Robin Gravel's Blog

Yeah, that's a known issue with floating point numbers. It happens on all computers, I first experienced it on a Commodore PET in 1980.

 

What Every Computer Scientist Should Know About Floating-Point Arithmetic

I never knew about this issue on computers. Thanks for sharing this link with us.

In most cases the BASIC interpreter of that generation didn't have floating point hardware and was written to be memory efficient. So BASIC may not have actually implemented true floating point, and certainly not to the IEEE specifications.

 

But let's look at the question in terms of what a typical floating point (using 32 bit mantissas) would do:

 

682.80 - 587.34

682.80 = $1.55666666 * 2^9

587.34 = $1.25AB851F * 2^9

difference = $0.2FBAE147 * 2^9 = 95.45999991893768310546875