<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 14 (filtered medium)">
<style><!--
/* Font Definitions */
@font-face
{font-family:Wingdings;
panose-1:5 0 0 0 0 0 0 0 0 0;}
@font-face
{font-family:Wingdings;
panose-1:5 0 0 0 0 0 0 0 0 0;}
@font-face
{font-family:Calibri;
panose-1:2 15 5 2 2 2 4 3 2 4;}
@font-face
{font-family:Tahoma;
panose-1:2 11 6 4 3 5 4 4 2 4;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{margin:0cm;
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;}
p
{mso-style-priority:99;
mso-margin-top-alt:auto;
margin-right:0cm;
mso-margin-bottom-alt:auto;
margin-left:0cm;
font-size:12.0pt;
font-family:"Times New Roman","serif";}
p.MsoAcetate, li.MsoAcetate, div.MsoAcetate
{mso-style-priority:99;
mso-style-link:"Balloon Text Char";
margin:0cm;
margin-bottom:.0001pt;
font-size:8.0pt;
font-family:"Tahoma","sans-serif";}
span.BalloonTextChar
{mso-style-name:"Balloon Text Char";
mso-style-priority:99;
mso-style-link:"Balloon Text";
font-family:"Tahoma","sans-serif";}
span.EmailStyle20
{mso-style-type:personal;
font-family:"Calibri","sans-serif";
color:#1F497D;}
span.EmailStyle21
{mso-style-type:personal;
font-family:"Calibri","sans-serif";
color:#1F497D;}
span.EmailStyle22
{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:612.0pt 792.0pt;
margin:72.0pt 72.0pt 72.0pt 72.0pt;}
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 Arnaud,<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">Great, I committed the assert improvement as r229400.<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">/Jonas<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"><o:p> </o:p></span></p>
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Arnaud A. de Grandmaison [mailto:arnaud.degrandmaison@arm.com]
<br>
<b>Sent:</b> den 16 februari 2015 11:11<br>
<b>To:</b> Jonas Paulsson; Lang Hames<br>
<b>Cc:</b> Commit Messages and Patches for LLVM<br>
<b>Subject:</b> RE: [PBQP] setEdgeCosts() (was PBQP bugfix)<o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal"><o:p> </o:p></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Hi Jonas,<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">The assert is all about making sure we had choice when spilling a node --- but it does not attempt to question how good the choice is. I think
your proposal to enhance the assert to all nodes which have gone thru the AllocatableNodes set is good to have.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Regarding the second point, which is about what to do when several options for a node have the same minimal cost, the situation is a bit more complicated.
The approach we have today is a bit too simple, but I do not think always preferring not to spill is a better approach. I think the right approach would be, in case of ties, to ensure the selection for this node will not prevent a good selection in the neighbouring
nodes, i.e. a local decision at the node should not prevent global cost improvement. This can be handled either at the reduction stage (improve reduction order), or at the backpropagate stage (do some look-ahead). FYI, this is a problem we already have, without
even speaking of spilling nodes.<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">--<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud<o:p></o:p></span></p>
<p class="MsoNormal"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"><o:p> </o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Jonas Paulsson [<a href="mailto:jonas.paulsson@ericsson.com">mailto:jonas.paulsson@ericsson.com</a>]
<br>
<b>Sent:</b> 16 February 2015 10:25<br>
<b>To:</b> Lang Hames; Arnaud De Grandmaison<br>
<b>Cc:</b> Commit Messages and Patches for LLVM<br>
<b>Subject:</b> RE: [PBQP] setEdgeCosts() (was PBQP bugfix)<o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal"><span lang="EN-GB"><o:p> </o:p></span></p>
<p class="MsoNormal"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Hi Lang and Arnaud,<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">What about the case where all cost elements are equal and non-infinite – would it then make sense to prefer a register allocation over spilling? Right now the
node always gets spilled in that case and I am just curious if this is a decision that has been taken?<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">/Jonas
<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"><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"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Lang Hames [<a href="mailto:lhames@gmail.com">mailto:lhames@gmail.com</a>]
<br>
<b>Sent:</b> den 13 februari 2015 19:19<br>
<b>To:</b> Arnaud A. de Grandmaison<br>
<b>Cc:</b> Jonas Paulsson; Commit Messages and Patches for LLVM<br>
<b>Subject:</b> Re: [PBQP] setEdgeCosts() (was PBQP bugfix)<o:p></o:p></span></p>
<p class="MsoNormal"><o:p> </o:p></p>
<div>
