<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=us-ascii">
<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:11.0pt;
font-family:"Calibri",sans-serif;}
a:link, span.MsoHyperlink
{mso-style-priority:99;
color:#0563C1;
text-decoration:underline;}
a:visited, span.MsoHyperlinkFollowed
{mso-style-priority:99;
color:#954F72;
text-decoration:underline;}
span.EmailStyle17
{mso-style-type:personal-compose;
font-family:"Calibri",sans-serif;
color:windowtext;}
.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="#0563C1" vlink="#954F72">
<div class="WordSection1">
<p class="MsoNormal">I have a loop with convergent instructions with a loop count of 1024. I use pragma to specify unroll count to be 32. However, the loop was unrolled by 512, which results in very long compilation time.<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">In tryToUnrollLoop, there is<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal"> // If the loop contains a convergent operation, the prelude we'd add<o:p></o:p></p>
<p class="MsoNormal"> // to do the first few instructions before we hit the unrolled loop<o:p></o:p></p>
<p class="MsoNormal"> // is unsafe -- it adds a control-flow dependency to the convergent<o:p></o:p></p>
<p class="MsoNormal"> // operation. Therefore restrict remainder loop (try unrollig without).<o:p></o:p></p>
<p class="MsoNormal"> //<o:p></o:p></p>
<p class="MsoNormal"> // TODO: This is quite conservative. In practice, convergent_op()<o:p></o:p></p>
<p class="MsoNormal"> // is likely to be called unconditionally in the loop. In this<o:p></o:p></p>
<p class="MsoNormal"> // case, the program would be ill-formed (on most architectures)<o:p></o:p></p>
<p class="MsoNormal"> // unless n were the same on all threads in a thread group.<o:p></o:p></p>
<p class="MsoNormal"> // Assuming n is the same on all threads, any kind of unrolling is<o:p></o:p></p>
<p class="MsoNormal"> // safe. But currently llvm's notion of convergence isn't powerful<o:p></o:p></p>
<p class="MsoNormal"> // enough to express this.<o:p></o:p></p>
<p class="MsoNormal"> if (Convergent)<o:p></o:p></p>
<p class="MsoNormal"> UP.AllowRemainder = false;<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">Later in computeUnrollCount, there is<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal"> // 2nd priority is unroll count set by pragma.<o:p></o:p></p>
<p class="MsoNormal"> unsigned PragmaCount = UnrollCountPragmaValue(L);<o:p></o:p></p>
<p class="MsoNormal"> if (PragmaCount > 0) {<o:p></o:p></p>
<p class="MsoNormal"> UP.Count = PragmaCount;<o:p></o:p></p>
<p class="MsoNormal"> UP.Runtime = true;<o:p></o:p></p>
<p class="MsoNormal"> UP.AllowExpensiveTripCount = true;<o:p></o:p></p>
<p class="MsoNormal"> UP.Force = true;<o:p></o:p></p>
<p class="MsoNormal"> if (UP.AllowRemainder &&<o:p></o:p></p>
<p class="MsoNormal"> getUnrolledLoopSize(LoopSize, UP) < PragmaUnrollThreshold)<o:p></o:p></p>
<p class="MsoNormal"> return true;<o:p></o:p></p>
<p class="MsoNormal"> }<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">Because UP.AllowRemainder is false, the unroll count specified by pragma is ignored. Later on, computeUnrollCount calculates an unroll count of 512.<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">Is this a bug? Essentially, this disables unroll count specified by pragma for any loops containing convergent operations, even though the unroll count divides the trip count.<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">Thanks.<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal">Sam<o:p></o:p></p>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
</body>
</html>