Skip to content

Ensure we run VF2PostLayout when needed in optimization level 1#9941

Merged
jakelishman merged 2 commits into
Qiskit:mainfrom
mtreinish:add-property_set_option-check-map
Apr 11, 2023
Merged

Ensure we run VF2PostLayout when needed in optimization level 1#9941
jakelishman merged 2 commits into
Qiskit:mainfrom
mtreinish:add-property_set_option-check-map

Conversation

@mtreinish
Copy link
Copy Markdown
Member

@mtreinish mtreinish commented Apr 11, 2023

Summary

This commit fixes an issue where we were incorrectly not running
VF2PostLayout in the one case where it should be run. Optimization level
1 is more involved than higher optimization levels because for backwards
compatibility it runs multiple different passes to try and find a
perfect layout, first starting with TrivialLayout, then VF2Layout, and
only if those 2 fail to find a perfect match do we run SabreLayout (or
DenseLayout and StochasticSwap if control flow are present). We only
should run VF2PostLayout if we're running the heuristic layout pass.
However, when we integrated routing into SabreLayout the checking for
whether we found a pefect layout in the layout stage stopped working as
expected.

To fix this issue a new flag is added to the CheckMap pass to specify an
alternative property set field to store the results of the check in. The
underlying issue was that the property set field between the earlier
check that we found a perfect layout and the later check whether we
should run a standalone routing pass. To fix this the new flag is used
to separate the results from the multiple CheckMap calls. and then we
only condition the VF2PostLayout condition on the results of the perfect
layout check and not the later routing check.

Details and comments

Fixes #9936

@mtreinish mtreinish added priority: high Changelog: Added Add an "Added" entry in the GitHub Release changelog. Changelog: Fixed Add a "Fixed" entry in the GitHub Release changelog. labels Apr 11, 2023
@mtreinish mtreinish added this to the 0.24.0 milestone Apr 11, 2023
@mtreinish mtreinish requested review from a team and nonhermitian as code owners April 11, 2023 13:37
@qiskit-bot
Copy link
Copy Markdown
Collaborator

Thank you for opening a new pull request.

Before your PR can be merged it will first need to pass continuous integration tests and be reviewed. Sometimes the review process can be slow, so please be patient.

While you're waiting, please feel free to review other open PRs. While only a subset of people are authorized to approve pull requests for merging, everyone is encouraged to review open pull requests. Doing reviews helps reduce the burden on the core team and helps make the project's code better for everyone.

One or more of the the following people are requested to review this:

@coveralls
Copy link
Copy Markdown

coveralls commented Apr 11, 2023

Pull Request Test Coverage Report for Build 4670573548

  • 10 of 10 (100.0%) changed or added relevant lines in 2 files are covered.
  • 1 unchanged line in 1 file lost coverage.
  • Overall coverage decreased (-0.0007%) to 85.432%

Files with Coverage Reduction New Missed Lines %
qiskit/transpiler/passes/synthesis/unitary_synthesis.py 1 95.14%
Totals Coverage Status
Change from base Build 4668920973: -0.0007%
Covered Lines: 67745
Relevant Lines: 79297

💛 - Coveralls

This commit fixes an issue where we were incorrectly not running
VF2PostLayout in the one case where it should be run. Optimization level
1 is more involved than higher optimization levels because for backwards
compatibility it runs multiple different passes to try and find a
perfect layout, first starting with TrivialLayout, then VF2Layout, and
only if those 2 fail to find a perfect match do we run SabreLayout (or
DenseLayout and StochasticSwap if control flow are present). We only
should run VF2PostLayout if we're running the heuristic layout pass.
However, when we integrated routing into SabreLayout the checking for
whether we found a pefect layout in the layout stage stopped working as
expected.

To fix this issue a new flag is added to the CheckMap pass to specify an
alternative property set field to store the results of the check in. The
underlying issue was that the property set field between the earlier
check that we found a perfect layout and the later check whether we
should run a standalone routing pass. To fix this the new flag is used
to spearate the results from the multiple CheckMap calls. and then we
only condition the VF2PostLayout condition on the results of the perfect
layout check and not the later routing check.

