0fe8bc7355
D5b MILESTONE: SWA+sink merge works! cos 0.969
...
- Run FMHA twice (compressed KV + SWA KV) with normalized O + LSE
- Merge with sink weights in Python
- LSE err=0.0, merge cos=0.969 PASS
- Update STAGE_D.md: D5b done, D5c/D5d are optimizations
2026-05-23 21:36:26 +00:00
3891f00b9a
D5b: Use normalized O + LSE for merge (correct formula), always output LSE
2026-05-23 21:35:40 +00:00
34125fa61c
D5b: Clean up merge test - stable formula for both ref and kernel
2026-05-23 21:33:45 +00:00
2c5799c8d8
D5b: Use reference per-row LSE for proper O normalization
2026-05-23 21:31:52 +00:00
0a0877c9bc
D5b: Fix kernel_obj reference
2026-05-23 21:30:59 +00:00
369d677c2c
D5b: Fix syntax error
2026-05-23 21:30:00 +00:00
6a47015e85
D5b: Debug reference formula mismatch, add numerically stable merge
2026-05-23 21:19:25 +00:00
023217e7b2
D5b: Python SWA+sink merge test
...
- Run FMHA twice (compressed KV + SWA KV, normalize=False)
- Merge with sink weights in Python
- Verify end-to-end correctness vs FP32 reference
2026-05-23 21:18:06 +00:00
6edb7a91a7
Update STAGE_D.md: D5a done, CG-2/CG-3 status updated, tOrP0 offset rule added
2026-05-23 21:16:52 +00:00
d02069094a
D5a: Fix LSE formula - lse = ln(row_sum) + row_max * ln(2)
...
row_max is in scale_log2 domain, need to convert to natural log domain.
attn_max = row_max * ln(2), so lse = ln(row_sum) + row_max * ln(2).
2026-05-23 21:15:14 +00:00
b0474c0ca9
D5a: Use tensor indexing for LSE write
2026-05-23 21:13:52 +00:00
aba8f111d5
D5a: Use cute.store for LSE write
2026-05-23 21:13:07 +00:00
ba288cfbd3
D5a: Fix LSE - compute row_max_safe from final row_max, remove mLSE None check
2026-05-23 21:12:29 +00:00
a331c8816a
D5a: Fix - add normalize param to __init__
2026-05-23 21:11:37 +00:00
cb9ebf33fa
D5a: Add normalize flag + LSE output
...
- normalize=True (default): O = softmax(P) @ V (existing behavior)
- normalize=False: O = P @ V (un-normalized) + lse = log(row_sum) + row_max
- LSE tensor passed as optional parameter
- Test includes D5a normalize=False verification with LSE comparison
- Cleaned up SMEM-P debug prints and broken make_tiled_copy_C code
- hd=64 TMEM-P regression: cos 0.973 PASS
2026-05-23 21:10:40 +00:00
f1912c01cb
D1.3: Use const_expr if for tOrP0 compile-time selection
2026-05-23 21:07:58 +00:00
61a19b69a7
D1.3: Pre-compute tOrP0_offset in _setup, use const_expr for compile-time selection
2026-05-23 21:07:10 +00:00
7f250cd4c8
D1.3: Use const_expr for tOrP0 offset (compile-time conditional)
2026-05-23 21:06:16 +00:00
e2ba21df46
D1.3: Use MLIR-compatible expression for tOrP0 offset (same as Stage C)
2026-05-23 21:05:12 +00:00
8d15e5e141
D1.3: Initialize tOrP0 before conditional for CuTeDSL scoping
2026-05-23 21:03:53 +00:00
911a80e721
D1.3: Fix tOrP0 for SMEM-P - skip make_tensor when offset is 0
...
CuTeDSL doesn't support OpResult + int. When offset is 0 (SMEM-P),
just use tOrP directly.
2026-05-23 21:03:00 +00:00
295e5a8c2f
D1.3: Fix tOrP0 offset - scale FP32 columns to BF16 elements
...
tmem_p0_offset is in FP32 columns, but tOrP uses BF16 elements.
Offset = p0_offset * (32/16) = p0_offset * 2.
2026-05-23 21:02:04 +00:00
f3f2ab4b50
D1.3: Fix CuTeDSL scoping - define tOrP0 unconditionally with p0 offset
2026-05-23 21:01:18 +00:00
27f336ad54
D1.3: Fix critical bug - add TMEM column offset for P0 in PV GEMM
...
