[CP2K-user] [CP2K:22402] Re: Error in BS calculations with HSE06 functional: required libint was not linked with CP2K

[Megaa M]

Hi Frederick,

Thank you for the advice. I'll ask the admin team see if they could install 
version 2026.2.

Cheers,
Weiyi

在2026年8月13日星期四 UTC+10 18:15:12<Frederick Stein> 写道：

> Hi Weiyi,
> Sorry, you need version 2026.2 instead of version 2026.1. The fix was 
> merged soon after the release of 2026.1. Otherwise, I do not know what is 
> going on with MKL. It could be an invalid value which cannot be 
> investigated without any further information or a debug build. Because CP2K 
> was not compiled with support for COSMA, you also cannot try a different 
> library.
> Best,
> Frederick
>
> Megaa M schrieb am Donnerstag, 13. August 2026 um 08:57:08 UTC+2:
>
>> Hi Frederick,
>>
>> Thanks for pointing out. I've added the output and log files in the 
>> attachment.
>>
>> When I used the EPS_DEFAULT = 1.0E-11, the error message showed up with  
>> `CPASSERT failed`. For more details, please refer to the document named 
>> 'EPS_DEFAULT_E-11_cp2k_out.txt'. Then I decreased the parameter to 
>> EPS_DEFAULT = 1.0E-13, there was no error messages in the out file 
>> ('EPS_13_cp2k_out.txt'), but all cores failed within MKL as you mentioned 
>> in the previous group chat. Could you please share some suggestions about 
>> that? 
>>
>> Many thanks !
>>
>> Cheers,
>> Weiyi
>>
>>
>> 在2026年8月12日星期三 UTC+10 16:33:54<Stein, Dr. Frederick> 写道：
>>
>>> Hi Weiyi, 
>>> Do you also have a CP2K output file? Just throwing a `CPASSERT failed` 
>>> is 
>>> not a useful error message (we used it everywhere in the code) and 
>>> according 
>>> to the stack trace, all cores fail within MKL. 
>>> Best, 
>>> Frederick 
>>>
>>> Am Tue, 11 Aug 2026 21:17:48 -0700 (PDT) schrieb Megaa M 
>>> <chrisz... at gmail.com>: 
>>> > Hi Frederick, 
>>> > 
>>> > I've switched to cp2k/2026.1 and run the input scripts. Unfortunately, 
>>> it 
>>> > still returned error like "CPASSERT failed". Then I tried to decrease 
>>> > EPS_DEFAULT from 1.0E-11 to 1.0E-13 as suggested. It can run but 
>>> terminate 
>>> > at some point. 
>>> > 
>>> >For your reference, I've added the log files in the attachment. Please 
>>> have 
>>> > a look. 
>>> > 
>>> > Just wondering that do you have any advice on this? 
>>> > 
>>> > Cheers, 
>>> > Weiyi 
>>> > 
>>> > 在2026年8月11日星期二 UTC+10 14:07:40<Megaa M> 写道： 
>>> > 
>>> >> Hi Frederick, 
>>> >> 
>>> >> Thank you for the advice. Will try on 2026 version then. 
>>> >> 
>>> >> Cheers, 
>>> >> Weiyi 
>>> >> 
>>> >> 在2026年8月10日星期一 UTC+10 18:01:26<Frederick Stein> 写道： 
>>> >> 
>>> >>> Dear Megaa, 
>>> >>> This bug is known and was fixed in version 2026.1. So I recommend 
>>> you to 
>>> >>> switch to version 2026.1 or 2026.2. If this is not possible, try a 
>>> >>> different number of MPI ranks or a different value for EPS_DEFAULT 
>>> (see 
>>> >>> 
>>> https://groups.google.com/g/cp2k/c/y_tap__2q4M/m/3LBtjlWdEQAJ?utm_medium=email&utm_source=footer 
>>> >>> ). 
>>> >>> Best, 
>>> >>> Frederick 
>>> >>> 
>>> >>> Megaa M schrieb am Montag, 10. August 2026 um 05:48:32 UTC+2: 
>>> >>> 
>>> >>>> Hi Frederick 
>>> >>>> 
