wu-arch/kernels
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

sgemm_impl: Mark threadblock_barrier convergent

Browse Source 
Thank you Chris Lattner
This commit is contained in:
Hansung Kim
2024-09-08 22:49:38 -07:00
parent 714b9f501e
commit 1f51f7f9d4
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/regression/sgemm_tcore/sgemm_impl.hpp
View File

@@ -536,7 +536,8 @@ wmma_store(const int thread_in_warp, const int warp_col, const int warp_row,
asm volatile ("wmma_store_finish_%=:" :: ); asm volatile ("wmma_store_finish_%=:" :: );
} }
inline void threadblock_barrier(const uint32_t barrier_id, const uint32_t count) { __attribute__((convergent)) inline void
threadblock_barrier(const uint32_t barrier_id, const uint32_t count) {
vx_fence(); vx_fence();
vx_barrier(barrier_id, count); vx_barrier(barrier_id, count);
} }