Re: [Pw_forum] QE-GPU problems.

[Oliver Generalao]

Hi Konrad,

Nvidia GTX 1060 is sm_61 and per https://github.com/fspiga/QE-GPU, it
supports up to sm_60, based on  my experience QE-GPU will compile with
sm_61 however when you run pw-gpu.x it will throw errors, hence there
should be a slight modification on the QE-GPU source code to include
sm_61, it should be easy to do that.
I am using Nvidia GTX 1070 and have been working on it with QE-GPU for
2 months now. The only catch, which Mr Spiga always points out, is
that the DP performance on GTX cards  is very slow , which is like
1/32 (or 1/24) of the Single Precision



On 12/29/16, Konrad Gruszka  wrote:
> Deer community,
>
> Recently I'm trying to compile and run QE-GPU version on my new Cuda
> capable card. Unfortunately after many attempts the result is poor.
>
> I have Nvidia GTX 1060 (with Pascal architecture, 1280 cuda units). I've
> managed to compile GPU PWSCF with sm_60 (pascal) support but when trying
> to run any calculation I get:
>
> Program received signal SIGFPE: Floating-point exception - erroneous
> arithmetic operation.Backtrace for this error: #0 0x7F84D59C4E08
>
> #1 0x7F84D59C3F90
> #2 0x7F84D4CB74AF
> #3 0x63B2C5 in newd_cuda_
> #4 0x639B9F in newq_compute_gpu_ at newq_compute_gpu.f90:122
> #5 0x597B1F in __dfunct_MOD_newd at newd.f90:262
> #6 0x4E2B87 in init_run_ at init_run.f90:101
> #7 0x4081DB in run_pwscf_ at run_pwscf.f90:78
> #8 0x408049 in pwscf at pwscf.f90:30 #9 0x7F84D4CA282F
>
> The configure options were:
>
> ./configure --enable-parallel --enable-cuda --enable-openmp
> --with-cuda-dir=/usr/local/cuda-8.0/ --with-internal-blas
> --with-internal-lapack --without-magma --with-phigemm
> FFT_LIBS=/mnt/fast/fftw-3.3.4/.libs/libfftw3.a --with-gpu-arch=pascal
>
> For now I'm trapped here, not knowing what to do. Is it possible to run
> QE-GPU not only on specialized computing devices like e.g. K20 at all?
> How to manage this?
>
> Konrad Gruszka
>
>


-- 
Oliver B. Generalao

M.S. Physics student
Structure and Dynamics Group
National Institue of Physics
University of the Philippines
Diliman, Quezon City
Trunkline: +63-2-981-8500
Mobile: +63-927-4033966
___
Pw_forum mailing list
Pw_forum@pwscf.org
http://pwscf.org/mailman/listinfo/pw_forum

Re: [Pw_forum] Problem: different output with identical input

[Pietro Delugas]

Dear Pablo

Final Coordinates and Job Done do not mean that  relaxation reached 
convergence you may have reached the max number of steps and relaxation 
stopped.


Check few lines before "Begin Final coordinates" you should find 
something like "bfgs converged in  scf cycles"


Did you try to use the same relaxation run for the 2 different first 
steps ? Do final results coincide in this case ?



On 28/12/2016 19:01, Pablo García Risueño wrote:

Dear Stefano

Thank you very much for your reply. I answer to your questions below.

2016-12-28 16:29 GMT+01:00 Stefano de Gironcoli >:


Dear Pablo Garcia Risueño

I'm not sure I understand the problem.
the two final positions for ecut=30 are identical to my eyes.
the two final positions for ecut=60 differ by about 10-5 A. !
If the property you are interested in depends so strongly on the
atomic positions you better learn how to live with it. Even assuming
that you are using perfect xc functional, perfect pseudopotentials and
converged cutoff and k-points (which likely you don't), your
calculation is neglecting zero point energy, thermal expansion,
quantum nature of H motion, ...
/
/

