On 2005-05-29 01:33:43 -0600, Roger Sayle wrote: > I apologise for coming into this argument late. I'll admit that I > haven't even caught up on the entire thread, but an interesting > relevant article that may or may not have already been mentioned is: > > http://web.archive.org/web/20040409144725/http://www.naturalbridge.com/floatingpoint/intelfp.html
I mentioned it here: Date: Fri, 27 May 2005 14:42:32 +0200 From: Vincent Lefevre <[EMAIL PROTECTED]> To: gcc@gcc.gnu.org Subject: Re: GCC and Floating-Point (A proposal) Message-ID: <[EMAIL PROTECTED]> > Admittedly on many IA-32 systems there's little difference between > using FSIN vs calling the OS's libm's sin function, as glibc and > microsoft's runtimes (for example) themselves use the x87 intrinsics. > GCC, however, is not to know this and assumes that the user might > provide a high-precision library, such as Lefevre's perfect O.5ulp > implementation. [It's nice to see him join this argument! :)] Well, I'm just one of the authors of MPFR. Concerning the runtime libraries for math functions in IEEE double precision, that partly provide correct rounding, I know: * IBM's MathLib, on which the glibc is based (for Athlon 64, Opteron, PowerPC, Alpha and PA-RISC). Does rounding-to-nearest only. URL: ftp://www-126.ibm.com/pub/mathlib/ * Arenaire's Crlibm. URL: https://lipforge.ens-lyon.fr/projects/crlibm/ * Sun's libmcr. URL: http://www.sun.com/download/products.xml?id=41797765 * MPFR does correct rounding in multiple precision, but a wrapper could be written for the double precision (and possibly other precisions for the*f and *l variants). Of course, this would be quite slow as MPFR wasn't written for such kind of things, but some users may still be interested. -- Vincent Lefèvre <[EMAIL PROTECTED]> - Web: <http://www.vinc17.org/> 100% accessible validated (X)HTML - Blog: <http://www.vinc17.org/blog/> Work: CR INRIA - computer arithmetic / SPACES project at LORIA