> http://napier.ncl.ac.uk/HSLA/LNSIntro/

I forgot I'd written that, sorry for not mentioning it in original
posting. :)

The LNS has much research interest, later this year the world's first full
32b microprocessor using it will be manufactured. It is indeed true that
for long words, a large amount of memory would be required to perform
add/sub, which is why nobody does it like that. In practice, you can
approximate the functions fa() and fs() mentioned on the above page.
There's a link to more references there.

Remember that this is now 6 years old, there have been several LNS
advances since then.

Finally, i repeat, for 8b words, tables are easy to deal with. 12b is
relatively easy. 16b is doable in a micro, but I used an approximation.
Also - if anybody does work with these, and _especially_ if they make any
kind of product with the LNS, I would like to hear from you. There's only
a few LNS researchers around the world, and it is our goal to expose how
ineffective floating-point is :) News of any applications is appreciated!

Part of the problem has been that in order to challenge floating-point on
its own playing field, we have techniques for 32b, and have neglected the
probably more useful area of uCs and short-word applications where some
real math could be a big bonus.

have fun, i might make a list about this stuff one day...

ed c

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads