<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.EmailStyle17
{mso-style-type:personal-reply;
font-family:"Calibri",sans-serif;
color:#1F497D;}
.MsoChpDefault
{mso-style-type:export-only;
font-family:"Calibri",sans-serif;}
@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">It’s true, I am working on this. I have committed the initial patch to add constrained intrinsics for the basic FP operations. This has the desired effect of
preventing optimizations that would violate strict FP semantics with regard to rounding mode and exception status, but it also prevents many safe optimizations. Later this year I’ll be going through the code base and trying to teach various optimizations
to handle the constrained intrinsics safely.<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">Nothing has been added to clang yet to generate these intrinsics, though I have some rough code to do so that’s just waiting for an implementation of the much
harder task of figuring out when such restrictions are needed. If anyone did have a front end that generated these intrinsics, everything is in place to get them through to code generation (though they currently become at least theoretically unrestricted
again after ISel). I have an experimental pass that converts all basic FP operations to the constrained versions and I’ve been able to successfully compile and run real-world FP-using applications with that pass in place.<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">I’m currently working on a patch to add constrained versions of the standard library FP intrinsics (sin, cos, pow, etc.).<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">If anyone is interested in getting pieces of the work-in-progress I’ve mentioned above to test drive, let me know. I’d be happy to share.<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">-Andy<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"><a name="_MailEndCompose"><span style="font-size:11.0pt;font-family:"Calibri",sans-serif;color:#1F497D"><o:p> </o:p></span></a></p>
<p class="MsoNormal"><a name="_____replyseparator"></a><b><span style="font-size:11.0pt;font-family:"Calibri",sans-serif">From:</span></b><span style="font-size:11.0pt;font-family:"Calibri",sans-serif"> Sanjay Patel [mailto:spatel@rotateright.com]
<br>
<b>Sent:</b> Wednesday, March 15, 2017 8:16 AM<br>
<b>To:</b> Samuel Antão <samuelfantao@gmail.com>; Kaylor, Andrew <andrew.kaylor@intel.com><br>
<b>Cc:</b> llvm-dev <llvm-dev@lists.llvm.org>; acjacob@us.ibm.com<br>
<b>Subject:</b> Re: [llvm-dev] Speculative execution of FP divide Instructions - Call-Graph Simplify<o:p></o:p></span></p>
<p class="MsoNormal"><o:p> </o:p></p>
<div>
<div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">> - is there any target for which fp division does not have side effects?
<o:p></o:p></span></p>
</div>
<p class="MsoNormal" style="margin-bottom:12.0pt"><br>
Yes - all of them. This goes back to the fact that clang/llvm do not support changing the FPENV:<br>
<a href="https://bugs.llvm.org/show_bug.cgi?id=8100" target="_blank">https://bugs.llvm.org/show_bug.cgi?id=8100</a><o:p></o:p></p>
</div>
<div>
<p class="MsoNormal" style="margin-bottom:12.0pt">There has been some effort to change that recently, so maybe this is (partly) fixed? (cc'ing Andy Kaylor)<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal">These reports may also provide info / answers / work-arounds:<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><a href="https://bugs.llvm.org/show_bug.cgi?id=6050" target="_blank">https://bugs.llvm.org/show_bug.cgi?id=6050</a><br>
<a href="https://bugs.llvm.org/show_bug.cgi?id=24343">https://bugs.llvm.org/show_bug.cgi?id=24343</a><o:p></o:p></p>
</div>
<div>
<div>
<p class="MsoNormal"><a href="https://bugs.llvm.org/show_bug.cgi?id=24818" target="_blank">https://bugs.llvm.org/show_bug.cgi?id=24818</a><o:p></o:p></p>
<div>
<p class="MsoNormal" style="margin-bottom:12.0pt"><o:p> </o:p></p>
<div>
<p class="MsoNormal">On Wed, Mar 15, 2017 at 3:41 AM, Samuel Antão via llvm-dev <<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a>> wrote:<o:p></o:p></p>
<blockquote style="border:none;border-left:solid #CCCCCC 1.0pt;padding:0in 0in 0in 6.0pt;margin-left:4.8pt;margin-right:0in">
<div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">Hi all,</span><o:p></o:p></p>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"><o:p> </o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">I came across an issue caused by the Call-Graph Simplify Pass. Here is a a small repro:<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"><o:p> </o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">```<o:p></o:p></span></p>
</div>
<div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">define double @foo(double %a1, double %a2, double %a3) #0 {<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">entry:<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> %a_mul = fmul double %a1, %a2<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> %a_cmp = fcmp ogt double %a3, %a_mul<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> br i1 %a_cmp, label %a.then, label %a.end<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"><o:p> </o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">a.then: <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> %a_div = fdiv double %a_mul, %a3<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> br label %a.end<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"><o:p> </o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">a.end:<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> %a_factor = phi double [ %a_div, %a.then ], [ 1.000000e+00, %entry ]<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> ret double %a_factor<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">}<o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">```<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">Here, the conditional is guarding a possible division by zero. However if I run CGSimplify on this I get:<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">```<o:p></o:p></span></p>
</div>
<div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">define double @foo(double %a1, double %a2, double %a3) local_unnamed_addr #0 {<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">entry:<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> %a_mul = fmul double %a1, %a2<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> %a_cmp = fcmp olt double %a_mul, %a3<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> %a_div = fdiv double %a_mul, %a3<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> %a_factor = select i1 %a_cmp, double %a_div, double 1.000000e+00<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> ret double %a_factor<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">}<o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">``` <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">This will cause a FP arithmetic exception, and the application will get a SIGFPE signal. The code that drives the change in the optimizer relies on `llvm::isSafeToSpeculativelyExecute` to decide whether the
division should be performed speculatively. Right now, this function returns true. One could do something similar to integer divisions and add a case there (this solved the issue for me):<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">```<o:p></o:p></span></p>
</div>
<div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">diff --git a/lib/Analysis/ValueTracking.cpp b/lib/Analysis/ValueTracking.cpp<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">index 1761dac..c61fefd 100644<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">--- a/lib/Analysis/ValueTracking.cpp<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+++ b/lib/Analysis/ValueTracking.cpp<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">@@ -3352,6 +3352,21 @@ bool llvm::isSafeToSpeculativelyExecute(const Value *V,<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> // The numerator *might* be MinSignedValue.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> return false;<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> }<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ case Instruction::FDiv:<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ case Instruction::FRem:{<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ const Value *Denominator = Inst->getOperand(1);<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ // x / y is undefined if y == 0<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ // The denominator is not a constant, so there is nothing we can do to prove<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ // it is non-zero.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ if (auto *VV = dyn_cast<ConstantVector>(Denominator))<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ Denominator = VV->getSplatValue();<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ if (!isa<ConstantFP>(Denominator))<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ return false;<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ // The denominator is a zero constant - we can't speculate here.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ if (m_AnyZero().match(Denominator))<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ return false;<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ return true;<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">+ }<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> case Instruction::Load: {<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> const LoadInst *LI = cast<LoadInst>(Inst);<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"> if (!LI->isUnordered() ||<o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">```<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">I did my tests using a powerpc64le target, but I couldn't find any target specific login involved in this transform. In any case, I wanted to drop the questions: <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"><o:p> </o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">- is there any target that would benefit from speculative fp divisions? <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">- is there any target for which fp division does not have side effects? <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"><o:p> </o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">If not, I can go ahead and post the patch above for review.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt"><o:p> </o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">Many thanks!<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal"><span style="font-size:9.5pt">Sam<o:p></o:p></span></p>
</div>
</div>
</div>
<p class="MsoNormal" style="margin-bottom:12.0pt"><br>
_______________________________________________<br>
LLVM Developers mailing list<br>
<a href="mailto:llvm-dev@lists.llvm.org" target="_blank">llvm-dev@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev</a><o:p></o:p></p>
</blockquote>
</div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
</div>
</div>
</div>
</div>
</body>
</html>