Frealign mp version crashing
Forums
I was testing frealign_v8_mpfftw with 4 CPUS as I wrote my earlier note and knew that it wasn't running through the data any faster than frealign_v8_fftw would. However, it has now finished the per particle search and seems to be burning all the CPU of two processors but accomplishing nothing. It has been sitting in the state for 5 or 10 minutes by now. I also probably should have mentioned that I am using the pre-compiled versions that came in the download...
The _mpfftw and _mp versions
The _mpfftw and _mp versions of Frealign were pre-compiled using the Portland compiler and static flag -Bstatic_pgi. This means that PGI libraries are statically linked, while system libraries are dynamic linked. These versions of Frealign appear to run without problems on many systems but, as your example shows, fail on some systems. If you can recompile the binaries on your system, I assume they will work.
Furthermore, there appears to be a problem with the _mpfftw version of Frealign. This version does not run properly on some of our systems, either. The fftw libraries do not appear to give much of an speed advantage. The most time-consuming subroutines of Frealign are the interpolation routines, which do not use FFTs. I would, therefore, recommend not using the fftw versions. The fftw option was an experiment and will not be supported anymore in future versions.