legacysurvey / legacypipe / 9f0ff447-1dac-471f-af76-e9c6438f78ad

52%

main: 52%

push

circleci

Updated runbrick MP init method and logic for how GPU workers are
assigned to blobs.  Blobs + sub-blobs are now sorted.  If no sub-blobs,
sort remains by npix.  If sub-blobs are used, all blobs + sub-blobs are
sorted by number of sources.  Then the top N are assigned a GPU priority
flag for N GPU workers (ngpu * threads-per-gpu).  Those GPU workers will
pull whatever the next blob is off the queue as before after finishing
their priority blob.

2571 of 5840 branches covered (44.02%)

9899 of 18931 relevant lines covered (52.29%)

0.52 hits per line