ViperEkura 9aca62c26c perf: remove per-element sentinel checks in softmax + fix stale comments ... - Replace 4*NC8 per-element -FLT_MAX comparisons with 2 row-level pn guards; masked entries naturally underflow via expf(-FLT_MAX - nm) ≈ 0; pn only guards all-masked-row edge where nm == -FLT_MAX (exp(0)=1 not 0); ~1-3% speedup on prefill (verified via standalone CUDA bench); correctness verified: prefill 4/4, decode 3/3, paged decode 13/13 - Fix stale comments: remove false 'pre-scale Q' claim, correct occupancy numbers, remove phantom sQ from smem description		2026-07-12 15:45:08 +08:00
..
attn_common.h	feat: paged decode attention with split-KV (scalar + MMA)	2026-07-11 18:14:51 +08:00
attn_decode.cu	perf: double-buffer KV pipeline and Q direct-to-register in decode	2026-07-12 14:14:54 +08:00
attn_decode_split_kv.cuh	refactor: template AttentionParams, rename .cuh to .h	2026-07-11 11:03:14 +08:00
attn_decode_split_kv_mma.cuh	perf: double-buffer KV pipeline and Q direct-to-register in decode	2026-07-12 14:14:54 +08:00
attn_entry_utils.cuh	refactor: template AttentionParams, rename .cuh to .h	2026-07-11 11:03:14 +08:00
attn_mma_utils.cuh	perf: remove per-element sentinel checks in softmax + fix stale comments	2026-07-12 15:45:08 +08:00
attn_paged_decode.cu	perf: double-buffer KV pipeline and Q direct-to-register in decode	2026-07-12 14:14:54 +08:00
attn_paged_decode_split_kv.cuh	feat: paged decode attention with split-KV (scalar + MMA)	2026-07-11 18:14:51 +08:00
attn_paged_decode_split_kv_mma.cuh	perf: double-buffer KV pipeline and Q direct-to-register in decode	2026-07-12 14:14:54 +08:00
attn_prefill.cu	perf: remove per-element sentinel checks in softmax + fix stale comments	2026-07-12 15:45:08 +08:00
attn_prefill_split_q.cuh	refactor: template AttentionParams, rename .cuh to .h	2026-07-11 11:03:14 +08:00
attn_prefill_split_q_mma.cuh	perf: remove per-element sentinel checks in softmax + fix stale comments	2026-07-12 15:45:08 +08:00