<div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, Aug 10, 2021 at 11:54 AM Wei Mi via Phabricator <<a href="mailto:reviews@reviews.llvm.org">reviews@reviews.llvm.org</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">wmi added a comment.<br>
<br>
> This extra pass turns to be positive in performance -- I'm seeing additional 0.6% - 0.8% improvement.<br>
<br>
That is a good improvement. I guess the improvement is because regalloc is profile sensitive and it is benefited from preciser fs profile information after you insert another pass of profile loading before regalloc?<br></blockquote><div>Yes. I guess so. </div><div>That's actually a bit surprising to me as I did not think RA was so sensitive to profile quality.</div><div>But the improvement is real as I can constantly reproduce in every run.</div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
<br>
CHANGES SINCE LAST ACTION<br>
<a href="https://reviews.llvm.org/D104579/new/" rel="noreferrer" target="_blank">https://reviews.llvm.org/D104579/new/</a><br>
<br>
<a href="https://reviews.llvm.org/D104579" rel="noreferrer" target="_blank">https://reviews.llvm.org/D104579</a><br>
<br>
</blockquote></div></div>