[PATCH] D21521: [X86][SSE] Add cost model for BSWAP of vectors

[Simon Pilgrim via llvm-commits]

RKSimon added a comment.

In http://reviews.llvm.org/D21521#462314, @mkuper wrote:

> I assume the costs are based on the lowering we have in test/CodeGen/X86/bswap-vector.ll ?
>  In any case, this LGTM.
>
> (Out of curiosity, do you see this get hit in practice?)


Thanks Michael - yes the costs are based off the codegen from bswap-vector.ll

I have a couple of examples where this patch (along with enabling the loadcombine pass) can help vectorize the loading/conversion of big endian float data.


Repository:
  rL LLVM

http://reviews.llvm.org/D21521