Redundant calls to Vec::reserve are not optimized out
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 119k
- Forks
- 16.1k
- PR merge metrics
- PR metrics pending
Description
pub fn reservy(x: &mut Vec<u8>) {
x.reserve(10);
x.reserve(10);
}
Currently compiles to:
define void @example::reservy::hfd53462e17d1ffb0(ptr noalias nocapture noundef align 8 dereferenceable(24) %x) unnamed_addr #2 !dbg !239 {
start:
%0 = getelementptr inbounds i8, ptr %x, i64 16, !dbg !241
%len = load i64, ptr %0, align 8, !dbg !241
%self = load i64, ptr %x, align 8, !dbg !248
%_6 = sub i64 %self, %len, !dbg !258
%_4 = icmp ult i64 %_6, 10, !dbg !262
br i1 %_4, label %bb2, label %bb4, !dbg !263
bb2:
tail call fastcc void @"alloc::raw_vec::RawVec<T,A>::reserve::do_reserve_and_handle::ha1fc33f26a8560a9"(ptr noalias noundef nonnull align 8 dereferenceable(16) %x, i64 noundef %len), !dbg !264
%len1.pre = load i64, ptr %0, align 8, !dbg !265
%self2.pre = load i64, ptr %x, align 8, !dbg !268
%.pre = sub i64 %self2.pre, %len1.pre, !dbg !275
%1 = icmp ult i64 %.pre, 10, !dbg !278
br i1 %1, label %bb3, label %bb4, !dbg !279
bb4:
ret void, !dbg !280
bb3:
tail call fastcc void @"alloc::raw_vec::RawVec<T,A>::reserve::do_reserve_and_handle::ha1fc33f26a8560a9"(ptr noalias noundef nonnull align 8 dereferenceable(16) %x, i64 noundef %len1.pre), !dbg !282
br label %bb4, !dbg !282
}
I found this while looking at the existing optimizer hints we have in RawVec. They are placed to optimize out redundant reserve calls, except that the hint is actually in the outlined part of reserve because of this wrapper that I wrote a while ago: https://github.com/rust-lang/rust/blob/edc4dc337b28a389dbd87b3825d7929381edeb8f/library/alloc/src/raw_vec.rs#L338-L343
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start in library/alloc/src/raw_vec.rs around the wrapper and optimizer hints at lines 338-343. Reproduce the shown reservy example and inspect its generated LLVM to understand why the second reserve remains. Done means redundant reserve calls are optimized out while the reserve behavior remains correct.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- compilers, performance
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100