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


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: libjava configure question


Tom Tromey writes:
 > >>>>> "Andrew" == Andrew Haley <aph@redhat.com> writes:
 > 
 > Andrew> One other thing: I don't think we can do this in general and meet even
 > Andrew> the more relaxed requirements of java.lang.Math, because the x86 has a
 > Andrew> restricted range of allowable arguments to its trig functions.
 > 
 > I thought the Math restrictions were designed to explicitly allow use
 > of the FPU on x86.

Yes, but not without some wrapper code.  The crucial issue here is
full range reduction, and the floating-point instructions don't do
that by themselves.

 > BTW, with 1.5 we will have more opportunities to use builtins here.
 > For floating point there is at least cbrt; for integer types there
 > is popcount and (I think) a few more bit-twiddling operations.
 > 
 > Andrew> Try this with -ffast-math:
 > 
 > I thought -ffast-math turned on some unsafe optimizations.  The
 > manual is really unclear about what it does though.  Are our FP
 > builtins conditional on -ffast-math?

In this particular case, the use of floating-point trig instructions
is enabled by -ffast-math.

Andrew.


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