ViperEkura b8b439b713 perf: fuse decode combine kernel to single-pass online-rescale reduction ...

- Replace 3-scan loops (mstar, lstar, acc) with 1-pass online rescale
- Halves __expf calls (num_splits vs 2*num_splits) and ml_part re-reads
- Mathematically equivalent, no change to o_part traffic or output

2026-07-11 00:24:09 +08:00

..

attn_common.cuh

refactor: rename gqa_* to attn_*, split-KV for all decode paths

2026-07-10 23:35:14 +08:00

attn_decode_split_kv_mma.cuh

perf: post-multiply attention scale in float instead of pre-scaling Q in bf16

2026-07-11 00:13:32 +08:00

attn_decode_split_kv.cuh

perf: fuse decode combine kernel to single-pass online-rescale reduction

2026-07-11 00:24:09 +08:00

attn_decode.cu

refactor: rename gqa_* to attn_*, split-KV for all decode paths

2026-07-10 23:35:14 +08:00

attn_entry_utils.cuh

refactor: rename gqa_* to attn_*, split-KV for all decode paths

2026-07-10 23:35:14 +08:00

attn_mma_utils.cuh

refactor: rename gqa_* to attn_*, split-KV for all decode paths

2026-07-10 23:35:14 +08:00

attn_prefill_split_q_mma.cuh

perf: post-multiply attention scale in float instead of pre-scaling Q in bf16

2026-07-11 00:13:32 +08:00

attn_prefill_split_q.cuh

refactor: rename gqa_* to attn_*, split-KV for all decode paths

2026-07-10 23:35:14 +08:00

attn_prefill.cu

refactor: rename gqa_* to attn_*, split-KV for all decode paths

2026-07-10 23:35:14 +08:00