Migrate interpolation backend to FastInterpolations.jl - #307
Merged
Conversation
… related documentation
…s, magnetics, and sol modules
…ct tests for its behavior
…tions.jl Migrate the last Interpolations.jl call sites to FastInterpolations and remove the dependency: - ρ_interpolant: cubic_spline_interpolation → FI.cubic_interp (ExtendExtrap) - avgZinterpolator: Gridded(Linear)+Flat → FI.linear_interp (ClampExtrap) - q_95 expression: linear_interpolation → FI.linear_interp - add _deduplicate_knots! (math.jl) replacing Interpolations.deduplicate_knots! for the SOL P(r)/r(P) interpolants - drop Interpolations from Project.toml and src/physics.jl Extend the interpolation characterization tests with avgZ Flat-extrapolation and _deduplicate_knots! contract checks. Full interpolation suite green.
… FastInterpolations
…r FastInterpolations
… FastInterpolations Migrate the last direct DataInterpolations calls to FI and remove the dependency: - sol.jl: CubicSpline → FI.cubic_interp (find_levels_from_P r(ψ)) - sources.jl: LinearInterpolation+Constant → FI.linear_interp+ClampExtrap (total_sources) - signal.jl: linear up-sampling → FI.linear_interp (smooth_by_convolution) - remove DataInterpolations import (physics.jl) and dependency (Project.toml) Extend test/runtests_interp1d.jl with smooth_by_convolution, total_sources, and find_levels_from_P characterization tests covering these sites.
…-shot API for direct querying
There was a problem hiding this comment.
Pull request overview
This PR migrates IMAS’s interpolation backend from Interpolations.jl/DataInterpolations.jl to FastInterpolations.jl, updating core physics workflows and adding characterization tests to preserve behavior within tolerances while improving performance.
Changes:
- Replace 1D/2D interpolation construction, evaluation, gradient, and extrapolation semantics with FastInterpolations equivalents across physics and expressions code.
- Update helper APIs (
cubic_interp1d,linear_interp1d) and remove legacy interpolation dependencies fromProject.toml. - Add new test suites covering interpolation-heavy workflows and 1D interpolation helpers, and wire them into
runtests.jl.
Reviewed changes
Copilot reviewed 19 out of 20 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| test/runtests.jl | Adds new interpolation-focused test files to the default test run. |
| test/runtests_interpolations.jl | New characterization/regression tests for interpolation-backed equilibrium/profile workflows. |
| test/runtests_interp1d.jl | New characterization/regression tests for IMAS 1D interpolation helpers and consumers. |
| src/signal.jl | Replaces DataInterpolations-based upsampling with FastInterpolations linear interpolation. |
| src/physics/thermal_loads.jl | Switches cubic interpolation usage to the updated one-shot helper API. |
| src/physics/sources.jl | Replaces DataInterpolations constant extrapolation with FastInterpolations clamp extrapolation. |
| src/physics/sol.jl | Updates interpolant types and knot deduplication; adjusts interpolant evaluation style. |
| src/physics/profiles.jl | Replaces Interpolations gridded linear+flat extrapolation with FastInterpolations linear+clamp. |
| src/physics/magnetics.jl | Updates interpolant argument types to FastInterpolations’ abstract interpolant. |
| src/physics/interferometer.jl | Updates segment interpolation evaluation to use FastInterpolations-backed interpolants. |
| src/physics/fluxsurfaces.jl | Updates interpolant argument types and uses new one-shot cubic interpolation call site(s). |
| src/physics/fields.jl | Replaces Interpolations.gradient with FastInterpolations gradient API; updates types/imports. |
| src/physics/equilibrium.jl | Replaces cubic spline interpolant builders with FastInterpolations cubic interpolation + extrapolation. |
| src/physics/currents.jl | Switches cubic interpolation call sites to the updated one-shot helper API. |
| src/physics.jl | Drops imports of legacy interpolation packages. |
| src/math.jl | Re-implements cubic_interp1d / linear_interp1d on FastInterpolations and adds one-shot overloads; adds _deduplicate_knots!. |
| src/IMAS.jl | Imports FastInterpolations as FI for module-wide use. |
| src/expressions/onetime.jl | Updates cubic interpolation usage to new helper overloads (vectorized evaluation). |
| src/expressions/dynamic.jl | Replaces Interpolations linear interpolation with FastInterpolations. |
| Project.toml | Removes DataInterpolations/Interpolations deps and adds FastInterpolations (with compat entry). |
Comments suppressed due to low confidence (1)
src/physics/magnetics.jl:57
PSI_interpolant.(r, z)is broadcasting over scalarr/z, which adds unnecessary broadcast overhead; the interpolant is already callable on scalars.
r = sum(pos.r for pos in loop.position) / length(loop.position)
z = sum(pos.z for pos in loop.position) / length(loop.position)
psi = PSI_interpolant.(r, z)
set_time_array(loop.flux, :data, time0, psi)
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
mgyoo86
changed the title
jmcclena
reviewed
Jun 23, 2026
jmcclena
left a comment
jmcclena
left a comment
Member
There was a problem hiding this comment.
Looks good to me. If the FUSE regressions pass, let's merge it.
…ent seam Downstream packages (FUSE/FRESCO) pass Interpolations.jl interpolants into IMAS, which broke after the FastInterpolations migration restricted args to ::FI.AbstractInterpolant (MethodError in find_magnetic_axis and trace_simple_surfaces!->Br_Bz; FUSE CI run 28040722813). Restore compatibility without re-adding Interpolations as a dependency: - drop ::FI.AbstractInterpolant on PSI_interpolant args in fields/fluxsurfaces/sol/magnetics (scalar eval itp(r,z) is backend-agnostic) - funnel the only FI-specific coupling (gradient) through _psi_gradient in fields.jl: FI -> FI.gradient (analytic, FI numerics unchanged); otherwise -> parentmodule(typeof(itp)).gradient (resolves the owning module from the object, so no Interpolations dependency is declared) - Br_Bz is the single gradient call site; Bp/trace_*/field! route through it fluxsurfaces_cubic.jl internal helpers left as-is (always FI, FI.value_gradient). Verified: runtests_interpolations.jl golden values green; smoke test with a real Interpolations cubic interpolant passes find_magnetic_axis and Br_Bz.
Member
Author
|
@jmcclena It seems this PR is safe to merge. Thanks! |
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
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
Replace Interpolations.jl and DataInterpolations.jl with FastInterpolations.jl for all interpolation in IMAS, and drop both legacy dependencies from
Project.toml.Behavior is preserved within interpolation tolerance; performance improves across the board with no regressions.
Performance
@benchmark(BenchmarkTools, minimum time), master worktree vs this branch, identical public-API callson the bundled samples (D3D 65×65, omas 17×17). All comparisons are pure backend swaps — same algorithm,
only the interpolation library differs.
Interpolation primitives
cubic_interp1dbuildlinear_interp1dbuild+eval ×400PSIeval (50×50 grid)Br_Bz(∇ψ) ×200avgZ×1000ψ_interpolantbuildρ_interpolantbuild+evalReal IMAS routines
Jtor_2_Jpar(currents)line_averagetrace_surfacesflux_surfacesNotes
Interpolations.Line()→ExtendExtrap(),Flat()→ClampExtrap().CubicFitboundary condition reproduces cubic polynomials exactly; on theproduction equilibrium grids the migrated results match the previous ones within the test tolerances.