[Octopus-notify] svn commit: r2825 - trunk/testsuite/finite_systems_3d by acastro

[octopus-notify at tddft.org]

      Author: acastro
        Date: Tue Apr 10 15:11:58 2007
New Revision: 2825
   Changeset: http://www.tddft.org/trac/octopus/changeset/2825

Modified:
   trunk/testsuite/finite_systems_3d/14-h2o_pol_lr.02_h2o_dyn_pol.inp
   trunk/testsuite/finite_systems_3d/14-h2o_pol_lr.03_h2o_dyn_pol_cmplx.inp

Log:
This was another case in which putting a too large tolerance criterion
for some parameter in one of the testsuite input files led to
different results depending on which compiler flags were used.

This way, intel 9.1 with optimization flags on ia32 architecture
produced a failure of the test, whereas most other choices produced
the correct result.

Decreasing PolConvAbsDens from 1.0e-3 to 1.0e-4 leads to correct
results no matter the options (it seems).

I tried digging in the code to find out where the numbers diverged
more notably, but I could not find a specific place.