Among interesting topics in HAKMEM, there is an item about a fixed point: Item 174 (Bill Gosper and Stuart Nelson)
So what's a modern variation of it? A simple minded one is on conversion between int32_t
(or int64_t
) and float
(or double
, respectively). Although IEEE 754 does not specify the memory layout of its floating-point representation, we can do an experiment for the problem in a particular processor anyway.
Unfortunately the IA-32 case has no non-trivial solution except obvious 0
. See a discussion about the same problem.