GPU updates by kylebeggs · Pull Request #89 · JuliaMeshless/RadialBasisFunctions.jl

kylebeggs · 2026-02-19T02:46:41Z

add Adapt, device kwarg

Thread `device` kwarg through all operator constructors and weight- building paths. Add Adapt.adapt_structure for RadialBasisOperator and Interpolator to enable GPU array conversion. Introduce _to_cpu helper for KDTree compatibility, _solve_system! dispatch for symmetric vs generic solvers, and an informative error when GPU stencil solve is attempted. Clean up verbose argument lists in execution.jl.

codecov · 2026-02-19T02:56:22Z

Codecov Report

❌ Patch coverage is 93.65079% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/operators/operators.jl	84.21%	3 Missing ⚠️
src/operators/operator_algebra.jl	83.33%	1 Missing ⚠️

Files with missing lines	Coverage Δ
src/RadialBasisFunctions.jl	`100.00% <ø> (ø)`
src/interpolation.jl	`100.00% <100.00%> (ø)`
src/operators/directional.jl	`100.00% <100.00%> (ø)`
src/solve/api.jl	`100.00% <100.00%> (ø)`
src/solve/assembly.jl	`97.48% <100.00%> (+0.09%)`	⬆️
src/solve/execution.jl	`99.06% <100.00%> (+0.02%)`	⬆️
src/utils.jl	`81.81% <100.00%> (+3.24%)`	⬆️
src/operators/operator_algebra.jl	`80.00% <83.33%> (-20.00%)`	⬇️
src/operators/operators.jl	`84.48% <84.21%> (+0.22%)`	⬆️

