<div dir="ltr"><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Feb 12, 2014 at 4:05 PM, Nadav Rotem <span dir="ltr"><<a href="mailto:nrotem@apple.com" target="_blank">nrotem@apple.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="">On Feb 12, 2014, at 3:54 PM, Owen Anderson <<a href="mailto:resistor@mac.com">resistor@mac.com</a>> wrote:<br>
<br>
> Author: resistor<br>
> Date: Wed Feb 12 17:54:07 2014<br>
> New Revision: 201275<br>
><br>
> URL: <a href="http://llvm.org/viewvc/llvm-project?rev=201275&view=rev" target="_blank">http://llvm.org/viewvc/llvm-project?rev=201275&view=rev</a><br>
> Log:<br>
> Remove a very old instcombine where<br>
<br>
</div>For vectors this optimization is always a win. Can you please bring it back and add a check that the comparison type is a vector?</blockquote></div><br>Do you have test cases which regressed? I find it odd that I saw no regressions from any of the heavily auto-vectorized benchmarks on x86-64.</div>
</div>