[Webkit-unassigned] [Bug 235114] REGRESSION (iOS 15): Massive variation in microbenchmark results when tab returns to foreground

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=235114

--- Comment #6 from Chris Dumez <cdumez at apple.com> ---
The process priorities look similar when the benchmark is slow and fast:

- Slow:
  491   9 (FGSupport)   13       15MB            46MB  1000MB                  ---    Y     Y     -    35711  com.apple.WebKit.WebContent

        policy: POLICY_TIMESHARE
                timeshare max  priority: 63
                timeshare base priority: 46
                timeshare cur  priority: 46
                timeshare depressed: NO, prio -1
        requested policy:
                req thread qos: THREAD_QOS_USER_INTERACTIVE, relprio: -1
                req workqueue/pthread overides:
                        req legacy qos override: THREAD_QOS_UNSPECIFIED
                        req workqueue qos override: THREAD_QOS_UNSPECIFIED
                req kernel overides:
                        req kevent overrides: THREAD_QOS_UNSPECIFIED
                        req workloop servicer override: THREAD_QOS_UNSPECIFIED
                req turnstiles sync promotion qos: THREAD_QOS_UNSPECIFIED, user promotion base pri: 0
                req latency qos: LATENCY_QOS_TIER_UNSPECIFIED
                req thruput qos: THROUGHPUT_QOS_TIER_UNSPECIFIED
                req darwin BG: NO  
                req internal/external iotier: THROTTLE_LEVEL_TIER0 (IMPORTANT) / THROTTLE_LEVEL_TIER0 (IMPORTANT)
                req other: 
        effective policy:
                eff thread qos: THREAD_QOS_USER_INTERACTIVE
                eff thread qos relprio: -1
                eff promotion qos: THREAD_QOS_USER_INTERACTIVE
                eff latency qos: LATENCY_QOS_TIER_0
                eff thruput qos: THROUGHPUT_QOS_TIER_0
                eff darwin BG: NO
                eff iotier: THROTTLE_LEVEL_TIER0 (IMPORTANT)
                eff other: 

- Fast:
  493   9 (FGSupport)   13       16MB            48MB  1000MB                  ---    Y     Y     -    35877  com.apple.WebKit.WebContent

policy: POLICY_TIMESHARE
                timeshare max  priority: 63
                timeshare base priority: 46
                timeshare cur  priority: 46
                timeshare depressed: NO, prio -1
        requested policy:
                req thread qos: THREAD_QOS_USER_INTERACTIVE, relprio: -1
                req workqueue/pthread overides:
                        req legacy qos override: THREAD_QOS_UNSPECIFIED
                        req workqueue qos override: THREAD_QOS_UNSPECIFIED
                req kernel overides:
                        req kevent overrides: THREAD_QOS_UNSPECIFIED
                        req workloop servicer override: THREAD_QOS_UNSPECIFIED
                req turnstiles sync promotion qos: THREAD_QOS_UNSPECIFIED, user promotion base pri: 0
                req latency qos: LATENCY_QOS_TIER_UNSPECIFIED
                req thruput qos: THROUGHPUT_QOS_TIER_UNSPECIFIED
                req darwin BG: NO  
                req internal/external iotier: THROTTLE_LEVEL_TIER0 (IMPORTANT) / THROTTLE_LEVEL_TIER0 (IMPORTANT)
                req other: 
        effective policy:
                eff thread qos: THREAD_QOS_USER_INTERACTIVE
                eff thread qos relprio: -1
                eff promotion qos: THREAD_QOS_USER_INTERACTIVE
                eff latency qos: LATENCY_QOS_TIER_0
                eff thruput qos: THROUGHPUT_QOS_TIER_0
                eff darwin BG: NO
                eff iotier: THROTTLE_LEVEL_TIER0 (IMPORTANT)
                eff other: 

I don't think this is a process priority issue. I am curious what JSC engineers think about this. I guess I can try and compare profiles next.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20220112/16894657/attachment.htm>