R-ScaLAPACK General Documentation

Back to: Main page, Individual functions, ASPECT.

Requirements:

Installation

Usage

Original R versionR-ScaLAPACK versionNotes
> solve(x,y) > sla.solve(x,y) Uses a single processor
> sla.solve(x,y,2) Uses two processors
Note that the R-ScaLAPACK library must be loaded before it can be used.

> load("data.R")
> solve(x,y)
> library(scalapack)
> load("data.R")
> sla.solve(x,y,4)
A longer program showing the loading of the library and using solve with 4 processes.
Optional arguments are available for tweaking

> load("data.R")
> qr(x)
> library(scalapack)
> load("data.R")
> sla.qr(x, NPROWS=16, NPCOLS=8, MB=64)
Runs QR decomposition on a 16 x 8 process grid (128 processors); uses a 64x64 block size
The process grid can be saved so that it doesn't have to be specified each time


> load("data.R")
> qr(x)
> library(scalapack)
> .RscalaGrid=c(16,8)
> load("data.R")
> sla.qr(x)
Runs QR decomposition on a 16 x 8 process grid; uses default block size

FAQ

Why?
Why not?



Written by David Bauer   January 17, 2004
Using: Vim
Valid HTML 4.0