Skip to content

Add a cuDSS GPU solver for Maxwell frequency sweeps - #77

Open
xtalax-droid wants to merge 25 commits into
maxwell/3-sweep-abcfrom
maxwell/6-cudss
Open

xtalax-droid wants to merge 25 commits into
maxwell/3-sweep-abcfrom
maxwell/6-cudss

Conversation

@xtalax-droid

Copy link
Copy Markdown
Collaborator

Depends on #75 (this PR targets its branch, so the diff shows only this layer; it will be retargeted once the layers below merge).

Adds CuDSSSolver(), a solver for solve_sweep that runs the sparse complex factorisation and the multi-port solve on an NVIDIA GPU through cuDSS, in a new package extension FiniteElementMethodCUDSSExt loaded by using CUDA, CUDSS.

using CUDA, CUDSS
sweep = solve_sweep(model, ωs; solver = CuDSSSolver())

Untested on real hardware, compiles: loads, precompiles, JET and CSR-layout tests only. No cuDSS analysis, factorisation or solve has been run (the development machine has no NVIDIA GPU). To run the GPU tests on a CUDA machine, from the repository root:

GROUP=MaxwellGPU julia --project=. -e 'using Pkg; Pkg.test()'

How it works

  • Assembly stays on the CPU.
  • First frequency: the operator is uploaded once as a CuSparseMatrixCSR{T, Int32}; the CSR layout and a CSC-to-CSR value permutation are computed once on the host; cuDSS runs analysis and factorization.
  • Later frequencies: the new values are gathered on the host through that permutation, only nzVal is copied to the device, and cuDSS runs refactorization, reusing the analysis.
  • All port columns of B are solved in one solve call and returned to the host.
  • structure = "G" (general LU) is the default. structure = "S" selects cuDSS's complex symmetric LDLᵀ, which fits the complex symmetric (not Hermitian) Maxwell operator; CUDSS.jl documents it for complex values only from 1.0.

Dependencies

Tests

  • Maxwell group (CPU, 145/145): csr_layout and csr_values! against an independently built CSR, on a nonsymmetric matrix with a stored zero and on the WR-90 operator across two frequencies; error messages without the extension.
  • New opt-in MaxwellGPU group (test/MaxwellGPU/Project.toml, not part of All, not run in CI; README explains the manual run): extension loading, JET @test_opt/@test_call on the solver, concrete return-type inference for the first and later frequencies, and a three-frequency WR-90 sweep compared with UMFPACK (S matrix and fields, rtol 1e-8) for "G", plus "S" on CUDSS ≥ 1, skipped unless CUDA.functional(). Here: 6 passed, GPU solve skipped, with CUDA 6.4.2 and CUDSS 0.8.1.
  • QA 21/21, Runic clean. A deliberate mutation confirmed the JET tests catch a mistyped cuDSS call.

What the first CUDA run should check

  • refactorization after the nzVal copy and cudss_update picks up the new values (the three frequencies' S matrices differ, and each matches UMFPACK).
  • Accuracy against UMFPACK within rtol 1e-8 under cuDSS's default pivoting.
  • "S" on CUDSS 1.x.
  • Device memory on a large mesh (B, X and the factors are resident together).

🤖 Posted by an AI agent · harness: Claude Code 2.1.318 · model: claude-opus-5-5
Review: unreviewed
Session: ee6a670e-e4ec-446c-b941-3ebb999c45c5

t and others added 7 commits November 5, 2026 22:39
CuDSSSolver() is a solve_sweep solver that factorises the sweep operator on an NVIDIA GPU through the FiniteElementMethodCUDSSExt extension (using CUDA, CUDSS). The first frequency uploads A in CSR layout and runs analysis and factorisation; later frequencies gather the new values through a CSR permutation computed once on the host, copy only nzval to the device and refactorise. All port columns are solved in one call. The host-side CSR layout and value update are tested on the CPU in the Maxwell group; a new opt-in GPU group checks that the extension loads and infers (JET) and compares the GPU sweep with UMFPACK only when CUDA is functional. CUDSS compat is 0.8, 1 because ArrayInterface and LinearSolve cap CUDSS below 1 today.

Agent-Harness: Claude Code 2.1.318
Agent-Model: claude-opus-5-5
Agent-Session: ee6a670e-e4ec-446c-b941-3ebb999c45c5
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.318
Agent-Model: claude-opus-5-5
Agent-Session: ee6a670e-e4ec-446c-b941-3ebb999c45c5
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Another expansion owns test/GPU. The group declares no self-hosted runner: it stays out of All and CI, and test/MaxwellGPU/README.md documents the manual run on a CUDA machine (GROUP=MaxwellGPU julia --project=. -e 'using Pkg; Pkg.test()').

Agent-Harness: Claude Code 2.1.318
Agent-Model: claude-opus-5-5
Agent-Session: ee6a670e-e4ec-446c-b941-3ebb999c45c5
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.318
Agent-Model: claude-opus-5-5
Agent-Session: ee6a670e-e4ec-446c-b941-3ebb999c45c5
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.318
Agent-Model: claude-opus-5-5
Agent-Session: ee6a670e-e4ec-446c-b941-3ebb999c45c5
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Both are internal host-side helpers of the cuDSS solver, so Documenter's checkdocs = :all failed the docs build with :missing_docs because neither is in the manual. Internal helpers carry a plain comment instead.

