Add impls of ProgramNode for various math operations by ihincks · Pull Request #16107 · Qiskit/qiskit

ihincks · 2026-04-29T20:04:05Z

This PR closes #16031 by implementing ProgramNode for various math-y structs.

Add
Subtract
Multiply
Divide
Remainder
Power

PR Stack

AI/LLM disclosure

I didn't use LLM tooling, or only used it privately.
I used the following tool to help write this PR description:
I used the following tool to generate or modify code: Claude 4.6

qiskit-bot · 2026-04-29T20:04:10Z

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

@Qiskit/terra-core

coveralls · 2026-04-29T20:33:58Z

Coverage Report for CI Build 33116403395

Warning

Build has drifted: This PR's base is out of sync with its target branch, so coverage data may include unrelated changes.
Quick fix: rebase this PR. Learn more →

Coverage decreased (-0.5%) to 87.74%

Details

Coverage decreased (-0.5%) from the base build.
Patch coverage: 147 uncovered changes across 3 files (818 of 965 lines covered, 84.77%).
2610 coverage regressions across 48 files.

Uncovered Changes

File	Changed	Covered	%
crates/providers/src/math_nodes/reduction.rs	577	492	85.27%
crates/providers/src/math_nodes/bitwise.rs	209	164	78.47%
crates/providers/src/math_nodes/binary.rs	172	155	90.12%
Total (5 files)	965	818	84.77%

Coverage Regressions

2610 previously-covered lines in 48 files lost coverage.

Top 10 Files by Coverage Loss	Lines Losing Coverage	Coverage
crates/circuit/src/dag_circuit.rs	534	86.0%
crates/circuit/src/parameter/symbol_expr.rs	449	78.35%
qiskit/qpy/binary_io/circuits.py	245	63.72%
crates/qpy/src/circuit_reader.rs	230	74.53%
crates/qpy/src/value.rs	139	73.54%
crates/circuit/src/circuit_data.rs	130	86.84%
crates/cext/src/circuit.rs	90	82.63%
crates/circuit/src/operations.rs	86	83.78%
qiskit/transpiler/passes/synthesis/hls_plugins.py	84	85.94%
crates/synthesis/src/multi_controlled/mcx.rs	64	90.5%

Coverage Stats


Relevant Lines:	131079
Covered Lines:	115009
Line Coverage:	87.74%
Coverage Strength:	982075.61 hits per line

💛 - Coveralls

qiskit-bot · 2026-05-16T11:07:42Z

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

@Qiskit/terra-core

ihincks · 2026-07-24T17:28:56Z

I also think we could use a bit more testing just to cover cases common cases (both positive and negative) for all the operations.

I've added a few more tests now, and in general I think coverage is decent. If there are any specific missing spots you see, please let me know and I'll add more.

ihincks · 2026-07-24T17:46:14Z

a sharp edge when creating a program that causes unhandleable crashes

This lead me to also realize that we should handle % 0 and / 0 better. Right now, in the float case, you get NaNs or infs. This is fine, not ideal, but not a runtime error, and some of the data might be recoverable if only some elements of an array have this.

However, for integer types, we panic. We should instead do what NumPy does, and return 0 on % 0, and return NaN/inf on / 0, which means we need to cast to real.

I'll do a new PR for this.

mtreinish

Overall I think this is almost good to go. The only thing that stood out to me is the array operations in the reduction module are allocating a bit more than is necessary. I think we can reduce the amount of allocations by mutating some of the intermediate arrays inplace. I left comments on some spots that I saw but I'm not sure I caught them all.

I also left some inline nits on removing the static lifetime from all the functions that return string. It's not necessary as it's implied by returning an inline string literal and it doesn't really convey anything useful.

mtreinish · 2026-08-25T21:58:46Z

+                let n = a.shape()[self.axis] as f64;
+                let mean = (a.sum_axis(Axis(self.axis)) / Complex::new(n, 0.0))
+                    .insert_axis(Axis(self.axis));
+                let sq_mod = (a - &mean).mapv(|c| c.re * c.re + c.im * c.im);


I think we can probably do mapv_inplace here too and avoid another allocation since the subtraction should be allocating a new array already too. But it might require a bit large of a change than is easy inline.

I applied your other mapv_inplace suggestions, but then ended up reverting because they were wrong for a trivial reason: stds/vars of a complex vector are real, so we can't put a float into a complex slot in-place.

