Arithmetic problems
Fri Dec 31 20:54:00 GMT 1999

In a message dated 12/30/1999 2:26:52 AM Pacific Standard Time, writes:

> an odd and inconsistent feature of gcc
>  on Linux (Intel)
>An option could be provided, allowing the user
    to force gcc to let intermediate results be rounded to
    double precision.

glibc has been providing function calls for this purpose, for the i686 
processors.  While the method changed from glibc-1 to glibc-2, I prefer this 
to the schemes of commercial compilers, where the treatment changes from one 
release to another, with no documentation or user option.


More information about the Gcc-bugs mailing list