Design and performance characterization of electronic structure calculations on massively parallel supercomputers: a case study of GPAW on the Blue Gene/P architecture

摘要

Density function theory (DFT) is the most widely employed electronic structure method because of itsrnfavorable scaling with system size and accuracy for a broad range of molecular and condensed-phasernsystems. The advent of massively parallel supercomputers has enhanced the scientific community’s abilityrnto study larger system sizes. Ground-state DFT calculations on ～10~3 valence electrons using traditionalrnO(N~3) algorithms can be routinely performed on present-day supercomputers. The performance characteristicsrnof these massively parallel DFT codes on >10~4 computer cores are not well understood. The GPAWrncode was ported an optimized for the Blue Gene/P architecture. We present our algorithmic parallelizationrnstrategy and interpret the results for a number of benchmark test cases.