[x86][clang] Missed optimization for memcpy/memmove on AVX512 targets when using [[assume(size <= 64)]]
Open
backend:X86
missed-optimization
- Dominant language
- LLVM
- Stars
- 40.5k
- Forks
- 18.7k
- PR merge metrics
- PR metrics pending
Description
When you have [[assume(size <= 64]] while the size is still variable
the compiler still emits a jmp/call to memcpy/memmove
when masked load/stores could be used
https://clang.godbolt.org/z/Pa8qj7TTW
Contributor guide
Assessment
This issue has not been assessed yet.