Did sampling profiling
In browser IO thread, PostPendingTask and malloc consume significant time (8% and 7%)
In renderer, found no single hot-spot
Revived tzik’s experiment code
Saw 5~10% improvement, legacy IPC based though
Resumed working on modulepreload (CL)
All CLs reviewed. Will send I2I+S and then land all.
[yhirano] Had a discussion about URLLoader interface decoupling. We will use the unified interface.
[kinuko] Created a patch to completely migrate to per-frame URLLoaderFactory in the renderer, under review
Out-of-Blink CORS (tyoshino, toyoshim)
Cleanup of existing WebCORS implementations
Basic unit tests for CORSURLLoader is added
Working on CORS check code moving
Origin Manifest (dhausknecht, mkwst, tyoshino, kinuko)
Landing ThrottlingURLLoader changes to support the usage
Discussed where and how response URL should be handled or informed between multiple redirects. One of the idea is to include it in ResourceResponseHead, but we’re going to go with smaller change in ThrottlingURLLoader first.
Preload + SRI (vogelheim, hiroshige, japhet, kouhei)
Having a lot of discussions. hiroshige@ preparing a summary doc.
V8 Cache Heuristics
[kouhei] landed UMA
[toyoshim] WebKit platform build file split
[hiroshige] Introduced XMLParserScriptRunner, refactoring doc.write intervension
More predictable MemoryCache (japhet, hiroshige, kinuko)
[kinuko] Sent out updates and shared the command-line flag to try it out on loading-dev@ and also with some external developers
WebFonts (ksakamoto, toyoshim)
Intervention v2 is 100% launched. Cleanup works have also finished.
[ksakamoto] Caused memory regression in webview CC. Investigating
ResourceLoadScheduler (toyoshim, ksakamoto)
Just waiting for field trial results