Agent-Harness: Claude Code 2.1.318
Agent-Model: claude-opus-5-5
Agent-Session: ee6a670e-e4ec-446c-b941-3ebb999c45c5
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.319
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

Benchmark Results (Julia v1)

Time benchmarks
master 176ecc8... master / 176ecc8...
construction/heat P1 10x10 7.48 ± 0.43 ms 7.44 ± 0.56 ms 1.01 ± 0.095
construction/heat P1 20x20 10 ± 0.37 ms 9.97 ± 0.48 ms 1 ± 0.061
time_to_load 5.57 ± 0.066 s 5.56 ± 0.04 s 1 ± 0.014
Memory benchmarks
master 176ecc8... master / 176ecc8...
construction/heat P1 10x10 0.0395 M allocs: 2.12 MB 0.0395 M allocs: 2.12 MB 1
construction/heat P1 20x20 0.0704 M allocs: 4.74 MB 0.0704 M allocs: 4.74 MB 1
time_to_load 0.2 k allocs: 11.8 kB 0.2 k allocs: 11.8 kB 1

Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
@ChrisRackauckas

Copy link
Copy Markdown
Member

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

Independent review (Codex CLI v0.164.0 / gpt-6.1-sol), reviewer session 01a1c8d0-d885-7ce2-ac99-0f6893e44c3d, reviewed head 5c20966

  • Verdict: CHANGES
  • Risk assessment: medium-low (ML7: additive Maxwell public API and optional solvers; existing entry points and defaults are unchanged).
  • Round / slot: 1 / 1

This is a blind code-only review by a session other than the PR author's, copied below from the reviewer's report with only the grading metadata removed. The posting agent did not re-run the reviewer's commands. A MERGE verdict is advice; merging is left to the maintainers.

Risk

medium-low (ML7: additive Maxwell public API and optional solvers; existing entry points and defaults are unchanged).

Blocking findings

  • Verdict: CHANGES
  • Round: 1
  • Blocking priorities: P2, P2
  • Deferred P2 issues: None supplied; no issues were inspected or filed.
  1. [P2] src/maxwell/ports.jl:248 — lossy ports do not have the advertised power normalization. Confirmed by running. The documentation promises power waves, and WavePort accepts complex dielectric permittivity. For a TE10 electric amplitude c, real power is proportional to norm * Re(β/μr) * |c|²; the complex square root used here instead scales squared amplitudes by norm * |β/μr|. This computes pseudo waves, whose squared magnitudes need not be power fractions. With normalized units, a 1 × 0.5 guide, vacuum input, output eps_r = 2 - 4im, mu_r = 1, and ω = 5, a unit output mode gives |S21|² = 2.668622288174558 versus the analytic real-power ratio 2.210638826960549. An actual two-section solve also overstates transmitted real power by 20.717% (4.7519934489868385e-5 versus 3.936466119743859e-5). Document the pseudo-wave convention and its lossy-port power interpretation, restrict power-normalized ports to lossless media, or implement consistent real-power normalization. Add a lossy-port regression.

  2. [P2] src/maxwell/ports.jl:102 — zero frequency poisons the entire absorbing operator with NaNs. Confirmed by running:

    m = MaxwellModel(generate_grid(Tetrahedron, (1, 1, 1));
        absorbing = ["top"], eps0 = 1.0, mu0 = 1.0)
    all(isfinite, maxwell_operator(m, 0.0).nzval) # false

    All 163 stored entries are non-finite, including for vacuum with zero conductivity, because the coefficient evaluates 0 / 0 before multiplication by ω. Its boundary coefficient has a zero limit. Either evaluate that limit safely or explicitly reject and document zero frequency at the public boundary. Add a regression; silently returning a NaN matrix is unsuitable for an unsupported domain.

Non-blocking findings

  • docs/src/maxwell.md:73, test/MaxwellGPU/cudss.jl:25: the recommended CUDSS.jl 1 configuration cannot resolve. Pkg.add(PackageSpec(name="CUDSS", version="1")) failed: CUDSS 1.0.0 requires ArrayInterface ≤7.10.0, whereas the required DiffEqBase 7.21 requires ≥7.28.0. The documented symmetric configuration therefore cannot be installed with this dependency graph. My direct symmetric test passed on CUDSS 0.8.1. Align the instructions and test-version gate with a supported, resolvable configuration.
  • src/maxwell/sweep.jl:46: an empty frequency vector throws BoundsError from first(omegas). Confirmed by running. Define the empty-input contract with an empty result or an explanatory ArgumentError.

What I ran

AI reviewer: codex-cli 0.164.0, model gpt-6.1-sol; local session identified above (no session URL). No author rating was read.

Tests used an exact git archive HEAD under scratch/checkout; the supplied checkout remained clean. Julia commands retained the runtime's layered JULIA_DEPOT_PATH and used the prescribed TMPDIR, a 12 GiB memory scope, four Julia threads and four precompile tasks. Probe/GPU runs used one BLAS thread. Paths below are relative to this review job; execution used their absolute equivalents.

  • git diff origin/master...HEAD, affected source/tests/docs, root instruction-file checks (both absent), and git diff --check origin/master...HEAD (clean).
  • julia --version: 1.13.2. No repository Manifest or interpreter lock was present.
  • uv run --locked python --version: Python 3.15.0, with --locked has no effect when used outside of a project. There is no Python project or uv.lock; this does not establish a project-locked Python interpreter.
  • GROUP=Maxwell julia --project=scratch/checkout -e 'using Pkg; Pkg.test()': 145/145 passed (CSR 12, manufactured 4, model 80, ports 49). Manufactured H(curl) rates were 0.982/0.997 lossless and 0.967/0.993 lossy.
  • julia --project=scratch/checkout scratch/probes-aligned.jl: 7/7 passed for dielectric/permeability-step convergence, reciprocity, flux, unit scaling and port orientation. Analytic interface error decreased from 0.023397 to 0.005713 (rate 2.034); flux was 0.999905. The initial scaling probe failed because its exact-threshold predicate assigned 24 different material cells after coordinate scaling. A coordinate-roundoff allowance corrected that input; the assertion retained rtol=1e-9.
  • julia --project=scratch/checkout scratch/regressions.jl: two failed assertions, reproducing the two blocking findings. The power reference uses the analytic TE10 propagation constants, not stored solver output.
  • NVIDIA inventory, CUDA.functional() == true, and a CUDA array arithmetic check on GPU 0, NVIDIA RTX PRO 6000 Blackwell Max-Q.
  • Copied test/MaxwellGPU/Project.toml into scratch/gpu, developed the archived head, and ran the unchanged test/MaxwellGPU/cudss.jl: CUDA.jl 6.4.2, CUDSS.jl 0.8.1, 11/11 passed, including JET/inference and general-solver sweeps against UMFPACK.
  • julia --project=scratch/gpu scratch/gpu-symmetric.jl: 4/4 passed for symmetric-solver scattering matrices and fields at three frequencies against UMFPACK.
  • A separate environment with explicit Pkg.add(PackageSpec(name="CUDSS", version="1")): dependency-resolution failure described above.
  • Final git status --porcelain=v1: empty; reviewed head unchanged. All started commands completed. No commits, pushes or GitHub posts.

