Like @collate but automatically regroups jobs which were a result of a previous @subdivide / @split (even after intervening @transform )
This is the only way job trickling can work without stalling the pipeline: We would know how many jobs were pending for each @recombine job and which jobs go together.
Like
@collatebut automatically regroups jobs which were a result of a previous@subdivide/@split(even after intervening@transform)This is the only way job trickling can work without stalling the pipeline: We would know how many jobs were pending for each
@recombinejob and which jobs go together.