search

dstndstn / astrometry.net

Astrometry.net -- automatic recognition of astronomical images
http://astrometry.net
Other
650 stars 184 forks source link

solve-field: Progressive Slowdown Observed #295

Closed michaelbaisch closed 2 months ago

michaelbaisch commented 2 months ago

Greetings,

While monitoring the command line output of the locally running solve-field command, I noticed something peculiar: it appears to slow down with each subsequent field object set, or possibly with each attempt that fails to find a solution. This slowdown becomes obvious in the screencast. The indices for file objects 1-10 run much faster than those for field objects 11-20 and beyond. The index files towards the end of the 1-10 run are already slower, likely because they are larger, but this is minor compared to the significant slowdown towards the end observed with the 11-20 objects. Given that each batch contains 10 objects, I see no obvious reason for such a marked decrease in processing speed.

Is this expected behavior, or is there potential for optimization to improve this?

https://github.com/dstndstn/astrometry.net/assets/1899691/57f1269d-c5f1-459f-8624-cb89fe27f217

macOS 14.4.1, M3 Max solve-field: Revision 0.94

dstndstn commented 2 months ago

This is expected. When you add more field objects (stars), it tries all combinations of those stars plus the previous stars, so the number of combinations goes up very very quickly. If you look for the astrometry.cfg config file and turn on the "inparallel" mode and show more debugging output (add "-v" to solve-field), it should print the number of quadrangles tried and matched.