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

DEBUG: hardcoded Int32(1) to test if TMA can read tile 1

Browse Source
This commit is contained in:
biondizzle
2026-05-22 20:34:21 +00:00
parent 79ebe20a39
commit 215282971c
1 changed files with 3 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
tests/unit/test_fmha_v3_stage_c.py
View File

@@ -224,18 +224,11 @@ class FmhaV3StageCMulti:
cute.copy(tma_q, tAgQ[(None, Int32(0))], tAsQ[(None, qh.index)], tma_bar_ptr=qh.barrier)
qp.tail()
kvp.reset()
# Try using the pipeline state count (kh.count) as the coordinate.
# This is what the CUTLASS reference's "mode 1" does — the pipeline
# index IS the GMEM tile index for 2-stage pipelines with 2 KV tiles.
# For more tiles, we need a separate counter.
# But first, let's test if the coordinate matters at all by using
# Int32(1) for the second tile when n=256.
# DEBUG: Use constant Int32(1) to test if TMA can read from tile 1 at all
for kt in range(n_kv_tiles):
kvh = kvp.acquire_and_advance()
# Use a hardcoded coord to test if TMA even reads different tiles
coord = Int32(kt) # Should be 0, 1, 2, ... but might be constant-folded
cute.copy(tma_k, tBgK[(None, coord)], tBsK[(None, kvh.index)], tma_bar_ptr=kvh.barrier)
cute.copy(tma_v, tVgV[(None, coord)], tVsV[(None, kvh.index)], tma_bar_ptr=kvh.barrier)
cute.copy(tma_k, tBgK[(None, Int32(1))], tBsK[(None, kvh.index)], tma_bar_ptr=kvh.barrier)
cute.copy(tma_v, tVgV[(None, Int32(1))], tVsV[(None, kvh.index)], tma_bar_ptr=kvh.barrier)
kvp.tail()
# ===== MMA warp =====