[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
tolerances in mappers.cc
From: |
Ben Abbott |
Subject: |
tolerances in mappers.cc |
Date: |
Wed, 27 Aug 2008 20:14:47 -0400 |
On Aug 26, 2008, at 11:29 PM, John W. Eaton wrote:
On 26-Aug-2008, Ben Abbott wrote:
I notice that single
| precision tests often make use of "eps" rather than "eps('single')".
| Would it be good form to change all single precision tests to
refer to
| eps('single')?
Yes, that would probably be the right thing to do.
I notice that some of the tolerances I've examined are rather "liberal".
For example, each of the below pass for me.
1184 %!assert(log ([1, e, e^2]), [0, 1, 2], eps);
1185 %!assert(log ([1, e, e^2]), [0, 1, 2], sqrt (eps));
As I have some nights to myself for a few weeks, I'll first
"single"ize the tests in mappers.cc which respect single precision but
use refer to "eps" for the tolerance. After that I'm happy to look at
each of tests tolerances to determine if they are extremely loose.
Thoughts?
Ben