<p class="MsoNormal">Hi Jonas, Arnaud,<o:p></o:p></p>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">That assert sounds good to me: If this node was conservatively allocatable then there must be some finite-cost register option available during back-propagation.<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
<div>
<p class="MsoNormal">Cheers,<o:p></o:p></p>
</div>
<div>
<p class="MsoNormal">Lang.<o:p></o:p></p>
</div>
</div>
<div>
<p class="MsoNormal"><o:p> </o:p></p>
<div>
<p class="MsoNormal">On Fri, Feb 13, 2015 at 4:27 AM, Arnaud A. de Grandmaison <<a href="mailto:arnaud.degrandmaison@arm.com" target="_blank">arnaud.degrandmaison@arm.com</a>> wrote:<o:p></o:p></p>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I removed the assert with r229103 (for now). The condition for the assert should indeed
be that the node is spilled because there was no register available. I still have to think where is the best place to have this assert, but it seems to me this should be in backpropagate. Something like (in pseudo code):</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">If (G.getNodeMetaData(NId).isConservativelyAllocatable()) {</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> assert(v.getLength() > 1 && “node should have some register options”) ;</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> assert(!v[1 .. v.getLenght()-1].hasOnlyInfinities() && “A spillable only node is not
conservatively allocatable”);</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">}</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">
<a href="mailto:llvm-commits-bounces@cs.uiuc.edu" target="_blank">llvm-commits-bounces@cs.uiuc.edu</a> [mailto:<a href="mailto:llvm-commits-bounces@cs.uiuc.edu" target="_blank">llvm-commits-bounces@cs.uiuc.edu</a>]
<b>On Behalf Of </b>Arnaud A. de Grandmaison<br>
<b>Sent:</b> 13 February 2015 11:04<br>
<b>To:</b> 'Jonas Paulsson'; Lang Hames</span><span lang="EN-GB"><o:p></o:p></span></p>
<div>
<div>
<p class="MsoNormal"><span lang="EN-GB"><br>
<b>Cc:</b> Commit Messages and Patches for LLVM<br>
<b>Subject:</b> RE: [PBQP] setEdgeCosts() (was PBQP bugfix)<o:p></o:p></span></p>
</div>
</div>
</div>
</div>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I’ll remove the assert until we can come up with a finer grained assert.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Jonas Paulsson [<a href="mailto:jonas.paulsson@ericsson.com" target="_blank">mailto:jonas.paulsson@ericsson.com</a>]
<br>
<b>Sent:</b> 13 February 2015 08:41<br>
<b>To:</b> Lang Hames<br>
<b>Cc:</b> Arnaud De Grandmaison; Commit Messages and Patches for LLVM<br>
<b>Subject:</b> RE: [PBQP] setEdgeCosts() (was PBQP bugfix)</span><span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Hi,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I’m thinking we could have an assert that a spilled conservatively node isn’t a “forced spill”. It
isn’t wrong to choose the spill option per what you just wrote, but it is wrong if there are no registers available, right?</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">/Jonas</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Lang Hames [<a href="mailto:lhames@gmail.com" target="_blank">mailto:lhames@gmail.com</a>]
<br>
<b>Sent:</b> den 12 februari 2015 19:46<br>
<b>To:</b> Jonas Paulsson<br>
<b>Cc:</b> Arnaud A. de Grandmaison; Commit Messages and Patches for LLVM<br>
<b>Subject:</b> Re: [PBQP] setEdgeCosts() (was PBQP bugfix)</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Hi Jonas,<span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">You're right: Just because a node is conservatively allocatable doesn't mean it must be allocated a register. I hadn't thought this assert all the way through. ;)<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">As an example of where spilling may be cheaper: imagine the solver has proved (via some chain of R0/RI/RII reductions) that allocating a register to this node will cause a more
expensive node to be spilled. In that case it's better to spill this node and leave the register available for the node further down the stack.<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Arnaud - do you mind removing your assert?<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Cheers,<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Lang.<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">On Wed, Feb 11, 2015 at 10:13 AM, Jonas Paulsson <<a href="mailto:jonas.paulsson@ericsson.com" target="_blank">jonas.paulsson@ericsson.com</a>> wrote:<span lang="EN-GB"><o:p></o:p></span></p>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Hi Arnaud,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Your patch worked great for me, thank you.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I however ran into a test case which triggered the assert. It turns out that the spill cost and all