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

github-actions · 2026-02-19T02:58:55Z

Benchmark Results

	main	`06f4efc`...	main / `06f4efc`...
Directional	2.4 ± 0.13 ms	2.52 ± 0.089 ms	0.95 ± 0.061
Directional (per point)	2.43 ± 0.13 ms	2.52 ± 0.097 ms	0.962 ± 0.064
Gradient	8.51 ± 0.32 ms	9.09 ± 0.23 ms	0.935 ± 0.043
MonomialBasis/dim=1/deg=0	0.045 ± 0.012 μs	0.044 ± 0.011 μs	1.02 ± 0.37
MonomialBasis/dim=1/deg=1	0.0729 ± 0.019 μs	0.0779 ± 0.019 μs	0.936 ± 0.33
MonomialBasis/dim=1/deg=2	0.0811 ± 0.018 μs	0.0651 ± 0.018 μs	1.25 ± 0.45
MonomialBasis/dim=2/deg=0	0.0393 ± 0.0031 μs	24.6 ± 1 ns	1.59 ± 0.14
MonomialBasis/dim=2/deg=1	0.0338 ± 0.011 μs	0.0378 ± 0.011 μs	0.894 ± 0.41
MonomialBasis/dim=2/deg=2	0.0391 ± 0.012 μs	0.0448 ± 0.012 μs	0.874 ± 0.35
MonomialBasis/dim=3/deg=0	0.0349 ± 0.012 μs	0.0386 ± 0.012 μs	0.904 ± 0.42
MonomialBasis/dim=3/deg=1	0.045 ± 0.012 μs	0.0445 ± 0.012 μs	1.01 ± 0.37
MonomialBasis/dim=3/deg=2	0.0452 ± 0.012 μs	0.0459 ± 0.011 μs	0.984 ± 0.35
Partial	2.58 ± 0.14 ms	2.73 ± 0.095 ms	0.943 ± 0.06
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 0/0/∂	9.76 ± 0.2 ns	9.9 ± 0.18 ns	0.986 ± 0.027
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 0/0/∂²	10.2 ± 0.079 ns	10.1 ± 0.19 ns	1.01 ± 0.021
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 0/0/∇	17 ± 0.061 ns	17.1 ± 0.07 ns	0.991 ± 0.0054
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 0/0/∇²	18.1 ± 0.061 ns	18.6 ± 0.061 ns	0.975 ± 0.0046
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 1/1/∂	9.73 ± 0.13 ns	9.91 ± 0.16 ns	0.982 ± 0.021
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 1/1/∂²	10.1 ± 0.17 ns	10.2 ± 0.18 ns	0.988 ± 0.024
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 1/1/∇	17.2 ± 0.14 ns	17.1 ± 0.07 ns	1 ± 0.0092
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 1/1/∇²	18.1 ± 0.07 ns	18.6 ± 0.069 ns	0.975 ± 0.0052
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 2/2/∂	9.71 ± 0.16 ns	9.9 ± 0.18 ns	0.981 ± 0.024
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 2/2/∂²	10.2 ± 0.08 ns	10.1 ± 0.19 ns	1.01 ± 0.021
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 2/2/∇	17 ± 0.061 ns	17.1 ± 0.07 ns	0.991 ± 0.0054
RBF/Gaussian, exp(-(ε*r)²)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 2/2/∇²	18.1 ± 0.07 ns	18.7 ± 0.17 ns	0.97 ± 0.0097
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 0/0/∂	6.27 ± 0.12 ns	6.32 ± 0.01 ns	0.992 ± 0.019
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 0/0/∂²	14.2 ± 0.07 ns	14.2 ± 0.089 ns	0.999 ± 0.008
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 0/0/∇	8.7 ± 0.07 ns	8.54 ± 0.18 ns	1.02 ± 0.023
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 0/0/∇²	15.8 ± 0.08 ns	15.7 ± 0.051 ns	1.01 ± 0.0061
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 1/1/∂	6.81 ± 0.011 ns	6.32 ± 0.01 ns	1.08 ± 0.0024
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 1/1/∂²	14.2 ± 0.02 ns	14.2 ± 0.021 ns	1 ± 0.002
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 1/1/∇	8.72 ± 0.06 ns	8.6 ± 0.09 ns	1.01 ± 0.013
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 1/1/∇²	15.8 ± 0.07 ns	15.7 ± 0.12 ns	1.01 ± 0.0089
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 2/2/∂	6.33 ± 0.08 ns	6.32 ± 0.001 ns	1 ± 0.013
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 2/2/∂²	14.2 ± 0.021 ns	14.2 ± 0.1 ns	0.999 ± 0.0072
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 2/2/∇	8.72 ± 0.06 ns	8.53 ± 0.18 ns	1.02 ± 0.023
RBF/Inverse Multiquadrics, 1/sqrt((r*ε)²+1)
├─Shape factor: ε = 1
└─Polynomial augmentation: degree 2/2/∇²	15.8 ± 0.07 ns	15.7 ± 0.06 ns	1.01 ± 0.0059
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 0/0/∂	3.73 ± 0.01 ns	3.73 ± 0.009 ns	1 ± 0.0036
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 0/0/∂²	4.7 ± 0.01 ns	4.7 ± 0.011 ns	1 ± 0.0032
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 0/0/∇	5.69 ± 0.02 ns	5.61 ± 0.04 ns	1.01 ± 0.0081
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 0/0/∇²	3.11 ± 0.001 ns	3.11 ± 0 ns	1 ± 0.00032
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 1/1/∂	3.73 ± 0.01 ns	3.72 ± 0.01 ns	1 ± 0.0038
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 1/1/∂²	4.7 ± 0.01 ns	4.7 ± 0.01 ns	1 ± 0.003
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 1/1/∇	5.69 ± 0.02 ns	5.61 ± 0.04 ns	1.01 ± 0.0081
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 1/1/∇²	3.11 ± 0.001 ns	3.11 ± 0 ns	1 ± 0.00032
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 2/2/∂	3.73 ± 0.01 ns	3.72 ± 0.01 ns	1 ± 0.0038
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 2/2/∂²	4.7 ± 0.01 ns	4.7 ± 0.01 ns	1 ± 0.003
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 2/2/∇	5.69 ± 0.029 ns	5.61 ± 0.039 ns	1.01 ± 0.0087
RBF/Polyharmonic spline (r³)
└─Polynomial augmentation: degree 2/2/∇²	3.11 ± 0.001 ns	3.11 ± 0 ns	1 ± 0.00032
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 0/0/∂	4.27 ± 0.01 ns	4.27 ± 0.01 ns	1 ± 0.0033
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 0/0/∂²	5.56 ± 0.02 ns	5.54 ± 0.02 ns	1 ± 0.0051
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 0/0/∇	7.11 ± 0.3 ns	7.12 ± 0.01 ns	0.999 ± 0.042
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 0/0/∇²	4.27 ± 0.01 ns	4.28 ± 0.01 ns	0.998 ± 0.0033
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 1/1/∂	4.27 ± 0.01 ns	4.27 ± 0.01 ns	1 ± 0.0033
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 1/1/∂²	5.56 ± 0.02 ns	5.54 ± 0.02 ns	1 ± 0.0051
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 1/1/∇	7.11 ± 0.26 ns	7.12 ± 0.011 ns	0.999 ± 0.037
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 1/1/∇²	4.27 ± 0.01 ns	4.28 ± 0.01 ns	0.998 ± 0.0033
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 2/2/∂	4.27 ± 0.01 ns	4.27 ± 0.01 ns	1 ± 0.0033
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 2/2/∂²	5.56 ± 0.02 ns	5.54 ± 0.019 ns	1 ± 0.005
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 2/2/∇	7.11 ± 0.3 ns	7.11 ± 0.19 ns	1 ± 0.05
RBF/Polyharmonic spline (r¹)
└─Polynomial augmentation: degree 2/2/∇²	4.27 ± 0.01 ns	4.28 ± 0.01 ns	0.998 ± 0.0033
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 0/0/∂	4.96 ± 0.01 ns	5.27 ± 0.01 ns	0.941 ± 0.0026
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 0/0/∂²	4.96 ± 0.01 ns	4.96 ± 0.01 ns	1 ± 0.0029
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 0/0/∇	6.11 ± 0.07 ns	6.09 ± 0.05 ns	1 ± 0.014
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 0/0/∇²	3.11 ± 0.01 ns	3.11 ± 0.01 ns	1 ± 0.0046
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 1/1/∂	4.96 ± 0.009 ns	5.27 ± 0.01 ns	0.941 ± 0.0025
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 1/1/∂²	4.96 ± 0.01 ns	4.96 ± 0.001 ns	1 ± 0.002
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 1/1/∇	6.11 ± 0.08 ns	6.09 ± 0.06 ns	1 ± 0.016
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 1/1/∇²	3.11 ± 0.01 ns	3.11 ± 0.009 ns	1 ± 0.0043
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 2/2/∂	4.96 ± 0.011 ns	5.27 ± 0.01 ns	0.941 ± 0.0027
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 2/2/∂²	4.96 ± 0.01 ns	4.96 ± 0.01 ns	1 ± 0.0029
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 2/2/∇	6.11 ± 0.07 ns	6.09 ± 0.041 ns	1 ± 0.013
RBF/Polyharmonic spline (r⁵)
└─Polynomial augmentation: degree 2/2/∇²	3.11 ± 0.01 ns	3.11 ± 0.01 ns	1 ± 0.0046
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 0/0/∂	10.2 ± 0.25 ns	10.1 ± 0.25 ns	1 ± 0.035
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 0/0/∂²	5.18 ± 0.019 ns	4.96 ± 0.001 ns	1.04 ± 0.0038
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 0/0/∇	12.5 ± 0.07 ns	12.5 ± 0.07 ns	1 ± 0.0079
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 0/0/∇²	8.06 ± 0.09 ns	8.05 ± 0.081 ns	1 ± 0.015
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 1/1/∂	10.2 ± 0.25 ns	10.1 ± 0.25 ns	1 ± 0.035
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 1/1/∂²	4.72 ± 0.04 ns	4.96 ± 0.001 ns	0.952 ± 0.0081
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 1/1/∇	12.5 ± 0.071 ns	12.5 ± 0.07 ns	1 ± 0.008
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 1/1/∇²	8.58 ± 0.14 ns	8.06 ± 0.08 ns	1.06 ± 0.02
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 2/2/∂	10.1 ± 0.25 ns	10.2 ± 0.24 ns	0.995 ± 0.034
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 2/2/∂²	4.72 ± 0.04 ns	4.96 ± 0.01 ns	0.952 ± 0.0083
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 2/2/∇	12.9 ± 0.1 ns	12.5 ± 0.069 ns	1.04 ± 0.0099
RBF/Polyharmonic spline (r⁷)
└─Polynomial augmentation: degree 2/2/∇²	8.06 ± 0.09 ns	8.05 ± 0.071 ns	1 ± 0.014
time_to_load	0.799 ± 0.0076 s	0.794 ± 0.0016 s	1.01 ± 0.0098

