<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
<title></title>
</head>
<style type="text/css">
<!--
#send_mail {
font-family: "Courier New";
background:#FFFFFF;
color:#000000;
}
#send_mail a {
color:#999999
}
#send_mail a:hover {
color:#0000FF
}
#send_mail div {
color:#000000;
font-family: "Courier New";
font-size:14px;
line-height:150%;
}
#send_mail .usersign {
line-height:100%;
}
-->
</style>
<body id="send_mail">
<div>Dear prof. P. Giannozzi:
<BR>
Thank you very much for your helpful and clear response! I appreciate very much!
<BR>
Best Wishes!
<BR>
Yours Sincerely
<BR>
L. F. Huang
<BR>
<BR>
</FONT><FONT color=#444444>> Date: Fri, 10 Dec 2010 12:33:20 +0100
<BR>
</FONT><FONT color=#444444>> From: Paolo Giannozzi
<BR>
</FONT><FONT color=#444444>>
<BR>
</FONT><FONT color=#444444>> lfhuang wrote:
<BR>
</FONT><FONT color=#444444>>
<BR>
</FONT><FONT color=#444444>> </FONT><FONT color=#444444>> it is understandable that ph.x requires more memory than pw.x.
<BR>
</FONT><FONT color=#444444>> </FONT><FONT color=#444444>> But how to estimate it to be qualitatively precise?
<BR>
</FONT><FONT color=#444444>>
<BR>
</FONT><FONT color=#444444>> sum the size of large arrays. I do not know any other way. If the
<BR>
</FONT><FONT color=#444444>> code runs, a call to "memstat" returns the size of dynamically
<BR>
</FONT><FONT color=#444444>> allocated memory (on operating systems that support it). Not sure
<BR>
</FONT><FONT color=#444444>> how much of the total memory is reported, though. Apparently the
<BR>
</FONT><FONT color=#444444>> question "how much memory does a process use" has become difficult
<BR>
</FONT><FONT color=#444444>> to answer in modern operating systems.
<BR>
</FONT><FONT color=#444444>>
<BR>
</FONT><FONT color=#444444>> </FONT><FONT color=#444444>> How can a 3GB processor breaks down when doing a ph.x job, when the
<BR>
</FONT><FONT color=#444444>> </FONT><FONT color=#444444>> requirement of it pw.x job is just 500MB?
<BR>
</FONT><FONT color=#444444>>
<BR>
</FONT><FONT color=#444444>> the phonon code uses more arrays and distributes less of them across
<BR>
</FONT><FONT color=#444444>> processors.
<BR>
</FONT><FONT color=#444444>>
<BR>
</FONT><FONT color=#444444>> </FONT><FONT color=#444444>> By the way, is it better to assign the N_pr to be m**2 (m=2,3,4...) than
<BR>
</FONT><FONT color=#444444>> </FONT><FONT color=#444444>> other values (e.g. 6, 10, 11, 14...) if I want to divide the R-&G-space
<BR>
</FONT><FONT color=#444444>> </FONT><FONT color=#444444>> in parallelization?
<BR>
</FONT><FONT color=#444444>>
<BR>
</FONT><FONT color=#444444>> I think it depends upon the specific communication hardware and
<BR>
</FONT><FONT color=#444444>> MPI implementation. Not sure though.
<BR>
</FONT><FONT color=#444444>>
<BR>
</FONT><FONT color=#444444>> Paolo
<BR>
</FONT><FONT color=#444444>> --
<BR>
</FONT><FONT color=#444444>> Paolo Giannozzi, Democritos and University of Udine, Italy
<BR>
<BR>
<BR>
------
<BR>
======================================================================
<BR>
L.F.Huang(黄良锋) DFT and phonon physics
<BR>
======================================================================
<BR>
Add: Research Laboratory for Computational Materials Sciences,
<BR>
Instutue of Solid State Physics,the Chinese Academy of Sciences,
<BR>
P.O.Box 1129, Hefei 230031, P.R.China
<BR>
Tel: 86-551-5591464-326(office)
<BR>
Fax: 86-551-5591434
<BR>
Our group: <A href=http://theory.issp.ac.cn target=_blank>http://theory.issp.ac.cn</A>
<BR>
======================================================================
<BR>
</div>
</body>
</html>