http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50032
--- Comment #4 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2011-08-12 18:28:56 UTC --- > Some bug reports for this performance issue date from 2008, so it seems we > won't see that fixed any time soon. IIRC the excuses were: (1) nobody use float (fortran does not exist indeed!-), (2) the results are "correctly rounded", (3) why don't you write the library yourself? If the linux users don't put some pressure on their favorite "vendor", the problem will never be fixed. I cannot say I really care since darwin provide a sensible library for float intrinsics (it seems that AMD is also providing one) also I have been bitten by the problem while helping a colleague to debug his code on our linux boxes.
