Unitary Synthesis of ChoiMixTableau for Diagonalisation (Again) #1320
Conversation
| ChoiMixTableau::col_key_t ctrl = tab.col_index_.right.at(op.first); | ||
| ChoiMixTableau::col_key_t trgt = tab.col_index_.right.at(op.second); | ||
| in_circ.add_op<Qubit>(OpType::CX, {ctrl.first, trgt.first}); | ||
| // DEBUG METHOD: Ignore this for coverage checks |
There was a problem hiding this comment.
This comment was more intended for any developer who is scanning through coverage reports to check for ways to improve their coverage. I don't know of any simple method for ignoring methods from gcovr - if you have any suggestions please let me know! I'd rather not have to lift this entire class structure up to a header and move this to test files, as the class is just a helper utility to perform a single method
There was a problem hiding this comment.
Ah I see. You should be able to use exclusion markers: https://gcovr.com/en/stable/guide/exclusion-markers.html .
There was a problem hiding this comment.
Thanks for the suggestion, I have added the markers and checked it ignores it locally. The CI errors seem to be from a quimb update which looks independent from this PR and will need some looking into
There was a problem hiding this comment.
Thanks for the suggestion, I have added the markers and checked it ignores it locally. The CI errors seem to be from a quimb update which looks independent from this PR and will need some looking into
Ah yes, we had this issue on the pytket-docs repo as well. Should get this fixed ASAP.
Description
#941 implemented a whole bunch of stuff: new features, bug fixes, reorganised code... and some unexpected regression for circuits of pauli gadgets. #1201 reverted the whole PR to temporarily fix the regression. This PR reintroduces the new features, bug fixes, and reorganised code without changing any of the observable behaviour for pytket users.
Related issues
#1199 and #1202.
Note that in #1202, the examples show that both the old and new synthesis methods had situational benefits over each other. The new behaviour would still be available (toggle the
allow_matching_finalparameter ofreduce_overlap_of_paulis), though analysis of which is better for which situation is left for future experimentation (likely best done through naively running with both and comparing effectiveness).Checklist