biondizzle/nvfp4-megamoe-kernel
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
1,066 Commits 1 Branch 17 Tags
bcfe52df0d0023efdb73acb40ee426bf56bc93b2
Commit Graph

211 Commits

Author SHA1 Message Date
biondizzle
828f81ca8c D1.4: Use cutlass.range(unroll=1) for k_sub loops in both TMA and MMA warps 2026-05-24 17:55:33 +00:00
biondizzle
29a4de9181 D1.4: Guard LSE computation with const_expr(not normalize) - fixes BF16 type mismatch in regression test 2026-05-24 15:11:39 +00:00
biondizzle
239f88af88 D1.4: Switch k_sub from cutlass.range to Python range (unrolled at trace time) 2026-05-24 15:10:28 +00:00
biondizzle
4ec88e035c D1.4: Fix tTMrO placeholder - define only inside const_expr block 2026-05-24 14:23:22 +00:00
biondizzle
301727a8c1 D1.4: Use cutlass.range loop for k_sub (reduce IR), guard O rescale with const_expr(n_kv_tiles>1) 2026-05-24 14:22:45 +00:00
biondizzle
59195c7d17 D1.4: Reduce pv_n_tile to 128 for hd=512 to fit SMEM budget (192KB) 2026-05-24 08:07:32 +00:00
biondizzle
f77feb6e0f D1: Unrolled k_sub path (hardcoded k_sub=0,1) to avoid cutlass.range IR explosion 2026-05-24 07:03:14 +00:00
biondizzle
555d953eff D1: Use cutlass.range for k_sub loops (CuTeDSL immutable handle) 2026-05-24 06:43:30 +00:00
biondizzle
978c789641 D1: Fix kvh scoping - define before loops, consume V via pipeline 2026-05-24 06:42:26 +00:00
biondizzle
c9eff9a60f D1: Fix kvb→kvh typo in PV GEMM 2026-05-24 06:41:25 +00:00
biondizzle
7acd5f10f3 D1: Remove qh.commit() - pipeline handles commit internally 2026-05-24 06:40:10 +00:00
biondizzle
cf6d7b1750 D1: TMA producer uses acquire_and_advance + commit (no wait_and_advance) 2026-05-24 06:38:15 +00:00
biondizzle
24ed5d311f D1: Use same pipeline API as working code (acquire_and_advance) for k_sub path 2026-05-24 06:36:19 +00:00
biondizzle
888af57f7b D1: Add PipelineState for k_sub TMA path 2026-05-24 05:02:17 +00:00
biondizzle
0db580f18a D1: Fix pipeline API for K sub-tile path (producer_acquire/commit) 2026-05-24 04:59:41 +00:00
biondizzle
9afef9ed7d D1: K sub-tile MMA path using pipeline barriers 2026-05-24 04:57:08 +00:00
biondizzle
8c82de7f61 D1: Fix TMA copies in k_sub path (no mbarrier, use cp_async wait) 2026-05-24 04:53:46 +00:00
biondizzle
8f0c9e0dad D1: Add K sub-tile loop for hd=512 (const_expr guarded, hd≤256 path unchanged) 2026-05-24 04:51:51 +00:00
biondizzle
9ed927f86a D1: Debug TMA partition shapes at hd=512 2026-05-24 04:43:12 +00:00
biondizzle
e58ec5ab41 D1: K sub-tiling - qk_mma_tiler K-dim = k_tile=256, SMEM fits at hd=512 2026-05-24 04:41:12 +00:00
biondizzle
1f38d54206 D1: Remove debug prints, clean up 2026-05-24 04:06:26 +00:00
biondizzle
bc97bf67ee D1: const_expr for sP layout selection (CuTeDSL) 2026-05-24 04:05:17 +00:00
biondizzle
95cafebad4 D1: Python if for sP layout (trace-time, not MLIR) 2026-05-24 04:04:27 +00:00
biondizzle
