Automatic generation of computation kernels for approximating elementary functions

A method of computing functions using polynomial-based approximation, comprising using at least one processing circuitry, the at least one processing circuitry comprises at least one interconnected computing grid, the at least one interconnected computing grid comprising:

• at least one interconnect network,
• a plurality of configurable data routing junctions, and
• a plurality of reconfigurable logical elements connectable by the plurality of configurable data routing junctions and the at least one interconnect network;
• the at least one processing circuitry is configured for executing a code, the code comprising:
• code instructions to compute a polynomial-based approximant approximating a function by executing at least one iteration, the at least one iteration comprising:
• computing the polynomial-based approximant using at least one scaled fixed-point unit according to a constructed set of coefficients,
• minimizing an approximation error of the computed polynomial-based approximant compared to the function while complying with at least one constraint selected from a group comprising at least:
• an accuracy, a compute graph size, a computation complexity, and a hardware utilization of the at least one processing circuitry,
• adjusting at least one of the coefficients of the set of coefficients in case the approximation error is incompliant with the at least one constraint and initiating another iteration, and
• code instructions to output the polynomial-based approximant with the adjusted set of coefficients for which the computed polynomial-based approximant complies with the at least one constraint to at least one processing circuitry configured to approximate the function by computing the polynomial-based approximant using the adjusted set of coefficients.