Gauge_configurations are generated by Markov Chain Monte Carlo using Metropolis-type-algorithms. To generate the random numbers they used a hybrid_Tausworthe.
Features of SIMULATeQCD
- currently SIMULATeQCD can generate pure SU(3) configurations and HISQ configurations with \(N_f=N_l+N_s\) (for N_l light and N_s strange) dynamical quarks.
- To enable simulation of 4 or more degenerate flavours, the determinant is split up into \(N_{pf}\) pseudofermion_fields since the rational_approximation requires \(N_{i}<4\). Note that \(N_{l}\) and \(N_{s}\) are input parameters of rational approximation, while \(N_{pf}\) is an input parameter of SIMULATeQCD. By default simulation runs with N_{pf}=1.
- By default dynamical configurations are generated at \(\mu_{B}=0\), but it is also possible to generate at pure imaginary \(\mu_{B}\).
- For pure gauge configurations generation (SU(3)) , standard Wilson gauge action is implemented with efficient sampling through heat_bath(HB) and over_relaxation (OR) updates.
- For dynamical fermion configuration generation , HISQ action is used along with RHMC algorithm.
- To carry out matrix inversions, conjugate gradient algorithm is used, for which multiple right hand sides(MRHS), multiple shifts, and mixed precision implementation are available to improve Performance.