Bug #43942 | Make SyncObject::backoff really backoff | ||
---|---|---|---|
Submitted: | 29 Mar 2009 20:53 | Modified: | 15 May 2009 16:05 |
Reporter: | Mark Callaghan | Email Updates: | |
Status: | Closed | Impact on me: | |
Category: | MySQL Server: Falcon storage engine | Severity: | S3 (Non-critical) |
Version: | 6.0.10 | OS: | Any |
Assigned to: | Kelly Long | CPU Architecture: | Any |
Tags: | F_PERFORMANCE |
[29 Mar 2009 20:53]
Mark Callaghan
[31 Mar 2009 2:29]
Kevin Lewis
Kelly confirms that the current Falcon backoff code is affectively a no-op because the ++a; inside a fixed size loop is optimized into a single addition. Kelly agreed to provide a solution that will provide performance gains under higher concurrency without costing very much under low concurrency.
[31 Mar 2009 17:34]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/70938 3088 Kelly Long 2009-03-31 Fix for Bug #43942 - keep compiler optimization from removing a 'noop' loop that we use for a delay Initial tests with SyncObjectTest show this is a gain in most situations. I'll run some DBT2 tests to see what the impact will be on benchmarks.
[2 Apr 2009 17:38]
Bugs System
Pushed into 6.0.11-alpha (revid:hky@sun.com-20090402144811-yc5kp8g0rjnhz7vy) (version source revid:k.long@sun.com-20090331173027-orwhbrmd7ima50og) (merge vers: 6.0.11-alpha) (pib:6)
[15 May 2009 16:05]
MC Brown
Internal/test fix. No changelog entry required.