|
| From: | Frantisek Kluknavsky |
| Subject: | Re: [Bug-gsl] ode-initval2 testsuite failure |
| Date: | Tue, 22 Jan 2013 10:35:36 +0100 |
| User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/17.0 Thunderbird/17.0 |
On 01/21/2013 06:11 PM, Tuomo Keskitalo wrote:
test_extreme_problems() gives some kind of evidence that ode-initval2 has methods that can tackle these kinds of numerical nightmares.
Maybe it is an accident rather than a positive evidence. Adding a few more bits of mantissa is enough to crash the test. Or is the simulation just too long, tolerance too narrow? Does msbdf contain some trick that relies on strict IEEE 754 conformance and occasional added precision ruins its design?
| [Prev in Thread] | Current Thread | [Next in Thread] |