other options were equal and non-infinite. minIndex() then returned 0 - the spill-option, and so a conservatively allocable node got spilled even though it did not have to.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I am now wondering if it could actually be that it is ok to spill a conservatively allocable node
if that is cheaper than selecting a register option? Or perhaps it is an error to spill if a register is available during the back propagation?</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">/Jonas
</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Arnaud A. de Grandmaison [mailto:<a href="mailto:arnaud.degrandmaison@arm.com" target="_blank">arnaud.degrandmaison@arm.com</a>]
<br>
<b>Sent:</b> den 11 februari 2015 10:13<br>
<b>To:</b> 'Lang Hames'; Jonas Paulsson</span><span lang="EN-GB"><o:p></o:p></span></p>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><br>
<b>Cc:</b> Commit Messages and Patches for LLVM<br>
<b>Subject:</b> RE: [PBQP] setEdgeCosts() (was PBQP bugfix)<span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
</div>
</div>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Committed (with the change to isSpillable) @ r228816</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Jonas, you may want to give it a try and see if it fixes the issue you had with your
target.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">--</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Lang Hames [<a href="mailto:lhames@gmail.com" target="_blank">mailto:lhames@gmail.com</a>]
<br>
<b>Sent:</b> 10 February 2015 20:06<br>
<b>To:</b> Arnaud De Grandmaison<br>
<b>Cc:</b> Commit Messages and Patches for LLVM<br>
<b>Subject:</b> Re: [PBQP] setEdgeCosts() (was PBQP bugfix)</span><span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Hi Arnaud,<o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">This looks good to me, with one small change: Since OptimallyReducible nodes can still be spilled, isSpillable should be:<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-family:"Courier New"">bool isSpillable() const {</span><span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-family:"Courier New"">return RS == NotProvablyAllocatable || RS == OptimallyReducible; }</span><span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Cheers,<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Lang.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;margin-bottom:12.0pt"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
</div>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">On Tue, Feb 10, 2015 at 9:12 AM, Arnaud A. de Grandmaison <<a href="mailto:arnaud.degrandmaison@arm.com" target="_blank">arnaud.degrandmaison@arm.com</a>> wrote:<o:p></o:p></span></p>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">This is an updated version of the patch, where an assertion now enforces that a node
never gets downgraded.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I think I can commit it as there are now assertions checking all invariants (only NotProvablyAllocatable
nodes can be spilled + a node’s ReductionState cannot be downgraded).</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">--</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">
<a href="mailto:llvm-commits-bounces@cs.uiuc.edu" target="_blank">llvm-commits-bounces@cs.uiuc.edu</a> [mailto:<a href="mailto:llvm-commits-bounces@cs.uiuc.edu" target="_blank">llvm-commits-bounces@cs.uiuc.edu</a>]
<b>On Behalf Of </b>Arnaud A. de Grandmaison<br>
<b>Sent:</b> 09 February 2015 23:44<br>
<b>To:</b> 'Lang Hames'<br>
<b>Cc:</b> Commit Messages and Patches for LLVM<br>
<b>Subject:</b> RE: [PBQP] setEdgeCosts() (was PBQP bugfix)</span><span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Given that nodes can never go back, why not just checking the NodeMetadata as in the
attached updated patch (see NodeMetadata::isSpillable()): only nodes from the NotProvablyAllocatable set can be spilled. Anyother case from the ReductionState enum is an error.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Now, my sentence above starts with ‘given’… Yet another assertion to check
</span><span lang="EN-GB" style="font-size:11.0pt;font-family:Wingdings;color:#1F497D">J</span><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">. Or not, as this is how the algorithm is written, and checking this
is a bit of an overkill (I think). It could however be worth adding a comment to promote / moveToConservativelyAllocatableNodes & moveToOptimallyReducibleNodes to document this assumption.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">--</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Lang Hames [<a href="mailto:lhames@gmail.com" target="_blank">mailto:lhames@gmail.com</a>]
<br>
<b>Sent:</b> 09 February 2015 19:47<br>
<b>To:</b> Arnaud De Grandmaison<br>
<b>Cc:</b> Jonas Paulsson; Commit Messages and Patches for LLVM<br>
