floating point multiply == very very slow ?

[jaguard]

[Kzinti]

[hm]

[refractor]

[Dan East]

[hm]

[jaguard]

I was affraid of that .

Ok, I have to deal with FP math. So, did anyone actually checked how exactly slow floating operations are?

1. What about floating add and sub? How slow they are?

2. What is better - translate for example 4 floating variables to fixed, multiply them altogether, then translate them back or simply multiply them using floating multiplication?

[xPeterx]

Translation from floating point to fixed point usually involves a floating point
mul so you're better off using only fixed point whenever lots of data needs to be processed.

Converting back and forth between floating point and fixed point only makes
sense when you're doing it less frequently.

But there's no reason to be afraid of fixed point.
Just know your bits and what kind of precision you need. That's all

[jaguard]

[xPeterx]

[refractor]

[xPeterx]

[Kzinti]

[abiemann]

[HTK]