MathGroup Archive 1996

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Inverse of Large Matrices

  • To: mathgroup at smc.vnet.net
  • Subject: [mg3450] Re: Inverse of Large Matrices
  • From: rmj at leland.Stanford.EDU (Roger M. Jones)
  • Date: Fri, 8 Mar 1996 01:30:49 -0500
  • Organization: Stanford University, CA 94305, USA
  • Sender: owner-wri-mathgroup at wolfram.com

In inverting large (412 )matrices I have noticed a large
difference between memory usage in LinearSolve &
Inverse.  Specifically I find for the below cases:

1: res1=LinearSolve[a,rhs];
2: res2=Inverse[a].rhs;

(here a and rhs are 412 X 412 dimensioned matrices)
 
case 1 repeatedly applied consumes .7 MB more memory
on each application, whereas case 2 is almost constant
(or small at the very least).  I need to use method 1
as this is faster than case 2.  Further, I need to apply
this many times (>2000) and so this is very important
consideration.  Case 1 rapidly runs out of memory of 
course.


Does LinearSolve perhaps generate hidden variables 
which are not cleared?  Any help offered will be
very gratefully received.


Regards,

Roger Jones
(rmj at leland.stanford.edu)
-- 
Regards,

Roger Jones

(rmj at leland.stanford.edu)

==== [MESSAGE SEPARATOR] ====


  • Prev by Date: Re: Q: Problems with package legend.m
  • Next by Date: Nonlinear fitting of a surface????
  • Previous by thread: Re: Q: Problems with package legend.m
  • Next by thread: Re: Inverse of Large Matrices