<b>Subject:</b> Re: [PBQP] setEdgeCosts() (was PBQP bugfix)</span><span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Hi Arnaud,<o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">About the assert - sorry I haven't applied that yet. When I was thinking of shuffling nodes back and forward between sets it seemed expensive: The verification
set would have to be updated too, which would be expensive and not particularly helpful for readability. That's no longer an issue (since your patch never moves things back to less-allocatable sets), but I think we can still do better with the assert: We can
expand the reduction stack to hold a NodeID plus an enum that tells us how the node was reduced (i.e. this was pushed to the stack from the conservatively-allocatable set). Then, when we rebuild the graph and assign options we can assert that conservatively-allocatable
nodes are never spilled. How does that sound?<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Cheers,<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Lang.<o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">On Mon, Feb 9, 2015 at 10:07 AM, Arnaud A. de Grandmaison <<a href="mailto:arnaud.degrandmaison@arm.com" target="_blank">arnaud.degrandmaison@arm.com</a>> wrote:<o:p></o:p></span></p>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I had nothing specific in mind, I am just not ruling out what can happen if the reduction
rules get changed.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I think it is good to check that no spilled node comes from the OptimalyReducible or
ConservativelyAllocatable sets though, as this would indicate something failed.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I’ll update my patch later today.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">--</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Lang Hames [mailto:<a href="mailto:lhames@gmail.com" target="_blank">lhames@gmail.com</a>]
<br>
<b>Sent:</b> 09 February 2015 18:28</span><span lang="EN-GB"><o:p></o:p></span></p>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"><br>
<b>To:</b> Arnaud De Grandmaison<br>
<b>Cc:</b> Jonas Paulsson; Commit Messages and Patches for LLVM<br>
<b>Subject:</b> Re: [PBQP] setEdgeCosts() (was PBQP bugfix)<o:p></o:p></span></p>
</div>
</div>
</div>
</div>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Hi Arnaud,<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">I'm not sure which other effects you had in mind, but in the back of my mind I was still thinking about mutable PBQP graphs. If we had those then we'd need to
move nodes back out of the optimally-reducible/conservatively-allocatable sets when spill code is inserted. Since we don't have mutable graphs though, I think what you're doing seems safe. I'll run it over the test-suite on X86 when I get to the office.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Cheers,<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Lang.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"><br>
Sent from my iPad<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;margin-bottom:12.0pt"><span lang="EN-GB"><br>
On Feb 9, 2015, at 8:14 AM, Arnaud A. de Grandmaison <<a href="mailto:arnaud.degrandmaison@arm.com" target="_blank">arnaud.degrandmaison@arm.com</a>> wrote:<o:p></o:p></span></p>
</div>
<blockquote style="margin-top:5.0pt;margin-bottom:5.0pt">
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Will do.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">My patch only prevents “transient” effects, but does not catch other cases, which should
arguably not happen.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Lang Hames [<a href="mailto:lhames@gmail.com" target="_blank">mailto:lhames@gmail.com</a>]
<br>
<b>Sent:</b> 09 February 2015 17:00<br>
<b>To:</b> Arnaud De Grandmaison<br>
<b>Cc:</b> Jonas Paulsson; Commit Messages and Patches for LLVM<br>
<b>Subject:</b> Re: [PBQP] setEdgeCosts() (was PBQP bugfix)</span><span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Hi Arnaud,<o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">I was about to teach handleReconnectEdge to move things back NotConservativelyAllocatable (and from OptimallyReducible back up too). Your solution is much nicer.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">Please commit away.<o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">- Lang.<o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB">On Mon, Feb 9, 2015 at 4:58 AM, Arnaud A. de Grandmaison <<a href="mailto:arnaud.degrandmaison@arm.com" target="_blank">arnaud.degrandmaison@arm.com</a>> wrote:<o:p></o:p></span></p>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Hi Jonas,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">I gave a look at this over the week-end, and I believe the attached patch can fix the
issue you described. </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">The patch modifies handleSetCosts so that it only promotes the nodes to the OptimallyReducible
