Parallel version

[mariavd]

This is not an issue but a question. I compiled with the --omp flag and submitted a calculation of the current density in 3D on 8 cores, specifying

export OMP_NUM_THREADS=8
export OMP_STACKSIZE=16M

Is it supposed to print that it indeed is running on 8 cores? The job is rather big, and ETA does not fit into the format specifier, so I won't be able to tell if it indeed ran faster than serial.

[heikef]

You can also add --blas to speed it up when compiling.

[heikef]

#170

[bast]

Right, the timing is not correct in case of OMP. As to BLAS, it is very important to use optimized blas, most probably MKL but for this we need to fix GIMIC issues on Intel.

[mariavd]

Ok, thanks. Possibly it might be useful to print an extra line with the number of threads in use but this is of course just a suggestion.

[mariavd]

Just to let you know that I added a line to print the ETA in hours in my fork. It looks like this:

 INFO: Estimated CPU time for single core calculation:  4781999.10 sec (1328.3 h )

I could make a pull request in case you find it useful. You don't have to, of course, since it's barely significant.

[bast]

I think this is useful.

[mariavd]

This issue should have been closed by #178. My fault that I didn't mark it.