Dear Li, part of ph.x can exploit some acceleration that comes from linear algebra but not as much as pw.x. If the work scheduled is going to continue as expected we will be able to deliver a first version of ph.x properly accelerated in one month/one month and half. I remember I said something like the end of this year, it is still likely possible. Stay tuned.
Cheers, Filippo On Nov 26, 2012, at 11:15 PM, Chengyang Li <li.chengyang at wmich.edu> wrote: > Dear Filippo > > I have a question about QE- GPU , does it include ph.x now? Last time I heard > ph.x will be used in the end of this year. > > Chengyang > > From: "Filippo Spiga" <spiga.filippo at gmail.com> > To: q-e-gpgpu at qe-forge.org > Cc: "PWSCF Forum" <Pw_forum at pwscf.org> > Sent: Monday, November 26, 2012 1:01:45 PM > Subject: [Pw_forum] QE-GPU-5.0.2 (build1) is ready for download > > Dear all, > > the QE-GPU-5.0.2 (build1) is ready for download. This version is fully > compatible with QE 5.0.2 released few days ago. > > The QE-GPU package works like a plug-in for the standard Quantum ESPRESSO > distribution. The procedure to follow to deploy it is very simple and it is > also descrbed here: http://qe-forge.org/gf/project/q-e-gpu/ > > Frst you need to download and install Quantum ESPRESSO 5.0.2. Then you have > to download the file "QE-GPU-5.0.2-build1.tar.gz" and place it in the > root/main directory of your QE installation (the same directory from which > you usually run the configure). At this point, by untarring > QE-GPU-5.0.2-build1.tar.gz, a directory called "GPU" will appear. Next steps > are quite easy: > > $ cd GPU > $ ./configure --enable-parallel --enable-openmp --enable-cuda > --with-gpu-arch=<your-GPU-capability> --with-cuda-dir=<your-CUDA-path> > --enable-phigemm --enable-magma > $ cd .. > $ make -f Makefile.gpu pw-gpu > > All the executables will appear under "bin/". By run "make -f Makefile.gpu" > you will see what packages can exploit entirely or partially some GPU > acceleration. Remember that PWscf (and so NEB) are the two code with the > large amount of walltime ported to GPU. Here the link for the direct > download: > http://qe-forge.org/gf/project/q-e-gpu/frs/?action=FrsReleaseView&release_id=126 > > For additional details please read the README.GPU file under the "GPU/" > directory. For any feedback or problem please subscribe the q-e-gpu > mailing-list (link: http://qe-forge.org/mailman/listinfo/q-e-gpgpu ) > > Happy Computing. > Filippo > > > QE-GPU-5.0.2 ~ short Change-Log : > - two "modes": normal and fast-cuda > - revised GPU memory management > - phiGEMM 2.0.0rc1 > - brand new MAGMA 1.3.0 > - better support for GPU with cc_30, cc_35 (TESLA K20, K20x, K10, GTX6xx) > - improved configure and XK7 support > - better GPU stream management > - corrected weird bug in addusdens > - batched FFT in serial execution temporary disabled (it will be restored in > the next build) > > -- > Mr. Filippo SPIGA, M.Sc. > http://filippospiga.me ~ skype: filippo.spiga > > ?Nobody will drive us out of Cantor's paradise.? ~ David Hilbert > > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum > > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum -- Mr. Filippo SPIGA, M.Sc. http://filippospiga.me ~ skype: filippo.spiga ?Nobody will drive us out of Cantor's paradise.? ~ David Hilbert -------------- next part -------------- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20121127/1fdcb9d4/attachment-0001.html
