[CP2K-user] [CP2K:12286] Reproducing results cross-platform

Timo van Opstal tvano... at limebv.nl
Mon Oct 7 07:35:50 UTC 2019


This is still an open issue, but I'm working around it by avoiding the 
precompiled OS X binaries. Maybe there is an uninitialised variable 
somewhere.

On Monday, September 30, 2019 at 9:45:53 AM UTC+2, Timo van Opstal wrote:
>
> Thank you Matthias! The OSX version still misbehaves after these updates, 
> I've also checked convergence in CUTOFF and REL_CUTOFF, (in the same vein 
> as the tutorial) and it seems that my relative grid error is below 10^-4 in 
> energy for those settings, so that should be OK I think?
>
> On Sunday, September 29, 2019 at 3:49:50 PM UTC+2, Matthias Krack wrote:
>>
>> Hi
>>
>>  
>>
>> Try again with a smaller EPS_DEFAULT value, e.g. 1.0E-12 instead of 
>> 1.0E-7, a larger CUTOFF value, at least 300 (Ry) instead of 200, and a 
>> larger REL_CUTOFF value, at least 40 instead of 30, to get numerically more 
>> meaningful results.
>>
>>  
>>
>> Matthias
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20191007/816b139e/attachment.htm>


More information about the CP2K-user mailing list