MathGroup Archive 2004

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

Search the Archive

The unstable LinearSolve

  • To: mathgroup at smc.vnet.net
  • Subject: [mg47779] The unstable LinearSolve
  • From: "Qihong Wu" <wu62 at purdue.edu>
  • Date: Mon, 26 Apr 2004 02:41:49 -0400 (EDT)
  • Organization: Purdue University
  • Sender: owner-wri-mathgroup at wolfram.com

Hi,  Guys

I have an amazing question about LinearSolve[A,b].
There is a 60k*60k complex matrix A in the form of SparseArray, and a
Complex Vector b, Now I want to solve A X=b by LinearSolve[A,b]. It will
take ~4 seconds to solve it by my IBM T40 2373-4cu laptop (PM1.5G/512M),
this is a good result I think as shown in the new features of Mathematica 5.
But the nightmare is that it takes ~50 second to finish it by a Dell
Optiplex GX 260 2.8GHz/1024M. The Dell speed is almost 10 times slower than
IBM Laptop. Does it make sense under Same platform (Both of them is
Mathematica 5/Win Xp)? Does anybody has the same experience? Any comments
are very appreciated.

Btw: What type should I use for a matrix in the form of SparseArray in the
Compile function. The elements in the Matrix are complex, when I said
{A,_Complex,2} in the compile function, Mathematica 5 always said the type
was wrong. Why? What type should I use in Compile? Thank you.



  • Prev by Date: Re: Parametric Plot
  • Next by Date: Re: Parametric Plot
  • Previous by thread: RE: FindMaximum doesn't work?
  • Next by thread: Re: The unstable LinearSolve