[molpro-user] MOLPRO GPU Support.
Martin Diefenbach
diefenbach at chemie.uni-frankfurt.de
Mon Mar 7 14:18:02 GMT 2011
While following this thread I was wondering whether there
were also options for producing GPU code for ATI/AMD architectures.
Are there any plans/projects to porting the GPU code to openCL?
Best wishes,
Martin
--
Dr. Martin Diefenbach
Institut für Anorganische Chemie der
Johann Wolfgang Goethe-Universität
Campus Riedberg, Frankfurt am Main
diefenbach at chemie.uni-frankfurt.de
On Mon, 7 Mar 2011, Andy May wrote:
> Jyh-Shyong,
>
> You need to specify -cuda option for configure.
>
> Also, most of the other options you specify are the default. I would
> suggest to use simply:
>
> ./configure -cuda -gfortran -gcc
>
> and even -gfortran and -gcc are only really needed if you actually have
> other Fortran or C compilers installed on the system.
>
> Best wishes,
>
> Andy
>
>
> On 05/03/11 10:16, Jyh-Shyong wrote:
>> Hi,
>>
>> I installed Molpro 2010.1 on our GPU-enabled cluster, for serial
>> version, I used command:
>>
>> ./configure -gfortran -gcc -i8 -nvcc -x86_64
>>
>> to generate file CONFIG, the compilation completed without error.
>> However, when I ran the test case
>> in directory testjobs (in this case, glypep1_vdz_gpu.test)
>>
>> ...
>> Checking input...
>> Passed
>> 1
>>
>>
>> *** PROGRAM SYSTEM MOLPRO ***
>> Copyright, University College Cardiff
>> Consultants Limited, 2008
>>
>> Version 2010.1 linked 5 Mar 2011
>> 11:44:23
>>
>>
>> **********************************************************************************************************************************
>> LABEL * glypep1
>> Linux-2.6.31.14-0.6-desktop/gc3(x86_64) 64 bit serial
>> version DATE: 05-Mar-11 TIME: 17:56:04
>> **********************************************************************************************************************************
>>
>> Patch level: 19
>> **********************************************************************************************************************************
>>
>> *** No errors in glypep1_vdz_gpu.test. (skipped)
>> **********************************************************************************************************************************
>>
>> Am I missing something? No other output was generated.
>>
>> The OS is OpenSuSE 11.2, CUDA version is 3.2, the computing node has 4
>> C1070 cards.
>> I also built the parallel version with GPU-enabled with command:
>>
>> ./configure -gfortran -gcc -i8 -nvcc -mpp -mppbase /opt/ga501
>>
>> and I have not test it yet. Thanks for any suggestion and advice.
>>
>> Jyh-Shyong Ho, Ph.D.
>> Research Scientist
>> National Center for High Performance Computing
>> Hsinchu, Taiwan, ROC
>>
>> _______________________________________________
>> Molpro-user mailing list
>> Molpro-user at molpro.net
>> http://www.molpro.net/mailman/listinfo/molpro-user
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
>
More information about the Molpro-user
mailing list