<html><head><style type="text/css"><!-- DIV {margin:0px;} --></style></head><body><div style="font-family:times new roman,new york,times,serif;font-size:10pt">
<div style="font-family: times new roman,new york,times,serif; font-size: 10pt;">
<div style="font-family: times new roman,new york,times,serif; font-size: 10pt;">
<div style="font-family: times new roman,new york,times,serif; font-size: 10pt;">
<div style="font-family: times new roman,new york,times,serif; font-size: 12pt;">
<div style="font-family: times new roman,new york,times,serif; font-size: 12pt;"><div>Hello Mr Julian Hall,<br><br>We are pleased to see you on this position.<br><br>Your ideas about Clp acceleration on GPGPU's seem challenging and interesting.<br>But lack of documentation makes it very difficult and slow to be
integrated into the development process. And the source code itself is
hard to understand.<br><br>I have got experience of parallel systems implementation and GPGPU programming and I would like join to CLP acceleration.<br>I used to try to do some experiments by myself, but the lack of documentation describing mathematical details and source code organization did not allow to make CLP to run really in parallel, I implemented on GPGPU just some low-level functions with matrix factorizations.<br><br>I have got the documentation which is refereed in the Clp source
code:<br> <font size="2"><span style="font-style: italic;"><br style="font-style: italic;"></span><span style="font-style: italic;"> Forrest and
Tomlin, Implementing the <span class="yshortcuts" id="lw_1265726399_2">simplex method</span> for</span><br style="font-style: italic;"><span style="font-style: italic;"> the Optimization Subroutine
Library - <span class="yshortcuts" id="lw_1265726399_3">IBM Systems Journal</span> 1992</span><br><br style="font-style: italic;"><span style="font-style: italic;"> Gill, Murray, Saunders, Wright A Practical Anti-Cycling</span><br style="font-style: italic;"><span style="font-style: italic;"> Procedure for <span class="yshortcuts" id="lw_1265726399_4">Linear and Nonlinear Programming</span> SOL report 1988</span></font><br><br>But that documentation mainly discourse numerical issues and corner cases.<br><br>So if some development specification or other document about the source code organization would be available, I believe many strong guys would be interested to become a part of Clp acceleration process.<br><br>Best Regards,<br>Armen Nahapetyan</div><div style="font-family: times new roman,new york,times,serif; font-size: 12pt;"><br><div style="font-family: arial,helvetica,sans-serif;
font-size: 13px;"><font face="Tahoma" size="2"><hr size="1"><b><span style="font-weight: bold;">From:</span></b> Julian Hall <J.A.J.Hall@ed.ac.uk><br><b><span style="font-weight: bold;">To:</span></b> clp@list.coin-or.org<br><b><span style="font-weight: bold;">Sent:</span></b> Sat, February 6, 2010 3:27:57
AM<br><b><span style="font-weight: bold;">Subject:</span></b> [Clp] Clp - Under new management<br></font><br>
Dear All,<br><br>As John announced earlier in the week, I have taken over as Project<br>Manager for Clp. This is an exciting opportunity to contribute to the <br>premier open-source LP solver.<br><br>First of all I'd like to express my appreciation for all that John has <br>contributed to computational linear optimization in general, and <br>COIN-OR in particular.<br><br>Taking over a necessarily large and complex code such as Clp is a <br>significant challenge and, until I get a reasonable mastery of it, I <br>cannot pretend that (any) bugs will be fixed with John's alacrity. In <br>this, and other aspects of Clp management, I will be aided by Lou <br>Hafer and Matthew Saltzman.<br><br>As for code development, in the short to medium term, my aim will be <br>to work with Ted Ralphs and the Cbc team to achieve a clear <br>distinction between Cbc and Clp. This will allow the two projects to <br>be
managed with greater independence, but with no loss in performance.<br><br>Beyond that, given that I have my own simplex solver that is <br>competitive with Clp, there may well be some computational techniques <br>that I can introduce to Clp to improve its performance.<br><br>In the longer term, I would expect to develop Clp so that it can <br>exploit multi-core and GPGPU architectures - but don't expect linear <br>speed-up!<br><br>These are merely my ideas: I'll be keen to hear from users of <br>developments that they would like to see.<br><br>Since I'm new to the COIN community (and am doubtless unknown to most<br>recipients of this message) let me close by introducing myself!<br><br>Having been in Edinburgh since 1990, when I completed my PhD at Dundee <br>under Roger Fletcher, I've worked almost exclusively on computational <br>techniques for<br>the simplex method. I have developed my own high performance
serial <br>industrial-strength simplex solver that has been the basis of <br>consultancy work and underpinned some very labour-intensive work on <br>parallel simplex. These are some of the reasons why I haven't <br>published much---although I did win the 2005 COAP prize for a paper on <br>"hyper-sparsity". Note that the Edinburgh group also includes <br>computational IPM experts Jacek Gondzio and Andreas Grothey, so we <br>cover high performance solution of LP/QP problems pretty well!<br><br>Have a good weekend,<br><br>Julian Hall<br>--<br>Dr. J. A. Julian Hall, Senior Lecturer, School of Mathematics,<br>University of Edinburgh, JCMB, King's Buildings, EDINBURGH, EH9 3JZ, UK.<br>Room: 6221 Phone: [+44](131) 650 5075 Email: <a rel="nofollow" ymailto="mailto:J.A.J.Hall@ed.ac.uk" target="_blank" href="mailto:J.A.J.Hall@ed.ac.uk">J.A.J.Hall@ed.ac.uk</a><br><span>Fax: [+44](131) 650 6553 Web:
<a target="_blank" href="http://www.maths.ed.ac.uk/hall">http://www.maths.ed.ac.uk/hall</a></span><br><br>The University of Edinburgh is a charitable body, registered in<br>Scotland, with<br>registration number SC005336.<br><br><br><br><br><br><br>-- <br>The University of Edinburgh is a charitable body, registered in<br>Scotland, with registration number SC005336.<br><br><br><br>_______________________________________________<br>Clp mailing list<br><a rel="nofollow" ymailto="mailto:Clp@list.coin-or.org" target="_blank" href="mailto:Clp@list.coin-or.org">Clp@list.coin-or.org</a><br><span><a target="_blank" href="http://list.coin-or.org/mailman/listinfo/clp">http://list.coin-or.org/mailman/listinfo/clp</a></span><br></div></div>
</div>
</div>
</div>
</div>
</div>
<!-- cg32.c2.mail.ac4.yahoo.com compressed/chunked Tue Feb 9 01:41:38 PST 2010 -->
</div><br>
</body></html>