>>> >>>> Thank you for the advice. 
>>> >>>> I tried to run the regtest as you suggested, it runs successfully. 
>>> I 
>>> >>>> checked my input scripts and I guess the error showed up was 
>>> because I used 
>>> >>>> different basis sets in the PBE and HSE06 input scipts. 
>>> >>>> 
>>> >>>> Then I changed the basis set used in the PBE calculation, make it 
>>> aligns 
>>> >>>> with the HSE06 script. However, at this time it gave me errors like 
>>> >>>> "CPASSERT failed". I've uploaded the log files in the attachment. 
>>> Please 
>>> >>>> have a look. 
>>> >>>> 
>>> >>>> Just wondering that do you have any idea about this error? 
>>> >>>> 
>>> >>>> I reckon this happened was probably because I added the auxiliary 
>>> basis 
>>> >>>> set ADMM in the HSE06 basis set. So I deleted the ADMM basis set 
>>> and then 
>>> >>>> it returned some errors like "AUX_FIT basis set is not defined." I 
>>> >>>> guess to do this calculation, the ADMM basis set is compulsory. 
>>> >>>> 
>>> >>>> Please let me know how you feel. 
>>> >>>> 
>>> >>>> Cheers, 
>>> >>>> Weiyi 
>>> >>>> 在2026年8月7日星期五 UTC+10 23:01:18<Frederick Stein> 写道： 
>>> >>>> 
>>> >>>>> Hi Weiyi, 
>>> >>>>> Something seems off. Have you run both calculations with exactly 
>>> the 
>>> >>>>> same executable? Can you show both output files? This kind of 
>>> behavior 
>>> >>>>> (runs smoothly without restart file and fails with missing 
>>> libraries with 
>>> >>>>> the given error message with restart file) cannot happen because 
>>> CP2K needs 
>>> >>>>> to call Libint in both cases and both code required for this 
>>> behavior paths 
>>> >>>>> exclude each other for the same executable. It must have been two 
>>> different 
>>> >>>>> executables (conditional compilation). 
>>> >>>>> I have just tried with a current development version to restart a 
>>> HF 
>>> >>>>> calculation (tests/QS/regtest-kp-hfx-ri-2/diamond_gapw_tc.inp) 
>>> from the 
>>> >>>>> corresponding PBE calculation and was successful. Can you try 
>>> that? 
>>> >>>>> Best, 
>>> >>>>> Frederick 
>>> >>>>> 
>>> >>>>> Megaa M schrieb am Donnerstag, 6. August 2026 um 07:51:51 UTC+2: 
>>> >>>>> 
>>> >>>>>> Hi Frederick, 
>>> >>>>>> 
>>> >>>>>> It seems that the admin team has done the regression test during 
>>> the 
>>> >>>>>> installation of the package. They excuted: 
>>> >>>>>> make ARCH=Linux-x86-64-gnu-nci VERSION=psmp test 
>>> TESTOPTS+=\"--mpiranks 
>>> >>>>>> 4 --ompthreads 1 --maxtasks 4 --timeout 900\" make 
>>> >>>>>> ARCH=Linux-x86-64-gnu-nci VERSION=psmp test 
>>> TESTOPTS+=\"--mpiranks 2 
>>> >>>>>> --ompthreads 1 --maxtasks 4 --timeout 900\" 
>>> >>>>>> 
>>> >>>>>> Just wondering that do you have any advise for that? 
>>> >>>>>> 
>>> >>>>>> Many thanks ! 
>>> >>>>>> 
>>> >>>>>> Cheers, 
>>> >>>>>> Weiyi 
>>> >>>>>> 
>>> >>>>>> 在2026年8月5日星期三 UTC+10 19:22:49<Frederick Stein> 写道： 
>>> >>>>>> 
>>> >>>>>>> Dear Weiyi, 
>>> >>>>>>> I am not quite sure. Officially, CP2K supported Libint version 
>>> 2.6.0 
>>> >>>>>>> at that time. It is also not clear why the calculation should 
>>> fail with 
>>> >>>>>>> restart file but not without because the libint wrapper should 
>>> be called in 
