[cfe-commits] r138411 - /cfe/trunk/lib/Headers/avxintrin.h

Bruno Cardoso Lopes bruno.cardoso at gmail.com
Tue Aug 23 16:46:48 PDT 2011

On Tue, Aug 23, 2011 at 4:40 PM, Eric Christopher <echristo at apple.com> wrote:
> On Aug 23, 2011, at 4:37 PM, Bruno Cardoso Lopes wrote:
>> Oh, forgot to mention, is going to be added on the llvm side, for
>> matching the intrinsic! :D
> Should be able to match the IR coming out of clang as different?

Actually, that could happen for several equivalent shuffles. Not in
this case though, because it's a really weird one. Right now, this
specific intrinsic can't match the 256-bit shufps instruction, because
the mask recognition isn't there yet (which is coming next w/ the
testcase), so it's emitting poor shuffles now.

Bruno Cardoso Lopes

More information about the cfe-commits mailing list