/Yes, my concern is not that the results are not perfect (of course 
DFT makes non-negligible errors). My concern is that fully identical 
input files run in the same machine and with the same executable 
provide different results./


As for your specific question. In principle if you tighten enough
the convergence the minima should tend to converge toward a common
value. However the stopping criteria will make the two calculations
stop as soon they are satisfied. In your case as soon as each force
component is lower than 10-6 Ry/au and the energy does not change more
that 10-8. Did the calculation complete ? 10-8 for the energy is
easily reachable but 10-6 for the forces looks to me very demanding.
/
/

/Yes, the calculations do finish. I read 'Begin final coordinates' 
(...) and 'JOB DONE' at the end of the output files. My system is very 
small (just 26 atoms and 56 electrons)./


  Anyhow... even if the relaxation did complete successfully two
equivalent calculations may still differ. If you run your calculations
with exactly the same input on exactly the same machine with exactly
the same mpi setup the results should be identical 


/This is exactly what I do./

..but why doing
twice exactly the same calculation ?. 

/The 'relax' calculation is the first step of a complicated process. I 
run two different calculations because the 6th step of the process is 
different. I obtained very different results at the end of the full 
process; then I tried to trace back the differences. And then I saw 
that despite the fact that the 1st step is identical for two different 
runs, the results of the 1st step are different. This made me think 
that there might be a bug, or a wrong parameter in my input files./


If however you change anything
in the input (Davidson vs CG, starting wavefunction options, mixing
mode, BFGS-related parameters) or parallel computational setup the
relaxation history will be slightly different and the final
configurations will differ by a certain amount allowed by the
tolerance defined by your stopping criteria.

/I changed nothing, this is why I am surprised with the results.../


Thank you very much for your attention. Best regards.


  HTH

stefano

Quoting Pablo García Risueño >:

> Dear professor
>
> Thank you very much for your reply. The differences are
important in this
> case, my final calculated quantities are very sensitive to these
optimized
> positions. Should I change any of the ***conv_thr variables, or
other
> variable, to have the same result for same inputs?
>
> Is there any random number in the algorithm of relax
calculations, can we
> be sure that different outputs with the same input are not due
to any bug?
>
> Thank you very much. Best regards.
>
> 2016-12-28 14:10 GMT+01:00 Paolo Giannozzi
>:
>
>> The differences you find are very small. Nothing to worry about
in my
>> opinion.
>>
>> By the way: Davidson diagonalization is typically faster than
CG; do not
>> specify incompatible options in K_POINTS (gamma or automatic,
not both;
>> gamma should be used unless you have a good reason not to)
>>
>> Paolo
>>
>> Il 28/dic/2016 01:46 PM, "Pablo García Risueño"
>
>> ha scritto:
>> >
>> > Dear Espresso community
>> >
>> > I have one problem that is important for me; it is somewhat
surprising.
>> I run 

[Pw_forum] Pwgui for gw

[Uttam Paliwal]

Hi
Can pwgui6.0 be used for general calculations. If yes then please guide.
___
Pw_forum mailing list
Pw_forum@pwscf.org
http://pwscf.org/mailman/listinfo/pw_forum

Re: [Pw_forum] QE-GPU problems.

[Rolly Ng]

Hi Konrad,

My experience with QE-GPU is that it works only with double precision 
supported GPUs but not single precision.


I am using TitanZ and Intel compiler, it works fast. Here is my ./configure

./configure CC=icc F90=ifort F77=ifort MPIF90=mpiifort --enable-parallel 
--disable-openmp --without-scalapack --enable-cuda --with-gpu-arch=sm_35 
--with-cuda-dir=/usr/local/cuda-6.5 --with-magma --with-phigemm 
--with-pinned-mem LDFLAGS="-L/usr/lib64/ -lstdc++"


I doubt if GTX 1060 supports DP? If it is SP than it does not work for 
QE-GPU or it may be very slow.


Rolly

On 12/29/2016 03:56 PM, Konrad Gruszka wrote:


Deer community,

Recently I'm trying to compile and run QE-GPU version on my new Cuda 
capable card. Unfortunately after many attempts the result is poor.


