Hi all,
The memory benchmarks on IsRustFastYet
(http://huonw.github.io/isrustfastyet/mem/) have been on a bit of a
hiatus, but they're back thanks to cmr's hard-working hardware (as well
as a little Rust program that we collaborated on to allow him to build
600+ of bors' merges automagically: https://github.com/huonw/multibuilder)
Since the beginning of August when IRFY dropped out, there have been 4
regressions (in the commits benchmarked so far), in chronological order:
- #8235 (500 MB!), which introduced the trait based visitor as a start
for replacing the @fn one (this is actually filed as
https://github.com/mozilla/rust/issues/8589)
- #8527 (60 MB), "Visitor trait rewrite, step 1." continuing the
(necessary) work above
- #9278 (60 MB), "Invert --cfg debug to --cfg ndebug"
- #9261 (130 MB), "Refactor the logging system for fewer allocations" (I
believe acrichto may be investigating this one already)
Highlighted:
http://huonw.github.io/isrustfastyet/mem/#2fd3c07,77739a7,1434b4b,2a7be1c
Due to the large backlog, we're not taking accurate time measurements,
just memory recordings, so the red dashed line is missing for most of
the new ones.
Huon
https://github.com/mozilla/rust/pull/8235
https://github.com/mozilla/rust/pull/8527
https://github.com/mozilla/rust/pull/9278
https://github.com/mozilla/rust/pull/9261
_______________________________________________
Rust-dev mailing list
Rust-dev@mozilla.org
https://mail.mozilla.org/listinfo/rust-dev