or ConservayivelyAllocatable sets once the metadata have been fully updated. And while there, I renamed handleSetCosts to handleUpdateCosts to better reflect what the method does.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Cheers,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">--</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D">Arnaud</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB" style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<div style="border:none;border-left:solid blue 1.5pt;padding:0cm 0cm 0cm 4.0pt">
<div>
<div style="border:none;border-top:solid #B5C4DF 1.0pt;padding:3.0pt 0cm 0cm 0cm">
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Jonas Paulsson [mailto:<a href="mailto:jonas.paulsson@ericsson.com" target="_blank">jonas.paulsson@ericsson.com</a>]
<br>
<b>Sent:</b> 06 February 2015 12:22<br>
<b>To:</b> Lang Hames; Arnaud De Grandmaison<br>
<b>Cc:</b> <a href="mailto:llvm-commits@cs.uiuc.edu" target="_blank">llvm-commits@cs.uiuc.edu</a><br>
<b>Subject:</b> [PBQP] setEdgeCosts() (was PBQP bugfix)</span><span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Hi Lang,<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">The patch seems good to me on my target as well, thank you.<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">However, the bug#3 I wrote about earlier is now still merely hidden. The problem is that setEdgeCosts() temporarily disconnects two nodes, moves them to ConservativelyAllocatableNodes
because their DeniedOpts have now been decreased by one, and then reconnects them. DeniedOpts gets increased again during reconnection, and the isConservativelyAllocatable() test would then fail, but the nodes have already been moved.<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">I resend the test case again along with reverts of patches that hide the bug (for the test case).<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">To investigate:<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Revert the two patches that hide the bug (see below), or just apply the attached patches that do the same thing. (To see that there is an error, also apply the patch with an assert
that no nodes in coservatively allocatables gets the spill option).<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Then run llc with the attached test case:<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">llc pbqp_reduced.ll -mtriple=aarch64-none-linux-gnu -mcpu=cortex-a57 -mattr=+neon -optimize-regalloc -regalloc=pbqp<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">(The assertion triggers, because Node 2 was in conservatively allocatables and yet spilled.)<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">The error is made *<b>during applyR2() of node 18</b>*.<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Debug dumps (patch for debug dumps not provided, sent earlier):<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">handleDisconnectEdge(9, 2) : DeniedOpts 10 -> 9<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">NId 9(%vreg15, GPR64common) moved to conservatively-allocatables.<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">handleDisconnectEdge(2, 9) : DeniedOpts 10 -> 9<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">NId 2(%vreg4, GPR64common) moved to conservatively-allocatables.<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Popped NId 2(%vreg4, GPR64common) , all edge costs added:<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">2.002748e+01 inf inf inf inf inf inf inf inf inf inf ** selection: 0<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">/Jonas<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:black"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">Attachments:</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">commit 4a44259e80c04c340349ed8f32bcbc8ffdb89b52</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">Author: Jonas Paulsson <<a href="mailto:jonas.paulsson@ericsson.com" target="_blank">jonas.paulsson@ericsson.com</a>></span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">Date: Fri Feb 6 10:34:35 2015 +0100</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> Revert "[PBQP Regalloc] Pre-spill vregs that have no legal physregs."</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">
</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> This reverts commit d54450ef63998effdb19476c4e4c6c3f0a8c5f50.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">commit 9ada754d86920eba6cbda342e643e235788d139b</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">Author: Jonas Paulsson <<a href="mailto:jonas.paulsson@ericsson.com" target="_blank">jonas.paulsson@ericsson.com</a>></span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">Date: Fri Feb 6 10:30:38 2015 +0100</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> Revert "[PBQP] Fix transposed worst row/column check in handleAdd/RemoveNode in the PBQP"</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">
