Embedded Multitasking

20 Chapter 2

it only requires an 8 × 8 multiply and 16 × 16 multiply in its library.

However, this does creates two inefficiencies; one, it wastes two data

memory locations holding values that will always be zero and, two,

it wastes execution cycles on 8 additional bits of multiply which will

always result in a zero.

The more efficient solution is to create a custom 8 × 16 multiply

routine. This saves the 2 data bytes and eliminates the wasted execution

time spent multiplying the always-zero MSB of the 8-bit value. Also,

because the routine can be optimized now to use an 8-bit multiplicand,

the routine will actually use less program memory as ...