From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joern Rennecke To: pderbysh@usa.net (Paul Derbyshire) Cc: egcs@egcs.cygnus.com Subject: Re: Bug in libm or libstdc++. Date: Mon, 01 Mar 1999 08:30:00 -0000 Message-id: <199903011630.QAA00110@phal.cygnus.co.uk> In-reply-to: <3.0.6.32.19990227220629.008862a0@pop.globalserve.net> from Paul Derbyshire at "Feb 27, 99 10:06:29 pm" References: <3.0.6.32.19990227220629.008862a0@pop.globalserve.net> X-SW-Source: 1999-03/msg00013.html > I may be mathematically inclined but I have none of the > numerical-computation background necessary to know how to implement these > things in an optimized way. Best I could probably do is a slowly converging > Taylor series for these things. You ened a volunteer with a bit more > knowledge in techniques of rapid numerical computation of trig functions. When you want to use a Polynom approximation, you should rather use a Tschebyscheff polynom. Look it up in a mathematical enceclopedia. If you just want to get something working, you can skip the proofs and go straight for the recipes ;-) I happen to have written sqrt functions when I worked on an XFmode / TFmode fp-bit.c . ( The project got stuck for lack of priority. ) I'll send it to you in private email. From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joern Rennecke To: pderbysh@usa.net (Paul Derbyshire) Cc: egcs@egcs.cygnus.com Subject: Re: Bug in libm or libstdc++. Date: Wed, 31 Mar 1999 23:46:00 -0000 Message-ID: <199903011630.QAA00110@phal.cygnus.co.uk> References: <3.0.6.32.19990227220629.008862a0@pop.globalserve.net> X-SW-Source: 1999-03n/msg00013.html Message-ID: <19990331234600.iJC22fOJyBoaagxbI3b58lTvuwnhApgBrqhFf2Pqllc@z> > I may be mathematically inclined but I have none of the > numerical-computation background necessary to know how to implement these > things in an optimized way. Best I could probably do is a slowly converging > Taylor series for these things. You ened a volunteer with a bit more > knowledge in techniques of rapid numerical computation of trig functions. When you want to use a Polynom approximation, you should rather use a Tschebyscheff polynom. Look it up in a mathematical enceclopedia. If you just want to get something working, you can skip the proofs and go straight for the recipes ;-) I happen to have written sqrt functions when I worked on an XFmode / TFmode fp-bit.c . ( The project got stuck for lack of priority. ) I'll send it to you in private email.