[PATCH] D35897: [X86][AVX512] Add masked MOVS[S|D] patterns

[Guy Blank via Phabricator via llvm-commits]

guyblank added inline comments.


================
Comment at: lib/Target/X86/X86InstrAVX512.td:4005
 
+def : Pat<(f32 (X86selects (scalar_to_vector (and GR8:$mask, (i8 1))), (f32 FR32X:$src1), (f32 FR32X:$src2))),
+          (COPY_TO_REGCLASS (VMOVSSZrrk (COPY_TO_REGCLASS FR32X:$src2, VR128X),
----------------
craig.topper wrote:
> Does this extend past 80 columns?
you're right, thanks!


https://reviews.llvm.org/D35897