Move 2q block-consolidation logic to qiskit-synthesis#15912
Move 2q block-consolidation logic to qiskit-synthesis#15912Cryoris merged 1 commit intoQiskit:mainfrom
qiskit-synthesis#15912Conversation
|
One or more of the following people are relevant to this code:
|
1 similar comment
|
One or more of the following people are relevant to this code:
|
jakelishman
force-pushed
the
circuit-free-qi/5
branch
from
1ebe456 to
037b1c7
Compare
jakelishman
added
on hold
This is dependent on both the circuit IRs _and_ the `quantum-info` objects. All of these functions are primarily used in synthesis, which makes `qiskit-synthesis` a relatively natural place for them; they're not _strictly_ about immediate synthesis of a matrix into circuit objects, but they are very much along the path to that.
jakelishman
force-pushed
the
circuit-free-qi/5
branch
from
037b1c7 to
32e13ea
Compare
ShellyGarion
left a comment
ShellyGarion
left a comment
There was a problem hiding this comment.
Thanks for the effort! Looks good!
two_qubit.rs is used mostly in the synthesis code for 2-qubit and QSD decomposition, but also in some tranpiler passes (consolidate_blocks.rs). does this dependency makes sense?
|
Yeah, |
ShellyGarion
left a comment
ShellyGarion
left a comment
There was a problem hiding this comment.
LGTM. thanks for the effort!
There was a problem hiding this comment.
This file also contains 1-qubit logic, the filename doesn't really reflect that -- but I also don't see a better name right now since it's a collection of quite some functions. What do you think of just moving it to mod.rs? Or do you have another idea?
An overkill idea could be to further split it lol
There was a problem hiding this comment.
Yeah, I didn't realise that it had other public 1q logic in it til something like the last compiler-error fix I did. Maybe it might be better to follow-up? (I mean, the file before was called convert_2q_block_matrix, so it's not exactly like the situation got worse lol)
There was a problem hiding this comment.
Yeah it's also a mild complaint since it's not public anyways... sure we can do it in a follow up (or later when we get annoyed bit it at some point) 😄
4 participants
This is dependent on both the circuit IRs and the
quantum-infoobjects. All of these functions are primarily used in synthesis, which makesqiskit-synthesisa relatively natural place for them; they're not strictly about immediate synthesis of a matrix into circuit objects, but they are very much along the path to that.