Avoid overhead for synthesized nodes lookup #13424
Merged
Conversation
This file contains 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
After Qiskit#12550 a hash implementation was added to the implementation of DAGOpNode to be able to have identical instances of dag nodes used be usable in a set or dict. This is because after Qiskit#12550 changed the DAGCircuit so the DAGOpNode instances were just a python view of the data contained in the nodes of a dag. While prior to Qiskit#12550 the actual DAGOpNode objects were returned by reference from DAG methods. However, this hash implementation has additional overhead compared to the object identity based version used before. This has caused a regression in some cases for high level synthesis when it's checking for nodes it's already synthesized. This commit addresses this by changing the dict key to be the node id instead of the node object. The integer hashing is significantly faster than the object hashing.
mtreinish
requested review from
alexanderivrii,
ShellyGarion and
a team
as code owners
|
One or more of the following people are relevant to this code:
|
|
Thank you for this addition! This makes more sense. Is this improvement noticeable when benchmarking runtime? |
|
I did a quick asv run and it didn't flag anything as an improvement or a regression. But it is noticeable in some benchpress benchmarks it has a noticeable improvement for example running |
mtreinish
added
the
stable backport potential
Pull Request Test Coverage Report for Build 11800658258Details
💛 - Coveralls |
kevinhartman
approved these changes
Nov 12, 2024
| @@ -382,7 +382,7 @@ def _run( | |||
|
|
|||
| # If the synthesis changed the operation (i.e. it is not None), store the result. | |||
| if synthesized is not None: | |||
| synthesized_nodes[node] = (synthesized, synthesized_context) | |||
| synthesized_nodes[node._node_id] = (synthesized, synthesized_context) | |||
There was a problem hiding this comment.
This looks fine in this case since synthesized_nodes is only used with one DAG (and not e.g. shared by the recursive control flow block handling).
But this optimization is something we should be careful when applying in other places, since the semantics of uniqueness in keys shifts from being global to local within a specific DAG (which means that keys will clobber each other if node indices from more than a single DAG are used in the same map). This was responsible for a bug we had in the visualization code after the initial DAG port to Rust.
mergify bot
pushed a commit
that referenced
this pull request
Nov 12, 2024
After #12550 a hash implementation was added to the implementation of DAGOpNode to be able to have identical instances of dag nodes used be usable in a set or dict. This is because after #12550 changed the DAGCircuit so the DAGOpNode instances were just a python view of the data contained in the nodes of a dag. While prior to #12550 the actual DAGOpNode objects were returned by reference from DAG methods. However, this hash implementation has additional overhead compared to the object identity based version used before. This has caused a regression in some cases for high level synthesis when it's checking for nodes it's already synthesized. This commit addresses this by changing the dict key to be the node id instead of the node object. The integer hashing is significantly faster than the object hashing. (cherry picked from commit 8c6ad02)
github-merge-queue bot
pushed a commit
that referenced
this pull request
Nov 12, 2024
After #12550 a hash implementation was added to the implementation of DAGOpNode to be able to have identical instances of dag nodes used be usable in a set or dict. This is because after #12550 changed the DAGCircuit so the DAGOpNode instances were just a python view of the data contained in the nodes of a dag. While prior to #12550 the actual DAGOpNode objects were returned by reference from DAG methods. However, this hash implementation has additional overhead compared to the object identity based version used before. This has caused a regression in some cases for high level synthesis when it's checking for nodes it's already synthesized. This commit addresses this by changing the dict key to be the node id instead of the node object. The integer hashing is significantly faster than the object hashing. (cherry picked from commit 8c6ad02) Co-authored-by: Matthew Treinish <[email protected]>
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
After #12550 a hash implementation was added to the implementation of DAGOpNode to be able to have identical instances of dag nodes used be usable in a set or dict. This is because after #12550 changed the DAGCircuit so the DAGOpNode instances were just a python view of the data contained in the nodes of a dag. While prior to #12550 the actual DAGOpNode objects were returned by reference from DAG methods. However, this hash implementation has additional overhead compared to the object identity based version used before. This has caused a regression in some cases for high level synthesis when it's checking for nodes it's already synthesized. This commit addresses this by changing the dict key to be the node id instead of the node object. The integer hashing is significantly faster than the object hashing.
Details and comments