In reply to Somu:
This is a general problem with representing decimal numbers in a fractional binary format. You may want to scale your number by 1000 as you increment it, and then scale it back where you reference it.
See What Every Computer Scientist Should Know About Floating-Point Arithmetic