What requires an FPU?

[SpeedGeek]

FPU version is more accurate but slower than integer version!
See the readme for mpega.library here:
http://fi.aminet.net/util/libs/mpega_library.readme
 
However, for apps which actually require Floating Point support hardware FPU is certainly faster than Floating Point emulation.

[SpeedGeek]

You realize this is a different code base than the libmad MPEGA library linked before? Just because the original integer MPEGA.library accuracy is lower doesn't mean that this would be some kind of generic rule. The original interger MPEGA.library just sucks. From the libmad MPEGA.library readme:

Code: [Select]

- Very high quality decoding (better than the FPU version of the original).

Also from the libmad MPEGA.library:
- Very low CPU usage (integer original is still faster on 68k though).
 
The point I was trying to make is that integer math is almost always faster than floating point math on 68K. The execution speed of integer or floating point instructions is completely irrelevant to accuracy.