<html><head><meta http-equiv="Content-Type" content="text/html charset=windows-1252"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;"><br><div><div>On Jul 30, 2013, at 4:10 PM, Guo, Xiaoyi <<a href="mailto:Xiaoyi.Guo@amd.com">Xiaoyi.Guo@amd.com</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div lang="EN-US" link="blue" vlink="purple" style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px;"><div class="WordSection1" style="page: WordSection1;"><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Hi Andy,<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Thanks very much for looking into the problem.<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">In this particular test case, it seems most of the time is spent in the sorting, not the grouping.<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Later, I realized that it seems in this test case most of the expressions to be compared have different length. I tried the following change in compare() when the LHS and RHS’s types are the same:<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">===================================================================<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">--- lib/Analysis/ScalarEvolution.cpp (revision 187379)<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">+++ lib/Analysis/ScalarEvolution.cpp (working copy)<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">@@ -585,6 +585,9 @@<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> case scAddExpr:<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> case scMulExpr:<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> case scSMaxExpr:<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> case scUMaxExpr: {<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> const SCEVNAryExpr *LC = cast<SCEVNAryExpr>(LHS);<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> const SCEVNAryExpr *RC = cast<SCEVNAryExpr>(RHS);<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> // Lexicographically compare n-ary expressions.<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> unsigned LNumOps = LC->getNumOperands(), RNumOps = RC->getNumOperands();<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">+ if (LNumOps != RNumOps) {<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">+ return (int)LNumOps - (int)RNumOps;<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">+ }<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> for (unsigned i = 0; i != LNumOps; ++i) {<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> if (i >= RNumOps)<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> return 1;<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">And the compile time is cut down from 45s to 1s.</span></div></div></div></blockquote><div><br></div>Committed r187475. Thanks!</div><div>-Andy</div><div><br><blockquote type="cite" dir="auto"><div lang="EN-US" link="blue" vlink="purple" style="font-family: Helvetica; font-size: 12px; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: auto; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; widows: auto; word-spacing: 0px; -webkit-text-stroke-width: 0px;"><div class="WordSection1" style="page: WordSection1;"><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"><o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">This will give different sorting result than the original algorithm. However, it looks like that shouldn’t be a problem according to this comment before the switch statement in compare();<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> // Aside from the getSCEVType() ordering, the particular ordering<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> // isn't very important except that it's beneficial to be consistent,<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> // so that (a + b) and (b + a) don't end up as different expressions.<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Does this solution seem ok?<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">If the above solution seems ok, that solves the problem for cases when most of the time the expressions to be compared have different lengths. However, the problem still exists if the expressions to be compared are large, similar, and have the same length. Maybe I’ll leave that to later when there’s a test case for such situations?<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Thanks,<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Xiaoyi<o:p></o:p></span></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span></div><div><div style="border-style: solid none none; border-top-color: rgb(181, 196, 223); border-top-width: 1pt; padding: 3pt 0in 0in;"><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><b><span style="font-size: 10pt; font-family: Tahoma, sans-serif;">From:</span></b><span style="font-size: 10pt; font-family: Tahoma, sans-serif;"><span class="Apple-converted-space"> </span>Andrew Trick [<a href="mailto:atrick@apple.com">mailto:atrick@apple.com</a>]<span class="Apple-converted-space"> </span><br><b>Sent:</b><span class="Apple-converted-space"> </span>Tuesday, July 30, 2013 2:20 PM<br><b>To:</b><span class="Apple-converted-space"> </span>Guo, Xiaoyi<br><b>Cc:</b><span class="Apple-converted-space"> </span><a href="mailto:LLVMdev@cs.uiuc.edu">LLVMdev@cs.uiuc.edu</a>; Dan Gohman<br><b>Subject:</b><span class="Apple-converted-space"> </span>Re: [LLVMdev] creating SCEV taking too long<o:p></o:p></span></div></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div><div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">On Jul 29, 2013, at 4:08 PM, Guo, Xiaoyi <<a href="mailto:Xiaoyi.Guo@amd.com" style="color: purple; text-decoration: underline;">Xiaoyi.Guo@amd.com</a>> wrote:<o:p></o:p></div></div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><br><br><o:p></o:p></div><div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Hi,</span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">We have a benchmark where there are 128 MAD computations in a loop. (See the attached IR.) Creating SCEVs for these expressions takes a long time, making the compile time too long. E.g., running opt with the “indvars” pass only takes 45 seconds.</span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">It seems that the majority of the time is spent in comparing the complexity of the expression operands to sort them.</span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">I realize that the expression grows to be really large towards the end of the loop.</span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">I don’t know of all the uses of the built SCEV. But I image it won’t be very useful for such complex expressions. Yet, it’s making the compile time much longer.</span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">So I’m wondering if it would make sense to abort the creation of SCEV when the expression gets really complex and large. Or is there any way to further optimize the performance of SCEV building for such cases.</span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);"> </span><o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><span style="font-size: 11pt; font-family: Calibri, sans-serif; color: rgb(31, 73, 125);">Thanks in advance for any response.</span><o:p></o:p></div></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><p class="MsoNormal" style="margin: 0in 0in 12pt; font-size: 12pt; font-family: 'Times New Roman', serif;">Nice test case. I tried printing the SCEV… oops. I haven’t seen a case this bad, but I know you’re not the first to run into this problem.<o:p></o:p></p></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">There are two steps in GroupByComplexity, sorting (std::sort) and grouping (N^2).<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">The sort calls SCEVComplexityCompare::compare() which can make multiple recursive calls for nodes with multiple operands. This looks like it could be a disaster for expressions that are not strictly trees--exponential in the size of the DAG.<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">If you just have a very large tree with many similar looking subexpressions, then I’m not sure what to do except cut it into reasonable subtrees. <o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">AFAICT, it’s not just sorting that’s a problem but also grouping? Also, I think the shear depth of the createSCEV recursion is itself a problem.<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">I don’t see any reason not to limit the size of SCEV expressions, but I also don’t have a brilliant idea for how to do it at the moment (other than the obvious depth cutoff).<o:p></o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;"><o:p> </o:p></div></div><div><div style="margin: 0in 0in 0.0001pt; font-size: 12pt; font-family: 'Times New Roman', serif;">-Andy</div></div></div></div></blockquote></div><br></body></html>