Backport LLVM changes to disable deferred inlining #92110
Conversation
|
r? @cuviper (rust-highfive has picked a reviewer for you, use r? to override) |
|
rust-highfive
added
the
S-waiting-on-review
|
We want to backport either this or #91190 to beta. |
nagisa
added
beta-nominated
|
cc @nikic |
|
Declining to beta-backport in favor of PR #91190. |
|
@rustbot label: -beta-nominated |
rustbot
removed
the
beta-nominated
|
@bors r+ rollup=never |
|
📌 Commit 2b821f2 has been approved by |
bors
added
S-waiting-on-bors
|
Note that this is an LLVM backport, not a beta backport (the PR wording confused me at first) |
|
☀️ Test successful - checks-actions |
|
Finished benchmarking commit (77497c7): comparison url. Summary: This change led to large relevant mixed results 🤷 in compiler performance.
If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf. Next Steps: If you can justify the regressions found in this perf run, please indicate this with @rustbot label: +perf-regression |
Mark drop calls in landing pads `cold` instead of `noinline` Now that deferred inlining has been disabled in LLVM (rust-lang#92110), this shouldn't cause catastrophic size blowup. I confirmed that the test cases from rust-lang#41696 (comment) still compile quickly (<1s) after this change. ~Although note that I wasn't able to reproduce the original issue using a recent rustc/llvm with deferred inlining enabled, so those tests may no longer be representative. I was also unable to create a modified test case that reproduced the original issue.~ (edit: I reproduced it on CI by accident--the first commit timed out on the LLVM 12 builder, because I forgot to make it conditional on LLVM version) r? `@nagisa` cc `@arielb1` (this effectively reverts rust-lang#42771 "mark calls in the unwind path as !noinline") cc `@RalfJung` (fixes rust-lang#46515) edit: also fixes rust-lang#87055
Mark drop calls in landing pads `cold` instead of `noinline` Now that deferred inlining has been disabled in LLVM (rust-lang#92110), this shouldn't cause catastrophic size blowup. I confirmed that the test cases from rust-lang#41696 (comment) still compile quickly (<1s) after this change. ~Although note that I wasn't able to reproduce the original issue using a recent rustc/llvm with deferred inlining enabled, so those tests may no longer be representative. I was also unable to create a modified test case that reproduced the original issue.~ (edit: I reproduced it on CI by accident--the first commit timed out on the LLVM 12 builder, because I forgot to make it conditional on LLVM version) r? `@nagisa` cc `@arielb1` (this effectively reverts rust-lang#42771 "mark calls in the unwind path as !noinline") cc `@RalfJung` (fixes rust-lang#46515) edit: also fixes rust-lang#87055
Fixes #91128
I was thinking of how to best add the test case from the issue, and I think rust perf infrastructure would probably be the best place for something like it.