| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
feat: Add support for tags in benchmark comparisons and graph data - Updated the API to include runtime targets and tags in benchmark info. - Modified the graph loading functions to accept tags as parameters. - Enhanced the comparison interfaces to include tag_a and tag_b for compile and runtime benchmarks. - Updated the frontend components to allow selection of tags for comparisons. - Adjusted the backend logic to handle tags in comparison queries and data retrieval. - Ensured default tags are used where no specific tag is provided. | 4 个月前 | |
Fix collector heartbeat suspending during long benchmark jobs The heartbeat loop performed two unbounded awaits per tick (pool connection acquisition and the UPDATE query). A pooled Postgres connection can go half-open while sitting idle for hours during a long benchmark; is_valid only checked is_closed(), which does no network round-trip, so the heartbeat task could pick up a dead connection and hang forever, permanently freezing the loop. Increasing runtime worker threads never helped because the job future runs on the block_on thread and does not compete with the heartbeat task for workers. - Bound each heartbeat round with a 30s timeout so a stuck round logs a warning and the loop keeps ticking; use MissedTickBehavior::Delay to avoid tick bursts after a stall. - Probe pooled Postgres connections with SELECT 1 (5s timeout) in is_valid, matching the SQLite behaviour, so stale connections are discarded on reuse instead of being handed out. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> | 1 个月前 | |
Correct multiple repos enqueue and comparison frontend | 2 个月前 | |
move Point and Interpolate to database crate | 2 年前 | |
Improve error message | 2 个月前 | |
Refactor runtime comparison handling and enhance perf-record detail display - Removed unused imports and streamlined the JSON renderer registration in CompareScreen. - Updated metric handling in RuntimeJsonTestCase to use the Metric enum directly. - Enhanced Metric enum with PartialOrd and Ord implementations for better sorting. - Introduced sorting for JSON comparisons by benchmark name in the frontend. - Added functionality to expand perf-record rows for detailed function-level diffs in runtime comparisons. - Created a new PerfRecordDetail component to display detailed performance metrics. - Updated API response structures to include metric information in JSON comparisons. - Improved diff summary handling in the backend to utilize structured diff summaries. - Added CSS styles for improved visual representation of performance changes in the frontend. | 4 个月前 | |
fmt | 2 个月前 | |
add cross-machine comparison functionality with new API endpoints and frontend components Co-authored-by: ywxtcwh<ywxtcwh@gmail.com> | 3 个月前 |