2ec464824e
Try flattening sP and rP_bf16_qk with group_modes to fix rank mismatch
2026-05-23 19:13:59 +00:00
0e8f8ca2cf
Add debug prints for SMEM-P partition layouts
2026-05-23 19:13:07 +00:00
99566ef8a3
Fix SMEM-P: use BF16 copy atom and BF16 source with QK C-fragment layout
2026-05-23 19:12:13 +00:00
f77264f684
Fix SMEM-P copy: use tcgen05.copy.St32x32bOp with Float32 and copy from rP_words (Float32) not rP_bf16
2026-05-23 19:11:08 +00:00
65a52d20fd
Fix rP scope issue: use rP_bf16.iterator instead of rP.iterator
2026-05-23 09:36:22 +00:00
d5d259b214
Fix duplicate else: line in SMEM-P block
2026-05-23 09:35:47 +00:00
5aee06b991
SMEM-P: Use QK C-fragment layout instead of TMEM layout to fix rank mismatch
2026-05-23 09:35:24 +00:00
8ffd1154fb
Fix sP_2d definition for tSMEM_CPYsP
2026-05-23 09:34:50 +00:00
ba810e96c1
Remove debug print lines referencing deleted sP_2d
2026-05-23 09:34:09 +00:00
1b8bf4996f
Remove duplicate sP_2d line causing indentation error
2026-05-23 09:33:40 +00:00
7a06723cef
SMEM-P: Implement rank mismatch fix by reshaping source tensor
2026-05-23 09:33:24 +00:00
21be1d4355
Add more debug prints for sP shapes
2026-05-23 09:26:30 +00:00
69f458ac30
Add debug prints to SMEM-P path to understand rank mismatch
2026-05-23 09:25:48 +00:00
f6b43227e5
Fix SMEM-P copy rank mismatch (use rP_bf16 directly instead of group_modes)
2026-05-23 09:21:13 +00:00
83a7dd0679
D1.3: Implement SMEM-P path (write P to SMEM via tiled_smem_copy instead of zeroing sP)
2026-05-23 09:20:37 +00:00
d9349bd127
Stage D1: Multi-PV-tile support for hd>256 (tcgen05 MMA max N=256)
2026-05-23 09:04:01 +00:00
568840ffc9
Fix v_fmha layout to use pv_n_tile instead of head_dim for multi-PV-tile support
2026-05-23 09:02:01 +00:00
6934b9b4b6
D1.4: Add pv_n_tile and n_pv_tiles for multi-PV-tile support (tcgen05 MMA max N=256)
2026-05-23 09:00:18 +00:00
4eccbb05c1
shit carmine left dangling
2026-05-23 06:55:22 +00:00
cdfafe50de
fix: add SwiGLU clamping to fused kernel (paper §4.2.3, CG-1)
...
The fused SwiGLU kernel stored swiglu_limit but never applied it.
Paper §4.2.3: gate capped at swiglu_limit, linear clamped to [-limit, +limit].
Non-fused reference path already applies clamping correctly.
Fix: add fmin/fmax clamping in FP32 before BF16 conversion.
