Array processing with fused multiply/add instruction.

摘要

Described is an array-processing method comprising: performing an inverse transform procedure on an input data array having a dimensionality of m, where m is an integer greater than or equal to one, the input data array being a scaled discrete cosine transform of data representative of a physical variable, the inverse transform procedure being an inverse of an m-dimensional discrete cosine transform scaled according to a predetermined scaling matrix Q, the inverse transform procedure including the following operations: computing a descaling constant and a seqence of descaling/inverse-transform matrices incorporating elements of the scaling matrix Q; multiplying one point of the input data array by the descaling constant; matrix multiplying the result by the sequence of descaling/inverse-transform matrices in turn, the descaling/inverse-transform matrices being such that each matrix multiplication involves fused signed multiply/add operations exclusively.