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

sgemm_tcore: Fix smem allocation for non-dma

Browse Source
This commit is contained in:
Hansung Kim
2024-06-15 01:28:27 -07:00
parent cfb6ae4a91
commit 85cace9524
1 changed files with 1 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/regression/sgemm_tcore/kernel.cpp
View File

@@ -542,8 +542,7 @@ void kernel_body(int task_id, kernel_arg_t *__UNIFORM__ arg) {
// "static" shared memory allocation. This would determine threadblock
// occupancy of a single cluster
float *sharedmem_per_threadblock =
(float *)DEV_SMEM_START_ADDR + (GEMMINI_DMA ? 2 /*double-buffer*/ : 1) *
(2 * BM * BK) *
(float *)DEV_SMEM_START_ADDR + 2/*overkill for non-dma*/ * (2 * BM * BK) *
threadblock_id_in_cluster;
thread_block_gemm(arg, tid_in_threadblock, threads_per_threadblock,