<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:w="urn:schemas-microsoft-com:office:word" xmlns:m="http://schemas.microsoft.com/office/2004/12/omml" xmlns="http://www.w3.org/TR/REC-html40"><head><meta http-equiv=Content-Type content="text/html; charset=utf-8"><meta name=Generator content="Microsoft Word 15 (filtered medium)"><style><!--
/* Font Definitions */
@font-face
{font-family:"Cambria Math";
panose-1:2 4 5 3 5 4 6 3 2 4;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0in;
margin-bottom:.0001pt;
font-size:12.0pt;
font-family:"Times New Roman",serif;}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:blue;
text-decoration:underline;}
a:visited, span.MsoHyperlinkFollowed
{mso-style-priority:99;
color:purple;
text-decoration:underline;}
span.apple-converted-space
{mso-style-name:apple-converted-space;}
span.EmailStyle18
{mso-style-type:personal;
font-family:"Calibri",sans-serif;
color:#1F497D;}
span.EmailStyle19
{mso-style-type:personal;
font-family:"Calibri",sans-serif;
color:#1F497D;}
span.EmailStyle20
{mso-style-type:personal-reply;
font-family:"Calibri",sans-serif;
color:#1F497D;}
.MsoChpDefault
{mso-style-type:export-only;
font-size:10.0pt;}
@page WordSection1
{size:8.5in 11.0in;
margin:1.0in 1.0in 1.0in 1.0in;}
div.WordSection1
{page:WordSection1;}
--></style><!--[if gte mso 9]><xml>
<o:shapedefaults v:ext="edit" spidmax="1026" />
</xml><![endif]--><!--[if gte mso 9]><xml>
<o:shapelayout v:ext="edit">
<o:idmap v:ext="edit" data="1" />
</o:shapelayout></xml><![endif]--></head><body lang=EN-US link=blue vlink=purple><div class=WordSection1><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Hi,<o:p></o:p></span></p><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'><o:p> </o:p></span></p><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Can the owner for SROA please take a look into the attached patch?<o:p></o:p></span></p><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span><o:p></o:p></p></div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Also if you are fine with the patch could you please commit it since I do not have commit rights?</span><o:p></o:p></p></div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span><o:p></o:p></p></div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Thanks,</span><o:p></o:p></p></div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Mandeep</span><o:p></o:p></p></div><div><div style='border:none;border-top:solid #E1E1E1 1.0pt;padding:3.0pt 0in 0in 0in'><div><p class=MsoNormal><b><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>From:</span></b><span class=apple-converted-space><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span></span><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Owen Anderson [<a href="mailto:resistor@mac.com">mailto:resistor@mac.com</a>]<span class=apple-converted-space> </span><br><b>Sent:</b><span class=apple-converted-space> </span>Wednesday, January 21, 2015 8:18 PM<br><b>To:</b><span class=apple-converted-space> </span>Mandeep Singh Grang<br><b>Cc:</b><span class=apple-converted-space> </span>llvm-commits; <a href="mailto:apazos@codeaurora.org">apazos@codeaurora.org</a><br><b>Subject:</b><span class=apple-converted-space> </span>Re: Question about commit d4748bbd497b550a4e5db246c6708fcd6de542da</span><o:p></o:p></p></div></div></div><div><p class=MsoNormal> <o:p></o:p></p></div><div><p class=MsoNormal>Ok, patches welcome. :-)<o:p></o:p></p></div><div><div><p class=MsoNormal> <o:p></o:p></p></div></div><div><div><p class=MsoNormal>—Owen<o:p></o:p></p></div></div><div><div><p class=MsoNormal> <o:p></o:p></p></div><div><blockquote style='margin-top:5.0pt;margin-bottom:5.0pt'><div><div><p class=MsoNormal>On Jan 21, 2015, at 8:00 PM, Mandeep Singh Grang <<a href="mailto:mgrang@codeaurora.org"><span style='color:purple'>mgrang@codeaurora.org</span></a>> wrote:<o:p></o:p></p></div></div><div><p class=MsoNormal> <o:p></o:p></p></div><div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Yes, in the scenario where we have an intrinsic and the following condition is<span class=apple-converted-space> </span><b>true</b>:</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>else if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(U->getUser()))</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>But the following condition in<span class=apple-converted-space> </span><b>false</b>:</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>if (II->getIntrinsicID() != Intrinsic::lifetime_start &</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> II->getIntrinsicID() != Intrinsic::lifetime_end)</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>In this case, all other subsequent else-if’s are skipped and the logic falls through to the “return true” and we see a lot of “ins” instructions being generated.</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>But I think it should evaluate other else if’s since some of them may meet the condition and we may return false.</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>We saw degradation in AArch64 cortex-a57 code due to this.</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>We modified the code on our end to the following:</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>IntrinsicInst *II = dyn_cast<IntrinsicInst>(U->getUser());</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'><start of if else clauses></span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>…</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>else if (II && II->getIntrinsicID() != Intrinsic::lifetime_start &</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> II->getIntrinsicID() != Intrinsic::lifetime_end)</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'><other else if’s></span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Making this change get us back the lost performance.