317d3209ac D1: Tiny 4-mode sP placeholder for TMEM-P path 2026-05-24 04:03:28 +00:00
biondizzle
8d565efd7a D1: Conditional sP allocation (saves 64KB SMEM for TMEM-P at hd=256) 2026-05-24 04:02:02 +00:00
biondizzle
65aaec5df0 D1: Fix sP dummy allocation 2026-05-24 04:00:19 +00:00
biondizzle
86064a2ffa D1: Skip sP allocation when use_smem_p=False (saves 64KB at hd=256) 2026-05-24 03:59:27 +00:00
biondizzle
9bab7fe659 D1: Fix syntax (separate kv_stage line) 2026-05-24 03:58:12 +00:00
biondizzle
15eef33728 D1: Reduce kv_stage to 1 at hd>128 to avoid SMEM overflow 2026-05-24 03:55:44 +00:00
biondizzle
e288fea062 D1: FIX qk_mma_tiler K-dim = head_dim (was hardcoded to 64, broke hd>64) 2026-05-24 03:53:19 +00:00
biondizzle
b408fa4718 D1: Print qk_ik in _setup 2026-05-24 03:51:40 +00:00
biondizzle
aa8b145035 D1: Add more debug prints (QK/PV mode2 sizes) 2026-05-24 03:49:55 +00:00
biondizzle
073119e9ca D1: Add sP shape debug print 2026-05-24 03:46:27 +00:00
biondizzle
44b72cc570 D1: Fix SMEM-P (coordinate store), LSE (FP32), add TMEM-P-only test 2026-05-24 03:27:14 +00:00
biondizzle
1effbd9484 D1: Fix SMEM-P - coordinate-indexed store (replaces make_tiled_copy_C) 2026-05-24 03:24:44 +00:00
biondizzle
0105037d99 feat: SMEM-P make_tiled_copy_C + zero-fill dest tensor 2026-05-24 03:23:53 +00:00
biondizzle
143a9ae5c7 feat: SMEM-P using make_tiled_copy_C(qk_mma) approach 2026-05-24 03:22:57 +00:00
biondizzle
8eef860947 fix: LSE type mismatch Float32→BFloat16 2026-05-24 03:20:26 +00:00
biondizzle
304429ee3e D1.5: Fix SMEM-P - use coordinate-indexed store (same proven pattern) 2026-05-24 03:19:32 +00:00
biondizzle
b471579140 D1.5: Always output un-normalized O + LSE (epilogue_tma_store only, no TMEM round-trip normalize) 2026-05-24 03:18:38 +00:00
biondizzle
940141a480 feat: SMEM-P using make_tiled_copy_A from PV MMA 2026-05-24 03:16:34 +00:00
biondizzle
641af108c3 fix: fence_proxy not fence 2026-05-24 02:45:01 +00:00
biondizzle
6b47cea90f D1.5: Replace broken make_cotiled_copy SMEM-P with coordinate-indexed store 2026-05-24 02:43:42 +00:00
biondizzle
45dd35ee92 Merge branch 'master' of ssh://sweetapi.com:2222/biondizzle/nvfp4-megamoe-kernel 2026-05-24 02:41:39 +00:00
biondizzle
4e593f1cad D1.5: Use tCtO_fake layout for epilogue_tma_store (needs STAGE dim) 2026-05-24 02:41:32 +00:00
biondizzle
4a2b48dc63 feat: SMEM-P using make_cotiled_copy (one-row-per-thread) 2026-05-24 02:41:19 +00:00
biondizzle
dd7eae0c1c D1.5: Remove duplicate tTMrO definition (keep unconditional one) 2026-05-24 02:40:40 +00:00
biondizzle
769181e503 fix: reorder tTMrO definition after tTMEM_LOADcO 2026-05-24 02:36:08 +00:00
biondizzle
05e53b53a2 fix: tTMrO scoping + restore SMEM-P coordinate write 2026-05-24 02:34:53 +00:00
biondizzle
d75f6c2afa D1.5: Move tTMrO def before softmax loop (CuTeDSL scoping) 2026-05-24 02:32:51 +00:00