ViperEkura cb51a3587b perf: optimize fused FP8 GEMM kernel

2026-08-18 19:56:15 +08:00

..

attn_common.h

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_decode_split_kv_mma.cuh

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_decode_split_kv.cuh

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_decode.cu

refactor: tidy attention params and launcher interfaces

2026-08-09 20:52:06 +08:00

attn_dispatchers.cuh

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_entry_utils.cuh

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_layout_policies.cuh

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_mma_utils.cuh

perf: bypass L1 for attention tile loads

2026-08-15 21:23:20 +08:00

attn_paged_decode.cu

refactor: tidy attention params and launcher interfaces

2026-08-09 20:52:06 +08:00

attn_paged_prefill.cu

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_prefill_split_q_mma.cuh

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_prefill_split_q.cuh

perf: precompute ragged Q tile scheduling

2026-08-16 23:32:46 +08:00

attn_prefill.cu

refactor: tidy attention params and launcher interfaces

2026-08-09 20:52:06 +08:00

attn_warp_utils.cuh

perf: extend MMA decode to arbitrary GQA ratio, add launch bounds, vectorize combine

2026-07-27 00:35:34 +08:00

fp8_mm.cu

perf: optimize fused FP8 GEMM kernel

2026-08-18 19:56:15 +08:00

rotary_emb.cu

refactor: standardize packed 3d inference

2026-08-16 13:24:02 +08:00