The softmax warps store P at tmem_p0_offset=32. PV MMA must read from
the same offset. tOrP0 was missing the offset, causing PV to read from
TMEM column 0 (where S is) instead of column 32 (where P is).
This was the root cause of NaN/zeros in D1 tests.
2026-05-23 21:00:29 +00:00
71c70e688f
D1.3: Revert to d1.3-pre-sm100-helpers baseline for testing
2026-05-23 20:58:06 +00:00
36f49e574c
D1.3: DIAGNOSTIC - test epilogue_tma_store raw PV without any round-trips
2026-05-23 20:57:13 +00:00
9e158dfc9f
D1.3: Remove NO-OP round-trip, keep normalize + epilogue_tma_store
2026-05-23 20:56:13 +00:00
2e7fdf3a08
D1.3: Full correction_epilog with TMA store, normalize in reg before SMEM write
...
One-way trip: TMEM->reg (normalize) ->SMEM->GMEM
Replicates epilogue_tma_store logic with normalize step added
Uses CUTLASS helpers for correct layout handling
2026-05-23 20:54:23 +00:00
b05ff9deb4
D1.3: Apply transform_partitioned_tensor_layout before epilogue helpers
2026-05-23 20:52:42 +00:00
dced6fa11e
D1.3: Replace NO-op TMEM round-trip with correction_epilog using epilogue_tmem_copy_and_partition + epilogue_smem_copy_and_partition
...
- Remove hand-constructed TMEM round-trips (3% layout mismatch error)
- Use CUTLASS get_tmem_load_op + get_smem_store_op paired atoms
- One-way trip: TMEM -> reg (normalize) -> SMEM -> GMEM
- SMEM-P path: zero-fill stub (proper copy TBD)
- Keep per-tile O rescale atoms for n>128 support
2026-05-23 20:50:23 +00:00
e677bf7833
SMEM-P: test permutation 4 (swap m↔n2)
2026-05-23 20:20:07 +00:00
c88ca96139
SMEM-P: add iterator offset debug print
2026-05-23 20:19:22 +00:00
6629fe57cf
SMEM-P: add tCrP debug print, reset permute to 0
2026-05-23 20:14:32 +00:00
3bbc9a5a86
SMEM-P: test permutation 1 (swap m↔n0)
2026-05-23 20:14:21 +00:00
e2514f292c
auto: pre-test commit
2026-05-23 20:13:47 +00:00
67e8d6dc70
SMEM-P: add debug_permute flag for coordinate permutation testing
2026-05-23 20:13:44 +00:00
be67295b71
SMEM-P: disable debug flags, revert to original mapping
2026-05-23 20:12:26 +00:00
4d428c6619
SMEM-P: fix scoping error, disable debug_p_one, enable debug_swap_mn
2026-05-23 20:11:50 +00:00
afa4b8a746
SMEM-P: add debug_swap_mn flag to test swapped coordinate mapping
2026-05-23 20:10:39 +00:00
aac08b57c4
SMEM-P: add debug_p_one flag to write constant P=1.0
2026-05-23 20:09:47 +00:00
dd0079b9dc
auto: pre-test commit
2026-05-23 20:08:31 +00:00
0bcdfa1dc9
SMEM-P: implement two-phase softmax with normalization before SMEM write
2026-05-23 20:08:29 +00:00
e4e63b0331
Revert TMEM layout change (caused error)
2026-05-23 20:07:02 +00:00
b450a82b4c
SMEM-P: try using PV A-operand layout directly for TMEM-P
2026-05-23 20:06:23 +00:00
74b3cc041d
SMEM-P: debug compute offset for known coordinate
2026-05-23 20:05:02 +00:00
5cfde8d3db
SMEM-P: add crd2idx debug attempt
2026-05-23 20:04:28 +00:00
98e0428840
SMEM-P: test pattern based on fragment indices (k,j)
2026-05-23 20:03:22 +00:00
a713cf55b4
auto: pre-test commit
2026-05-23 20:02:35 +00:00
8ca07722ae
SMEM-P: add shape debug prints
2026-05-23 20:02:32 +00:00
d008505330
SMEM-P: add debug prints for coordinates
2026-05-23 20:00:33 +00:00
