Next Silicon Ltd

Daniel Khankin

An apparatus for computing functions using polynomial-based approximation, comprising one or more processing circuitries configured for computing a polynomial-based approximant approximating a function by executing one or more iterations. Each iteration comprising computing the polynomial-based approximant using scaled fixed-point unit(s) according to a constructed set of coefficients, minimizing an approximation error of the computed polynomial-based approximant compared to the function while complying with one or more constraints selected from a group comprising at least: an accuracy, a compute graph size, a computation complexity, and a hardware utilization of the processing circuitry(s), adjusting one or more of the coefficients in case the approximation error is incompliant with the constraint(s) and initiating another iteration. The polynomial-based approximant and its adjusted set of coefficients for which the computed polynomial-based approximant complies with the constraint(s) may be output to one or more processing circuitries configured to approximate the function by computing the polynomial-based approximant.

Automatic generation of computation kernels for approximating elementary functions

A method of generating automatically architecture-specific algorithms, comprising receiving an architecture independent algorithm and one or more algorithm parameters defining at least a target processing architecture and a format of an output of an architecture-specific algorithm implementing the received algorithm, determining automatically a functionality of the algorithm by analyzing the algorithm, selecting one or more architecture-specific computing blocks of the target processing architecture according to the functionality of the algorithm and the algorithm parameter(s) wherein each computing block is dynamically reconfigurable in runtime and associated with (1) simulation code simulating its functionality, and (2) execution code executing its functionality, testing an emulated architecture-specific algorithm constructed using the simulation code of the selected architecture-specific computing block(s) to verify compliance with the algorithm parameter(s), and, responsive to successful compliance verification, generating automatically an architecture-specific code segment implementing the architecture-specific algorithm based on the execution code of the selected architecture-specific computing blocks.

Growing community of inventors

Beer Sheva, Israel

Daniel Khankin