Open sebastiangomezlopez opened 11 months ago
Hi @sebastiangomezlopez, pycbc_multi_inspiral
already includes the FFT and processing scheme CLI options you mention. You can verify this using the --help
option. You will not see those options explicitly added in the pycbc_multi_inspiral
executable script, because they are added by module-specific functions, specifically scheme.insert_processing_option_group(parser)
and fft.insert_fft_option_group(parser)
.
Your other point about the help strings is valid though.
I'm opening this new issue because:
The pycbc_multi_inspiral executable is currently lacking a more thorough explanation of several argaparse arguments related to the chi-squared statistic, in particular around this line of code. There are no "help" kwargs
Moreover, it would be nice to add several argparse.add_argument, for several fftw options that we'd like to take a look into from the performance side, so that the following flagged options become usable in the future:
--fft-measure-level -> (int [0,1,2,3]) --fftw-import-system_wisdom -> (flag) --fftw-input-float-wisdom-file -> (str) --fftw-input-double-wisdom-file -> (str) --fftw-threads-backend -> (str ['openmp','pthreads', 'unthreaded'])
argparse.add_argument("--processing-scheme", help=help = " takes as input the processing method " " as name:N. Where name can be either" " mkl or cpu, and N is the number of threads" " It can be set to cuda for GPU processing")
Since we are exploring the performance of mkl with multiple threads and plan to take a look into GPU stuff in the future, it'd be nice if the help string of processing scheme described how one can select cpu, mkl and cuda, as well as the syntax for selecting multiple threads when using cpu or mkl.