Re: differences between implementations of IEEE-754 basic operators

[Nick Maclaren <nmm1@xxxxxxxxxxxxx>]

David Monniaux <David.Monniaux@xxxxxxx> wrote:
> Maybe I should have explained why I was asking those questions.

It helps to focus the debate, to be sure!

> Since we're looking at assembly-level code we do not have to worry about 
> funny business in compilers or libraries. (In the past I had warned them 
> about possible issues between C source and what actually happens on the 
> FPU.)        

Actually, the bigger problem is finding out what, if anything, the C
standard specifies what the source does!  But, since this is assembler,
that is still irrelevant.

> * Do not look at FPU flags except for vanilla comparisons.

I don't understand!  It may not matter.

You have left out at least one important constraint:

    They do not set either underflow or inexact trapping modes, and
    do not look at those flags at all.

Given all that, my understanding is that most architectures will give
bitwise exact results.  The IA64 may be an exception, as will most
supercomputers.

Also note that many embedded chips (past? present?) set hard underflow
by default or support only that.  Obviously, there is a simple binary
choice between competently implemented hard and soft underflows; the
vendors who implement hybrids should be rejected as not knowing what
they are on about.


Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email:  nmm1@xxxxxxxxx
Tel.:  +44 1223 334761    Fax:  +44 1223 334679