I have Nvidia GTX 1060 (with Pascal architecture, 1280 cuda units). 
I've managed to compile GPU PWSCF with sm_60 (pascal) support but when 
trying to run any calculation I get:


Program received signal SIGFPE: Floating-point exception - erroneous 
arithmetic operation.Backtrace for this error: #0 0x7F84D59C4E08


#1 0x7F84D59C3F90
#2 0x7F84D4CB74AF
#3 0x63B2C5 in newd_cuda_
#4 0x639B9F in newq_compute_gpu_ at newq_compute_gpu.f90:122
#5 0x597B1F in __dfunct_MOD_newd at newd.f90:262
#6 0x4E2B87 in init_run_ at init_run.f90:101
#7 0x4081DB in run_pwscf_ at run_pwscf.f90:78
#8 0x408049 in pwscf at pwscf.f90:30 #9 0x7F84D4CA282F

The configure options were:

./configure --enable-parallel --enable-cuda --enable-openmp 
--with-cuda-dir=/usr/local/cuda-8.0/ --with-internal-blas 
--with-internal-lapack --without-magma --with-phigemm 
FFT_LIBS=/mnt/fast/fftw-3.3.4/.libs/libfftw3.a --with-gpu-arch=pascal


For now I'm trapped here, not knowing what to do. Is it possible to 
run QE-GPU not only on specialized computing devices like e.g. K20 at 
all? How to manage this?


Konrad Gruszka



___
Pw_forum mailing list
Pw_forum@pwscf.org
http://pwscf.org/mailman/listinfo/pw_forum


--
PhD. Research Fellow,
Dept. of Physics & Materials Science,
City University of Hong Kong
Tel: +852 3442 4000
Fax: +852 3442 0538

___
Pw_forum mailing list
Pw_forum@pwscf.org
http://pwscf.org/mailman/listinfo/pw_forum

Re: [Pw_forum] PAW Pseudo for Mn

[Antimo Marrazzo]

On 12/27/2016 08:46 PM, Mostafa Youssef wrote:

Dear Mohammadreza,


Have a look at this link:

http://materialscloud.org/sssp/


Also here:
http://qe-forge.org/gf/project/pslibrary/


Regards,
Mostafa

On 12/28/2016 11:49 PM, mohammadreza hosseini wrote:

Dear all I need PAW pseudo for Mn . I could not find in pseudo web page this 
type. How can I obtain it?
Maybe you could look more carefully to those 2 websites. For instance, 
the SSSP efficiency pseudopotential is a PAW pseudopotential from 
pslibrary 0.3.1. At the very center of the SSSP homepage 
http://materialscloud.org/sssp/ there is a link to download all the 
pseudopotentials of the SSSP efficiency library 
http://materialscloud.org/sssp/pseudos/SSSP_eff_PBE.tar.gz .

Do you find it now?

HTH
Antimo

--
Antimo Marrazzo
Doctoral Assistant
EPFL STI IMX THEOS
ME-D2 1019 (Bâtiment ME)
Station 9
CH-1015 Lausanne (Switzerland)

___
Pw_forum mailing list
Pw_forum@pwscf.org
http://pwscf.org/mailman/listinfo/pw_forum

Re: [Pw_forum] effective mass calculation

[alberto]

Hi,
I continue the discussion about calculate effective carrier masses. During
my search on web, I found a script named 'pw_nscf_2_EIGENVAL' in this page
https://www.mail-archive.com/pw_forum@pwscf.org/msg27908.html. Has anyone
tried it?


regards

Al

2016-12-28 17:00 GMT+01:00 alberto :

> Hi,
> I would calculate effective carrier masses at level SR-DFT and at the
> level SOC-DFT. It's possible with quantum espresso?
> On web I found only a script for VASP or CRYSTAL
>
>
> regards
>
> Al
>
___
Pw_forum mailing list
Pw_forum@pwscf.org
http://pwscf.org/mailman/listinfo/pw_forum