[ROCM][Perf] Add ROCM_SEGMENTED_ATTN attention-backend works across RDNA3/3.5/4 (#59132)

Signed-off-by: root <jeffaw99@hotmail.com>
Signed-off-by: big-yellow-duck <jeffaw99@hotmail.com>
Co-authored-by: Tan Pin Siang <1716735+tanpinsiang@users.noreply.github.com>
Co-authored-by: vllmellm <190700713+vllmellm@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
This commit is contained in:
big_yellow_duck
2026-10-06 07:53:47 +00:00
committed by GitHub
co-authored by Tan Pin Siang vllmellm Codex
parent 15c1cc12ac
commit edfccc9469
9 changed files with 4431 additions and 0 deletions
@@ -220,6 +220,13 @@ def kernel_warmup(worker: "Worker", *, process_local_only: bool = False):
if process_local_only:
return
if current_platform.is_rocm():
from vllm.model_executor.warmup.rocm_segmented_attn_autotune_warmup import (
rocm_segmented_attn_autotune_warmup,
)
rocm_segmented_attn_autotune_warmup(worker)
flashinfer_sparse_mla_decode_autotune_warmup(worker)
deepseek_v4_sparse_mla_attention_warmup(worker)
File diff suppressed because it is too large Load Diff