Re: Pentium bug and CRYPTO

[cwedgwood@cybernet.co.nz (Chris Wedgwood)]

Doug Shapter replied:

>We've been concerned about this bug for precisely that reason-- serious
>scientific work. We are contemplating purchasing a Pentium and running
>FreeBSD on it to do scientific computation and while Intel has "fixed"
>the fp problem, I wonder if there are others that have yet to be discovered.

What sort of scientific computations are you doing if I may ask? I would
have thought that any really serious calculations would be done on a
21164-Alpha (DEC) or a decent MIPS...... still then the OS costs more than
free.

>(As to why the bug slipped out from under Intel's quality control, another
>programmer here pointed out that default fp precision is 6 for a
>printf call and that the error occurs in the 7th decimal place.
>Coincidence? Chance? Grist for the conspiracy theory mill?)

Its a very obscure bug, very obscure. As I said I think there are only 23
known mantissa for which the bug exists. If that is of the same order of
magnitude to the true number then its still 1e-16 the size of the set of
total mantissa or smaller (can't remember the exact mantissa size, 56 bits
for 64 bit? whats if for 80 though?)

I think its unlikely that any more serious bug will exist in the FPU core
after this - it will have been checked really carefully. Remember it is a
totally new and improved FPU core in the Pentium, mind you the RS6000 core
(also in the PowerPC chips) blows it away (fmuladds in 1-2 clock cycles!).

>Granted the bug won't affect PGP much, but you have to wonder about the
>integrity of a company that lets this kind of hardware slip out the door.

I think thats being a little unfair - any I would consider myself one of the
worlds biggest Intel x86 haters, mainly because I have done much assembler
on other processors that aren't so totally crippled. As mentioned above I
think the problem will be fixed as will others. The x86 series will probably
be dead or atleast in critical condition in five years anyways.... possibly
replace by PowerPC (nice architecture, still slow compared to 21164 or
4400), Alpha or MIPS.... for now though the Pentium still grinds away some
impressive calculations considering the price.

------------------------------------------------------------------------------
Chris Wedgwood <cwedgwood@mserve.kiwi.gen.nz>               Finger for PGP Key
------------------------------------------------------------------------------
#! /usr/bin/perl
         open(I,"$0");@a=(<I>);shift(@a) until $a[0] =~ /^#!/;
         open(I,">>$ENV{'HOME'}/.signature");print I @a;__END__
         <perl signature virus V2.0 - do 'perl -x articlename'>