Dear Filippo,

Thank you for the instruction. I think at first, I compile QE without GPU so 
pw.x was generated.

Then, I unzip and put the GPU folder into QE directory and then recompile, 
somehow I mess up the files when I tried to make clean the GPU compilation.

So what I finally did is, remove the directory, untar QE and unzip the GPU 
folder to QE, then I rerun,
./configure CC=icc F90=ifort F77=ifort MPIF90=mpiifort --enable-parallel 
--disable-openmp --without-scalapack --enable-cuda 
--with-gpu-arch=sm_20--with-cuda-dir=/usr/local/cuda-6.5 --with-magma 
--with-phigemm
make �Cj 12 all

Then I found pw.x appeared in PW/src and ./check-pw.x.j runs ok and completed.

I would like to run GPU on the examples in PW/examples. How can I do that? 
Should I modify the run_all_examples script?

Regards,
Rolly

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf 
Of Filippo Spiga
Sent: 2015年10月28日 16:19
To: PWSCF Forum
Subject: Re: [Pw_forum] QE-GPU 14.10.0 with intel compliers

On Oct 27, 2015, at 5:29 PM, Dr. NG Siu Pang <[email protected]> wrote:
> I used ./configure CC=icc F90=ifort F77=ifort MPIF90=mpiifort
> ./PW/tests/check-pw.x.j runs all okay

Weird, in another email you said it was not working.


> I have copied GPU to the espresso-5.2.0 folder and I noticed that for
> parallel configuration, I need to do $ cd GPU $ ./configure
> --enable-parallel --enable-openmp --with-scalapack \
>   --enable-cuda --with-gpu-arch=sm_35 \
>   --with-cuda-dir=<full-path-where-CUDA-is-installed> \
>   --without-magma --with-phigemm
> $ cd ..
> $ make -f Makefile.gpu pw-gpu
>
> These looks like using non-intel compliers, so can I use the following intead?
> $ cd GPU
> $ ./configure CC=icc F90=ifort F77=ifort MPIF90=mpiifort \
> --enable-parallel --enable-openmp --with-scalapack \ --enable-cuda
> --with-gpu-arch=sm_35 \
> --with-cuda-dir=<full-path-where-CUDA-is-installed> \ --without-magma
> --with-phigemm


This is the server with 8 GPUs C2050? As I mentioned in the other email, 
"--with-gpu-arch" must be "sm_20" otherwise GPU code will fail. If you have 8 
GPU into the same server, no point of having OpenMP and if you run on a single 
server probably no point of having ScaLAPACK.

Try:

$ ./configure CC=icc F90=ifort F77=ifort MPIF90=mpiifort --enable-parallel 
--disable-openmp --without-scalapack --enable-cuda 
--with-gpu-arch=sm_20--with-cuda-dir=<full-path-where-CUDA-is-installed> 
--with-magma --with-phigemm

You must make sure mpirun handle bindings to cores and scatter across sockets 
(if more than one) properly otherwise performance may sucks. Or, if I guess 
correctly,

export I_MPI_PIN=on
export MPI_PIN_PROCESSOR_LIST=all:map=bunch
mpirun -genvall -np 8 ./pw.x -input ...

HTH

--
Mr. Filippo SPIGA, M.Sc.
Quantum ESPRESSO Foundation
http://fspiga.github.io ~ skype: filippo.spiga

*****
Disclaimer: "Please note this message and any attachments are CONFIDENTIAL and 
may be privileged or otherwise protected from disclosure. The contents are not 
to be disclosed to anyone other than the addressee. Unauthorized recipients are 
requested to preserve this confidentiality and to advise the sender immediately 
of any error in transmission."
_______________________________________________
Pw_forum mailing list
[email protected]
http://pwscf.org/mailman/listinfo/pw_forum


Disclaimer: This email (including any attachments) is for the use of the 
intended recipient only and may contain confidential information and/or 
copyright material. If you are not the intended recipient, please notify the 
sender immediately and delete this email and all copies from your system. Any 
unauthorized use, disclosure, reproduction, copying, distribution, or other 
form of unauthorized dissemination of the contents is expressly prohibited.

_______________________________________________
Pw_forum mailing list
[email protected]
http://pwscf.org/mailman/listinfo/pw_forum

Reply via email to