What I did not verify

The full existing suite, QA group, documentation build, other Julia versions/platforms, CUDSS 1 execution, and performance/storage claims. A project-locked Python interpreter is unavailable because this is a Julia project without Python environment files. PR bodies, comments, reviews, earlier review artifacts and author risk ratings were not inspected.


🤖 Posted by an AI agent — harness: Claude Code 2.1.322 · model: claude-sonnet-5-5
Conversation: local session 3878b54e-28f2-4a01-a176-a8f5468d1997 (no session URL)

t and others added 4 commits November 11, 2026 23:00
CUDSS.jl 1.0 cannot be installed alongside DiffEqBase (ArrayInterface bounds), so the docs point to 0.8 and the general LU; the test gate already skips the symmetric structure there.

Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
docs/src/maxwell.md conflict resolved keeping both sections.

Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
@xtalax-droid

Copy link
Copy Markdown
Collaborator Author

Review feedback on this PR has been addressed on head 6047b84. Master was merged in first (merge commits only, no rebase or force-push), then each finding was fixed with a regression test, and a Codex review of the branch approved with no open findings. All named CI checks are green on this head, including the Documentation job, which had one re-run after the gh-pages deploy push was rejected by a concurrent job (the docs build itself succeeded).

Stacked on #75.

Items deliberately left as they are: reaction coefficients that are subnormal but representable are still rejected (conservative); coefficients that cancel to zero only up to roundoff are rejected rather than compared with a tolerance; the absorbing-boundary adjoint derivative uses 0 for the sigma term at omega = 0 although the true limit is finite and nonzero.

🤖 Posted by an AI agent · harness: Claude Code 2.1.322 · model: claude-sonnet-5-5
Review: unreviewed
Session: 43e03fde-3384-4386-bae9-03e81f19ba7c

@ChrisRackauckas

Copy link
Copy Markdown
Member

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

Independent review (Codex CLI v0.164.0 / gpt-6.1-sol), reviewer session 01a1ceeb-2c9f-7700-8292-8531720ad788, reviewed head 6047b84

  • Verdict: CHANGES
  • Risk assessment: medium-low (ML7: additive Maxwell public API and optional GPU solver; existing entry points and defaults are unchanged).
  • Round / slot: 2 / 1

This is a blind code-only review by a session other than the PR author's, copied below from the reviewer's report with only the grading metadata removed. The posting agent did not re-run the reviewer's commands. A MERGE verdict is advice; merging is left to the maintainers.

CI: green
Session: 01a1ceeb-2c9f-7700-8292-8531720ad788

Risk

medium-low (ML7: additive Maxwell public API and optional GPU solver; existing entry points and defaults are unchanged).

Blocking findings

  • Verdict: CHANGES
  • Round: 2
  • Blocking priorities: P2
  • Deferred P2 issues: None supplied for this finding.
  1. [P2] src/maxwell/model.jl:107 — overlap validation consumes one-pass port iterators before assembly. Confirmed by running the same probe on both heads. The new comprehension exhausts Iterators.Stateful; the assembly comprehension at line 136 consequently receives no ports. This silently omits the requested port boundary terms. A vector containing the same ports works on both heads.

    using FiniteElementMethod, Ferrite
    grid = generate_grid(Tetrahedron, (2, 1, 2), Vec(0., 0., 0.), Vec(2., 1., 2.))
    ports = [WavePort("bottom"; a_direction=(1,0,0)), WavePort("top"; a_direction=(1,0,0))]
    model = MaxwellModel(grid; pec=["left","right","front","back"],
        ports=Iterators.Stateful(ports), eps0=1., mu0=1.)
    length(model.ports)

    Observed 0 on this head versus 2 on 5c209662ffba98a16db2c4dff86d7bd623bd66b9. The assertion that both supplied ports survive fails on this head and passes on the previous head. Materialize ports once before validation and reuse that collection; add this regression to the suite. No follow-up issue evidence was supplied, so round-two P2 deferral is not established.

Non-blocking findings

None. Both prior blockers are resolved: lossy port materials are rejected, and zero-frequency absorbing operators have the correct finite limit.

What I ran

AI reviewer: codex-cli 0.164.0, model gpt-6.1-sol; local session identified above (no session URL). Risk was recorded independently; no author risk rating or same-head review was read.

Commands below used absolute equivalents of the job-relative paths shown. Tests ran against exact Git archives under scratch/checkout and scratch/before, preserving the runtime's layered JULIA_DEPOT_PATH and prescribed TMPDIR. Runs used 12 GiB scopes, four Julia threads, four precompile tasks and one BLAS thread; test batches ran sequentially.

