6 Commits

Author	SHA1	Message	Date
biondizzle	b02e103ac0	Add c_simple GMEM tensor (non-dynamic) for SMEM accumulator TMA store	2026-05-27 05:33:30 +00:00
biondizzle	bf36979a8d	Use CUTLASS FMHA reference pattern for sC->GMEM TMA store (flat_divide + tma_partition)	2026-05-27 05:24:39 +00:00
biondizzle	97bc6d8d2f	Add c_direct GMEM tensor for direct writes in SMEM accumulator path	2026-05-27 05:15:47 +00:00
biondizzle	a858ed1c14	Fix test: normalize=False for un-normalized O comparison	2026-05-27 05:06:52 +00:00
biondizzle	3a7d87adba	Fix test_smem_acc: use keyword args for lse/row_sums	2026-05-27 04:54:23 +00:00
biondizzle	6a621bdf64	D1.5: SMEM accumulator FMHA kernel — one-way TMEM→REGS→SMEM, no round-trip ... TMEM round-trip (Ld32x32bOp/St32x32bOp) is FUNDAMENTALLY BROKEN. Even NO-OP (multiply by 1.0) corrupts data. New approach: - PV always ACCUMULATE=False (fresh TMEM each kt) - After pv_done_bar: one-way Ld32x32bOp load O_kt from TMEM→REGS - Coordinate-indexed SMEM accumulation: sO_acc = acc_scale * sO_acc + O_kt - sO_acc: FP32 [128, pv_n_tile] row-major (32KB at hd=64, 64KB at hd=128) - Final: normalize, cast BF16, write to sC, TMA store to GMEM	2026-05-27 04:53:40 +00:00