Fixes Qiskit#9936
@mtreinish mtreinish force-pushed the add-property_set_option-check-map branch from 429d4b0 to 6071f05 Compare April 11, 2023 14:57
Copy link
Copy Markdown
Member

@jakelishman jakelishman left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All the actual code looks good, there's just some issues in the tests where the comments about the assertions don't actually match the assertions.

Comment thread test/python/transpiler/test_preset_passmanagers.py Outdated
Comment thread test/python/transpiler/test_preset_passmanagers.py
Comment thread test/python/transpiler/test_preset_passmanagers.py Outdated
@mtreinish
Copy link
Copy Markdown
Member Author

Sigh, that's what I get for copy and pasting tests while I was in meetings. Let me remove those inaccurate comments.

@jakelishman jakelishman enabled auto-merge April 11, 2023 18:35
@jakelishman jakelishman added this pull request to the merge queue Apr 11, 2023
Merged via the queue into Qiskit:main with commit 0a50118 Apr 11, 2023
giacomoRanieri pushed a commit to giacomoRanieri/qiskit-terra that referenced this pull request Apr 16, 2023
…it#9941)

* Ensure we run VF2PostLayout when needed in optimization level 1

This commit fixes an issue where we were incorrectly not running
VF2PostLayout in the one case where it should be run. Optimization level
1 is more involved than higher optimization levels because for backwards
compatibility it runs multiple different passes to try and find a
perfect layout, first starting with TrivialLayout, then VF2Layout, and
only if those 2 fail to find a perfect match do we run SabreLayout (or
DenseLayout and StochasticSwap if control flow are present). We only
should run VF2PostLayout if we're running the heuristic layout pass.
However, when we integrated routing into SabreLayout the checking for
whether we found a pefect layout in the layout stage stopped working as
expected.

To fix this issue a new flag is added to the CheckMap pass to specify an
alternative property set field to store the results of the check in. The
underlying issue was that the property set field between the earlier
check that we found a perfect layout and the later check whether we
should run a standalone routing pass. To fix this the new flag is used
to spearate the results from the multiple CheckMap calls. and then we
only condition the VF2PostLayout condition on the results of the perfect
layout check and not the later routing check.

Fixes Qiskit#9936

* Fix remove out of date comments
@mtreinish mtreinish deleted the add-property_set_option-check-map branch April 19, 2023 13:18
king-p3nguin pushed a commit to king-p3nguin/qiskit-terra that referenced this pull request May 22, 2023
…it#9941)

* Ensure we run VF2PostLayout when needed in optimization level 1

This commit fixes an issue where we were incorrectly not running
VF2PostLayout in the one case where it should be run. Optimization level
1 is more involved than higher optimization levels because for backwards
compatibility it runs multiple different passes to try and find a
perfect layout, first starting with TrivialLayout, then VF2Layout, and
only if those 2 fail to find a perfect match do we run SabreLayout (or
DenseLayout and StochasticSwap if control flow are present). We only
should run VF2PostLayout if we're running the heuristic layout pass.
However, when we integrated routing into SabreLayout the checking for
whether we found a pefect layout in the layout stage stopped working as
expected.

To fix this issue a new flag is added to the CheckMap pass to specify an
alternative property set field to store the results of the check in. The
underlying issue was that the property set field between the earlier
check that we found a perfect layout and the later check whether we
should run a standalone routing pass. To fix this the new flag is used
to spearate the results from the multiple CheckMap calls. and then we
only condition the VF2PostLayout condition on the results of the perfect
layout check and not the later routing check.

Fixes Qiskit#9936

* Fix remove out of date comments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Changelog: Added Add an "Added" entry in the GitHub Release changelog. Changelog: Fixed Add a "Fixed" entry in the GitHub Release changelog.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

VF2PostLayout not running in Optimization Level 1

4 participants