From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 7613 invoked by alias); 13 Mar 2007 05:03:55 -0000 Received: (qmail 7604 invoked by uid 22791); 13 Mar 2007 05:03:54 -0000 X-Spam-Check-By: sourceware.org Received: from alnrmhc12.comcast.net (HELO alnrmhc12.comcast.net) (206.18.177.52) by sourceware.org (qpsmtpd/0.31) with ESMTP; Tue, 13 Mar 2007 05:03:48 +0000 Received: from hippogriff.homeunix.org ([75.71.67.96]) by comcast.net (alnrmhc12) with ESMTP id <20070313050346b1200nnpdae>; Tue, 13 Mar 2007 05:03:46 +0000 Received: by hippogriff.homeunix.org (Postfix, from userid 1000) id 47ECC36398; Mon, 12 Mar 2007 23:03:50 -0600 (MDT) Date: Tue, 13 Mar 2007 05:03:00 -0000 From: Patrick Alken To: gsl-discuss@sourceware.org Subject: generalized eigensystems Message-ID: <20070313050349.GA22421@hippogriff.physics.drexel.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.4.2.2i Mailing-List: contact gsl-discuss-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gsl-discuss-owner@sourceware.org X-SW-Source: 2007-q1/txt/msg00044.txt.bz2 Hello all, The current CVS contains support for generalized eigensystems. All the functions are documented in eigen.texi. I have tested the code for randomized and integer matrix systems up to N = 1000 and compared to lapack and things look good. All my development/testing was done on amd32/linux so if anyone is willing to test it out on other archs I'd appreciate it. There is a test program in eigen/testgen.c which will generate matrices, compute eigenvalues, and test them against lapack's output. The header of testgen.c explains how to compile it. Once its compiled, you can just do: ./testgen -n 100 -z to test N = 100 matrices and compute the schur vectors and test them too. Sometimes it will output a small discrepancy with lapack (say around a 10^{-5} difference for an eigenvalue) and most of these I have tracked down to an initial near-singular or ill-conditioned matrix input in which case lapack's output can't be trusted either. But you can send along any output from the test program to me and I'll have a look at it. Thanks, Patrick Alken