_amdgcn_readfirstlane(
index_t interm_idx_nr0 = amd_wave_read_first_lane(
intermediate_tile_id *
BlockShape::Block_Nr0); // intermediate_tile_id * Block_N / (N in W)
index_t interm_idx_kr1 = __builtin_amdgcn_readfirstlane(
index_t interm_idx_kr1 = amd_wave_read_first_lane(
intermediate_tile_id *
BlockShape::Block_Kr1); // intermediate_tile_id * Block_N / (N in W)