biondizzle/nvfp4-megamoe-kernel
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

debug: test chunk 1 only

Browse Source
This commit is contained in:
biondizzle
2026-05-30 07:00:14 +00:00
parent 72779e7f71
commit 3dbd3c5e7f
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
dsv4/kernels/attention/fmha_6warp_tma_multirow_multitile.cuh
View File

@@ -131,7 +131,9 @@ fmha_6warp_tma_multirow_multitile_kernel(FmhaTmaMultiRowMultiTileParams params)
int phase = 0; // reset mbarrier phase per hd_chunk
// DEBUG: skip hd_chunk > 0 to test just chunk 0
if (hd_chunk > 0) {
// DEBUG: skip hd_chunk 0 to test just chunk 1
bool skip_chunk = (hd_chunk == 0); // test chunk 1 only
if (skip_chunk) {
if (my_row_active) {
for (int d = 0; d < HD_CHUNK; d++)
o_head[my_row * HD + hd_chunk_start + d] = f32_to_bf16(0.0f);