[CP2K-user] [CP2K:22060] Re: Issue with EPS_DEFAULT Sensitivity: Why Does 1.0E-12 Cause CPASSERT Failure?

[Frederick Stein]

Hi,
I found the culprit and will merge a bugfix today. It indeed seems to be a 
corner case of the Fortran standard/GCC compiler implementation.
Best,
Frederick

Frederick Stein schrieb am Montag, 12. Januar 2026 um 16:17:32 UTC+1:

> I have opened an issue on Github.
>
> yis... at 163.com schrieb am Montag, 12. Januar 2026 um 15:57:30 UTC+1:
>
>> Thank you for your outstanding work. I was able to run CP2K smoothly with 
>> EPS_DEFAULT = 1.0E-13 as well. I'm reporting this bug in the hope that CP2K 
>> will become even more user-friendly.
>>
>> 在2026年1月12日星期一 UTC+8 17:52:29<Frederick Stein> 写道：
>>
>>> I can reproduce it with the same number of MPI ranks but a lower number 
>>> resolves it. I guess you hit a rare corner case of the implementation. I 
>>> still do not understand why it happens.
>>>
>>> yis... at 163.com schrieb am Samstag, 10. Januar 2026 um 12:12:55 UTC+1:
>>>
>>>> thank you very much
>>>>
>>>> 在2026年1月9日星期五 UTC+8 04:05:46<Frederick Stein> 写道：
>>>>
>>>>> I can reproduce it. I will investigate it further.
>>>>>
>>>>> Frederick Stein schrieb am Mittwoch, 7. Januar 2026 um 10:34:37 UTC+1:
>>>>>
>>>>>> Thank you, I will try to reproduce it.
>>>>>>
>>>>>> yis... at 163.com schrieb am Mittwoch, 7. Januar 2026 um 03:54:00 UTC+1:
>>>>>>
>>>>>>> compiled toolchain using `--with-cmake=system --with-gcc=install 
>>>>>>> --with-openmpi=install --with-sirius=no`. The host OS is Rocky Linux 8.10. 
>>>>>>>
>>>>>>> 在2026年1月7日星期三 UTC+8 09:46:21<yis... at 163.com> 写道：
>>>>>>>
>>>>>>>> Regarding this issue, I've provided feedback to the administrators 
>>>>>>>> of the supercomputing center. The SCC admin has now shown me this 
>>>>>>>> diagram—I'm not sure if it fully addresses your question.[image: 
>>>>>>>> 微信图片_20260107093904_62_790.png]
>>>>>>>> 在2026年1月7日星期三 UTC+8 00:15:53<Frederick Stein> 写道：
>>>>>>>>
>>>>>>>>> Ok, serife-less fonts are not a good choice here... Please try `module 
>>>>>>>>> list` after you have loaded all modules to run CP2K.
>>>>>>>>>
>>>>>>>>> yis... at 163.com schrieb am Dienstag, 6. Januar 2026 um 16:37:15 
>>>>>>>>> UTC+1:
>>>>>>>>>
>>>>>>>>>> Enter "module ls" in the command window, and its output is shown 
>>>>>>>>>> in the following figure.
>>>>>>>>>>
>>>>>>>>>> [image: module Is.png]
>>>>>>>>>> 在2026年1月6日星期二 UTC+8 22:18:39<Frederick Stein> 写道：
>>>>>>>>>>
>>>>>>>>>>> If you use the module system, can you provide the result of 
>>>>>>>>>>> `module ls` (or similar command to display the currently loaded modules). 
>>>>>>>>>>> It would be great to have some more information on how CP2K was built. The 
>>>>>>>>>>> behavior you observed appears as a bug in the code or the 
>>>>>>>>>>> compiler/libraries.
>>>>>>>>>>> yis... at 163.com schrieb am Dienstag, 6. Januar 2026 um 14:59:43 
>>>>>>>>>>> UTC+1:
>>>>>>>>>>>
>>>>>>>>>>>> I am using a supercomputer provided by a commercial 
>>>>>>>>>>>> supercomputing center. The CP2K software I am using was compiled by the 
>>>>>>>>>>>> center's staff. Is there anything I can help with?
>>>>>>>>>>>>
>>>>>>>>>>>> 在2026年1月5日星期一 UTC+8 17:56:03<Frederick Stein> 写道：
>>>>>>>>>>>>
>>>>>>>>>>>>> Hi,
>>>>>>>>>>>>> This could be a bug. Did you compile CP2K yourself? If yes, 
>>>>>>>>>>>>> with which compiler?
>>>>>>>>>>>>> Best,
>>>>>>>>>>>>> Frederick
>>>>>>>>>>>>>
>>>>>>>>>>>>> yis... at 163.com schrieb am Dienstag, 30. Dezember 2025 um 
>>>>>>>>>>>>> 17:43:29 UTC+1:
>>>>>>>>>>>>>
>>>>>>>>>>>>>> Dear CP2K developers,
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I would like to report a potential issue regarding the 
>>>>>>>>>>>>>> keyword EPS_DEFAULT. During my tests, I found that its value 
>>>>>>>>>>>>>> affects the calculation in the following way:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>    - 
>>>>>>>>>>>>>>    
>>>>>>>>>>>>>>    With EPS_DEFAULT = 1.0E-12, the program fails with a 
>>>>>>>>>>>>>>    “CPASSERT failed” error.
>>>>>>>>>>>>>>    - 
>>>>>>>>>>>>>>    
>>>>>>>>>>>>>>    With EPS_DEFAULT = 1.0E-11 or 1.0E-13, the calculation 
>>>>>>>>>>>>>>    proceeds normally into the iteration process.
>>>>>>>>>>>>>>    
>>>>>>>>>>>>>> I have attached the input and output files for each test 
>>>>>>>>>>>>>> case. The naming convention is as follows:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>    - 
>>>>>>>>>>>>>>    
>>>>>>>>>>>>>>    1.0E-12_*: corresponds to EPS_DEFAULT = 1.0E-12
>>>>>>>>>>>>>>    - 
>>>>>>>>>>>>>>    
>>>>>>>>>>>>>>    1.0E-11_*: corresponds to EPS_DEFAULT = 1.0E-11
>>>>>>>>>>>>>>    - 
>>>>>>>>>>>>>>    
>>>>>>>>>>>>>>    1.0E-13_*: corresponds to EPS_DEFAULT = 1.0E-13
>>>>>>>>>>>>>>    
>>>>>>>>>>>>>> Thank you for your attention and for all your work on CP2K. I 
>>>>>>>>>>>>>> look forward to your feedback.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>

-- 
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+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/45b0ed1c-8e93-4b52-92f3-9683f6daae7bn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260123/530ddc06/attachment.htm>