Method to compute sliding window block sum using instruction based selective horizontal addition in vector processor

Abstract

This invention forms a block sum of picture elements employing a vector dot product instruction to sum packed picture elements and the mask producing a vector of masked horizontal picture element. The block sum is formed from plural horizontal sums via vector single instruction multiple data (SIMD) addition.

What is claimed is: 
     
         1 . A method for forming a block sum of picture elements comprising the steps of:
 precalculating at least one mask dependent upon a vector element size and a horizontal size of the block;   supplying packed picture elements of a row of the block and a mask of the at least one bit to a vector dot product unit;   employing the vector dot product unit to form a vector dot product of the packed picture elements and the mask producing a vector of masked horizontal element sums;   summing vectors of plural masked horizontal element sums to produce block sums via vector single instruction multiple data (SIMD) addition.   
     
     
         2 . The method of  claim 1 , wherein:
 said step of supplying packed picture elements includes
   recalling a vector size number of picture elements from memory starting with a row beginning,   recalling a further vector size number of picture elements immediately following a prior recalling, and   upon reaching an end of a row, recalling a vector size number of picture elements from memory starting with a row beginning of a next row for an entire image.   
   
     
     
         3 . The method of  claim 1 , wherein:
 said step of summing vectors of plural masked horizontal element sums includes
 summing a first two plural masked horizontal element sums forming an intermediate sum, and
 summing a next masked horizontal element sum with the intermediate sum forming a next intermediate sum for each masked horizontal element sum up to a vertical