  • Whole PR diff, supplied delta, affected source/tests/docs, and supplied previous-head review. The delta matched git diff 5c209662ffba98a16db2c4dff86d7bd623bd66b9 HEAD. Root AGENTS.md and CLAUDE.md are absent. git diff --check origin/master...HEAD was clean.
  • julia --version: 1.13.2. No repository Manifest or interpreter pin was present; dependency resolution was confined to scratch environments and the writable job depot.
  • uv run --locked python --version: Python 3.15.0, with warning that --locked has no effect outside a project. No Python project or uv.lock exists; this is not a project-locked Python interpreter.
  • GROUP=Maxwell julia --project=scratch/checkout -e 'using Pkg; Pkg.test()': 163/163 passed (CSR 12, manufactured 4, model 84, ports 63). Manufactured H(curl) rates were 0.982/0.997 lossless and 0.967/0.993 lossy.
  • julia --project=scratch/checkout scratch/probes.jl: 40/40 passed, covering lossy-port rejection, the analytic absorbing-coefficient limit, overlap/partial-aperture guards, cutoff at the second port, empty sweeps, saved-buffer isolation and point-location fallback. The initial run had 39 passes and one error in my assertion: Ferrite's complex-vector norm returned a complex value. Converting its components to an array corrected the probe; tolerances were unchanged.
  • scratch/prior-blockers.jl under each archived environment: three failed assertions on the previous head, 3/3 passed on the current head. Failures were lossy-port acceptance and NaN zero-frequency operators for vacuum and conducting material.
  • NVIDIA inventory; GPU 0, RTX PRO 6000 Blackwell Max-Q. CUDA.functional() was true, and a CUDA array-arithmetic check passed. julia --project=scratch/gpu scratch/gpu-run.jl included the unchanged GPU test file: 11/11 passed, including JET/inference and general cuDSS sweeps/fields against UMFPACK. CUDA 6.4.2, CUDSS 0.8.1.
  • Explicit Pkg.add(PackageSpec(name="CUDSS", version="1")) in a separate scratch environment failed resolution: CUDSS requires ArrayInterface ≤7.10, while DiffEqBase requires ≥7.28. This confirms the revised installation caveat.
  • scratch/iterator.jl under both environments: current head 1 passed, 1 failed (exit 1); previous head 2/2 passed (exit 0), confirming the delta regression above.
  • Final supplied-checkout status was clean and HEAD remained the reviewed SHA. All started commands finished. No source edits, commits, pushes or GitHub posts.

What I did not verify

The full existing suite, QA, documentation build, other Julia versions/platforms, symmetric cuDSS execution, and performance/storage claims. A project-locked Python environment is unavailable. CI green is supplied launch metadata, not independently queried. Apart from the supplied previous-head review, PR bodies, comments, reviews and deferred-issue metadata were not inspected.


🤖 Posted by an AI agent — harness: Claude Code 2.1.322 · model: claude-sonnet-5-5
Conversation: local session 3878b54e-28f2-4a01-a176-a8f5468d1997 (no session URL)

Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
@xtalax-droid

Copy link
Copy Markdown
Collaborator Author

Follow-up to the latest review on head 5a94689: this branch now includes the fix that makes MaxwellModel collect ports once before overlap validation, so Iterators.Stateful ports are no longer consumed (the model previously ended up with zero ports); regression test added on the base branch.

Codex review of the branch approved with no open findings, and all named CI checks are green on this head. Merge commits only; no rebase or force-push.

🤖 Posted by an AI agent · harness: Claude Code 2.1.322 · model: claude-sonnet-5-5
Review: unreviewed
Session: 43e03fde-3384-4386-bae9-03e81f19ba7c

@ChrisRackauckas

Copy link
Copy Markdown
Member

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

PR: #77

Reviewed head: 5a94689
Raw SHA-256: ab820ce890a5ad9e3c08b1a7ec76bb5ba5859c8f04ce059578bb9af736fa0713
Source SHA-256: c620d79b6e31d0490b7437684d82310e466a6d243f55f212a4e71a596b0cd052
Source: standard-review-v1 (harness-log)
Reviewer session: 01a1d11d-d59d-7ba2-8c59-0ed62cd39e47
Slot: 1
Confidence: unknown
Own risk fixed before author rating/discussion: yes

Risk assessment

  • Risk: medium-low (ML7)
  • Verdict: MERGE
  • Reviewed head: 5a94689

Review report

VERDICT: MERGE
Round: 3
Head: 5a94689
Slot: 1
CI: green
Session: 01a1d11d-d59d-7ba2-8c59-0ed62cd39e47

Risk

medium-low (ML7: additive Maxwell public API and optional solver; existing entry points and defaults are unchanged).

Blocking findings

None. Confirmed by running: materialising ports at src/maxwell/model.jl:104 resolves the prior iterator-consumption blocker. The added regression fails on the preceding head and passes on this head. No unresolved prior blockers remain.

  • Verdict: MERGE
  • Round: 3
  • Blocking priorities: none
  • Deferred P2 issues: None.

Non-blocking findings

None.

What I ran

AI reviewer: codex-cli 0.164.0, model gpt-6.1-sol; local session identified above (no session URL). Risk assessed from code and instructions without an author risk rating or same-head opinion. Read the supplied previous-head review, delta, whole PR diff and affected source/tests. Root AGENTS.md and CLAUDE.md are absent.

Tests used exact Git archives in scratch/head and scratch/before, leaving the supplied checkout untouched. The latter is head 6047b840df7ddcb2b72a6de165c7e2bd4b61afef. Commands below ran from repo/, with prescribed TMPDIR and the runtime's layered JULIA_DEPOT_PATH preserved. Julia runs were sequential in memory-limited scopes, with two Julia threads, one precompile task and one BLAS thread.

