ieee expression control

[Stu Anderson 865-3595 (206)]

Is uniform expression evaluation in IEEE worthwhile?  Yes!  For years, I have 
found precise control of Intel floating-point chips to be invaluable for the
development/testing/auditing of elementary/special function software. 

Note that Borland's Turbo C/C++ (for PCs) has a function for setting the X87 
control word.  This function is non-portable and X87 specific.  However, it is 
better than using assembly language for control.  It would be nice to have 
portable usage control for common features (precision, round mode, exception
masking).

I also agree completely on exception processing: IEEE default or termination 
(which could be abort or Ada/C++ style exception handling).

By the way, has anyone defined "IEEE quad" precisely?  I could probably infer 
the definition if I knew the exponent bias.

With regard to other issues, there are plenty if you admit the existence of
"sort-of-IEEE" architectures.  Therefore, I assume you are restricting this 
discussion to implementations of the complete IEEE 754 standard.

Stu Anderson
--

--------------------------+---------------------------------
Stuart L. Anderson        |  slaaespresso.rt.cs.boeing.com
Boeing Computer Services  |  Who speaks for Boeing?  Not me!