Created attachment 548572 [details] perf raw data and ASCII call graph, xz-compressed Running Google's ROME demo, perf says that we spend roughly 10% time in the GC. Maybe that's normal, I don't know. The attached file is a xz-compressed archive containing the raw Perf data + ascii call graph attached. It still runs decently fast, but uses 2G of RAM. An about:memory capture of the same run is here: attachment 548567 [details] See also: bug 657115
The ROME demo does a lot of empty-chunk GCs, and I didn't see it do a single compartmental GC, which could make it spend more time than it really needs to.
Version: unspecified → Trunk
You need to log in before you can comment on or make changes to this bug.