Oxidize `UnitarySynthesis` pass #13141

ElePT · 2024-09-12T10:05:09Z

Summary

The code is an almost literal copy of the Python one (not very rusty), but it's a start.

Details and comments

TODO:

Add unit tests for qsd path and 1q synthesis path
Address drawer issue after calling dag_to_circuit (not reflected in unit tests)
Clean up comments
Review use of dag.push_back and see if it can be done with the new dag.apply_operation_back method

coveralls · 2024-09-12T12:12:15Z

Pull Request Test Coverage Report for Build 10922534515

Details

736 of 796 (92.46%) changed or added relevant lines in 7 files are covered.
160 unchanged lines in 4 files lost coverage.
Overall coverage increased (+0.4%) to 88.769%

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
crates/accelerate/src/two_qubit_decompose.rs	11	15	73.33%
crates/accelerate/src/unitary_synthesis.rs	709	765	92.68%

Files with Coverage Reduction	New Missed Lines	%
crates/accelerate/src/two_qubit_decompose.rs	1	91.68%
crates/qasm2/src/lex.rs	2	92.48%
crates/qasm2/src/parse.rs	6	96.23%
qiskit/transpiler/passes/synthesis/unitary_synthesis.py	151	62.15%

Totals
Change from base Build 10919208944:	0.4%
Covered Lines:	74129
Relevant Lines:	83508

💛 - Coveralls

…unitary-synthesis

ElePT · 2024-09-13T14:18:05Z

Performance is starting to look promising:

Benchmarks that have improved:

| Change   | Before [2fa6dd65] <main>   | After [739a5c84] <oxidize-unitary-synthesis-benchmarking-3>   |   Ratio | Benchmark (Parameter)                                              |
|----------|----------------------------|---------------------------------------------------------------|---------|--------------------------------------------------------------------|
| -        | 1.30±0.04s                 | 1.17±0.02s                                                    |    0.9  | utility_scale.UtilityScaleBenchmarks.time_qaoa('ecr')              |
| -        | 483±60ms                   | 375±2ms                                                       |    0.78 | utility_scale.UtilityScaleBenchmarks.time_square_heisenberg('cx')  |
| -        | 695±90ms                   | 476±20ms                                                      |    0.69 | utility_scale.UtilityScaleBenchmarks.time_square_heisenberg('cz')  |
| -        | 705±100ms                  | 481±10ms                                                      |    0.68 | utility_scale.UtilityScaleBenchmarks.time_square_heisenberg('ecr') |

SOME BENCHMARKS HAVE CHANGED SIGNIFICANTLY.
PERFORMANCE INCREASED.

ElePT · 2024-09-16T11:08:31Z

crates/accelerate/src/unitary_synthesis.rs

+                        let new_qargs = dag.qargs_interner.get(packed_instr.qubits).to_vec();
+                        let mut owned_instr = packed_instr.clone();
+                        owned_instr.qubits = out_dag.qargs_interner.insert_owned(new_qargs);
+                        let _ = out_dag.push_back(py, owned_instr);


This (and all other places where dag.push_back is used) could be replaced by #13143.

ElePT · 2024-09-16T13:35:42Z

Benchmarking status as of eb815d1 (more benchmarks show improvement but the improvement looks slightly smaller).

| Change   | Before [2fa6dd65] <unitary-test>   | After [266a22f0] <oxidize-unitary-synthesis>   |   Ratio | Benchmark (Parameter)                                              |
|----------|------------------------------------|------------------------------------------------|---------|--------------------------------------------------------------------|
| -        | 1.38±0.04s                         | 1.21±0.03s                                     |    0.87 | utility_scale.UtilityScaleBenchmarks.time_qaoa('cz')               |
| -        | 425±8ms                            | 361±8ms                                        |    0.85 | utility_scale.UtilityScaleBenchmarks.time_square_heisenberg('cx')  |
| -        | 566±10ms                           | 467±10ms                                       |    0.83 | utility_scale.UtilityScaleBenchmarks.time_square_heisenberg('cz')  |
| -        | 3.94±0.2s                          | 3.04±0.02s                                     |    0.77 | utility_scale.UtilityScaleBenchmarks.time_qv('ecr')                |
| -        | 4.21±0.1s                          | 3.15±0.04s                                     |    0.75 | utility_scale.UtilityScaleBenchmarks.time_qv('cz')                 |
| -        | 626±40ms                           | 456±4ms                                        |    0.73 | utility_scale.UtilityScaleBenchmarks.time_square_heisenberg('ecr') |

qiskit-bot · 2024-09-16T16:06:21Z

One or more of the following people are relevant to this code:

@Qiskit/terra-core
@kevinhartman
@levbishop
@mtreinish

mtreinish

I took a first pass through this, thanks for doing this @ElePT! I left some inline comments. The two high level ones that stuck out the most to me from my first pass is that this PR seems to make a lot of things public, some of them make sense like the dagcircuit methods, but others I'm skeptical we need to be reusing or accessing outside of their definition. Some of the internal attributes of structures for example.

