bjoernager/rust
bjoernager/rust
1
Fork 0
Code Issues Pull requests Activity

Auto merge of #96500 - SparrowLii:rpo, r=tmiasko

Browse source 
Reduce duplication of RPO calculation of mir

Computing the RPO of mir is not a low-cost thing, but it is duplicate in many places. In particular the `iterate_to_fixpoint` method which is called multiple times when computing the data flow.
This PR reduces the number of times the RPO is recalculated as much as possible, which should save some compile time.
This commit is contained in:
bors 2022-04-30 05:06:47 +00:00
commit 9a98c63b30
4 changed files with 96 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

4
compiler/rustc_const_eval/src/transform/promote_consts.rs
View file

@ -13,7 +13,7 @@
//! move analysis runs after promotion on broken MIR.
use rustc_hir as hir;
use rustc_middle::mir::traversal::ReversePostorder;
use rustc_middle::mir::traversal::ReversePostorderIter;
use rustc_middle::mir::visit::{MutVisitor, MutatingUseContext, PlaceContext, Visitor};
use rustc_middle::mir::*;
use rustc_middle::ty::cast::CastTy;
@ -170,7 +170,7 @@ impl<'tcx> Visitor<'tcx> for Collector<'_, 'tcx> {
pub fn collect_temps_and_candidates<'tcx>(
ccx: &ConstCx<'_, 'tcx>,
rpo: &mut ReversePostorder<'_, 'tcx>,
rpo: &mut ReversePostorderIter<'_, 'tcx>,
) -> (IndexVec<Local, TempState>, Vec<Candidate>) {
let mut collector = Collector {
temps: IndexVec::from_elem(TempState::Undefined, &ccx.body.local_decls),