</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Thanks,</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D'>Mandeep</span><o:p></o:p></p></div></div><div><div style='border:none;border-top:solid #E1E1E1 1.0pt;padding:3.0pt 0in 0in 0in'><div><div><p class=MsoNormal><b><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>From:</span></b><span class=apple-converted-space><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span></span><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Owen Anderson [<a href="mailto:resistor@mac.com"><span style='color:purple'>mailto:resistor@mac.com</span></a>]<span class=apple-converted-space> </span><br><b>Sent:</b><span class=apple-converted-space> </span>Wednesday, January 21, 2015 6:30 PM<br><b>To:</b><span class=apple-converted-space> </span>Mandeep Singh Grang<br><b>Cc:</b><span class=apple-converted-space> </span>llvm-commits;<span class=apple-converted-space> </span><a href="mailto:apazos@codeaurora.org"><span style='color:purple'>apazos@codeaurora.org</span></a><br><b>Subject:</b><span class=apple-converted-space> </span>Re: Question about commit d4748bbd497b550a4e5db246c6708fcd6de542da</span><o:p></o:p></p></div></div></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span><o:p></o:p></p></div></div><div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Yes, it is intentional. Do you have a counter example where it results in undesirable behavior?</span><o:p></o:p></p></div></div></div><div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span><o:p></o:p></p></div></div></div><div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>-Owen</span><o:p></o:p></p></div></div></div><div><p class=MsoNormal style='margin-bottom:12.0pt'><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'><br>On Jan 21, 2015, at 6:04 PM, Mandeep Singh Grang <<a href="mailto:mgrang@codeaurora.org"><span style='color:#954F72'>mgrang@codeaurora.org</span></a>> wrote:</span><o:p></o:p></p></div><blockquote style='margin-top:5.0pt;margin-bottom:5.0pt'><div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Hi Owen,</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>In the following commit, we observed that with your else-if condition all the subsequent else-if’s are skipped.</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Could you please confirm whether this was your intent?</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>commit d4748bbd497b550a4e5db246c6708fcd6de542da</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Author: Owen Anderson <<a href="mailto:resistor@mac.com"><span style='color:#954F72'>resistor@mac.com</span></a>></span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Date: Thu Aug 7 21:07:35 2014 +0000</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> Fix a case in SROA where lifetime intrinsics could inhibit alloca promotion. In</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> this case, the code path dealing with vector promotion was missing the explicit</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> checks for lifetime intrinsics that were present on the corresponding integer</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> promotion path.</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>+++ b/lib/Transforms/Scalar/SROA.cpp</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>@@ -1659,6 +1659,10 @@ static bool isVectorPromotionViableForSlice(</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> return false;</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> if (!I->isSplittable())</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> return false; // Skip any unsplittable intrinsics.</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>+ }<span class=apple-converted-space> </span><b>else if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(U->getUser())) {</b></span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><b><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>+ if (II->getIntrinsicID() != Intrinsic::lifetime_start &&</span></i></b><o:p></o:p></p></div></div><div><div><p class=MsoNormal><b><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>+ II->getIntrinsicID() != Intrinsic::lifetime_end)</span></i></b><o:p></o:p></p></div></div><div><div><p class=MsoNormal><b><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>+ return false;</span></i></b><o:p></o:p></p></div></div><div><div><p class=MsoNormal><b><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> }<span class=apple-converted-space> </span></span></i></b><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>else if (U->get()->getType()->getPointerElementType()->isStructTy()) {</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> // Disable vector promotion when there are loads or stores of an FCA.</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><i><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> return false;</span></i><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'> </span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Thanks,</span><o:p></o:p></p></div></div><div><div><p class=MsoNormal><span style='font-size:11.0pt;font-family:"Calibri",sans-serif'>Mandeep</span><o:p></o:p></p></div></div></div></blockquote></div></blockquote></div><div><p class=MsoNormal> <o:p></o:p></p></div></div><p class=MsoNormal><Fix_SROA_for_intrinsics.patch><o:p></o:p></p><p class=MsoNormal><o:p> </o:p></p></div></body></html>