Closed
Description
#69635 (a rollup) caused a small performance regression. Since the regression has persisted through the last half dozen perf runs, it is likely not spurious. There are no outstanding perf runs near the regression (2020-03-02), so I am fairly confident in this attribution. That rollup consists of the following PRs:
- Add documentation to compiler intrinsics #68682 (Add documentation to compiler intrinsics)
- Unrevert "Remove
checked_add
inLayout::repeat
" #69544 (Unrevert "Removechecked_add
inLayout::repeat
") - constify mem::forget #69617 (constify mem::forget)
- Rename
syntax
in librustc_ast/README.md #69622 (Renamesyntax
in librustc_ast/README.md) - stash API: remove panic to fix ICE. #69623 (stash API: remove panic to fix ICE.)
- Toolstate: Don't block beta week on already broken tools. #69624 (Toolstate: Don't block beta week on already broken tools.)
I would not expect any of them to negatively impact performance, although I did do a perf run with an unchecked_add
in Layout::repeat
to no avail (see #69679). The regression is only noticeable in clean/patched incremental builds, so I don't think this is high priority, but it would be nice to know what's going on here.