The other thing that I think is potentially a bigger issue especially for performance is around some of the typing choices and how that leads to a lot of conversions. The most concrete example is things are normalized on DAGCircuit between all the different decomposer, but this leads to creating an expensive object to be constructed when it'd be far more efficient to iterate over the synthesized sequence directly and add the gates directly to the dag.

crates/accelerate/src/target_transpiler/nullable_index_map.rs

crates/accelerate/src/two_qubit_decompose.rs

qiskit/transpiler/passes/synthesis/unitary_synthesis.py

mtreinish · 2024-09-16T17:35:55Z

crates/accelerate/src/unitary_synthesis.rs

+    let circuit_to_dag = imports::CIRCUIT_TO_DAG.get_bound(py);
+    let dag_to_circuit = imports::DAG_TO_CIRCUIT.get_bound(py);


These exist in rust now, see: https://github.com/Qiskit/qiskit/blob/main/crates/circuit/src/converters.rs so I'd suggest calling them directly. Especially if you're doing circuit_to_dag you can call the inner DAGCircuit constructor and avoid the python overhead when you've built a CircuitData directly in rust (not that I think that's the case here).

crates/accelerate/src/unitary_synthesis.rs

crates/circuit/src/dag_circuit.rs

crates/accelerate/src/unitary_synthesis.rs

mtreinish · 2024-09-16T19:49:36Z

crates/accelerate/src/unitary_synthesis.rs

+
+#[derive(Clone, Debug)]
+enum UnitarySynthesisReturnType {
+    DAGType(Box<DAGCircuit>),


Why does this need a box?

This was a suggestion from clippy, it raised a warning because of the size difference between variants. Apparently (citing the docs) "Enum size is bounded by the largest variant. Having one large variant can penalize the memory layout of that enum."

warning: large size difference between variants --> crates/accelerate/src/unitary_synthesis.rs:65:1 | 65 | / enum UnitarySynthesisReturnType { 66 | | DAGType(DAGCircuit), | | ------------------- the largest variant contains at least 776 bytes 67 | | TwoQSequenceType(TwoQubitUnitarySequence), | | ----------------------------------------- the second-largest variant contains at least 120 bytes 68 | | } | |_^ the entire enum is at least 776 bytes | = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#large_enum_variant = note: `#[warn(clippy::large_enum_variant)]` on by default help: consider boxing the large fields to reduce the total size of the enum | 66 | DAGType(Box<DAGCircuit>), | ~~~~~~~~~~~~~~~

Yeah, that's true DAGCircuit is huge compared to the Vec. But I guess what my concern here is that we're adding a layer of pointer indirection to an owned object that we're creating in the pass. I think maybe this is pointing to the larger concern I had around some of the switching logic here in that we're treating a DAGCircuit and sequence identically when there is an optimized path when we're working with the TwoQubitUnitarySequence because it's a simplified data type.

crates/accelerate/src/unitary_synthesis.rs

Co-authored-by: Matthew Treinish <[email protected]>

…unitary-synthesis

* Fix details after applying inline suggestions * Keep TwoQubitWeilDecomposition attributes private. Add getter. * Initialize new_blocks using size hint * Remove basis_set as it's not used if there is a target. * Use ref_qubits ([PhysicalQubit; 2]) instead of wire_map (IndexMap<Qubit, Physicalqubit>) * Define static GOODBYE_SET as suggested * Use ImportOnceCell for XXDecomposer and XXEmbodiments to avoid importing in a loop. * Set preferred_direction without making it mutable. * Fix check_goodbye * Privatize assets * Use the add_global_phase method instead of the private function. * Add qs_decomposition to imports * Simplify flip_bits * Use NormalOperation to pass around decomposer gate and params info * First attempt at attaching synth circuits directly * Second attempt at attaching synth circuits directly * Use edge set for coupling map * Avoid exposing internals from NullableIndexMap * Use unitary_to_gate_sequence_inner instead of optimize_1q_gates_decomposition. * Use concat! in long error message.

