I think Dhrystone is stressing the ALU (integer calculations)
Whetstone is the FPU (math co-processor, since the pentium they've been on the same chip though)
Floating point arithmatic is when you need to specify a number out of the normal range of the CPU. Sort of like Standard form in maths. (i.e. 6.02*10^23) It is stored in two parts. The mantissa, which stores the actual number, and the exponent that stores the position of the binary point.
For some more detailed information, I suggest you examine IEEE754:
http://grouper.ieee.org/groups/754/
That link will provide you with the basic resources.