My Community

Hi --

Today, it’s my first day in the VASP world. I’m currently a student and will be working with VASP 4.6. My job will be to determine what features I can rewrite so they will run faster on the GPU (CUDA C).

Since I’m new I have some questions that someone can give some direction on. My UNIX skills need a refresher too, but that is a different story.

From looking at the installation instructions [http://cms.mpi.univie.ac.at/vasp/vasp/I ... _VASP.html] I noticed it references these files.

“
vasp.4.X.X.tar.gz or vasp.4.X.X.tar.Z
vasp.4.lib.tar.gz or vasp.4.lib.tar.Z
benchmark.tar.gz
bench.Hg.tar.gz
“

My question is will I still be able to install and use VASP correctly if I only had the VASP 4.6 source files (I don’t think I have access to “vasp.4.lib.tar.gz�, “benchmark.tar.gz� and “bench.Hg.tar.gz�)? I don’t know what my school currently has, and I plan to install VASP on Ubunutu 10.10.

My second question are what are some good materials for me to read to learn how people actually use VASP, and what are the problems that are associated with VASP? I would assume that there are some features that people would like to run faster

Thanks for any input. It is much appreciative.

-Pete

Dear Pete,

we just submitted a paper with the title "Accelerating the blocked Davidson algorithm with Graphics Processing Units" where the use of NVIDIA's CUDA framework to accelerate VASP is described. The authors are S.Maintz, B. Eck and R. Dronskowski, the Journal is J.Chem.Theor.Comput. If wished I will announce the publication here. We are also right now preparing a website where the source for this work can be downloaded.


Concerning your questions, if they are still relevant:
First Question: if you have access to vasp.4.X.X.tar.gz then you should also have acess to the file vasp.4.lib.tar.gz. They are on the same site for download. The two bechmark files are not needed for installation but may contain ready to run examples to test the installation. I never used them.
Which flavour of linux or unix you are using doesn't really matter, its more a case of the tools installed.

Second Question: there surely is a simple answer to this: EVERYTHING should go faster. So, as a general advice: start accelerating the central parts everybody is using...like we did in our work.
Btw, how much experience in programming/fortan and DFT programs do you have?

Good luck,
Bernhard

Hi Beck,

Thanks for taking your time in providing a great response.

First off, it sounds you have done great GPU work recently so congratulations on that.

After I made my post, I did several searches and it seems that those benchmark libraries you used here and there.
Thanks for your thoughts to this. Also, I managed to find someone who had installed VASP before so I am going to work with him to get me set-up.

As I mentioned I am new to VASP so my plan is first get it installed and see what it can and can’t do. Reading the VASP documentation can only get you so far so I can’t wait for me to touch and feel the tool. To answer your question, I have never developed with Fortran before. My programming language background is C++, C, JAVA, Assembly and MIPS. Personally, my favorite is C++ since I have been using for roughly 9 years.

I didn’t know what a DFT program was so I looked it up. I would say my DFT is next to null even though I have worked with Mathematica and Maple before but when I did, it was more calculus base projects.

Thanks again for the response and especially providing some CUDA direction.

--Pete

P.S.

If it has any value, here are the VASP files that I was given. I will double check I was provided with everything.

aedens.f
all.table
asa.f
base.f
base.inc
brent.f
broyden.f
broyden.inc
chain.f
charge.f
chgloc.f
choleski2.F
cl_shift.f
constant.f
constant.inc
constrmag.F
constr_cell_relax.F
cube.inc
davidson.f
dfast.f
dgemmtest.F
dgemm_opt.f
dgemv.f
dipol.f
dos.f
dyna.f
dynbr.f
ebs.f
edtest.f
electron.f
elf.f
elpol.f
fft3dcray.F
fft3dessl+furth.F
fft3dessl.F
fft3dfujitsu.F
fft3dfurth.F
fft3dfurth_f77.F
fft3dlib.f
fft3dlib_f77.F
fft3dnec.f
fft3dsimple.F
fftmpi.f
fftmpiw.f
fftmpi_map.F
ffttest.f
fftw3d.f
fileio.f
files.txt
finite_diff.F
force.f
ftflop.f
hamil.f
hamil_rot.F
ini.f
jacobi.f
lattice.f
lattice.inc
lattlib.f
LDApU.F
lincom.table
main.f
main_mpi.f
Makefile
makefile.altrix
makefile.cray
makefile.dec
makefile.hp
makefile.hpux_itanium
makefile.linux_abs
makefile.linux_alpha
makefile.linux_efc_itanium
makefile.linux_gfortran
makefile.linux_ifc_ath
makefile.linux_ifc_opt
makefile.linux_ifc_P4
makefile.linux_pg
makefile.linux_pgi_opt
makefile.nec
makefile.rs6000
makefile.sgi
makefile.sp2
makefile.sun
makefile.t3d
makefile.t3e
makefile.vpp
makeparam.F
metagga.f
mgrid.f
mgrid.inc
mix.f
mkpoints.f
mkpoints.inc
mpi.f
mpimy.inc
music.f
nonl.f
nonl.inc
nonlr.f
nonlr.inc
opergrid.f
optics.f
optreal.f
paircorrection.F
param.inc
pardens.f
paw.f
pm.inc
poscar.f
poscar.inc
pot.f
potex1.f
potex2.f
preprocess
pseudo.f
pseudo.inc
radial.f
random.f
reader.f
readme
relativistic.F
rmm-diis.f
rot.f
rpro.table
scala.f
setex.f
setexch.f
setexm.inc
setlocalpp.F
shm.f
smart_allocate.F
sphpro.f
steep.f
step.f
stepprecor.F
stepver.f
stm.f
stufak.f
subrot.f
symbol.inc
symlib.f
symmetry.f
test.f
tet.f
tutor.f
us.f
wave.f
wave.inc
wave_mpi.f
wavpre.f
wavpre_noio.F
writer.f
xcgrad.f
xclib.f
xcspin.f
xml.f
xml_writer.F
zgemmtest.F
zgemm_opt.f