<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<p>Hi Francesco,</p>
<p>There are two things for QE-GPU,</p>
<p>1) Hardware, the Maxwell GPU is not recommended for QE since its
double precision is very slow. You shall look for Fermi (C207X) or
Kepler (Titan Black/Z, K40/80) cards for QE-GPU computation.</p>
<p>2) Software, cuda-6.5 works okay for QE-GPU v.530 on my servers
and I think it should also work for v.540. I also used Intel
complier to do compile QE-GPU.</p>
<p>Here are my config input,</p>
<p>./configure CC=icc F90=ifort F77=ifort MPIF90=mpiifort \
<br>
--enable-parallel --disable-openmp --without-scalapack \
<br>
--enable-cuda --with-gpu-arch=sm_20
--with-cuda-dir=/usr/local/cuda-6.5 \
<br>
--with-magma --with-phigemm --with-pinned-mem \
<br>
LDFLAGS="-L/usr/lib64/ -lstdc++" <br>
</p>
Then you can follow the steps in Filippo's webpage,
<a class="moz-txt-link-freetext" href="https://github.com/fspiga/QE-GPU">https://github.com/fspiga/QE-GPU</a><br>
<br>
Hope it helps,<br>
<br>
Rolly<br>
<br>
<div class="moz-cite-prefix">On 08/17/2016 05:08 PM, Francesco
Pelizza wrote:<br>
</div>
<blockquote cite="mid:57B429A8.80701@strath.ac.uk" type="cite">Hi
dear Users,
<br>
<br>
I am trying to install and run jobs on the GPU I have, a gtx 980
Ti nVidia.
<br>
I followed the official git web page to intall the GPU version.
But after several fails and a couple if days trying to solve. I
ask for help now.
<br>
<br>
The configuration seems to go well. The make command doesn't run
properly since is not finding some "ELPA/src" folder and not
installing PhiGemm and probably also magma is not getting
installed.
<br>
<br>
I am not sure If it is legitimate to ask here for that help.
<br>
<br>
But I bet someone before tryed and may solved my problems already.
<br>
<br>
The System is Ubuntu 15.04 since the installation of Cudatoolkit
is smooth.
<br>
Nvidia drivers are 352 since are the latest supported in 15.04.
<br>
Cudatoolkit 7.5. Computing capacity of the nvidia graphic card is
5.2.
<br>
<br>
Could it be that the architeture sm_35 is not appropriate for a
Maxwell GPU? What about using the TITAN X. I read someone
did...How?
<br>
<br>
Here attached:
<br>
- the command I used in cmd.txt file
<br>
- the configure output in conf.txt
<br>
- the make output in make.txt
<br>
- the error raised from make command in Error_make.txt
<br>
<br>
Please let me know if I am doing something wrong
<br>
<br>
Thank you very much
<br>
<br>
Francesco Pelizza
<br>
<br>
PhD Student in Strathclyde University
<br>
Glasgow
<br>
<br>
<fieldset class="mimeAttachmentHeader"></fieldset>
<br>
<pre wrap="">_______________________________________________
Pw_forum mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Pw_forum@pwscf.org">Pw_forum@pwscf.org</a>
<a class="moz-txt-link-freetext" href="http://pwscf.org/mailman/listinfo/pw_forum">http://pwscf.org/mailman/listinfo/pw_forum</a></pre>
</blockquote>
<br>
<pre class="moz-signature" cols="72">--
PhD. Research Fellow,
Dept. of Physics & Materials Science,
City University of Hong Kong
Tel: +852 3442 4000
Fax: +852 3442 0538</pre>
</body>
</html>