[PATCH] [AVX] Lower / fast-isel scalar FP selects into VBLENDV instructions (PR22483)
spatel at rotateright.com
Thu Mar 5 10:33:54 PST 2015
Hi Quentin -
Thanks for the prompt review!
I considered splitting this into 2 pieces as you suggested, but we would temporarily have a mess in test/CodeGen/X86/fast-isel-select-sse.ll. Instead of having a single set of CHECK lines, we'd have the old codegen for one case and the new codegen for the other case. This file would then have to get updated to the state you see in this patch upon the second commit. Do you see a way to avoid that?
More information about the llvm-commits