benchmarking Poly/ML & floating point

[Matthew Fluet]

> I don't think I actually need inline assembler for Real_equal, I just need a
> procedure call boundary.  Does the following seem OK?
> 
> Int Real_equal(Double x1, Double x2) {
> 	return x1 == x2;
> }
> 
> It used to be
> 
> #define Real_equal(x1,x2) ((x1) == (x2))

Yes, I think that's true.  But I'd still like to use the inline assembly
for the native backend, assuming I can determine the right tests for all
the operators.