Oleg is quite right that for small values, radix conversion by subtraction is not at all efficient.

However, the successive subtraction method when applied to large numbers does have the advantage of generating the digits in the right order i.e. most significant digit first.

Successive division by the radix is, however, one of the fastest possible methods. On each iteration, the remainder is the value of the next least significant digit - so you get the results least significant digit first.

For a 32-bit to BCD conversion, see here:

http://www.programmersheaven.com/zone5/cat.../32302.htm