Ben Kimock
Ben Kimock
@bors rollup=iffy Yeah I'm on a phone. We really are struggling here aren't we.
> since we're waiting anyway Hunh? Oh. The CI failure is hidden. Ugh
`eza`'s patch now dirties 32 CGUs instead of 16, and the microbenchmarks have an extra CGU. Regressions are as expected, the dubious CGU partitioning is a predictable outcome of compiling...
@bors try @rust-timer queue You can always ask for a perf run here: https://rust-lang.zulipchat.com/#narrow/channel/182449-t-compiler.2Fhelp/topic/perf.20run/with/543397764
I _think_ this is fixable if we lower the assignment of the uninit const to a store of an undef scalar. Previously we always lowered to a memcpy of an...
The fix that I described above ^ works fine for this "deinitialization" pattern of Scalars, so types like `u32`. And this is indeed clearly worse on nightly/beta than it is...
That's not right.
@bors try @rust-timer queue
@bors try @rust-timer queue