gcc/gcc config.gcc config/sparc/linux.h config ...
Toon Moene
toon@moene.indiv.nluug.nl
Thu Apr 4 12:25:00 GMT 2002
Brad Lucier wrote:
> Ah, I think I see now. This sets the rounding mode for the whole program,
> not just those parts that are compiled with -ffast-math, and this can
> screw up other parts of the program that were compiled assuming, e.g.,
> strict (or stricter ;-) IEEE compliance. Not good.
Another view on this issue (mine :-) is that *if a library needs a
certain fpu mode to have its computations succeed* it should save the
mode(s) in effect, set those needed and restore the modes in effect
before returning from the library routine.
Yes, this would be expensive - so is IEEE-conformance in general.
--
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)
More information about the Gcc-patches
mailing list