Global‐view coefficients: a data management solution for parallel quantum Monte Carlo applications

Abstract

SUMMARYQuantum Monte Carlo (QMC) applications perform simulation with respect to an initial state of the quantum mechanical system, which is often captured by using a cubic B-spline basis. This representation is stored as a read-only table of coefficients and accesses to the table are generated at random as part of the Monte Carlo simulation. Current QMC applications, such as QWalk and QMCPACK, replicate this table at every process or node, which limits scalability because increasing the number of processors does not enable larger systems to be run. We present a partitioned global address space approach to transparently managing this data using Global Arrays in a manner that allows the memory of multiple nodes to be aggregated. We develop an automated data management system that significantly reduces communication overheads, enabling new capabilities for QMC codes. Experimental results with QWalk and QMCPACK demonstrate the effectiveness of the data management system.

Noncovalent Interactions by Fixed-Node Diffusion Monte Carlo: Convergence of Nodes and Energy Differences vs Gaussian Basis-Set Size

“…Millions of CPU-hours are often invested in promising large-scale projects. ,,, It would be desirable to single out ways to reduce the related CPU costs to a maximum possible extent. This can be acomplished by algorithmic and methodological improvements − as well as by simplification of Ψ T .…”

Optimization and Parallelization of B-Spline Based Orbital Evaluations in QMC on Multi/Many-Core Shared Memory Processors

“…To conduct cutting edge scientific research on material science, QMCPACK has been deployed on the current generation of leadership supercomputers, Mira (IBM Blue Gene/Q) at Argonne National Laboratory and Titan (AMD Opteron CPUs and NVIDIA Tesla K20 GPU) at Oak Ridge National Laboratory. As the most performance-critical component in QMCPACK, 3D B-splines orbitals have been extensively optimized over the years [13] [14]. The highly optimized routines evaluating B-spline SPOs are implemented in QPX intrinsics [15] on BG/Q, SSE/SSE2 intrinsics on x86 and in CUDA [16] to maximize single-node performance.…”