This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: gcc handling of the sin, cos functions


In a message dated 11/19/99 4:11:08 PM Pacific Standard Time, 
khan@NanoTech.Wisc.EDU writes:

> > But then why does glibc define all those functions with inline asm?
>  > 
>  
>  Just so we can turn those off using -D__NO_MATH_INLINES ;-) I have found 
>  the inline versions to be rather lacking in stability for extreme values, 
>  at least on the versions that I have installed, for our FP-intensive 
>  modeling codes. So does GNU Octave for example. 
>  
>  Regards,
>  Mumit

That's why I came up with a sanitized version which passes the elefunt tests, 
and works both with cygwin and linux-glibc, as well as including the stack 
bug fixes, leaving out the functions which are so big that they run slower 
in-lined.  See
ftp://members.aol.com/n8tm/llibm.tgz


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]