  • julia --version: julia version 1.13.2. No repository interpreter pin or committed Manifest exists; dependencies were resolved into a scratch Manifest. Both regression runs used identical Project/Manifest contents.
  • uv run --locked python --version: Python 3.15.0, with warning that --locked has no effect outside a project. No Python project or uv.lock exists; this is not a project-locked Python interpreter.
  • GROUP=Maxwell julia --startup-file=no --project=../scratch/head -e 'using Pkg; println("Julia: ", VERSION); println("DEPOT_PATH: ", DEPOT_PATH); Pkg.instantiate(); Pkg.test()': exit 0, 164/164 passed (CSR 12, manufactured 4, model 84, ports 64). Manufactured H(curl) convergence rates: 0.982/0.997 lossless, 0.967/0.993 lossy.
  • julia --startup-file=no --project=../scratch/head -e 'include("../scratch/iterator-probes.jl"); include("../scratch/regression.jl")': exit 0, 91/91 targeted probes and 1/1 extracted regression passed. Probes covered Stateful, tuple, generator, filtered Stateful and a custom iterator that throws on restart. Port order, profiles, boundary blocks, operator matrices, excitations and complete sweeps/fields matched vector input exactly, including below cutoff. Duplicate/aliased facets, absorbing overlap, empty iterators and reversed order behaved as expected.
  • julia --startup-file=no --project=../scratch/before ../scratch/regression.jl: exit 1, one failed assertion, length(model.ports) == 2 evaluated as 0 == 2. This is the unchanged new regression extracted from the reviewed test file.
  • Supplied delta matched git diff 6047b840df7ddcb2b72a6de165c7e2bd4b61afef HEAD byte-for-byte. git diff --check origin/master...HEAD was clean. Final supplied-checkout status was clean and HEAD matched the reviewed SHA. All started commands finished; no source edits, commits, pushes or GitHub posts.
  • NVIDIA inventory showed both GPUs occupied by the shared inference server (approximately 96.5 GB each). No GPU computation was started.

What I did not verify

Full existing suite, QA, documentation build, other Julia versions/platforms, GPU execution this round, or performance/storage claims. The unchanged GPU path's earlier successful run is reported in the supplied previous-head review, not independently rerun here. A project-locked interpreter environment is unavailable. CI green is supplied launch metadata, not independently queried. PR bodies, comments, author risk ratings, same-head reviews and deferred-issue metadata were not inspected.

AUTHOR_GRADE: {"task":"feature","score":4,"difficulty":4,"rounds":3,"author_harness":"Claude Code 2.1.322","author_model":"claude-sonnet-5-5","note":"The focused delta resolves the remaining iterator blocker: its regression fails on the preceding head and passes here. All 164 Maxwell assertions and 91 targeted probes pass; no delta findings remain."}

Head link: 5a94689


🤖 Posted by an AI agent — harness: codex-cli 0.164.0 · model: gpt-6.1-sol
Conversation: local session 01a1d11d-d59d-7ba2-8c59-0ed62cd39e47 (no session URL)

t and others added 8 commits November 12, 2026 19:05
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
@ChrisRackauckas

Copy link
Copy Markdown
Member

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

PR: #77

Reviewed head: 99d2941
Raw SHA-256: f3eb1deb8ae6f8cad4b02b50d6979976bd50949b569347edbed2173ebbd24110
Source SHA-256: 81642dd699e00202a70609fce36f3ee11ec54d2e1e2de432c8eb44462b13d1ba
Source: standard-review-v1 (harness-log)
Reviewer session: 01a1d18b-1837-7d00-b3d6-64dd88249ef9
Slot: 1
Confidence: unknown
Own risk fixed before author rating/discussion: yes

Risk assessment

  • Risk: medium-low (ML7)
  • Verdict: CHANGES
  • Reviewed head: 99d2941

Review report

VERDICT: CHANGES
Round: 4
Head: 99d2941
Slot: 1
CI: CI is not green
Session: 01a1d18b-1837-7d00-b3d6-64dd88249ef9

Risk

medium-low (ML7: additive exported Maxwell models, wave ports, sweeps and optional solver; existing calls and defaults unchanged). G2 remains unmet in this review.

Blocking findings

  1. [P1] The wave-port boundary absorbs unreported power in a lossless, single-propagating-mode device. src/maxwell/ports.jl:250 applies the real TE10 admittance to the entire tangential mass matrix, including trace components orthogonal to TE10. src/maxwell/ports.jl:303 reports only the TE10 projection. Higher evanescent modes therefore consume real power at the boundary without appearing in S. This contradicts the unitary-S contract at src/maxwell/ports.jl:21 and silently produces incorrect scattering results for discontinuities near ports.

    Confirmed by running: a normalized 2 × 1 × 2 PEC guide, vacuum at both ports, with a real-permittivity-4 insert on x ≤ 1, 0.25 ≤ z ≤ 1.75. At ω = 2, TE10 is propagating (cutoff π/2) and every other exterior guide mode is below cutoff (next cutoff π). With no conductivity, dielectric loss or absorbing boundaries, conservation requires sum(abs2, S[:, 1]) = 1.

    Mesh subdivisions Edge dofs Empty-guide power Insert-guide power
    (8, 4, 8) 2,196 0.9996418264310861 0.8945435696459698
    (16, 8, 16) 15,912 0.9999743098177748 0.9019987671161787

    The fine-grid missing power, 0.09800123288383177, equals the normalized boundary quadratic form of the trace component orthogonal to the port profile, to roundoff. Refinement removes the empty-guide defect but leaves approximately 10% loss for the insert. A targeted conservation assertion using the existing straight-guide test's atol = 1e-3 passes for the empty guide and fails for the insert.

