|
From: | Evgeny Stambulchik |
Subject: | Re: [Bug-gsl] Negative a or b in gsl_sf_hyperg_2F1_e() |
Date: | Fri, 15 Jul 2005 01:51:50 +0300 |
User-agent: | Mozilla Thunderbird 1.0.2 (X11/20050317) |
Brian Gough wrote:
Evgeny Stambulchik writes: > Since the 2F1 series is finite for either a or b being negative, this > should be used irrespectively of whether the absolute values are below > 10.0 (why this specific value, BTW?). Have you done any comparisons or testing to confirm whether this is better?
Well, I mean - for (|a| > 10 || |b| > 10) the function (the original variant of it) refuses to work at all. So I'd say it's better with the patch ;-). Now, I didn't compare results for integer negative a or b below 10 (by abs. value), but don't think a sum of a few (< 10) terms can introduce any noticeable rounding errors.
Best regards, Evgeny
[Prev in Thread] | Current Thread | [Next in Thread] |