[Q-e-developers] hybrid functionals (CVS version of Q-E)

[josepht at chips.ncsu.edu]

Hello,

I just ran the 8-atom Si supercell tests on a Cray machine, using k=4,4,4 and
q=gamma grids, and it indeed works with timings similar to those you report.  I am
not sure why the test fails on the IBM machine (usually I find that the IBM machine
is more robust for pwscf) - is this worthwhile exploring with a different
compilation profile?

A test with 8-atom Si, k=4,4,4 and q=4,4,4, with nimages=4 is running successfully,
so parallelization "pools" over q indeed appears to work now (though I would suggest
using at least 8 processors).

I'm running some more tests and will report anything odd/interesting.  Thanks for
the feedback and fixes.

Joseph Turnbull
NC State University



> Thank you for your tests.
>
>> HSE "images" parallelization doesn't quite work
>
> it should work now. There was a rather stupid problem, I think: many images trying
to delete the same file.
>
>> In cases 6 and 10, the run terminates with "problems computing cholesky
decomposition" piped to 'CRASH'
>
> example 6 works for me on two different machines (my PC and a SP6); I am currently
trying to run example 10 on the sp6, but it is quite a big job.
>
>> I would first like to confirm which pseudopotentials from the repository should
be used
>
> it depends on what you want to test. If you want to test performances only, any
norm-conserving PP will do the job.
>
> P.
> --
> Paolo Giannozzi, Democritos and University of Udine, Italy
>