>>> >>>>>>> both cases. Have you or your admin team the regression test 
>>> suite after 
>>> >>>>>>> compilation? 
>>> >>>>>>> Best, 
>>> >>>>>>> Frederick 
>>> >>>>>>> 
>>> >>>>>>> Megaa M schrieb am Mittwoch, 5. August 2026 um 09:21:06 UTC+2: 
>>> >>>>>>> 
>>> >>>>>>>> Dear Community, 
>>> >>>>>>>> 
>>> >>>>>>>> I'm trying to calculate the band structure of a cubic bulk 
>>> material 
>>> >>>>>>>> using the HSE06 hybrid functional with CP2K 2025.2. 
>>> >>>>>>>> 
>>> >>>>>>>> With relaxed structure, I first performed a PBE energy 
>>> calculation 
>>> >>>>>>>> with k-point sampling and obtained a wavefunction file named 
>>> ENERGY-PBE.kp. 
>>> >>>>>>>> I then attempted to use this file as the initial wavefunction 
>>> guess for the 
>>> >>>>>>>> HSE06 calculation. 
>>> >>>>>>>> 
>>> >>>>>>>> However, when the HSE06 calculation reads the ENERGY-PBE.kp 
>>> file and 
>>> >>>>>>>> tries to run, CP2K terminates with the following error: 
>>> >>>>>>>> 
>>> >>>>>>>> "This CP2K executable has not been linked against the required 
>>> >>>>>>>> library libint." 
>>> >>>>>>>> 
>>> >>>>>>>> The complete output is attached as cp2k_out.txt. 
>>> >>>>>>>> 
>>> >>>>>>>> Interestingly, when I run the same HSE06 calculation without 
>>> reading 
>>> >>>>>>>> the PBE wavefunction file, the calculation starts successfully. 
>>> However, 
>>> >>>>>>>> convergence is much slower and the calculation requires 
>>> substantially more 
>>> >>>>>>>> computational time. 
>>> >>>>>>>> 
>>> >>>>>>>> I contacted the staff responsible for the HPC system, but they 
>>> were 
>>> >>>>>>>> unable to determine the cause of the error. They provided 
>>> details of the 
>>> >>>>>>>> installed CP2K and Libint configurations, which are attached as 
>>> >>>>>>>> "cp2k_2025.2_libint.txt". Based on the available information, 
>>> it seems that 
>>> >>>>>>>> the hfx related libint appears to absent (Tho the HSE06 
>>> calculation without 
>>> >>>>>>>> reading PBE.kp file can still run). 
>>> >>>>>>>> 
>>> >>>>>>>> Can you please advise whether this problem may be related to 
>>> the 
>>> >>>>>>>> CP2K compilation, the compatibility of the PBE wavefunction 
>>> file with the 
>>> >>>>>>>> HSE06 calculation, or a particular input setting? Is there 
>>> anything wrong 
>>> >>>>>>>> during the installation? Did they miss some files that were 
>>> supposed to be 
>>> >>>>>>>> downloaded? If yes, is there any way for them to get the files? 
>>> >>>>>>>> 
>>> >>>>>>>> Much appreciate for any guidance on this. 
>>> >>>>>>>> 
>>> >>>>>>>> Kind regards, 
>>> >>>>>>>> Weiyi 
>>> >>>>>>>> 
>>> >>>>>>> 
>>> > -- 
>>> > 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... at googlegroups.com. 
>>> > To view this discussion visit 
>>> >
>>> https://groups.google.com/d/msgid/cp2k/df8b39a6-df13-4bbe-b836-48edf68df54en%40googlegroups.com. 
>>>
>>>
>>> Dr. Frederick Stein 
>>> Scientific Computing Core 
>>> CASUS - Center for Advanced Systems Understanding, Görlitz 
>>> Helmholtz-Zentrum Dresden-Rossendorf e.V. (HZDR) 
>>>
>>

-- 
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/b24c891d-36bf-4c33-afd9-7b35d55eecb1n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260819/e67b30f4/attachment-0001.htm>