</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> This reverts commit 4bde7909b4eb2aba2cd62b79d8cca98d9af6692e.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">commit 45b838cc499a6af384c43fbc2d8c83c17b48f25e</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">Author: Jonas Paulsson <<a href="mailto:jonas.paulsson@ericsson.com" target="_blank">jonas.paulsson@ericsson.com</a>></span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">Date: Wed Jan 28 08:30:35 2015 +0100</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> Assert in PBQP that a node that selects '0' (spilled) was not pushed on node</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> stack as conservatively allocatable.</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:black">Test case: pbqp_reduced.ll</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="font-size:11.0pt;font-family:"Calibri","sans-serif";color:#1F497D"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif"">From:</span></b><span style="font-size:10.0pt;font-family:"Tahoma","sans-serif""> Lang Hames [<a href="mailto:lhames@gmail.com" target="_blank">mailto:lhames@gmail.com</a>]
<br>
<b>Sent:</b> den 3 februari 2015 07:19<br>
<b>To:</b> Jonas Paulsson; Arnaud A. de Grandmaison<br>
<b>Cc:</b> <a href="mailto:llvm-commits@cs.uiuc.edu" target="_blank">llvm-commits@cs.uiuc.edu</a><br>
<b>Subject:</b> Re: PBQP bugfix</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Hi Jonas,<span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Apologies for the delay - I've committed my fix for this in r227942.<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">I haven't had a chance to try it out on your ARM test case yet, but hope to soon. Please let me know if it doesn't fix the problem for your out-of-tree target.<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Cheers,<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Lang.<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">On Fri, Jan 30, 2015 at 11:10 AM, Lang Hames <<a href="mailto:lhames@gmail.com" target="_blank">lhames@gmail.com</a>> wrote:<span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Hi Jonas,<span lang="EN-GB"><o:p></o:p></span></p>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Thanks very much for tracking this down!<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">I think the best approach to fixing this is to have the PBQP register allocator recognize these nodes early and never represent them in the graph. Any matrix that has only one column
must be connected to a node that has only one solution, and this should be allocated up-front to simplify the solver's job.<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">I'll have a fix for this later today.<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Cheers,<span lang="EN-GB"><o:p></o:p></span></p>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">Lang.<span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<div>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">On Thu, Jan 29, 2015 at 9:25 AM, Jonas Paulsson <<a href="mailto:jonas.paulsson@ericsson.com" target="_blank">jonas.paulsson@ericsson.com</a>> wrote:<span lang="EN-GB"><o:p></o:p></span></p>
</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>
<div>
<div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="SV">Hi,</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="SV"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">This is a bugfix for the PBQP register allocator. The case for a one-column matrix must be handled in the MatrixMetadata constructor.<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto">See commit message for further explanation,<span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:#888888"> </span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:#888888">Jonas Paulsson</span><span lang="EN-GB"><o:p></o:p></span></p>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span style="color:#888888"> </span><span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;margin-bottom:12.0pt">_______________________________________________<br>
llvm-commits mailing list<br>
<a href="mailto:llvm-commits@cs.uiuc.edu" target="_blank">llvm-commits@cs.uiuc.edu</a><br>
<a href="http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits" target="_blank">http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits</a><span lang="EN-GB"><o:p></o:p></span></p>
</blockquote>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
</div>
</div>
</blockquote>
</div>
</div>
</div>
</div>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"><span lang="EN-GB"> <o:p></o:p></span></p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<p class="MsoNormal" style="mso-margin-top-alt:auto;mso-margin-bottom-alt:auto"> <span lang="EN-GB"><o:p></o:p></span></p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<p class="MsoNormal"><o:p> </o:p></p>
</div>
</div>
</div>
</body>
</html>