CP2K 5.1 / IC 2018 regtests

[Alfio Lazzaro]

Thanks! Actually, I see that there are some discussions on the Fortran 
group about ICC2018:

https://groups.google.com/d/msg/comp.lang.fortran/zWrmsDgJaLc/KK6E-VJdAQAJ

ICC2018 is buggy and Intel is working on it...

Alfio

Il giorno martedì 5 dicembre 2017 13:10:10 UTC+1, drahomirko ha scritto:
>
> Hi Alfio,
>
> thanx for suggestions. I have cp2k 5.1 working with IC 2017 alright. No 
> failed tests and 3 wrong results reported due to numerical accuracy 
> (10^-10) which is alright as well. I only posted *sopt because it was 
> fastest way to check on compiler. Already spent day and half on 
> StackOverflow to get the solution, but all the posts there are in regard to 
> change the code routines. In the meantime I will try to make it work with 
> IC 2018 and in case I am successful I will post solution. 
>
> On Tuesday, 5 December 2017 09:59:54 UTC, Alfio Lazzaro wrote:
>>
>> Hi Drahomir,
>> well, I don't have a clear solution to your problem, therefore I can give 
>> you only some suggestions...
>> First of all, I've checked in the CP2K dashboard and there are not tests 
>> based on IC 2018. I see that you are using SOPT version. The corresponding 
>> test with IC 2017 on the dashboard is:
>>
>> http://cp2k-www.epcc.ed.ac.uk/phi/sopt/regtest-0
>>
>> (on top of the log you can find the arch file that is used in the test).
>> I see the the arch file is pretty similar to yours (note that you don't 
>> need -D__INTEL and OBJECTS_ARCHITECTURE = machine_intel.o anymore).
>> A first suggestion would be to use -O0 -g (and remove all other options) 
>> and see if the code runs.
>> It would be also interesting to run under gdb and see where it is 
>> crashing.
>>
>> Alfio
>>
>>
>>
>> Il giorno lunedì 4 dicembre 2017 15:19:37 UTC+1, drahomirko ha scritto:
>>>
>>>
>>> To follow up, I am getting same errors with the CP2K 4.1 that was 
>>> working correctly with IC 2017, so this issue I suspect is due to compiler. 
>>> Any ideas how to fix this?
>>>
>>> Thanx
>>>
>>> Drahomir
>>>
>>> On Monday, 4 December 2017 12:43:57 UTC, drahomirko wrote:
>>>>
>>>>
>>>> Sirs,
>>>>
>>>> I am trying to compile CP2K 5.1 with IC/MKL 2018u1. Compilation runs 
>>>> alright, however I am getting number errors and runtime fails during 
>>>> regtesting.  The errors are (perhaps) due to compiler as they read:
>>>>
>>>> Error in 
>>>> `/calcs/cp2k-5.1/regtesting/Linux-x86-64-intel/sopt/../../..//exe/Linux-x86-64-intel/cp2k.sopt': 
>>>> corrupted double-linked list: 0x000000000c9c43d0 ***
>>>>
>>>> *** Error in 
>>>> `/calcs/cp2k-5.1/regtesting/Linux-x86-64-intel/sopt/../../..//exe/Linux-x86-64-intel/cp2k.sopt': 
>>>> double free or corruption (fasttop): 0x000000000c7db370 ***
>>>>
>>>> *** Error in 
>>>> `/calcs/cp2k-5.1/regtesting/Linux-x86-64-intel/sopt/../../..//exe/Linux-x86-64-intel/cp2k.sopt': 
>>>> free(): corrupted unsorted chunks: 
>>>>
>>>> followed by memory backtrace and runtime fail afterwards. 
>>>>
>>>>
>>>> I am only compiling basic functionality *sopt version at the time to 
>>>> get the compiler settings right.  The makefile I am attaching was working 
>>>> fine with CP2K 4.1 / IC 2017u4
>>>>
>>>> Suggestions are very welcome!
>>>>
>>>> Thank you.
>>>>
>>>> Drahomir
>>>>
>>>>
>>>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20171205/be7816ef/attachment.htm>