> I think it is possible to have a standard AND have good performance.
> The key is to adopt a standard that is based upon fixing a data format
> DESCRIPTION rather than fixing the data format itself. ...
Well, usually. It can also cause trouble by making it much harder to
issue reasonable diagnostics.
> ... Look at Lawrence Livermore Lab's PDB (Portable Data
> Base) product by Stewart Brown to see how it handles numeric format
> conversion. It uses a parameter-value language sufficient to
> characterize all integer and floating point formats that have been in
> use by past and present CPUs and which are ever likely to exist in
> future CPUs.
I think that you may be overstating the case a little! Can you provide
me with a reference to save me scouring the net, and I will see what I
can think of that it won't handle :-)
More seriously, this is probably feasible for all normal numbers in their
canonical form, some numbers in alternative forms, and a few selected
special values (infinities, NaNs etc.) But it is STILL necessary to
define a subset of possible 'values' that the model will support, and to
say that conversion of anything else is implementation dependent or even
undefined.
Nick Maclaren,
University of Cambridge Computer Laboratory,
New Museums Site, Pembroke Street, Cambridge CB2 3QG, England.
Email: nmm1@cam.ac.uk
Tel.: +44 1223 334761 Fax: +44 1223 334679