<html><head><meta http-equiv="content-type" content="text/html; charset=utf-8"></head><body dir="auto"><div>I think we should just apply this patch now as we iterate over the proper solution. We can then revert it.</div><div><br>On Aug 7, 2014, at 3:35 PM, Alexander Kornienko <<a href="mailto:alexfh@google.com">alexfh@google.com</a>> wrote:<br><br></div><blockquote type="cite"><div><div dir="ltr">Any news here? This keeps biting us.<br><div class="gmail_extra"><br></div><div class="gmail_extra">Maybe you can apply this patch first and work on a better solution after that?<br><br><div class="gmail_quote">
On Tue, Aug 5, 2014 at 6:23 PM, Artyom Skrobov <span dir="ltr"><<a href="mailto:Artyom.Skrobov@arm.com" target="_blank">Artyom.Skrobov@arm.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div lang="EN-GB" link="blue" vlink="purple"><div><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">So, the simplest way to restore the original behaviour seems to be<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"><u></u> <u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">Index: lib/Analysis/LiveVariables.cpp<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">===================================================================<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">--- lib/Analysis/LiveVariables.cpp (revision 214183)<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">+++ lib/Analysis/LiveVariables.cpp (working copy)<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">@@ -451,6 +451,7 @@<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> // Construct the dataflow worklist. Enqueue the exit block as the<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> // start of the analysis.<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> DataflowWorklist worklist(*cfg, AC);<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">+ worklist.quick_hack();<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> llvm::BitVector everAnalyzedBlock(cfg->getNumBlockIDs());<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> <u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> // FIXME: we should enqueue using post order.<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">Index: include/clang/Analysis/Analyses/DataflowWorklist.h<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">===================================================================<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">--- include/clang/Analysis/Analyses/DataflowWorklist.h (revision 214183)<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">+++ include/clang/Analysis/Analyses/DataflowWorklist.h (working copy)<u></u><u></u></span></p><p class="MsoNormal">
<span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">@@ -53,6 +53,7 @@<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> const CFGBlock *dequeue();<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> <u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> void sortWorklist();<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d">+ void quick_hack() { enqueuedBlocks.reset(); PO_I = PO_E; }<u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> };<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> <u></u><u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Courier New";color:#1f497d"> } // end clang namespace<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"><u></u> <u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">This has the effect of disabling the fallback post-order iteration.<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"><u></u> <u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">The reason for the performance regression, however, is related to the iteration order over the CFG: UninitializedValues expects to use the default (i.e. reverse) PostOrderCFGView iteration, whereas LiveVariables expects the opposite order, corresponding to using llvm::GraphTraits<llvm::Inverse<const CFG*> > > as the last parameter in typedef po_iterator.<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"><u></u> <u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">I’m currently thinking of a proper solution which would allow using PostOrderCFGView for iteration in either direction.<u></u><u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"><u></u> <u></u></span></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"><u></u> <u></u></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"><u></u> <u></u></span></p><div style="border:none;border-top:solid #b5c4df 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal"><b><span lang="EN-US" style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span lang="EN-US" style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Alexander Kornienko [mailto:<a href="mailto:alexfh@google.com" target="_blank">alexfh@google.com</a>] <br>
<b>Sent:</b> 05 August 2014 12:30</span></p><div class=""><br><b>To:</b> Artyom Skrobov<br><b>Cc:</b> <a href="mailto:cfe-commits@cs.uiuc.edu" target="_blank">cfe-commits@cs.uiuc.edu</a> Commits<br></div><div class=""><b>Subject:</b> Re: r214064 - Factoring DataflowWorklist out of LiveVariables and UninitializedValues analyses<u></u><u></u></div>
<p></p></div><div class=""><p class="MsoNormal"><u></u> <u></u></p><div><p class="MsoNormal">On another file from the same project the difference in my setup is 2 minutes before the patch vs. 50+ hours (I terminated the analyzer before it finished) after the patch. So it's a rather bad performance regression.<u></u><u></u></p>
<div><p class="MsoNormal" style="margin-bottom:12.0pt"><u></u> <u></u></p><div><p class="MsoNormal">On Tue, Aug 5, 2014 at 10:17 AM, Artyom Skrobov <<a href="mailto:Artyom.Skrobov@arm.com" target="_blank">Artyom.Skrobov@arm.com</a>> wrote:<u></u><u></u></p>
<div><div><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">Indeed, for the other version of options.c, the analysis (on my machine) takes 3 sec before the patch, and 6 min after the patch.</span><u></u><u></u></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">It doesn’t hang indefinitely though, and the output is the same in both cases; so what we have is a performance regression.</span><u></u><u></u></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"> </span><u></u><u></u></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">I’m now trying a fix.</span><u></u><u></u></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"> </span><u></u><u></u></p><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d"> </span><u></u><u></u></p>
<div style="border:none;border-top:solid #b5c4df 1.0pt;padding:3.0pt 0cm 0cm 0cm"><p class="MsoNormal"><b><span lang="EN-US" style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span lang="EN-US" style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Alexander Kornienko [mailto:<a href="mailto:alexfh@google.com" target="_blank">alexfh@google.com</a>] <br>
<b>Sent:</b> 04 August 2014 19:26<br><b>To:</b> Artyom Skrobov</span><u></u><u></u></p><div><p class="MsoNormal"><br><b>Subject:</b> Re: r214064 - Factoring DataflowWorklist out of LiveVariables and UninitializedValues analyses<u></u><u></u></p>
</div></div><div><p class="MsoNormal"> <u></u><u></u></p><div><div><p class="MsoNormal">On Mon, Aug 4, 2014 at 4:36 PM, Artyom Skrobov <<a href="mailto:Artyom.Skrobov@arm.com" target="_blank">Artyom.Skrobov@arm.com</a>> wrote:<u></u><u></u></p>
</div><div><div><blockquote style="border:none;border-left:solid #cccccc 1.0pt;padding:0cm 0cm 0cm 6.0pt;margin-left:4.8pt;margin-top:5.0pt;margin-right:0cm;margin-bottom:5.0pt"><div><div><p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">Are you sure this issue has anything to do with my patch?</span><u></u><u></u></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1f497d">I’ve just run “clang -cc1 -analyze -analyzer-checker=deadcode.DeadStores options.i” using a build of Clang pre-dating my patch by a couple of weeks; and the analysis hangs nevertheless.</span><u></u><u></u></p>
</div></div></blockquote><div><p class="MsoNormal"> <u></u><u></u></p></div><div><p class="MsoNormal">You're right. This specific file causes hangs regardless of your patch. Sorry for not verifying this.<u></u><u></u></p>
</div><div><p class="MsoNormal"> <u></u><u></u></p></div><div><p class="MsoNormal">However, I've got another version of this file, which only makes analyzer hang with your patch. Please try it.<u></u><u></u></p></div>
<div><p class="MsoNormal"> </p></div></div></div></div></div></div></div></div></div></div></div></div></div></blockquote></div>
</div></div>
</div></blockquote><blockquote type="cite"><div><span>_______________________________________________</span><br><span>cfe-commits mailing list</span><br><span><a href="mailto:cfe-commits@cs.uiuc.edu">cfe-commits@cs.uiuc.edu</a></span><br><span><a href="http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits">http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits</a></span><br></div></blockquote></body></html>