@KrisThielemans out of 39 unit tests, 26 work with two AcquisitionData objects (stored in file in 13 tests) currently of size 357 x 126 x 344 = 15 473 808 - with my_forward_projection.hs it takes 20 sec on my laptop

I guess the AcquisitionData tests still take a lot of time (hopefully only the InFile ones). It certainly makes sense to do it with smaller data, as it's just a waste of resources otherwise.

PET AcquisitionData can be created from the scanner name - is it possible to create a small enough one for testing? My guess is that max_ring_diff can help - what about span and view_mash_factor?

To change to smaller dataset it is sufficient to replace the self.image1 and self.image2 in all the test classes.

https://github.com/SyneRBI/SIRF/blob/master/src/xSTIR/pSTIR/tests/test_algebra.py

When I created the tests I just used some ImageData and AcquisitionData that was in the data directory. Clearly this is sub-optimal.

@KrisThielemans now takes 6 sec