Paraphrasing/plagiarizing the comment in fp-logic-replace.ll:
We can replace "scalar" FP-bitwise-logic with the optimal type instruction. Scalar SSE/AVX FP-logic instructions only exist in your imagination and/or the bowels of compilers, but float and double variants of FP-logic instructions are reality, and the float variant may be a shorter instruction depending on which flavor of vector ISA you have...so just prefer float all the time. Yay, x86!
Details
Details
Diff Detail
Diff Detail
Event Timeline
Comment Actions
Thanks, Craig!
For reference, this is part of the yak-shaving needed to fix:
https://llvm.org/bugs/show_bug.cgi?id=6137