|
From: | Robert T. Short |
Subject: | Bessel Functions and Thumbscrews |
Date: | Fri, 02 Mar 2012 09:58:37 -0800 |
User-agent: | Mozilla/5.0 (X11; Linux i686; rv:10.0.2) Gecko/20120216 Thunderbird/10.0.2 |
Following up on a post and bug report several weeks ago.I have attached a patch in which I added a bunch of tests to bessel{j,y,i,k}. The existing tests were fine, but the Amos code (libcruft/amos) uses different algorithms for different orders and arguments and the existing tests were for small values of the order and magnitude. I created a subset of the tables in Abramowitz and Stegun and compared the octave results to the tables. In particular for large order/arguments Amos used the asymptotic expansion and so one might expect some differences (and there are).
I also added tests to cover the bug that set me off on this mission - checking that negative integer orders give the correct answers.
Bob
bsltst.diff
Description: Text Data
[Prev in Thread] | Current Thread | [Next in Thread] |