Search for a command to run...
Maranatha.jl v1.2.2 Maranatha.jl is a research-oriented Julia framework for deterministic quadrature-based convergence analysis and continuum extrapolation on hyperrectangular domains. The package combines: multi-dimensional tensor-product quadrature derivative-based and refinement-based error-scale modeling weighted least $\chi^2$ fitting for extrapolation toward $h \to 0$ structured plotting, reporting, and reproducible archival output Unlike black-box adaptive integration libraries, Maranatha is designed for controlled numerical studies in which convergence structure, rule behavior, and error scaling are made explicit and reproducible. Current capabilities Deterministic tensor-product quadrature Multi-dimensional quadrature on rectangular domains $[a_1,b_1] × ⋯ × [a_d,b_d]$, including the hypercube special case $[a,b]^d$ Shared-scalar or axis-wise specification for: domain bounds quadrature rule endpoint boundary Rule-dispatched quadrature backends including: Newton–Cotes Gauss-family rules B-spline-based rules Multiple execution backends: serial CPU evaluation threaded subgrid CPU execution CUDA-based GPU execution Error-scale modeling Maranatha supports two complementary error-model families. Derivative-based residual models residual-order detection from rule structure multi-term residual modeling via nerr_terms direct and jet-based derivative pathways backend support through: ForwardDiff.jl TaylorSeries.jl FastDifferentiation.jl Enzyme.jl Refinement-based models derivative-free coarse-vs-refined quadrature comparison unified refinement dispatch across supported quadrature families support for axis-wise rule configurations when all active axes remain within the same quadrature family These models are designed to provide stable error-scale information for extrapolation workflows rather than strict certified error bounds. Continuum extrapolation Weighted least $\chi^2$ fitting for $h → 0$ Automatically selected exponent bases from rule-dispatched residual structure Axis-wise residual-power merging for mixed per-axis configurations Optional ff_shift control for suppressing unstable or vanishing leading orders Full covariance matrix output and covariance-aware uncertainty propagation Visualization, reporting, and workflow support Convergence plots with fitted uncertainty bands Datapoint-only and fitted-result reporting workflows Internal-note style report generation Structured JLD2/TOML output for reproducible studies Merge and filtering tools for saved datapoint runs TOML-driven execution and interactive configuration wizard support Highlights of this release This release reflects a substantial expansion and reorganization of the package compared with earlier releases. Notable improvements include: support for axis-wise quadrature configuration across domain bounds, rule, and boundary unified handling of axis-wise specifications across quadrature, error estimation, fitting, IO, plotting, and reporting strengthened refinement-based error-estimation support, including same-family axis-wise refinement workflows improved runner, utility, documentation, quadrature-dispatch, and error-estimation module organization performance improvements in the generic ND threaded subgrid backend reduced CUDA host-side launch-preparation overhead reduced allocation overhead in jet-based automatic-differentiation helper paths broader internal documentation and docstring coverage across newly factored helper layers Typical workflow A standard Maranatha workflow is: Run multi-resolution quadrature with run_Maranatha(...) Build an error-scale model using derivative-based or refinement-based estimation Fit the convergence data with least_chi_square_fit(...) Visualize and archive the results with the plotting and reporting tools Source code https://github.com/saintbenjamin/Maranatha.jl Documentation https://saintbenjamin.github.io/Maranatha.jl/