Provide some guidelines for how to write QNode passes #2483
Merged
Conversation
Contributor
|
Hello. You may have forgotten to update the changelog!
|
1 similar comment
Contributor
|
Hello. You may have forgotten to update the changelog!
|
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #2483 +/- ##
==========================================
+ Coverage 96.61% 96.76% +0.14%
==========================================
Files 156 156
Lines 16984 17016 +32
Branches 1657 1658 +1
==========================================
+ Hits 16409 16465 +56
+ Misses 421 403 -18
+ Partials 154 148 -6 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
joeycarter
approved these changes
Feb 19, 2026
Contributor
joeycarter
left a comment
joeycarter
left a comment
There was a problem hiding this comment.
Looks good! 💯
I have no notes for now, but may want to revisit next time I write a qnode pass. Thanks @dime10!
dime10
commented
Feb 19, 2026
dime10
commented
Feb 19, 2026
mudit2812
reviewed
Feb 19, 2026
Contributor
mudit2812
left a comment
mudit2812
left a comment
There was a problem hiding this comment.
Let me know if this is clear enough. We could add IR example if additional clarity is needed :)
We already have a google doc with an IR example, so it would be great to move it here if this is going to be a centralized guidance doc. Let me see if I can find the doc and send you a link on Slack.
paul0403
reviewed
Feb 24, 2026
paul0403
approved these changes
Feb 25, 2026
dime10
added a commit
that referenced
this pull request
Mar 6, 2026
…#2497) _Note I don't think this PR is a must, but while working on the [transform post-processing epic](https://app.shortcut.com/xanaduai/epic/107240) one of the ideas was to better formalize some of the attributes we use and might have implicit assumptions for. The verification introduced here is not very strong, but it is something. There are several older passes in Catalyst which currently don't follow the proposed notion of qnode functions (e.g. they might [not propagate the attribute](https://github.com/PennyLaneAI/catalyst/blob/main/mlir/lib/Gradient/Transforms/GradMethods/ParameterShift.cpp) in new functions they create, or they might also have [unusual structures](https://github.com/PennyLaneAI/catalyst/blob/main/mlir/lib/Mitigation/Transforms/MitigationMethods/Zne.cpp) like having MPs in a function but device init/quantum allocs in a different one)._ In this PR, the existing `qnode` attribute is renamed to `quantum.node` to indicate the dialect origin. Additionally, verification logic is introduced to formalize the notion of a "qnode" function (see also #2483), which is characterized by the presence of measurement process operations. These operations represent an implicit transition from quantum state manipulation to execution logic and classical post-processing. While quantum gate operations can appear in functions that are not qnodes, measurement processes should not as they have this "non-local" effect. The attribute helps identify functions with quantum execution semantics during transformations. The new attribute and verification logic is currently implemented during the "quantum kernel" phase only, i.e. before `inline-nested-modules` is called. All functions carrying the attribute ~must have at least one measurement process~ (I have removed this verification for now since Catalyst still uses MP-free QNodes as a feature), and all measurement processes appearing within a nested module (quantum kernel) must be contained in a func op with that attribute. To limit the scope of this PR the new attribute is replaced with the previous `qnode` attribute after lowering the quantum kernel module. A bug in the xDSL split-non-commuting pass is also fixed (required to pass the new verification). 9f75c0e [sc-107492]
multiphaseCFD
pushed a commit
that referenced
this pull request
Mar 6, 2026
…#2497) _Note I don't think this PR is a must, but while working on the [transform post-processing epic](https://app.shortcut.com/xanaduai/epic/107240) one of the ideas was to better formalize some of the attributes we use and might have implicit assumptions for. The verification introduced here is not very strong, but it is something. There are several older passes in Catalyst which currently don't follow the proposed notion of qnode functions (e.g. they might [not propagate the attribute](https://github.com/PennyLaneAI/catalyst/blob/main/mlir/lib/Gradient/Transforms/GradMethods/ParameterShift.cpp) in new functions they create, or they might also have [unusual structures](https://github.com/PennyLaneAI/catalyst/blob/main/mlir/lib/Mitigation/Transforms/MitigationMethods/Zne.cpp) like having MPs in a function but device init/quantum allocs in a different one)._ In this PR, the existing `qnode` attribute is renamed to `quantum.node` to indicate the dialect origin. Additionally, verification logic is introduced to formalize the notion of a "qnode" function (see also #2483), which is characterized by the presence of measurement process operations. These operations represent an implicit transition from quantum state manipulation to execution logic and classical post-processing. While quantum gate operations can appear in functions that are not qnodes, measurement processes should not as they have this "non-local" effect. The attribute helps identify functions with quantum execution semantics during transformations. The new attribute and verification logic is currently implemented during the "quantum kernel" phase only, i.e. before `inline-nested-modules` is called. All functions carrying the attribute ~must have at least one measurement process~ (I have removed this verification for now since Catalyst still uses MP-free QNodes as a feature), and all measurement processes appearing within a nested module (quantum kernel) must be contained in a func op with that attribute. To limit the scope of this PR the new attribute is replaced with the previous `qnode` attribute after lowering the quantum kernel module. A bug in the xDSL split-non-commuting pass is also fixed (required to pass the new verification). 9f75c0e [sc-107492]
kipawaa
pushed a commit
that referenced
this pull request
Mar 10, 2026
…#2497) _Note I don't think this PR is a must, but while working on the [transform post-processing epic](https://app.shortcut.com/xanaduai/epic/107240) one of the ideas was to better formalize some of the attributes we use and might have implicit assumptions for. The verification introduced here is not very strong, but it is something. There are several older passes in Catalyst which currently don't follow the proposed notion of qnode functions (e.g. they might [not propagate the attribute](https://github.com/PennyLaneAI/catalyst/blob/main/mlir/lib/Gradient/Transforms/GradMethods/ParameterShift.cpp) in new functions they create, or they might also have [unusual structures](https://github.com/PennyLaneAI/catalyst/blob/main/mlir/lib/Mitigation/Transforms/MitigationMethods/Zne.cpp) like having MPs in a function but device init/quantum allocs in a different one)._ In this PR, the existing `qnode` attribute is renamed to `quantum.node` to indicate the dialect origin. Additionally, verification logic is introduced to formalize the notion of a "qnode" function (see also #2483), which is characterized by the presence of measurement process operations. These operations represent an implicit transition from quantum state manipulation to execution logic and classical post-processing. While quantum gate operations can appear in functions that are not qnodes, measurement processes should not as they have this "non-local" effect. The attribute helps identify functions with quantum execution semantics during transformations. The new attribute and verification logic is currently implemented during the "quantum kernel" phase only, i.e. before `inline-nested-modules` is called. All functions carrying the attribute ~must have at least one measurement process~ (I have removed this verification for now since Catalyst still uses MP-free QNodes as a feature), and all measurement processes appearing within a nested module (quantum kernel) must be contained in a func op with that attribute. To limit the scope of this PR the new attribute is replaced with the previous `qnode` attribute after lowering the quantum kernel module. A bug in the xDSL split-non-commuting pass is also fixed (required to pass the new verification). 9f75c0e [sc-107492]
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
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.
[sc-107497]