[CP2K-user] [CP2K:13235] Re: CP2K 7.1-Cuda Bandgap and HF energies different from previous versions

[Fabian Ducry]

So far I could not reproduce the behaviour with any other system than the 
one provided by Andres. I tried 
 - QS/regtest-admm-1/CH3-BP-NONE.inp
 - QS/regtest-hybrid-4/CH4-PBE0_TC_LRC.inp
 - the Si-8 example from https://www.cp2k.org/howto:static_calculation
 - the Si-8 example with 2 Si replaced by Al
I took the subsys from these 4 examples and replaced the one in the input 
from Andres. All of them behave as they should with CUDA+OMP.

So it seems to be difficult to catch and maybe does not trigger very often?
That is probably why it does not show up in the regtests.

Best,
Fabian

On Monday, 4 May 2020 18:26:40 UTC+2, Ole Schütt wrote:
>
> Hi Leopold, 
>
> I agree that this is a serious issue that should have been caught by our 
> testing. 
> AFAIK, the performance tests on the dashboard do not check the computed 
> results. 
>
> However, we do have a daily CUDA regtest: 
> https://dashboard.cp2k.org/archive/cuda-pascal/index.html 
> It uses two OpenMP threads, which might not be enough to exceed the 
> tolerance thresholds? 
>
> -Ole  
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200504/e6ecf1b7/attachment.htm>