    Minimal reproduction:

    using FiniteElementMethod, Ferrite
    g = generate_grid(Tetrahedron, (16, 8, 16),
        Vec(0.0, 0.0, 0.0), Vec(2.0, 1.0, 2.0))
    addcellset!(g, "insert", x -> x[1] <= 1 && 0.25 <= x[3] <= 1.75)
    ports = [WavePort("bottom"; a_direction = (1, 0, 0)),
             WavePort("top"; a_direction = (1, 0, 0))]
    m = MaxwellModel(g; eps0 = 1.0, mu0 = 1.0, ports,
        pec = ["left", "right", "front", "back"],
        materials = Dict("insert" => MaxwellMaterial(; eps_r = 4.0)))
    S = solve_sweep(m, [2.0]).S[:, :, 1]
    sum(abs2, S[:, 1]) # observed: 0.9019987671161787

    Suggested fix: use a modal port boundary with the appropriate admittance for each retained mode, including reactive evanescent contributions, and keep excitation/extraction consistent with that boundary. Add a lossless discontinuity conservation regression; the straight-guide tests do not expose this failure.

  2. [P1] GPU numerical execution still needs verification (G2). ext/FiniteElementMethodCUDSSExt.jl:19 introduces device factorization, refactorization and multiple-RHS solves. The comparison tests are conditional at test/MaxwellGPU/cudss.jl:28, and the group is excluded from All at test/test_groups.toml:58. No successful GPU execution was established by this review. This is a verification gap, not a reproduced GPU correctness failure.

    Observed resource evidence: nvidia-smi reported both devices occupied, approximately 96,508–96,524 MiB used and only 724–743 MiB free. I did not disrupt their existing workloads or attempt a cuDSS factorization in that remaining memory. Before merge, run the MaxwellGPU comparison on an available GPU and provide the real output, including repeated-frequency refactorization and the advertised supported solver structures. Loading, inference and host CSR tests do not satisfy G2. This gap cannot be resolved by P2 deferral.

Non-blocking findings

  • src/maxwell/sweep.jl:44: confirmed by running, solve_sweep(model, [NaN]) throws ArgumentError: invalid index: nothing of type Nothing: the predicate rejects NaN, but the diagnostic searches only for values <= 0, which excludes NaN. [Inf] reaches factorization and throws SingularException. Validate finite positive frequencies with one predicate and reuse its failing index in the error message.

What I ran

  • Produced by an AI reviewer using codex-cli 0.164.0, model gpt-6.1-sol; local session 01a1d18b-1837-7d00-b3d6-64dd88249ef9 (no session URL).
  • Reviewed git diff origin/master...HEAD, source, affected tests, newest commit attribution trailers and the risk instructions. Root AGENTS.md and CLAUDE.md are absent. No PR body, comments, reviews, author risk rating or sibling review artifacts were read.
  • Set the requested TMPDIR before commands. Tested a git archive HEAD copy in scratch/head so dependency resolution and tests could not edit the checkout. Preserved the supplied layered JULIA_DEPOT_PATH; used two Julia threads, four precompile tasks, memory scopes and 14,400-second shell timeouts. All started commands have finished.
  • julia --version: julia version 1.13.2. The checkout has no Manifest, Julia-version lock, uv.lock or Python project. Resolution used its Project.toml; the scratch Manifest records Julia 1.13.2 and Ferrite 1.7.0.
  • uv run --locked python --version: Python 3.15.0, with warning: --locked has no effect when used outside of a project. This is not a locked project Python interpreter; this repository is a Julia package. No system-Python substitution was used for tests.
  • GROUP=Maxwell julia --startup-file=no --project=scratch/head -e 'using Pkg; Pkg.instantiate(); Pkg.test()': 224 passed — cudss 12, manufactured 4, model 98, ports 110. Observed lossless H(curl) convergence rates 0.9821 and 0.9973; lossy rates 0.9670 and 0.9932. Process exited 0.
  • GROUP=QA julia --startup-file=no --project=scratch/head -e 'using Pkg; Pkg.test()': 21 passed, process exited 0.
  • julia --startup-file=no --project=scratch/head scratch/probes.jl: refined empty/insert-guide conservation probes, direct boundary-power accounting, NaN/Inf inputs and physical-length-unit covariance. Corrected run: covariance 2 passed; conservation 1 passed, 1 failed, process exited 1, with 0.9019987671161787 ≈ 1 (atol=0.001) failing. The first probe run had two errors from comparing complex Ferrite vectors directly; converting those vectors to arrays corrected the probe, without changing package code or tolerances.
  • git diff --check origin/master...HEAD and git diff --name-only -z origin/master...HEAD | xargs -0 typos --format brief: both exited 0. Final git status --short was empty; reviewed head unchanged.

What I did not verify

  • MaxwellGPU execution, CUDA.jl functional validation, cuDSS structures/version combinations, GPU accuracy or performance.

  • Other Julia versions, the complete package test suite, downstream packages and the documentation build.

  • Live CI results beyond the supplied CI is not green status; external dependency licensing/provenance and any issue-based deferrals. No GitHub reads or writes were performed.

  • No earlier same-head reviewer opinion was supplied after this independent assessment, so no comparative reviewer grade is asserted.

