The configuration-interaction (CI) method, long a popular approach to describe quantum many-body systems, is cast as a very large sparse matrix eigenpair problem with matrices whose dimension can exceed one billion. Such formulations place high demands on memory capacity and memory bandwidth --- two quantities at a premium today. In this paper, we describe an efficient, scalable implementation, BIGSTICK, which, by factorizing both the basis and the interaction into two levels, can reconstruct the nonzero matrix elements on the fly, reduce the memory requirements by one or two orders of magnitude, and enable researchers to trade reduced resources for increased computational time. We optimize BIGSTICK on two leading HPC platforms --- the Cray XC30 and the IBM Blue Gene/Q. Specifically, we not only develop an empirically-driven load balancing strategy that can evenly distribute the matrix-vector multiplication across 256K threads, we also developed techniques that improve the performance of the Lanczos reorthogonalization. Combined, these optimizations improved performance by 1.3-8× depending on platform and configuration.

Good FPGA placement is crucial to obtain the best Quality of Results (QoR) from FPGA hardware. Although many published global placement techniques place objects in a continuous ASIC-like environment, FPGAs are discrete in nature, and a continuous algorithm cannot always achieve superior QoR by itself. Therefore, discrete FPGA-specific detail placement algorithms are used to improve the global placement results. Unfortunately, most of these detail placement algorithms do not have a global view. This paper presents a discrete "middle" placer that fills the gap between the two placement steps. It works like simulated annealing, but leverages various acceleration techniques. It does not pay the runtime penalty typical of simulated annealing solutions. Experiments show that with this placer, final QoR is significantly better than with the global-detail placer approach.

This paper presents a module generator which automatically generates and optimizes circuitry to satisfy constraints of speed, area and power. The user has complete control over the clock timing driving the circuitry and the area, width, or height of the resulting module. Unlike other programs that have been optimized for area and speed, this program supports more degrees of freedom and a broad range of circuit constructs, permitting a complete integrated circuit to be designed to meet the overall IC project objectives.