[Beowulf] [External] Re: HPCG

[Prentice Bisbal]

Thanks for the list. Most of them  I've already come across in my 
search, but the AMD HPC guide is one I haven't come across before that 
will definitely be handy.

--
Prentice

On 8/7/20 5:02 AM, Benson Muite wrote:
> Maybe the following are helpful:
>
> https://sx-aurora.github.io/posts/hpcg-tuning/
> https://www.hpcadvisorycouncil.com/pdf/HPCG_Analysis_POWER8.pdf
> https://link.springer.com/chapter/10.1007/978-3-030-50743-5_21
> https://ulhpc-tutorials.readthedocs.io/en/latest/parallel/hybrid/HPCG/
> https://www.intel.com/content/dam/www/public/us/en/documents/presentation/hpc-clusters-best-practices-performance-study.pdf 
>
> http://developer.amd.com/wp-content/resources/56420.pdf
> https://upcommons.upc.edu/bitstream/handle/2117/116642/1HPCG_shared_mem_implementation_tech_report.pdf?sequence=8&isAllowed=y 
>
> https://armkeil.blob.core.windows.net/developer/Files/downloads/hpc/files/Arm-HPC-UG-ISC18/GoingArm_SC18_BSC.pdf 
>
> https://dl.acm.org/doi/10.1145/3182177
>
> On 8/7/20 11:31 AM, Jim Cownie wrote:
>>> Source is unfortunately only accessible for SPEC members. 
>>
>> Spec HPG <https://www.spec.org/hpg/> (“High Performance Group”) 
>> benchmarks are available free to “Non-profit/educational” users:
>>
>>     /Non-profit/educational pricing for HPG suites (ACCEL, MPI2007,
>>     OMP2012)/
>>     The HPG benchmarks are available free of charge to organizations
>>     which qualify for the non-commercial license
>>     <https://www.spec.org/order.html#hpgcommercial> by submitting a
>>     request <https://www.spec.org/hpgdownload.html> for a license. As
>>     with all non-profit/educational licenses, the software is licensed
>>     to the organization rather than an individual.
>>
>> https://www.spec.org/order.html
>>
>> -- Jim
>> James Cownie <jcownie at gmail.com <mailto:jcownie at gmail.com>>
>> Mob: +44 780 637 7146
>>
>>> On 6 Aug 2020, at 20:32, Jan Wender <j.wender at web.de 
>>> <mailto:j.wender at web.de>> wrote:
>>>
>>> Hi Prentice,
>>>
>>> If all you want to compare is the performance of one CPU, then you 
>>> could use the SPEC benchmarks, SPECint and SPECfp. Both are 
>>> available for many CPUs at spec.org <http://spec.org>. Source is 
>>> unfortunately only accessible for SPEC members.
>>>
>>> Best, Jan
>>> -- 
>>> Jan Wender - j.wender at web.de <mailto:j.wender at web.de>
>>>
>>>> Am 05.08.2020 um 20:09 schrieb Prentice Bisbal via Beowulf 
>>>> <beowulf at beowulf.org <mailto:beowulf at beowulf.org>>:
>>>>
>>>> Beowulfers,
>>>>
>>>> Do any of you have any experience using HPCG as a benchmark. I'm 
>>>> trying to compare the performance of several different processors 
>>>> for an upcoming purchase. I've already run LINPACK, and now I'd 
>>>> like to run HPCG. It seems the only tuning parameter is the size of 
>>>> the local grid in the x,y,z dimensions.
>>>>
>>>> While the guidelines say to increase the gridsize until the job 
>>>> consumes 1/4 or more of RAM, my testing has shown that as the 
>>>> gridsize goes up, so does the performance,  and it keeps going up 
>>>> for me until I consume all the memory and the job gets killed by 
>>>> Slurm for exceeding memory requirements.
>>>>
>>>> I've been doing a lot of Google searching for how to tune HPCG for 
>>>> maximum results, and there are some papers for tuning HPCG for 
>>>> large supercomputers. In these cases, they use x,y,z dimensions 
>>>> that are not necessarily equal, but I don't understand how they 
>>>> determined to use these unique values for x,y,z.
>>>>
>>>> When I compare my HPL results to my HPCG results, I'm getting HPCG 
>>>> results that are 0.3 - 0.5% of HPL. On the HPCG Top500 list, most 
>>>> systems are getting 2-3% of HPL, so I'm off by an order of magnitude.
>>>>
>>>> -- 
>>>> Prentice
>>>>
>>>> _______________________________________________
>>>> Beowulf mailing list, Beowulf at beowulf.org 
>>>> <mailto:Beowulf at beowulf.org> sponsored by Penguin Computing
>>>> To change your subscription (digest mode or unsubscribe) visit 
>>>> https://beowulf.org/cgi-bin/mailman/listinfo/beowulf
>>>
>>> _______________________________________________
>>> Beowulf mailing list, Beowulf at beowulf.org 
>>> <mailto:Beowulf at beowulf.org> sponsored by Penguin Computing
>>> To change your subscription (digest mode or unsubscribe) visit 
>>> https://beowulf.org/cgi-bin/mailman/listinfo/beowulf
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
>> To change your subscription (digest mode or unsubscribe) visit 
>> https://beowulf.org/cgi-bin/mailman/listinfo/beowulf
>>
>
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
> To change your subscription (digest mode or unsubscribe) visit 
> https://beowulf.org/cgi-bin/mailman/listinfo/beowulf

-- 
Prentice Bisbal
Lead Software Engineer
Research Computing
Princeton Plasma Physics Laboratory
http://www.pppl.gov