However, you sent me down a mini rabbit hole that also addresses this comment: see aea04da, which is both faster (in some regims) and more memory efficient than the current implementation.

shape axis input original ms / peak lane ms slice ms dispatch ms / peak

[1024,1024] 0 8.0 MiB 1.72 / 12.0 MiB 1.69 0.58 0.58 / 0.02 MiB

[1024,1024] 1 8.0 MiB 1.66 / 12.0 MiB 1.26 1.94 1.26 / 0.02 MiB

[100,100000] 0 76.3 MiB 19.18 / 115.2 MiB 12.97 6.90 6.85 / 1.53 MiB

[100000,100] 1 76.3 MiB 22.05 / 115.2 MiB 9.21 33.02 8.68 / 1.53 MiB

[64,128,512] 1 32.0 MiB 9.32 / 48.2 MiB 7.57 5.41 4.95 / 0.50 MiB

[4096,4096] 0 128.0 MiB 29.44 / 192.0 MiB 95.37 8.47 8.25 / 0.06 MiB

[1000000,4] 0 30.5 MiB 107.18 / 45.8 MiB 50.01 101.35 49.75 / 0.00 MiB

[4,1000000] 1 30.5 MiB 5.87 / 45.8 MiB 5.15 76.61 5.13 / 0.00 MiB

[1000000] 0 7.6 MiB 1.88 / 11.4 MiB 1.27 54.03 1.27 / 0.00 MiB

[2,2000000] 0 30.5 MiB 11.00 / 61.0 MiB 11.16 8.26 6.35 / 30.5 MiB

[2000000,2] 1 30.5 MiB 42.72 / 61.0 MiB 13.65 12.72 12.32 / 30.5 MiB

[1024,1024] strided view 1 8.0 MiB 5.62 / 12.0 MiB 2.25 4.27 2.24 / 0.02 MiB

[2048,2048] transposed 0 32.0 MiB 7.67 / 48.0 MiB 5.22 15.11 5.22 / 0.03 MiB

Co-authored-by: Matthew Treinish <mtreinish@kortar.org>

This reverts commit 0dc1c02.

mtreinish

This LGTM now. Just one inline nit on the bare unreachable macros. It'd be good to add a reason that we think it's unreachable just for diagnostic when debugging an issue if it somehow got tripped and things panic.

Co-authored-by: Matthew Treinish <mtreinish@kortar.org>

This commit expands the test coverage of the reduction module to make sure all the code paths are exercised and validated. It was mostly dtype coverage for all the math nodes that was missing. The other piece was the optimizations added aea04da the accumulation path was not covered so a test was added and compared against the result from ndarray's var axis.

mtreinish

I pushed a commit to close the test coverage gaps in the reduction module in:
a801fc0

If this looks good to you feel free to enqueue this for merging.

ihincks requested a review from a team as a code owner April 29, 2026 20:04

ihincks added Rust This PR or issue is related to Rust code in the repository mod: providers Related to the backend and job abstractions labels Apr 29, 2026

ihincks marked this pull request as draft April 29, 2026 20:04

ihincks mentioned this pull request Apr 29, 2026

BackendV3 #15902

Open

ShellyGarion added this to the 2.5.0 milestone Apr 30, 2026

github-project-automation Bot added this to Qiskit 2.5 Apr 30, 2026

github-project-automation Bot moved this to Ready in Qiskit 2.5 Apr 30, 2026

ihincks force-pushed the program-node-math branch 2 times, most recently from b429b8d to a437668 Compare April 30, 2026 20:31

ihincks changed the title ~~Add impls of Program node for various math operations~~ Add impls of ProgramNode for various math operations May 1, 2026

ihincks force-pushed the program-node-math branch 4 times, most recently from 31adc4f to 768edb2 Compare May 16, 2026 11:07

ihincks marked this pull request as ready for review May 16, 2026 11:07

ihincks requested a review from mtreinish May 16, 2026 11:07

ihincks commented May 16, 2026

View reviewed changes

Comment thread crates/providers/src/math_nodes/bitwise.rs Outdated

ihincks commented May 16, 2026

View reviewed changes

Comment thread crates/providers/src/math_nodes/bitwise.rs Outdated

ihincks force-pushed the program-node-math branch 3 times, most recently from a590a50 to 2be78bc Compare May 21, 2026 19:02

ihincks mentioned this pull request May 23, 2026

Switch Tensor to use ArcArray instead of Array #16256

Merged

3 tasks

ihincks force-pushed the program-node-math branch 2 times, most recently from 73504ec to d590df2 Compare May 28, 2026 11:08

ihincks self-assigned this May 28, 2026

ihincks removed this from Qiskit 2.5 May 28, 2026

Merge branch 'main' into program-node-math

337dba8

ihincks force-pushed the program-node-math branch 2 times, most recently from 6b8b47b to 337dba8 Compare July 24, 2026 17:43

clippy

f3d7dbc

ihincks force-pushed the program-node-math branch from 1b0409c to f3d7dbc Compare July 24, 2026 18:19

ihincks requested a review from mtreinish August 25, 2026 13:50

mtreinish reviewed Aug 25, 2026

View reviewed changes

ihincks and others added 4 commits August 26, 2026 09:50

Apply suggestions from code review re: &str

2639c1b

Co-authored-by: Matthew Treinish <mtreinish@kortar.org>

Apply suggestions from code review re: inplace arary ops

0dc1c02

Co-authored-by: Matthew Treinish <mtreinish@kortar.org>

Revert "Apply suggestions from code review re: inplace arary ops"

