[Core] Reuse empty block lists whenever possible in KVCacheBlocks to mitigate GC costs (#24964)

Signed-off-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>
2025-10-14 12:58:43 -07:00
parent 82af928c41
commit acaa2c0a4a
5 changed files with 53 additions and 26 deletions
--- a/vllm/v1/core/sched/scheduler.py
+++ b/vllm/v1/core/sched/scheduler.py
@@ -421,9 +421,7 @@ class Scheduler(SchedulerInterface):
                # KVTransfer: WAITING reqs have num_computed_tokens > 0
                # after async KV recvs are completed.
                else:
-                    new_computed_blocks = (
-                        self.kv_cache_manager.create_empty_block_list()
-                    )
+                    new_computed_blocks = self.kv_cache_manager.empty_kv_cache_blocks
                    num_new_local_computed_tokens = 0
                    num_computed_tokens = request.num_computed_tokens