Bug 1927538 Comment 5 Edit History

Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.

(In reply to Andrew Creskey [:acreskey] from comment #2)
> 2 - don't bother breaking up the classification into events done on background and then main thread: just do it on the current thread
> https://hg.mozilla.org/try/rev/00e00f6adcd5494b178e448e544c1e90c48c00eb

And this idea also seems to make this worse -- [perf compare](https://treeherder.mozilla.org/perfherder/compare?originalProject=try&originalRevision=a6b65ef50eae501e5747fc2691dc0f42be9938ce&newProject=try&newRevision=28895910df88f497882d91edaa977a6d5552ccb2&page=1&framework=13&showOnlyImportant=1)

A few improvements to `HttpChannelAsyncOpenToTransactionPending`
• amazon perfstat-HttpChannelAsyncOpenToTransactionPending opt fission warm webrender, `11% improvement`, high confidence
• bing-search perfstat-HttpChannelAsyncOpenToTransactionPending opt cold fission webrender,  `6% improvement`, high confidence
• cnn perfstat-HttpChannelAsyncOpenToTransactionPending opt fission warm webrender, `13% improvement`, high confidence
• google-slides perfstat-HttpChannelAsyncOpenToTransactionPending opt fission warm webrender,  `9% improvement`, high confidence

And at least one visual metric improvement:
• netflix PerceptualSpeedIndex opt fission warm webrender,  `8.7% improvement`, high confidence

But too many regressions to list, visual metrics and networking perfstats.
(In reply to Andrew Creskey [:acreskey] from comment #2)
> 2 - don't bother breaking up the classification into events done on background and then main thread: just do it on the current thread
> https://hg.mozilla.org/try/rev/00e00f6adcd5494b178e448e544c1e90c48c00eb

And this idea also seems to make things worse -- [perf compare](https://treeherder.mozilla.org/perfherder/compare?originalProject=try&originalRevision=a6b65ef50eae501e5747fc2691dc0f42be9938ce&newProject=try&newRevision=28895910df88f497882d91edaa977a6d5552ccb2&page=1&framework=13&showOnlyImportant=1)

A few improvements to `HttpChannelAsyncOpenToTransactionPending`
• amazon perfstat-HttpChannelAsyncOpenToTransactionPending opt fission warm webrender, `11% improvement`, high confidence
• bing-search perfstat-HttpChannelAsyncOpenToTransactionPending opt cold fission webrender,  `6% improvement`, high confidence
• cnn perfstat-HttpChannelAsyncOpenToTransactionPending opt fission warm webrender, `13% improvement`, high confidence
• google-slides perfstat-HttpChannelAsyncOpenToTransactionPending opt fission warm webrender,  `9% improvement`, high confidence

And at least one visual metric improvement:
• netflix PerceptualSpeedIndex opt fission warm webrender,  `8.7% improvement`, high confidence

But too many regressions to list, visual metrics and networking perfstats.

Back to Bug 1927538 Comment 5