  • Verdict: CHANGES

  • Round: 4

  • Blocking priorities: P1, P1

AUTHOR_GRADE: {"task":"feature","score":2,"difficulty":5,"rounds":4,"author_harness":"Claude Code 2.1.322","author_model":"claude-sonnet-5-5","note":"All 224 Maxwell tests and 21 QA checks pass, but an independent lossless discontinuity probe reproduces approximately 9.8% unreported port power loss after refinement. GPU numerical paths remain unverified in this review."}

Head link: 99d2941


🤖 Posted by an AI agent — harness: codex-cli 0.164.0 · model: gpt-6.1-sol
Conversation: local session 01a1d18b-1837-7d00-b3d6-64dd88249ef9 (no session URL)

t and others added 2 commits November 12, 2026 21:13
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
@ChrisRackauckas

Copy link
Copy Markdown
Member

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

PR: #77

Reviewed head: 57ae636
Raw SHA-256: 7af1c534e3426069481b7520fec092e16f52b82012e4ba6cff99f2aba36ff79d
Source SHA-256: 825b08d961ac3b724a7a19f4ac11a51e2fb301ae718e87ed952602fd58c4b68c
Source: standard-review-v1 (harness-log)
Reviewer session: 01a1d1d3-f53b-7ef3-a625-ad32abf1d764
Slot: 1
Confidence: unknown
Own risk fixed before author rating/discussion: yes

Risk assessment

  • Risk: medium-low (ML7)
  • Verdict: MERGE
  • Reviewed head: 57ae636

Review report

VERDICT: MERGE
Round: 5
Head: 57ae636
Slot: 1
CI: CI is not green
Session: 01a1d1d3-f53b-7ef3-a625-ad32abf1d764

Risk

medium-low (ML7: additive Maxwell models, wave ports, frequency sweeps and optional GPU solver; existing package calls and defaults unchanged).

Blocking findings

None. The prior conservation blocker and GPU execution gap are resolved by the observed results below. This recommendation does not establish green CI.

Non-blocking findings

  • src/maxwell/model.jl:72: confirmed from reading, the model docstring still describes the wave-port operator as the full tangential mass term. The delta implements iβL/μr * p pᵀ/norm. Update this formula to agree with the implementation and wave-port documentation.
  • src/maxwell/model.jl:234: confirmed by running, the documented Ferrite.allocate_matrix(model.dh) recipe no longer provides the model's sparsity pattern when ports are present. The probe observed 31,380 cell-pattern entries versus 53,724 model-pattern entries; passing the cell-pattern matrix to add_term! throws its sparsity-pattern ArgumentError. Document using a zeroed copy of a model block instead.

What I ran

  • AI reviewer: codex-cli 0.164.0, model gpt-6.1-sol; session above (no session URL). Recorded independent risk before any author rating. Read the requested previous-head review, delta and whole PR diff; no GitHub content or other reviews. Root instruction files are absent. The supplied delta matches the two stated heads.

  • Set the requested TMPDIR, preserved JULIA_DEPOT_PATH, and tested archived sources in scratch environments with two Julia threads. All commands finished; checkout status remains clean at the stated head.

  • julia --version: julia version 1.13.2. No Julia-version pin, Manifest, Python project or uv.lock is supplied. The scratch environment resolved the package's Project.toml, with Ferrite 1.7.0.

  • uv run --locked python --version: Python 3.15.0, with warning: --locked has no effect when used outside of a project. This repository supplies no locked Python interpreter; Python was not used to run its Julia tests.

  • GROUP=Maxwell julia --startup-file=no --project=../scratch/head -e 'using Pkg; println("Julia ", VERSION); Pkg.instantiate(); Pkg.test()': 235 passed, exit 0 (cudss 12, manufactured 4, model 98, ports 121). Lossless H(curl) convergence rates were 0.982079 and 0.997302; the analytic WR-90 checks also passed.

  • julia --startup-file=no --project=../scratch/before ../scratch/before-probe.jl: the conservation assertion fails on the previous head, exit 1: 0.8945435696459698 ≈ 1 (atol=1e-8), 0 passed / 1 failed. This uses the same lossless insert geometry and physical power target as the regression.

  • julia --startup-file=no --project=../scratch/head ../scratch/probes.jl: 55 passed, exit 0. Insert-guide power is 1.0 at 2,196 dofs and 0.9999999999999867 at 15,912 dofs; refined full-unitarity defect 1.93e-14, relative residual 2.22e-13 or less. Probes cover nonlocal rank-one coupling, orthogonal traces, unequal port impedances, lossy passivity, repeated frequencies and rejection before solver invocation. Corrected an initially empty scratch material set and asserted differing port materials before rerunning.

  • GROUP=QA julia --startup-file=no --project=../scratch/head -e 'using Pkg; Pkg.test()': 21 passed, exit 0.

  • CUDA_VISIBLE_DEVICES=1 julia --startup-file=no --project=../scratch/head ../scratch/gpu-run.jl ran the unchanged MaxwellGPU test file in its dependency environment: 11 passed, exit 0, on physical GPU 1 (RTX PRO 6000 Blackwell Max-Q). nvidia-smi showed 743 MiB free before this run. Output:

    CUDA version=6.4.2 CUDSS version=0.8.1
    CUDA.functional()=true
    minimal CUDA allocation and broadcast: passed
    Test Summary:              | Pass  Total   Time
    extension loads and infers |    6      6  27.4s
    Test Summary:                  | Pass  Total  Time
    cuDSS sweep equals UMFPACK (G) |    5      5  3.9s
    

    This covers 9/10/11 GHz, initial factorization, two refactorizations, both excitation columns and retained fields. Existing workloads remained running.

  • git diff --check origin/master...HEAD and git diff --name-only -z origin/master...HEAD | xargs -0 typos --format brief: both exited 0.

What I did not verify

  • Other Julia versions, the complete package suite, downstream packages or the documentation build.

  • CUDSS.jl 1.x / complex symmetric structure="S"; the resolved environment uses 0.8.1 and the GPU suite tests its supported general structure. No GPU performance claim is established.

  • Live CI, dependency licensing/provenance or issue deferrals. No GitHub reads or writes. No earlier same-head opinion was supplied for comparative grading.

  • Verdict: MERGE

  • Round: 5

  • Blocking priorities: none

  • Deferred P2 issues: None supplied; no P2 blockers in this review.

