Quote:
Originally Posted by carrion
@bebbo
I can live without sinuses in the FPU but more questions arise...
What I'm coding at the moment is working fine with fixed point math (shifting back and forth for fixed precisions). So, can I expect any performance increase when using floats with FPU? Does it actually makes sense for a simple muls and divs?
I guess these questions are out of this thread subject so sorry for asking here but maybe you can somehow guide me. TIA
|
Depends on what you're doing and the required precision. You should be aware that the 060 doesn't support 64-bit integer multiplies (32x32 -> 64)/divisions(64/32 -> 32) in hardware, so those operations should be avoided (or changed to use the FPU) if possible. In practice this means that 16.16 fixpoint is not that attractive on 060 unless you're only adding/subtracting.
Probably better to start a new thread if you have further questions regarding optimizations. (As always measurement is king, but WinUAE is far from perfect for 060 in this regard).