[PATCH] [X86] tranform insertps to blendps when possible for better performance

[Chandler Carruth]

On Mon, Mar 2, 2015 at 3:47 PM, Sanjay Patel <spatel at rotateright.com> wrote:

> Updated patch to:
>
> 1. Only handle the low element to low element insertps case (immediate ==
> 0)
> 2. Added test case to confirm that we're only transforming the low element
> insertps case
> 3. Removed use of 'auto'
>

Thanks, and I'll look at the updated code once we sort out whether this can
be done directly during lowering with an adjustment to the lowering and/or
patterns for instruction selection.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150303/4ea427c5/attachment.html>