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR.
Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).

- Widen batch call from Vector to AbstractVector and use map for GPU parallelism - Add test that non-CPU device throws ArgumentError - Add breadcrumb comment in _construct_sparse for future GPU sparse conversion (#88)

On Julia 1.10/1.11, _construct_sparse return type is inferred as a Union, causing AugmentedRuleReturnError. Explicitly parameterize AugmentedReturn with the primal type from the Enzyme return annotation so type parameters match exactly.

…edOperator Replace `out[:, d] = op.weights[d] * x` with `mul!(view(...), ...)` in three _eval_op methods, matching the pattern already used in the in-place variants. Also use `similar` instead of explicit Array constructors for GPU-safe allocation.

Combine pre-computed weights directly instead of rebuilding from scratch via the unified constructor, which always produces CPU SparseMatrixCSC. This prevents scalar indexing errors when calling combined operators on CuArrays.

Avoids errors when collections are empty or on GPU arrays where first() may not be efficient.

kylebeggs added 7 commits February 19, 2026 08:52

format

1fdba15

docs

2fd3a18

fix: use eltype() instead of typeof(first()) in show methods

06f4efc

Avoids errors when collections are empty or on GPU arrays where first() may not be efficient.

kylebeggs merged commit b373516 into main Feb 20, 2026
24 of 26 checks passed

kylebeggs deleted the gpu branch February 20, 2026 20:02

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

GPU updates#89

GPU updates#89
kylebeggs merged 8 commits intomainfrom
gpu

kylebeggs commented Feb 19, 2026

Uh oh!

codecov bot commented Feb 19, 2026 •

edited

Loading

Uh oh!

github-actions bot commented Feb 19, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

kylebeggs commented Feb 19, 2026

Uh oh!

codecov bot commented Feb 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

github-actions bot commented Feb 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmark Results

Benchmark Plots

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

codecov bot commented Feb 19, 2026 •

edited

Loading

github-actions bot commented Feb 19, 2026 •

edited

Loading