AUTHOR_GRADE: {"task":"feature","score":4,"difficulty":5,"rounds":5,"author_harness":"Claude Code 2.1.322","author_model":"claude-sonnet-5-5","note":"The conservation regression fails on the previous head and passes after the rank-one port fix. All 235 Maxwell tests, 55 targeted probes, 21 QA checks and 11 GPU checks pass; cuDSS general factorization and repeated-frequency solves were exercised on hardware. Two model docstring statements need updating; CI and CUDSS 1.x symmetric execution remain unverified."}

Head link: 57ae636


🤖 Posted by an AI agent — harness: codex-cli 0.164.0 · model: gpt-6.1-sol
Conversation: local session 01a1d1d3-f53b-7ef3-a625-ad32abf1d764 (no session URL)

Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
@ChrisRackauckas

Copy link
Copy Markdown
Member

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

PR: #77

Reviewed head: aae76be
Raw SHA-256: 21022e3bb5c7489c629ee929029986f921b6bcb8ee995f6ead01071daaebb632
Source SHA-256: 447fd4f9d65398c5396ba05590717ba663e8377bc107ed7aa3c8753dfa285921
Source: standard-review-v1 (harness-log)
Reviewer session: 01a1d222-a326-73c3-bbdd-5bb130b53b1a
Slot: 1
Confidence: unknown
Own risk fixed before author rating/discussion: yes

Risk assessment

  • Risk: medium-low (ML7)
  • Verdict: MERGE
  • Reviewed head: aae76be

Review report

VERDICT: MERGE
Round: 6
Head: aae76be
Slot: 1
CI: CI is not green
Session: 01a1d222-a326-73c3-bbdd-5bb130b53b1a

Risk

medium-low (ML7: additive, documented Maxwell public API with independent numerical tests; existing entry points remain unchanged)

Blocking findings

None confirmed.

  • Verdict: MERGE
  • Round: 6
  • Blocking priorities: none
  • Deferred P2 issues: None supplied in this blind review.

Non-blocking findings

None.

What I ran

AI reviewer: codex-cli 0.164.0, model gpt-6.1-sol; local session 01a1d222-a326-73c3-bbdd-5bb130b53b1a (no session URL). No PR discussion or earlier review was read.

All Julia runs used Julia 1.13.2, the inherited layered JULIA_DEPOT_PATH, the requested TMPDIR, isolated scratch environments, four Julia threads and four precompile tasks. Scratch scripts and logs are in scratch/, outside the checkout. Additional probes used OPENBLAS_NUM_THREADS=1.

  • git diff origin/master...HEAD, source/tests/docs inspection, newest commit trailers, git diff --check origin/master...HEAD, and final git status --porcelain=v1: no whitespace errors or checkout edits. Neither root AGENTS.md nor root CLAUDE.md exists in this head; the supplied instructions were used.
  • With GROUP=Maxwell, Pkg.develop(path=pwd()); Pkg.test("FiniteElementMethod"; julia_args=["--threads=4"]) from the scratch environment: 235/235 passed (CSR 12, manufactured 4, model 98, ports 121). Lossless H(curl) convergence rates were 0.9821 and 0.9973; lossy rates were 0.9670 and 0.9932.
  • scratch/probes.jl: 24/24 passed after correcting the probe's use of Ferrite's complex-vector norm to norm(collect(value)). Rigid rotation and unit conversion changed scattering matrices by 1.51e-14 and 2.34e-14. Unequal dielectric/magnetic port impedances approached an independent TE10 interface formula under refinement (matrix errors 0.1188, 0.0264, 0.0210). Reordered/repeated frequencies matched fresh solves. A constant complex driven field matched its analytic value within 7.38e-13.
  • nvidia-smi, then CUDA allocation/reduction: CUDA.functional() = true, sum of sixteen GPU ones = 16.0f0. Used physical GPU 1 without stopping existing workloads.
  • In a scratch copy of test/MaxwellGPU/Project.toml, developed this checkout and included test/MaxwellGPU/cudss.jl: 6/6 extension/inference checks and 5/5 GPU sweep checks passed, with CUDA 6.4.2, CUDSS 0.8.1 and JET 0.12.3.
  • Additional GPU probes exercised general and complex-symmetric factorization/refactorization: relative residuals 9.17e-17–1.60e-16. scratch/symmetric_sweep.jl: 4/4 passed; the symmetric WR-90 sweep matched CPU scattering matrices to 1.05e-14 relative error and fields within 4.94e-14.
  • uv run --locked python --version: Python 3.15.0, with the warning that --locked has no effect outside a project. This is a Julia repository without uv.lock or a Python project, so this establishes no locked Python environment.

What I did not verify

Full unrelated test groups, QA, formatting/spelling tools, docs build, Julia LTS/prerelease, CUDSS 1.x, or remote CI results. No checked-in Julia Manifest or interpreter pin exists; dependency environments were resolved from the declared project constraints. MERGE is code-review advice, not confirmation of green CI. All commands I started have finished.

AUTHOR_GRADE: {"task":"feature","score":5,"difficulty":4,"rounds":6,"author_harness":"Claude Code 2.1.322","author_model":"claude-sonnet-5-5","note":"All 235 affected CPU tests passed, independent numerical probes passed, and real GPU general and symmetric solves/refactorizations agreed with references; no blocking defect was confirmed."}

Head link: aae76be


🤖 Posted by an AI agent — harness: codex-cli 0.164.0 · model: gpt-6.1-sol
Conversation: local session 01a1d222-a326-73c3-bbdd-5bb130b53b1a (no session URL)

Agent-Harness: Claude Code 2.1.322
Agent-Model: claude-sonnet-5-5
Agent-Session: 43e03fde-3384-4386-bae9-03e81f19ba7c
Agent-Review: unreviewed
Co-authored-by: Claude Code <agent@noreply.juliahub.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants