powerpc floating point

Hi all,

I am working on an embedded powerpc system and found that floating point
computation doesn't work. I enabled kernel math emulator when building
kernel image. I also tried to compile glibc with either --with-fp or
--without-fp, and compile applications with either -msoft-float or
without it. But none of combination (--w/o-fp vs. -msoft-float) works,
the floating computation result is just not right. Does anybody know the
what else I may do, either in kernel, toolchains or application?

Thanks.

Regards,

Jingyu Liu