I've set up Abinit 6.8.2 on my ubuntu 10.04 64bit system with MPI and CUDA enabled. It compiles and builds without errors, also almost all internal tests (make tests_acc) and (make tests_paral) run successful. Just one fails:
Code: Select all
./tutorespfn/tmp-test_Linux_20111004/report:
Case_elphon_5 failed (too large absolute error : 0.143740000000001 , accepted 2.0e-4 )If I look at the out file and compare with the reference file, there appears a significant difference the first time at these value: (<,> indicate the diff of course...)
Code: Select all
Output of the linewidths for the first point of each segment. Linewidths are given in Hartree.
Q point = 0.000000E+00 0.000000E+00 0.000000E+00 isppol = 1
Mode number Frequency (Ha) Linewidth (Ha) Lambda(q,n)
< 1 0.000000E+00 3.812308E-22 0.000000E+00
< 2 0.000000E+00 -6.022120E-23 0.000000E+00
< 3 0.000000E+00 4.213812E-22 0.000000E+00
---
> 1 0.000000E+00 -1.123744E-21 0.000000E+00
> 2 0.000000E+00 -1.163635E-21 0.000000E+00
> 3 0.000000E+00 -1.384445E-21 0.000000E+00
As I'm a system administrator I have very little understanding of the simulation itself. So I'd really appreciate any hint.
Is this significant? What's going wrong? ...
I checked the "Known_problem" file for anything similar, but there was nothing.
If this is the wrong section of the forum, please tell me as well, however I figured elphon is a DFT method, so I put it here.
Thanks a lot for your time and help
Philipp