A High-Level Approach to Synthesis of High-Performance Codes for Quantum Chemistry

Abstract

This paper discusses an approach to the synthesis of high-performance parallel programs for a class of computations encountered in quantum chemistry and physics. These computations are expressible as a set of tensor contractions and arise in electronic structure modeling. An overview is provided of the synthesis system, that transforms a high-level specification of the computation into high-performance parallel code, tailored to the characteristics of the target architecture. An example from computational chemistry is used to illustrate how different code structures are generated under different assumptions of available memory on the target computer.

Statistical Models for Empirical Search-Based Performance Tuning

“…• We implemented the linear least squares regression method as described in Section 4.2, Equation (6). Since the least squares fit is based on choosing the fit parameters to minimize the total square error between the execution time data and the model predictions, errors in the larger problem sizes will contribute more significantly to the total squared error than smaller sizes, and therefore tend to dominate the fit.…”

Efficient synthesis of out-of-core algorithms using a nonlinear optimization solver

“…The optimization presented in this paper has been developed in the context of the Tensor Contraction Engine (TCE) [4,12], a domain-specific compiler for ab initio quantum chemistry calculations. The TCE takes as input a high-level specSadayappan), choppell@iiitmk.ac.in (Venkatesh Choppella).…”

“…In such situations, it is necessary to develop out-of-core algorithms that explicitly orchestrate the movement of blocks of data between main memory and secondary disk storage. We are developing a program synthesis tool called the Tensor Contraction Engine (TCE) [4,3] to facilitate the development of parallel programs for this domain, by automatically transforming high-level tensor contraction expressions into efficient parallel programs.…”

Data Locality Optimization for Synthesis of Efficient Out-of-Core Algorithms

“…The code synthesis process in the Tensor Contraction Engine [1] involves multiple steps, including algebraic transformation, loop fusion and loop tiling. The loop fusion and loop tiling steps are coupled together.…”

“…This paper presents an approach to automatically synthesize efficient out-of-core algorithms in the context of the Tensor Contraction Engine (TCE) program synthesis tool [1,3,2,4]. The TCE targets a class of electronic structure calculations which involve many computationally intensive components expressed as tensor contractions.…”