diff options
author | Florian Hahn <flo@fhahn.com> | 2025-04-02 20:27:43 +0100 |
---|---|---|
committer | GitHub <noreply@github.com> | 2025-04-02 20:27:43 +0100 |
commit | 3bdf9a08804a5b424fd32fef3b0089f3a6db839d (patch) | |
tree | ef0723c1f2feb9e8808d237dfd8fe75c3e48898b /clang/lib/CodeGen | |
parent | 60efed3f2025cc9612570487dba5cb299ee28167 (diff) | |
download | llvm-3bdf9a08804a5b424fd32fef3b0089f3a6db839d.zip llvm-3bdf9a08804a5b424fd32fef3b0089f3a6db839d.tar.gz llvm-3bdf9a08804a5b424fd32fef3b0089f3a6db839d.tar.bz2 |
[EquivalenceClasses] Use SmallVector for deterministic iteration order. (#134075)
Currently iterators over EquivalenceClasses will iterate over std::set,
which guarantees the order specified by the comperator. Unfortunately in
many cases, EquivalenceClasses are used with pointers, so iterating over
std::set of pointers will not be deterministic across runs.
There are multiple places that explicitly try to sort the equivalence
classes before using them to try to get a deterministic order
(LowerTypeTests, SplitModule), but there are others that do not at the
moment and this can result at least in non-determinstic value naming in
Float2Int.
This patch updates EquivalenceClasses to keep track of all members via a
extra SmallVector and removes code from LowerTypeTests and SplitModule
to sort the classes before processing.
Overall it looks like compile-time slightly decreases in most cases, but
close to noise:
https://llvm-compile-time-tracker.com/compare.php?from=7d441d9892295a6eb8aaf481e1715f039f6f224f&to=b0c2ac67a88d3ef86987e2f82115ea0170675a17&stat=instructions
PR: https://github.com/llvm/llvm-project/pull/134075
Diffstat (limited to 'clang/lib/CodeGen')
0 files changed, 0 insertions, 0 deletions