On Mon, May 21, 2018 at 2:32 PM, Pawel Dziekonski <[email protected]>
wrote:

OpenMPI 2.1.3, GCC 7.3.0, netlib BLAS/LAPACK:
> ./configure --enable-openmp --enable-parallel --with-netlib
> CP:
> https://pastebin.com/x4dqkxCA


Small numerical differences. Different mathematical libraries and execution
modes may lead to slightly different results.


> PW:
> https://pastebin.com/hgjMuYCK


The "Langevin" test gives results that depend upon the number of processors
used and should be discarded.
The "Berry" and many "vdw" tests fail for reasons that is impossible to
know without looking at the outputs
All the remaining cases fall in the "small numerical differences" case.

Intel parallel studio xe 2016.3.067 (I'm aware of MKL bug):
>
./configure --enable-parallel
>
CP:
>
https://pastebin.com/8F49sL6u
>

Same as above for CP

PW:
>
https://pastebin.com/njFEmt0M
>

same as above for PW

BTW, when I start compilation by
>
make all
>
only some targets are build (mostly cp, nothing from pwall).
>
Is this expected?
>

no

Paolo
-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
_______________________________________________
users mailing list
[email protected]
https://lists.quantum-espresso.org/mailman/listinfo/users

Reply via email to