> As I know, the modulo (x%y) and division (x/y) instructions are fairly
slow on GBA.

Modulo and division are relatively slow on the AGB as it has no hardware
divide unit, however they are not so slow they are going to screw up
interrupt latency within your VBlank or any other interrupt.

The best thing is just to forget about the relative hit of divides and so on
unless you need to do something which uses divides intensively, such as 3D.
In which case you could consider going to a precalculated table based
reciprocal divide routine which uses a multiply and 1/divisor to increase
the speed of division at the cost of accuracy.

To be honest, I'd forget about it until your game / demo starts to slow down
and optimise then....

--James