mikros/rust
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Rollup merge of #77107 - bugadani:perf, r=oli-obk

Browse Source 
Enable const propagation into operands at mir_opt_level=2

Feature was added in #74507 but gated with `mir_opt_level>=3` because of compile time regressions. Let's see whether the LLVM 11 update solves that.

As the [perf results](https://github.com/rust-lang/rust/pull/77107#issuecomment-697668154) show, enabling this optimization results in a lot less regression as before.

cc @oli-obk

r? @ghost
This commit is contained in:
Jonas Schievink 2020-09-27 01:53:16 +02:00 committed by GitHub
commit 6b8fb3fbd8
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
compiler/rustc_mir/src/transform/const_prop.rs
View File

@ -1046,9 +1046,9 @@ fn visit_body(&mut self, body: &mut Body<'tcx>) {
fn visit_operand(&mut self, operand: &mut Operand<'tcx>, location: Location) {
self.super_operand(operand, location);
// Only const prop copies and moves on `mir_opt_level=3` as doing so
// currently increases compile time.
if self.tcx.sess.opts.debugging_opts.mir_opt_level >= 3 {
// Only const prop copies and moves on `mir_opt_level=2` as doing so
// currently slightly increases compile time in some cases.
if self.tcx.sess.opts.debugging_opts.mir_opt_level >= 2 {
self.propagate_operand(operand)
}
}
@ -1246,8 +1246,8 @@ fn fmt(&self, fmt: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
| TerminatorKind::InlineAsm { .. } => {}
// Every argument in our function calls have already been propagated in `visit_operand`.
//
// NOTE: because LLVM codegen gives performance regressions with it, so this is gated
// on `mir_opt_level=3`.
// NOTE: because LLVM codegen gives slight performance regressions with it, so this is
// gated on `mir_opt_level=2`.
TerminatorKind::Call { .. } => {}
}