Gaugefields.jl v1

Gaugefields.jl provides gauge-link fields and core algorithms for lattice QCD. The v1 API uses LatticeMatrices and JACC by default and represents a Dim-dimensional gauge configuration as a vector of Dim link fields.

Start here

For the main workflow, begin with the four-dimensional quick start. It covers configuration creation, measurements, heatbath, molecular dynamics, stout smearing, file I/O, MPI, and GPU execution.

import JACC
JACC.@init_backend

using Gaugefields

U = gauge_configuration(
    (16, 16, 16, 32);
    colors=3,
    halo=1,
    start=:hot,
    seed=1234,
)

println("plaquette = ", measure_plaquette(U))
println("Polyakov loop = ", measure_polyakov_loop(U))

The same API is used for two and three dimensions; see Two and three dimensions.

Manual structure

Backends

LatticeMatricesBackend() is the default. CPU threads, GPUs, MPI, and multi-GPU execution are selected through JACC and process_grid without accelerator-specific Gaugefields constructors.

LegacyBackend() is available only when an application deliberately needs the serial compatibility implementation.

Pre-v1 compatibility API

Programs written against earlier Gaugefields releases remain supported, but their constructors, backend flags, low-level HMC drivers, and historical examples are documented only on the Legacy API (compatibility) page. They are not the recommended interface for new programs.

Installation

In Julia package mode:

pkg> add Gaugefields JACC

Gaugefields v1 requires LatticeMatrices v1.1 or later within the compatibility bounds declared by the package. Add MPI directly to an application that imports it, and add the selected GPU package when requested by JACC.