Skip to content
Presubmit Benchmarks

Index active transfer schedules by block ID in RegisterActivePlan so KVCacheManagerBase::GetBlockChunks scales linearly O(N) instead of quadratically O(N²). #2705

Sign in to view logs

Index active transfer schedules by block ID in RegisterActivePlan so KVCacheManagerBase::GetBlockChunks scales linearly O(N) instead of quadratically O(N²).

Index active transfer schedules by block ID in RegisterActivePlan so KVCacheManagerBase::GetBlockChunks scales linearly O(N) instead of quadratically O(N²). #2705

Triggered via pull request October 30, 2026 00:11
@copybara-servicecopybara-service[bot]
synchronize #1397
test_1008056720
Status Success
Total duration 12m 35s
Artifacts 1

presubmit_benchmarks.yml

on: pull_request
Run Single Host Benchmarks  /  Validate registry security
47s
Run Single Host Benchmarks / Validate registry security
Run Single Host Benchmarks  /  Generate matrix
1m 12s
Run Single Host Benchmarks / Generate matrix
Matrix: Run Single Host Benchmarks / run_benchmarks
Run Single Host Benchmarks  /  Analyze results
59s
Run Single Host Benchmarks / Analyze results
Run Single Host Benchmarks  /  Bundle artifacts
1m 7s
Run Single Host Benchmarks / Bundle artifacts
Fit to window
Zoom out
Zoom in

Artifacts

Produced during runtime
Name Size Digest
artifacts-0cdd2c6d4cd77fa83bab1aaa61803c1c
3.57 KB
sha256:a1b70196840fca7817d629b164ab826d44e99cc495dc0723ec99c77dbb700815