rtl-power-fftw
rtl-power-fftw copied to clipboard
add wisdom file option
This just adds a wisdom file flag (-W or --wisdom) so that fftw wisdom may be cached to a user-specified file if desired. This is because the planning stage can take a long time for large single-shot spectra.
Hi cozzyd, thanks for the patch. Before I merge this change, I'd like to address two issues, namely:
-
The use of a wisdom file is a relatively infrequent requirement, and I don't think it warrants the use of a one-letter command line option (-W). I would prefer retaining only the long version (--wisdom). We have been discussing for a while that we were overzealous when initially designing the command line interface and used far too many short options which now seem rather useless (but we hesitate to remove them because we don't want to break existing scripts). So we would not want to "pollute" the namespace of short options any further.
-
The logic you have added seems to always write back the wisdom file, even if there is no new wisdom actually generated. My concern is that this could unnecessarily wear flash-based media (USB keys, SD cards in e.g. Raspberry Pi). Due to the accumulative nature of the FFTW wisdom, the simplest complete solution would be to separate the operation into two parameters: reading and writing (for example
--import-wisdom <file>
and--export-wisdom <file>
).
Hi Andrej,
On Mon, 2020-01-27 at 13:03 -0800, Andrej Lajovic wrote:
Hi cozzyd, thanks for the patch. Before I merge this change, I'd like to address two issues, namely: The use of a wisdom file is a relatively infrequent requirement, and I don't think it warrants the use of a one-letter command line option (-W). I would prefer retaining only the long version (--wisdom). We have been discussing for a while that we were overzealous when initially designing the command line interface and used far too many short options which now seem rather useless (but we hesitate to remove them because we don't want to break existing scripts). So we would not want to "pollute" the namespace of short options any further.
Sure that sounds reasonable!
The logic you have added seems to always write back the wisdom file, even if there is no new wisdom actually generated. My concern is that this could unnecessarily wear flash-based media (USB keys, SD cards in e.g. Raspberry Pi). Due to the accumulative nature of the FFTW wisdom, the simplest complete solution would be to separate the operation into two parameters: reading and writing (for example -- import-wisdom
and --export-wisdom ).
That's a reasonable concern. At the very least, I could add a check to
see if the file has actually changed (perhaps by first planning with
the FFTW_WISDOM flag?). Then we can have three options
--use-wisdom-file
Thanks, -Cosmin
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.