[PATCH] D10683: AVX-512 vector shuffle lowering

Simon Pilgrim llvm-dev at redking.me.uk
Fri Jul 24 10:00:47 PDT 2015

RKSimon added a comment.

In http://reviews.llvm.org/D10683#210761, @delena wrote:

> Added shuffle decoding for VPERMV and VPERMV3 nodes.
>  I can't add llc comments because the instructions load mask from memory.

You should be able to implement this like PSHUFB is handled in X86AsmPrinter::EmitInstruction.

Comment at: lib/Target/X86/X86ISelLowering.cpp:10529
@@ -10363,1 +10528,3 @@
 /// \brief Handle lowering of 8-lane 64-bit floating point shuffles.
Could you put lowerVectorShuffleWithUNPCK earlier in the source file so we can use it for the 128/256 bit shuffle lowering functions? You don't have to fix the other uses in this patch if you want to keep it focussed on AVX512.


More information about the llvm-commits mailing list