[Kernel] Optimize long-context MLA cache gathers (#51739)
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com> Co-authored-by: OpenAI Codex <noreply@openai.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
co-authored by
OpenAI Codex
Claude Opus 5
parent
65b7662d3f
commit
c76a425278
@@ -2574,7 +2574,7 @@ class MLACommonBaseImpl(MLAAttentionImpl[A], Generic[A]):
|
||||
# FP8 path: gather cache without dequantization
|
||||
ops.cp_gather_cache(
|
||||
src_cache=kv_c_and_k_pe_cache,
|
||||
dst=workspace,
|
||||
dst=workspace[:toks],
|
||||
block_table=block_table,
|
||||
cu_seq_lens=chunk.cu_seq_lens,
|
||||
batch_size=chunk.num_requests,
|
||||
@@ -2683,7 +2683,7 @@ class MLACommonBaseImpl(MLAAttentionImpl[A], Generic[A]):
|
||||
else:
|
||||
ops.cp_gather_cache(
|
||||
src_cache=kv_c_and_k_pe_cache,
|
||||
dst=workspace,
|
||||
dst=workspace[:toks],
|
||||
block_table=block_table,
|
||||
cu_seq_lens=padded_local_cu_seq_lens,
|
||||
batch_size=chunk.num_requests,
|
||||
|
||||
Reference in New Issue
Block a user