Experiences with re-engineering and parallelizing a high- T c superconductivity code

Abstract

Motivated by experiments on high- T c superconducting compounds and made possible by the power of present day supercomputers, a number of researchers have used the Quantum Monte Carlo method to gain more insight into Hubbard-like many particle models. We used the Projector Quantum Monte Carlo method (PQMC) [1,2] as a computational technique for studying the physical properties of many-electron and quantum-spin systems. This review gives a survey of the algorithms and techniques as well as of the methods to implement the PQMC scheme on modern parallel computers like the Intel Hypercube or IBM RS/6000 workstation clusters. In order to study systems of larger size or with more complex interactions, increasingly large amounts of computational power are required. Although based on one particular algorithm, we believe that our results should interest simulators with quite different algorithms and other parallel architectures as well.