<div dir="ltr"><br><div class="gmail_extra">Hi again Mr. Spiga, thanks for the reply.<br><br>I was able to compile with your help. What I did was:<br><br>1. configure the main QE 5.1 with --enable-openmp and --disable-parallel.<br>2. make all the QE 5.1<br>3. recompile BLAS with -fPIC<br>4. configure the GPU (from the git master branch) with --enable-openmp --disable-parallel --enable-cuda --with-gpu-arch=sm_30 --with-cuda-dir=/usr/local/cuda-6.5 --with-phigemm --without-magma<br>5. and make -f Makefile.gpu all-gpu<br><br>When i try to run pw-gpu, ph-gpu or neb-gpu I get the following error:<br><br>[<a href="http://vloc_psi_k_cuda_v1.cu:171">vloc_psi_k_cuda_v1.cu:171</a>] Allocation of GPU resources for the plan failed! Program exits... <br>Error condition encountered during test: exit status = 1<br><br>The cp-gpu is working. I checked during the run with nvidia-smi and it was using the GPU. Also the samples which come with CUDA work.<br><br>Maybe you have an idea about this error.<br><br>Thanks in advance and have a nice holiday,<br><br>Uri Argaman<br>Ben-Gurion University<br>Israel<br><br><div class="gmail_quote"><br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
Date: Mon, 22 Sep 2014 23:49:23 +0800<br>
From: Filippo Spiga <<a href="mailto:spiga.filippo@gmail.com" target="_blank">spiga.filippo@gmail.com</a>><br>
Subject: Re: [Pw_forum] QE-GPU compilation problems<br>
To: PWSCF Forum <<a href="mailto:pw_forum@pwscf.org" target="_blank">pw_forum@pwscf.org</a>><br>
Message-ID: <<a href="mailto:4925EFBE-FF35-4671-9310-F74CABF15719@gmail.com" target="_blank">4925EFBE-FF35-4671-9310-F74CABF15719@gmail.com</a>><br>
Content-Type: text/plain; charset=utf-8<br>
<br>
Dear Uri,<br>
<br>
On Sep 22, 2014, at 10:17 PM, Uri Argaman <<a href="mailto:argamanu@post.bgu.ac.il" target="_blank">argamanu@post.bgu.ac.il</a>> wrote:<br>
> I tried recompiling BLAS with -fPIC and after that it seems to be able to compile pw-gpu, neb-gpu, cp-gpu, although it doesn't put pw.x, neb.x etc in the bin directory, but trying to compile ph-gpu gives the following error:<br>
<br>
Ok, that missing flag is a case the configure should handle. I will fix this. I have no machine with GCC 4.8, I have to upgrade one and do some testing.<br>
<br>
About "pw.x" or "neb.x", QE-GPU makefile does not create them. If you need the CPU-only executable first run the QE configure in a "standard way" then run the QE-GPU one. In this way you get both type of executable (but yes, you need to do twice the procedure).<br>
<br>
Speaking about ph-gpu.x, this is a problem that has been already fixed in the git master branch. Try to checkout the latest git version of QE-GPU or please wait me to finalize a new QE-GPU version. As this point, without pushing too much in release new functionalities, I can create just one new minor version with a few important bug fixes related to the configure and makefile.<br>
<br>
F<br>
<br>
--<br>
Mr. Filippo SPIGA, M.Sc.<br>
<a href="http://filippospiga.info" target="_blank">http://filippospiga.info</a> ~ skype: filippo.spiga<br>
<br>
?Nobody will drive us out of Cantor's paradise.? ~ David Hilbert<br></blockquote></div><br></div></div>