[PATCH] D12561: [X86][SSE] Match zero/any extension shuffles that don't start from the first element

Quentin Colombet via llvm-commits llvm-commits at lists.llvm.org
Mon Sep 21 11:04:45 PDT 2015

qcolombet accepted this revision.
qcolombet added a comment.
This revision is now accepted and ready to land.

> Regarding the tests - the over use of PSHUFB can be a performance issue - you often have to load the shuffle mask (>5cy) and many targets (AMD + Intel) have a poor latency/throughput executing it (>icy).

Ah, that’s the catch, loading the mask! From the diff that wasn’t clear since we do not see the actual operands.

Alright, LGTM then!




More information about the llvm-commits mailing list