2026-05-23 06:32:54 +00:00
17b40eb3f8
STAGE_D.md: restructure with correctness gaps, TMEM budget, execution order
2026-05-23 06:31:37 +00:00
f30e7f5f43
D1.1: Fix make_fragment_A — use sP for SMEM source pv_mma
2026-05-23 06:04:44 +00:00
8b7b124295
D1.1: Fix PV A-operand construction — compile-time branch for TMEM vs SMEM
2026-05-23 06:03:27 +00:00
d7361fe9ae
D1.1: Add SMEM-P path behind use_smem_p flag (stub: zero sP)
2026-05-23 06:01:02 +00:00
e19a5d02d2
D1.0: Replace HEAD_DIM=64 with self.head_dim constructor parameter
2026-05-23 05:55:03 +00:00
a9adeb6e4a
Rename FmhaV3StageC → FmhaKernel — no dev stage artifacts in production API
2026-05-23 05:45:58 +00:00
0d3db00c44
Migrate Stage C kernel (proven cos 0.97) into module - exact copy, no modifications
2026-05-23 05:36:22 +00:00
a9449e9484
Fix TMEM-P offset calc: match Stage C with p_cols_fp32 from pv_mma_tiler[2]
2026-05-23 05:18:37 +00:00
88c8e49739
Add missing TMEM fence after P store in TMEM-P path
2026-05-23 05:17:45 +00:00
8deb987a3f
Fix p_cols_fp32: use pv_mma_tiler[2] (K-dim) not [1] (N-dim)
2026-05-23 05:16:19 +00:00
963c35ea29
Fix PV A-operand major mode: K for TMEM-P, a_major for SMEM-P
2026-05-23 05:14:08 +00:00
7f1febccf0
Fix CuTeDSL scoping: hoist P store vars out of if block
2026-05-23 05:12:30 +00:00
d15bb7b84a
Fix O rescale: use Stage C proven correction_rescale pattern
2026-05-23 05:10:46 +00:00
300482e40a
Fix tOrP0 indexing: 3-dim slice (None,None,kb) not 4-dim
2026-05-23 05:09:19 +00:00
0b277f4199
Fix CuTeDSL scoping: unconditionally define tOrP0 and tCrP
2026-05-23 05:08:10 +00:00
03fd055be0
Fix CuTeDSL variable scoping: define tOrP0 and tCrP in both branches
2026-05-23 05:07:30 +00:00
531f1a12b4
Fix p_tmem_s: use ComposedLayout from make_smem_layout_a, pass as kernel arg
2026-05-23 05:06:45 +00:00
c5ed9e3119
Consolidate FMHA stages A/B/C into unified kernel module with SMEM-P stub
2026-05-23 05:04:43 +00:00
0c7a69cf34
WIP: make_tiled_copy_C for P→SMEM
2026-05-23 03:56:56 +00:00
abecc0bd77
fix: cpasync.CopyOp for reg→SMEM
2026-05-23 03:54:49 +00:00
9611d9bc19
fix: CopyAtomUniversalOp
2026-05-23 03:52:47 +00:00
cfe21685d1
WIP: tiled copy for P→SMEM (zero fill)
2026-05-23 03:51:58 +00:00
373c395810
fix: cute.copy(dst, src) order
2026-05-23 03:51:00 +00:00
6996abcef3
fix: BFloat16 not Float32 for bf16 reg
2026-05-23 03:50:09 +00:00
3a5083baf2
WIP: P→SMEM write stub (zero fill, proper mapping TODO)
2026-05-23 03:49:05 +00:00
c5d20d5bd4
fix: partition_A not partition_S
2026-05-23 03:47:53 +00:00
be19dc5e39
fix: make_smem_layout_epi not make_epilogue_smem_layout
2026-05-23 03:47:09 +00:00
c7cbfe3675
WIP: SMEM P path for PV (compiles but P write not implemented)
2026-05-23 03:46:01 +00:00
49b3c2a5a2
D1: P store as BF16 using PV A-fragment layout (tOrP0)
...
Reverted tP to p_tmem_s.outer (needed for make_fragment_A profile).
P store now writes BF16 to TMEM using tOrP0's layout, matching PV A-fragment reads.
This fixes the layout mismatch at hd>64 where QK C-fragment composition
writes to different TMEM columns than PV A-fragment reads.
2026-05-23 03:42:07 +00:00
c99d78bf29
D1: align P store and PV A-fragment layouts via tP
...
Key insight: tP (PV A-fragment base) used p_tmem_s.outer layout,
but P store used QK C-fragment composition layout. These diverge at hd>64.
Fix: tP now uses the same QK C-fragment composition layout (tStP_layout)
as the P store. PV A-fragment is derived from tP, so it automatically
uses the same layout. No double-offset since tP includes P offset.
2026-05-23 03:40:10 +00:00
