Homec4science

fixing fftwmpi tests

Authored by frerot on Dec 1 2020, 21:23.

Description

fixing fftwmpi tests

It is unreasonable to test fft parallel vs sequential with high tolerance. The
errors I found where in places of the complex output where the transform should
be zero, but in both seq and par impl they are of the order of ~1e-13. The exact
value depends on the order of operations, and one should not expect exact
comparison of par and seq here .

Details

Committed
frerotDec 1 2020, 21:23
Pushed
frerotDec 1 2020, 21:26
Parents
rTAMAAS4022e3acd5e6: testing parallel pkr with hertz instead of sequential pkr
Branches
Unknown
Tags
Unknown