u need not floating point ,
u can use 32 bit math routines
to do pseudo floating point
for example for multiplying a variable with 1,25 u just multiply it by 125 and then divide by 100
that way u can achieve the same precision
u can find 32 bit math routines in this website
just follow the links