[PATCH] D18949: [AArch64] Cannonicalize commutative operands based on LSLFast

[Haicheng Wu via llvm-commits]

haicheng added a comment.

In http://reviews.llvm.org/D18949#396702, @flyingforyou wrote:

> I think this optimization will affect all ARM Architectures. Is this optimization also good for cortex-a57?
>
> Could you let me know which benchmarks have this pattern, please?


If compiling spec2006 targeting cortex-a57, this pattern happens in spec2006/gcc once, spec2006/h264ref twice, spec2006/perlbench once.  Rest of spec2006 remain unchanged.  As to the performance, after the change, spec2006/gcc -0.05%, spec2006/h264ref +0.12%, spec2006/perlbench +0.01% averaging from 5 iterations on a cortex-a57 device.


Repository:
  rL LLVM

http://reviews.llvm.org/D18949