The attached microbenchmark is based on the VanillaJS/Adding100Items part of Speedometer (the template engine). I get the following numbers: Nightly: 174 ms Chrome: 73 ms We spend a lot of time under RopeMatch resizing the Vector in StringSegmentRange. Pretty silly: giving this Vector an inline capacity of 8 improves our numbers to 65 ms or so.
I suspect this Vector used to have inline capacity at some point but the move to GCVector/StringVector regressed it...
Attachment #8847556 - Flags: review?(luke)
Pushed by email@example.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/f307a6f0809f Give StringSegmentRange stack Vector some inline capacity to avoid malloc/free overhead. r=luke
You need to log in before you can comment on or make changes to this bug.