It seems the program is not fully parallelized through OpenMP, just want to confirm I didn't misconfigure.
Hi, I find that the program isn't using all the OpenMP threads available in most of the time. If I specify OMP_NUM_THREADS = 4.
The number of data references issued from each thread for 'SWFFT/build.openmp/TestDfft 1 200 200 200' would be:
0 : 1250484419
1 : 261089358
2 : 264851092
3 : 271099842
I just want to make sure that I didn't misconfigure the program. FFTW library is compiled with --enable-openmp flag.
Regards, Chen