Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FormCalc 7

S. Agrawal, T. Hahn, E. Mirabella

TL;DR

FormCalc 7 advances one-loop calculation workflows by integrating analytic tensor reduction with Denner–Dittmaier TensRed, enabling selectable OPP unitarity methods, and linking MSSM initialization to FeynHiggs. It also enhances code generation with precision-agnostic RealType/ComplexType abstractions and moves toward C99 output, while adding command-line model initialization for flexibility. The Cuba library is now capable of automatic multi-core parallelization within FormCalc’s ecosystem, broadening scalable numerical integration capabilities. Collectively, these changes improve both the analytical handling of tensor integrals and the practical usability and performance of cross-section computations in complex theories such as the MSSM.

Abstract

We present additions and improvements in Version 7 of FormCalc, most notably analytic tensor reduction, choice of OPP methods, and MSSM initialization via FeynHiggs, as well as a parallelized Cuba library for numerical integration.

FormCalc 7

TL;DR

FormCalc 7 advances one-loop calculation workflows by integrating analytic tensor reduction with Denner–Dittmaier TensRed, enabling selectable OPP unitarity methods, and linking MSSM initialization to FeynHiggs. It also enhances code generation with precision-agnostic RealType/ComplexType abstractions and moves toward C99 output, while adding command-line model initialization for flexibility. The Cuba library is now capable of automatic multi-core parallelization within FormCalc’s ecosystem, broadening scalable numerical integration capabilities. Collectively, these changes improve both the analytical handling of tensor integrals and the practical usability and performance of cross-section computations in complex theories such as the MSSM.

Abstract

We present additions and improvements in Version 7 of FormCalc, most notably analytic tensor reduction, choice of OPP methods, and MSSM initialization via FeynHiggs, as well as a parallelized Cuba library for numerical integration.

Paper Structure

This paper contains 8 sections, 1 equation, 1 figure.

Table of Contents

  1. Introduction
  2. Analytic tensor reduction
  3. Unitarity methods (OPP)
  4. Improved code generation
  5. Command-line parameters for model initialization
  6. MSSM initialization via FeynHiggs
  7. Built-in parallelization in Cuba
  8. Summary

Figures (1)

  • Figure 1: Cuba speed-ups for a three-dimensional integral on an i7-2600 Linux system (2.6.37) with 4 real/8 virtual (hyperthreaded) cores. The vertical line at 4 cores marks the cross-over. Left column: 'easy' integrand, right column: 'hard' integrand. Top row: 'fast' integrand ($1\,\mu$sec), bottom row: 'slow' integrand ($1000\,\mu$sec per evaluation). Note that also in the one-core case a parallel version is used (one master, one worker), which explains why the timings normalized to the serial version are below 1, in the top row visibly so. The measured speed-ups are, if anything, on the conservative side as the i7 CPUs boast a feature named Turbo Boost which allows the CPU to scale up the frequency if not all cores are loaded, i.e. the serial version will likely have run at a somewhat higher CPU frequency.