The numbers for the benchmark test in https://github.com/davidthomas5412/PanglossNotebooks/blob/master/GroupMeeting12_5_16.ipynb are encouraging (factor of 200 speed-up) but you should show that the two codes get the same answers too. Maybe you can avoid computing correlation functions etc by computing the same set of predicted background galaxy shears with the two codes, and investigating the component-wise differences? They should be at least very tightly correlated if not actually equal. Thoughts?
The numbers for the benchmark test in https://github.com/davidthomas5412/PanglossNotebooks/blob/master/GroupMeeting12_5_16.ipynb are encouraging (factor of 200 speed-up) but you should show that the two codes get the same answers too. Maybe you can avoid computing correlation functions etc by computing the same set of predicted background galaxy shears with the two codes, and investigating the component-wise differences? They should be at least very tightly correlated if not actually equal. Thoughts?