0f95a1c

This reverts commit 0dc1c02.

improve memory and speed of complex reductions

aea04da

ihincks requested a review from mtreinish August 26, 2026 21:00

mtreinish previously approved these changes Aug 27, 2026

View reviewed changes

Comment thread crates/providers/src/math_nodes/reduction.rs Outdated

Comment thread crates/providers/src/math_nodes/reduction.rs Outdated

Comment thread crates/providers/src/math_nodes/reduction.rs Outdated

Apply suggestions from code review

eb2006d

Co-authored-by: Matthew Treinish <mtreinish@kortar.org>

mtreinish dismissed their stale review via eb2006d August 27, 2026 19:10

mtreinish previously approved these changes Aug 27, 2026

View reviewed changes

mtreinish dismissed their stale review via a801fc0 August 27, 2026 20:55

mtreinish previously approved these changes Aug 27, 2026

View reviewed changes

Fix rustfmt

04f5b27

mtreinish dismissed their stale review via 04f5b27 August 27, 2026 21:04

mtreinish approved these changes Aug 27, 2026

View reviewed changes

ihincks enabled auto-merge August 27, 2026 21:09

ihincks added this pull request to the merge queue Aug 27, 2026

Merged via the queue into Qiskit:main with commit 4017510 Aug 27, 2026
31 checks passed

ihincks deleted the program-node-math branch August 27, 2026 21:55

github-project-automation Bot moved this from Ready to Done in Qiskit 2.6 Aug 27, 2026

shape	axis	input	original ms / peak	lane ms	slice ms	dispatch ms / peak
`[1024,1024]`	0	8.0 MiB	1.72 / 12.0 MiB	1.69	0.58	0.58 / 0.02 MiB
`[1024,1024]`	1	8.0 MiB	1.66 / 12.0 MiB	1.26	1.94	1.26 / 0.02 MiB
`[100,100000]`	0	76.3 MiB	19.18 / 115.2 MiB	12.97	6.90	6.85 / 1.53 MiB
`[100000,100]`	1	76.3 MiB	22.05 / 115.2 MiB	9.21	33.02	8.68 / 1.53 MiB
`[64,128,512]`	1	32.0 MiB	9.32 / 48.2 MiB	7.57	5.41	4.95 / 0.50 MiB
`[4096,4096]`	0	128.0 MiB	29.44 / 192.0 MiB	95.37	8.47	8.25 / 0.06 MiB
`[1000000,4]`	0	30.5 MiB	107.18 / 45.8 MiB	50.01	101.35	49.75 / 0.00 MiB
`[4,1000000]`	1	30.5 MiB	5.87 / 45.8 MiB	5.15	76.61	5.13 / 0.00 MiB
`[1000000]`	0	7.6 MiB	1.88 / 11.4 MiB	1.27	54.03	1.27 / 0.00 MiB
`[2,2000000]`	0	30.5 MiB	11.00 / 61.0 MiB	11.16	8.26	6.35 / 30.5 MiB
`[2000000,2]`	1	30.5 MiB	42.72 / 61.0 MiB	13.65	12.72	12.32 / 30.5 MiB
`[1024,1024]` strided view	1	8.0 MiB	5.62 / 12.0 MiB	2.25	4.27	2.24 / 0.02 MiB
`[2048,2048]` transposed	0	32.0 MiB	7.67 / 48.0 MiB	5.22	15.11	5.22 / 0.03 MiB

Uh oh!

Conversation

ihincks commented Apr 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Stack

AI/LLM disclosure

Uh oh!

qiskit-bot commented Apr 29, 2026

Uh oh!

coveralls commented Apr 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Coverage Report for CI Build 33116403395

Coverage decreased (-0.5%) to 87.74%

Details

Uncovered Changes

Coverage Regressions

Coverage Stats

💛 - Coveralls

Uh oh!

qiskit-bot commented May 16, 2026

Uh oh!

Uh oh!

Uh oh!

ihincks commented Jul 24, 2026

Uh oh!

ihincks commented Jul 24, 2026

Uh oh!

mtreinish left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

mtreinish Aug 25, 2026

Choose a reason for hiding this comment

Uh oh!

ihincks Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

ihincks Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

mtreinish left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

mtreinish left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

ihincks commented Apr 29, 2026 •

edited

Loading

coveralls commented Apr 29, 2026 •

edited

Loading