Cores to use the H2O_64 benchmark

[Julita Inca Chiroque]

Hello cp2k group

Can I ask if there is a reason to plot the graph attached (found in the CP2K official website) since 8 cores to test the H2O_256.inp instead of 1?

I was trying to run the benchmark using 1 to 512 cores, and it crashes when I set 1, 2, and 4 cores with MPI. I was looking for the solution to fix this bug, but I don't have the right answer yet. I run this sample three times so far on a Broadwell architecture. 

Thanks in advance,

Julita Inca

[Thomas Kühne]

Dear Julita, 

this is a rather old benchmark conducted on „Jump“ at Jülich around 2005! 

For the largest systems benchmarks for small CPU number are most likely 

not available due to limited memory and/or unreasonable execution time. 

Cheers, 

Thomas

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAFP92tw4moPGAoQgdYMqDAKtQ9vqXKGp%2BqeKSAVUdhdZ5U38JA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.
<Screen Shot 2019-06-26 at 9.07.05 PM.png>

[Julita Inca Chiroque]

Thank you Thomas for the answer.

So, this means that I can start the benchmarking of the 256 water sample using 8 cores?

As you said, I experienced memory issues with MPI when I use 1, 2 and 4 with popt.

Thanks again Thomas,

Julita 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/47A3E28E-9CC2-4A32-82D8-850B63A6D480%40gmail.com.