Date: Sat, 24 Aug 2013 19:51:12 -0700 From: Steve Kargl <sgk@troutmask.apl.washington.edu> To: Bruce Evans <brde@optusnet.com.au> Cc: freebsd-numerics@FreeBSD.org Subject: Re: (2nd time) tweaks to erff() threshold values Message-ID: <20130825025112.GA56868@troutmask.apl.washington.edu> In-Reply-To: <20130825023029.GA56772@troutmask.apl.washington.edu> References: <20130822213315.GA6708@troutmask.apl.washington.edu> <20130823202257.Q1593@besplex.bde.org> <20130823165744.GA47369@troutmask.apl.washington.edu> <20130824202102.GA54981@troutmask.apl.washington.edu> <20130825023029.GA56772@troutmask.apl.washington.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Aug 24, 2013 at 07:30:29PM -0700, Steve Kargl wrote: > On Sat, Aug 24, 2013 at 01:21:02PM -0700, Steve Kargl wrote: > > On Fri, Aug 23, 2013 at 09:57:44AM -0700, Steve Kargl wrote: > > > On Fri, Aug 23, 2013 at 09:12:33PM +1000, Bruce Evans wrote: > > > > > >> The whole erf implementation is probably not the best way for > > >> float precision, but is good for understanding higher precisions. > > > > > > I'm fairly certain that the implementation of erff() is not very > > > efficient. The polynomials, used in the rational approximations, > > > are the same order as those used in the double precision approximation. > > > I'll check the polys when update my P(x)/Q(x) remes algorithm for > > > erfl and erfcl. > > > > I seem to be right (although I haven't iterated on the P's and Q's). > > > > Now, with pretty testing and (perhaps) better coefficients: > LMAO. I have not idea where "pretty" came from. s/pretty/better -- Steve
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130825025112.GA56868>