<div dir="ltr">It would be helpful if you can file a bug with a test case.<div><br></div><div>David</div><div><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Jul 16, 2015 at 5:43 PM, Lawrence <span dir="ltr"><<a href="mailto:lawrence@codeaurora.org" target="_blank">lawrence@codeaurora.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi, Daniel:<br>
<br>
Something interesting, even though your patch performed better for our precheckin perf run, it doesn't help the test I was looking at, for my case, revert your patch give the best results.<br>
<br>
I will revert internally for now, and looking for better solution in the future.<br>
<br>
Regards<br>
<br>
Lawrence Hu<br>
<span class="im HOEnZb"><br>
<br>
-----Original Message-----<br>
From: Lawrence [mailto:<a href="mailto:lawrence@codeaurora.org">lawrence@codeaurora.org</a>]<br>
Sent: Wednesday, July 15, 2015 9:36 PM<br>
To: 'Daniel Berlin'<br>
Cc: 'LLVM Developers Mailing List'<br>
</span><span class="im HOEnZb">Subject: RE: Register pressure mechanism in PRE or Smarter rematerialization/split/spiller/coalescing ?<br>
<br>
Hi, Daniel:<br>
<br>
Thanks, I tried that patch you provided, it is better than just disabling your previous patch, it has more improvements than degradations.<br>
<br>
Do you want to post that patch or you want me to do that?<br>
<br>
Regards<br>
<br>
</span><span class="im HOEnZb">Lawrence Hu<br>
<br>
-----Original Message-----<br>
From: Daniel Berlin [mailto:<a href="mailto:dberlin@dberlin.org">dberlin@dberlin.org</a>]<br>
</span><span class="im HOEnZb">Sent: Wednesday, July 15, 2015 1:36 PM<br>
To: Lawrence<br>
Cc: LLVM Developers Mailing List<br>
Subject: Re: Register pressure mechanism in PRE or Smarter rematerialization/split/spiller/coalescing ?<br>
<br>
</span><div class="HOEnZb"><div class="h5">On Wed, Jul 15, 2015 at 1:10 PM, Daniel Berlin <<a href="mailto:dberlin@dberlin.org">dberlin@dberlin.org</a>> wrote:<br>
> IMHO, This doesn't make a lot of sense to turn off this part on it's own.<br>
> I would just use the enable-pre flag to turn off scalar PRE, as it<br>
> will cause the same issue in other cases as well.<br>
> Is there some reason you aren't just doing that?<br>
> I suspect if this is a performance win, that would be as well.<br>
><br>
<br>
Ugh, actually, it should be a win with the following change:<br>
<br>
<br>
diff --git a/lib/Transforms/Scalar/GVN.cpp b/lib/Transforms/Scalar/GVN.cpp index 2c47a8a..a3387e3 100644<br>
--- a/lib/Transforms/Scalar/GVN.cpp<br>
+++ b/lib/Transforms/Scalar/GVN.cpp<br>
@@ -1767,7 +1767,7 @@ bool GVN::processNonLocalLoad(LoadInst *LI) {<br>
}<br>
<br>
// Step 4: Eliminate partial redundancy.<br>
- if (!EnablePRE || !EnableLoadPRE)<br>
+ if (!EnableLoadPRE)<br>
return false;<br>
<br>
return PerformLoadPRE(LI, ValuesPerBlock, UnavailableBlocks);<br>
<br>
<br>
<br>
<br>
This will disable Scalar PRE without disabling load PRE.<br>
<br>
<br>
(note, again, however, that load PRE can create exactly the same GEP situation you are referring to)<br>
<br>
<br>
_______________________________________________<br>
LLVM Developers mailing list<br>
<a href="mailto:LLVMdev@cs.uiuc.edu">LLVMdev@cs.uiuc.edu</a> <a href="http://llvm.cs.uiuc.edu" rel="noreferrer" target="_blank">http://llvm.cs.uiuc.edu</a><br>
<a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev" rel="noreferrer" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev</a><br>
</div></div></blockquote></div><br></div>