Floating-point questions

Hi all,

I just want to make sure that I understand correctly the behavior of floating-points on systems without FPU. There is 2 way for doing floating-point operations:

The first is to let the CPU generate exceptions on a FP operation call , which are handl e by the kernel that do the job in software, using integers only.

The second is to build your app using the -msoft-float compilation flag, making GCC to convert itself the FP operations to integer operations, using its own libraries.

Also, I've test the both methods with simple float operations, and I figured suprisingly that the first method is a bit faster (I though the overhead would have been larger than the second one).

Do I miss something? Is there anything I can do to optimize significantly float operations?

Thanks a lot,
Karl