ElePT force-pushed the oxidize-unitary-synthesis branch from 479fbe0 to e4ec3ff Compare September 12, 2024 10:20

mtreinish added performance Rust This PR or issue is related to Rust code in the repository mod: transpiler Issues and PRs related to Transpiler labels Sep 12, 2024

ElePT force-pushed the oxidize-unitary-synthesis branch from 8de52b5 to e11ffe6 Compare September 12, 2024 10:55

mtreinish added this to the 1.3.0 milestone Sep 12, 2024

ElePT added 2 commits September 12, 2024 13:47

Inital attempt at oxidizing unitary synthesis.

7528db9

Add cacache pygates feature to accelerate

1296234

ElePT force-pushed the oxidize-unitary-synthesis branch from e11ffe6 to 1296234 Compare September 12, 2024 11:48

ElePT added 4 commits September 13, 2024 12:22

Use sabre target to speed up preferred_direction

4456b36

Merge branch 'main' of https://github.com/Qiskit/qiskit into oxidize-…

b2b202c

…unitary-synthesis

Adapt code to changes in PackedInstruction

c49acc1

Use target in preferred_direction

739a5c8

Some cleanup

f527018

ElePT commented Sep 16, 2024

View reviewed changes

More cleanup

eb815d1

ElePT marked this pull request as ready for review September 16, 2024 16:06

ElePT requested review from alexanderivrii, ShellyGarion and a team as code owners September 16, 2024 16:06

ElePT changed the title ~~[WIP] Oxidize UnitarySynthesis pass~~ Oxidize UnitarySynthesis pass Sep 16, 2024

mtreinish reviewed Sep 16, 2024

View reviewed changes

Apply inline suggestions from Matt's code review

7498419

Co-authored-by: Matthew Treinish <[email protected]>

ElePT force-pushed the oxidize-unitary-synthesis branch 2 times, most recently from e560bf8 to 7498419 Compare September 18, 2024 10:08

Merge branch 'main' of https://github.com/Qiskit/qiskit into oxidize-…

594a697

…unitary-synthesis

ElePT added 2 commits September 18, 2024 12:09

Use dag.apply_operation_back in 1q case. Additional cleanup.

7fc0a8a

ElePT force-pushed the oxidize-unitary-synthesis branch from 607aa1f to 7fc0a8a Compare September 18, 2024 12:52

raynelfss mentioned this pull request Sep 18, 2024

Add cache_pygates feature to accelerate crate. #13183

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Oxidize `UnitarySynthesis` pass #13141

Oxidize `UnitarySynthesis` pass #13141

ElePT commented Sep 12, 2024 •

edited

Loading

coveralls commented Sep 12, 2024 •

edited

Loading

ElePT commented Sep 13, 2024

ElePT Sep 16, 2024 •

edited

Loading

ElePT commented Sep 16, 2024

qiskit-bot commented Sep 16, 2024

mtreinish left a comment

mtreinish Sep 16, 2024

mtreinish Sep 16, 2024

ElePT Sep 17, 2024 •

edited

Loading

mtreinish Sep 17, 2024

		let circuit_to_dag = imports::CIRCUIT_TO_DAG.get_bound(py);
		let dag_to_circuit = imports::DAG_TO_CIRCUIT.get_bound(py);

Oxidize UnitarySynthesis pass #13141

Are you sure you want to change the base?

Oxidize UnitarySynthesis pass #13141

Conversation

ElePT commented Sep 12, 2024 • edited Loading

Summary

Details and comments

coveralls commented Sep 12, 2024 • edited Loading

Pull Request Test Coverage Report for Build 10922534515

Details

💛 - Coveralls

ElePT commented Sep 13, 2024

ElePT Sep 16, 2024 • edited Loading

Choose a reason for hiding this comment

ElePT commented Sep 16, 2024

qiskit-bot commented Sep 16, 2024

mtreinish left a comment

Choose a reason for hiding this comment

mtreinish Sep 16, 2024

Choose a reason for hiding this comment

mtreinish Sep 16, 2024

Choose a reason for hiding this comment

ElePT Sep 17, 2024 • edited Loading

Choose a reason for hiding this comment

mtreinish Sep 17, 2024

Choose a reason for hiding this comment

Oxidize `UnitarySynthesis` pass #13141

Oxidize `UnitarySynthesis` pass #13141

ElePT commented Sep 12, 2024 •

edited

Loading

coveralls commented Sep 12, 2024 •

edited

Loading

ElePT Sep 16, 2024 •

edited

Loading

ElePT Sep 17, 2024 •

edited

Loading