Co-Exploration of NLA Kernels and Specification of Compute Elements in Distributed Memory CGRAs

Coarse Grained Reconfigurable Architectures (CGRA) are emerging as embedded application processing units in computing platforms for Exascale computing. Such CGRAs are distributed memory multi- core compute elements on a chip that communicate over a Network-on-chip (NoC). Numerical Linear Algebra (NLA) kernels are key to several high performance computing applications. In this paper we propose a systematic methodology to obtain the specification of Compute Elements (CE) for such CGRAs. We analyze block Matrix Multiplication and block LU Decomposition algorithms in the context of a CGRA, and obtain theoretical bounds on communication requirements, and memory sizes for a CE. Support for high performance custom computations common to NLA kernels are met through custom function units (CFUs) in the CEs. We present results to justify the merits of such CFUs.

Lower Bound Axisymmetric Limit Analysis Using Drucker-Prager Yield Cone and Simulation with Mohr-Coulomb Pyramid

This paper presents a lower bound limit analysis approach for solving an axisymmetric stability problem by using the Drucker-Prager (D-P) yield cone in conjunction with finite elements and nonlinear optimization. In principal stress space, the tip of the yield cone has been smoothened by applying the hyperbolic approximation. The nonlinear optimization has been performed by employing an interior point method based on the logarithmic barrier function. A new proposal has also been given to simulate the D-P yield cone with the Mohr-Coulomb hexagonal yield pyramid. For the sake of illustration, bearing capacity factors N-c, N-q and N-gamma have been computed, as a function of phi, both for smooth and rough circular foundations. The results obtained from the analysis compare quite well with the solutions reported from literature.