diff options
| author | Nikita Popov <npopov@redhat.com> | 2024-11-25 11:49:00 +0100 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-11-25 11:49:00 +0100 |
| commit | e477989a055f92f6ca63fc8f76929cde81d33e44 (patch) | |
| tree | d713e0e42cce4351abde42a08af622ead7b84e98 /llvm/lib/CodeGen/SelectOptimize.cpp | |
| parent | 612f8ec7ac5dcddd16fb027aad64e2e353faa528 (diff) | |
| download | llvm-e477989a055f92f6ca63fc8f76929cde81d33e44.zip llvm-e477989a055f92f6ca63fc8f76929cde81d33e44.tar.gz llvm-e477989a055f92f6ca63fc8f76929cde81d33e44.tar.bz2 | |
[InstCombine] Handle trunc i1 pattern in eq-of-parts fold (#112704)
Equality/inequality of the low bit can be represented by `(trunc (xor x,
y) to i1)`, possibly with an extra not. We have to handle this in the
eq-of-parts fold now that we no longer canonicalize this to a masked
icmp.
Proofs: https://alive2.llvm.org/ce/z/qidkzq
Fixes https://github.com/llvm/llvm-project/issues/110919.
Diffstat (limited to 'llvm/lib/CodeGen/SelectOptimize.cpp')
0 files changed, 0 insertions, 0 deletions