Does anybody know what the performance hit is for the Keil ARM floating point routines? Or where I might find it? Looking for cycles for a floating point add and multiply. I suspect because the ARM is 32 bits its not as bad as the 8 bitters implementation of floating point.