Eve library

Kalika Phase 113 Performance Benchmark Report

Guide

Revision 1

On this page

Generated: 2026-04-28T00:00:00.000Z

Scope#

Task 113.9.1.2 is implemented as the @kalika/core phase-113 benchmark gate:

  • 69 Kalika packages are represented by explicit benchmark specs.
  • Every task-required SOTA comparator is covered: Mathematica, FLINT, GAP, PARI/GP, xAct, FeynCalc, FORM, LAPACK, and PETSc.
  • Additional domain comparators are tracked where they are the right reference point: SUNDIALS, SINGULAR, Magma, SDPB, CVXPY, NLopt, CLASS, CAMB, Gmsh, Regina, SnapPy, and Gfan.
  • The suite records deterministic benchmark results, compares them to target thresholds, stores baselines, and flags regressions by metric direction.

Source Registry#

The benchmark target registry uses official project sources for comparator identity and domain assignment:

System Source
Mathematica https://www.wolfram.com/mathematica/
FLINT https://flintlib.org/
GAP https://www.gap-system.org/
PARI/GP https://pari.math.u-bordeaux.fr/
xAct https://xact.es/
FeynCalc https://feyncalc.github.io/
FORM https://www.nikhef.nl/~form/
LAPACK https://netlib.org/lapack/index.html
PETSc https://petsc.org/
SUNDIALS https://computing.llnl.gov/projects/sundials
SINGULAR https://www.singular.uni-kl.de/
Magma https://www.maths.cam.ac.uk/computing/software/Magma
SDPB https://github.com/davidsd/sdpb
CVXPY https://www.cvxpy.org/
NLopt https://nlopt.readthedocs.io/
CLASS https://lesgourg.github.io/class_public/class.html
CAMB https://camb.info/
Gmsh https://gmsh.info/
Regina https://regina-normal.github.io/
SnapPy https://snappy.computop.org/index.html
Gfan https://users-math.au.dk/jensen/software/gfan/gfan.html

Regression Gate#

The suite exports:

  • createPhase113PerformanceBenchmarkSuite()
  • runPhase113PerformanceBenchmarkSuite()
  • createPhase113BenchmarkBaselineStore()
  • trackPhase113BenchmarkRegressions()
  • summarizePhase113PerformanceBenchmarks()
  • phase113BenchmarkReportToMarkdown()

The corresponding Vitest coverage verifies complete module coverage, official source URLs, required comparator coverage, deterministic target passes, new-baseline recording, and regression failure detection.