2022-05-11: APP 2.0.0-beta 1 has been released !!!
Download links per platform:
Is this expected behavior? Single core performance on mosaic registration...
I have a very small data set: about 49 light frames from an ASI1600 camera. Some of them were captured using a different telescope. I am trying to register the subframes with the mosaic setting, but it seems like APP is doing the dynamic distortion calculation using just a single core and is running super slowly. It takes more than 5 hours to complete registration with DDC enabled. You can see in the screenshot only about 105% of the 8 core CPU is being used, which in macOS indicates single core operation (if 200% is being used, that's 2 cores of the CPU being used, and so on..). Is this expected?
This could be because of the process yes, only @mabula-admin can answer that, but APP does report to be using 7 threads. So if it isn't, it might be it doesn't do it at this stage internally. Lets see what Mabula says..
During the multi view registration, my multi-core PC is using all the threads I've said. However during this specific task of the registration where it can't find a region to map and is decreasing the "trust" - it could be that it is actually just using one thread - it seems that your mosaic doesn't have a region it can combine: the "Decreasing Trust Region" comes up (for me at least) when it is struggling to find areas of commonality in your mosaic.
In your Activity Monitor you can see the Thread Count for that APP process is 155. Obviously your PC doesn't have that many threads available, so APP is multi-threaded, but just isn't using it for this part of the calculation. And in this case, your data is very likely never going to combine into a Mosaic because there isn't an overlapping region.
I had the same scenario on Linux - I left a mosaic running thinking my regions were good, but it was perpetually looping with messages about "Decreasing Trust Region"
Here's Mosaic Registration under Linux showing all allocated cores running hot:
Update to my post above thinking it is a specific calculation based on no-overlapping data: I can concur/agree that the DDC calculation is using a single thread as per OP:
The bold thread is 99%, rest are idle