# Model-independent partial wave analysis using a massively-parallel fitting framework

## TL;DR

This paper extends the GooFit framework to perform model-independent partial wave analysis of three-body decays using GPU acceleration, enabling faster and more flexible amplitude extraction.

## Contribution

It introduces a GPU-accelerated, model-independent partial wave analysis method within GooFit, utilizing spline interpolation for S-wave amplitudes in three-body decays.

## Key findings

- GPU implementation is hundreds of times faster than CPU.
- The method allows flexible amplitude modeling with spline interpolation.
- Performance benchmarks across platforms demonstrate significant speedups.

## Abstract

The functionality of GooFit, a GPU-friendly framework for doing maximum-likelihood fits, has been extended to extract model-independent S-wave amplitudes in three-body decays such as $D^+ \to h^+h^+h^-$. A full amplitude analysis is done where the magnitudes and phases of the S-wave amplitudes are anchored at a finite number of $m^2(h^+h^-)$ control points, and a cubic spline is used to interpolate between these points. The amplitudes for P-wave and D-wave intermediate states are modeled as spin-dependent Breit-Wigner resonances. GooFit uses the Thrust library, with a CUDA backend for NVIDIA GPUs and an OpenMP backend for threads with conventional CPUs. Performance on a variety of platforms is compared. Executing on systems with GPUs is typically a few hundred times faster than executing the same algorithm on a single CPU.