V_SQRT

Abstract

Computing square-roots have a substantially high result latency on the Cray MPP. We have implemented an efficient software routine on Cray MPP to compute the square-roots of a double-precision real vector. This program written in CAM assembler language runs about seven times faster than the currently available library routine. The worst case deviation of the computed values using v_sqrt from those computed using the standard library routine is about one least significant bit.

DateOfInformation

Wed Apr 9 12:05:32 1997

Domain

Numerical Programs and Routines!Miscellaneous

Keyword

square root computation; numerical routine; architecture dependence

Name

V_SQRT

TargetEnvironment

Cray MPP, written in CAM assembler language

Webpage

http://sdcd.gsfc.nasa.gov/ESS/exchange/contrib/udaya/v_sqrt.html

ContactIs

Udaya A. Ranawake