Browse Prior Art Database

Improved Efficiency of Vector Sum Accumulations

IP.com Disclosure Number: IPCOM000103375D
Original Publication Date: 1990-Oct-01
Included in the Prior Art Database: 2005-Mar-17
Document File: 1 page(s) / 45K

Publishing Venue

IBM

Related People

Olson, RW: AUTHOR

Abstract

On computers, an arithmetic vector X is a set of numbers {X1,X2,....Xn}. The sum of X is defined as: SUM = X1 + X2 + X3 + ... + Xn

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 74% of the total text.

Improved Efficiency of Vector Sum Accumulations

      On computers, an arithmetic vector X is a set of numbers
{X1,X2,....Xn}. The sum of X is defined as:
   SUM = X1 + X2 + X3 + ... + Xn

      On vector computer this summation is usually implemented by
using partial sums.
                          PSUM1 = X1 + Xp+1 + X2p+1 + ...
                          PSUM2 = X2 + Xp+2 + X2p+2 + ...
                          PSUMp = Xp + Xp+p + X2p+p + ...

      Calculating the PSUM values is called accumulation of the
partial sums.  This may occur all at once or it may be done in
several steps at different times.  After all the partial sums are
complete, they are added together (summation of the partial sums) to
arrive at the final result.
   SUM = PSUM1 + PSUM2 + PSUM3 + ... + PSUMp

      Many vector computers use a vector register to hold the partial
sums during their accumulation and summation. Typically the number of
partial  sums, P (often 4-8), is much less than the number of spaces
in the vector register (often 64-256).  Because the chosen register
may also be needed for other purposes, it is advantageous to have
instructions which move the partial sum quantities into and out of
the register without disturbing the other spaces in the register.
For convenience these may be termed "load partial sums" and "store
partial sums" instructions.  With these instructions it is easy to:
s...