Add option to 1q euler decomposer for disabling unitary check #5928
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
levbishop
requested changes
Mar 4, 2021
There was a problem hiding this comment.
The idea of having a way to bypass consistency checks internally makes sense to me, but I prefer a different API for this kind of thing:
- Instead of a user-visible flag, can we have an internal
_decompose()method that skips the checks, and have the user-visible__call__do the checks and then go to_decomposefor the (dangerous) implementation - For the internal API can we have the implementation work only with ndarray, and push the
to_operator,to_matrix,np.asarraystuff for the user-facing friendly API (this should give a small additional perf boost)
516c5ec to
9a2711b
Compare
Yeah, that's a good idea and a pattern more consistent with the rest of qiskit too. I've pivoted to use the private method instead of the kwarg in: 9a2711b |
Previously when running the OneQubitEulerDecomposer the first step it would perform is a check that the input matrix is a 1 qubit unitary. However, in situations when we're explicitly generating a unitary matrix and know the input we're passing is unitary there is no need to run this check and it adds unecessary overhead to the process. This commit adds a new kwarg, check_unitary, to the __call__ method on the OneQubitEulerDecomposer which when set to False will skip the unitary check to avoid this overhead. The flag is then set when the decomposer is called from the Optimize1qGatesDecomposition, UnitarySynthesis passes, and TwoQubitBasisDecomposer to avoid the overhead when we know the input is unitary.
This commit pivots to using an inner private method that doesn't have the validation instead of having a kwarg to disable the unitary check. This means the end user doesn't have the option to disable the validation which could potential cause issues but for internal qiskit usage when we know the input is valid we just call the inner method instead of __call__.
9a2711b to
809cfd5
Compare
levbishop
approved these changes
Mar 5, 2021
There was a problem hiding this comment.
LGTM. If its not too much trouble, be nice to post perf numbers before/after to get a sense of how much this kind of thing matters - If this makes a big difference I'll fold the analog for TwoQubitBasisDecomposer into my PR #5827 changes
|
Running the same script as I did in: #5926 (comment) the results for me locally were:
|
This was referenced Mar 5, 2021
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
Previously when running the
OneQubitEulerDecomposerthe first stepit would perform is a check that the input matrix is a 1 qubit unitary.
However, in situations when we're explicitly generating a unitary matrix
and know the input we're passing is unitary there is no need to run this
check and it adds unnecessary overhead to the process. This commit adds a
new kwarg, check_unitary, to the
__call__method on theOneQubitEulerDecomposerwhich when set to False will skip the unitarycheck to avoid this overhead. The flag is then set
when the decomposer is called from the
Optimize1qGatesDecomposition,UnitarySynthesispasses, andTwoQubitBasisDecomposerto avoid theoverhead when we know the input is unitary.
Details and comments
This depends on #5926