Theos-Talk Archives (October 1997 Message tl00315)

Bart Lidofsky wrote:
> 
> Vincent Beall wrote:
> > What a bunch of nonsense! Bart the chip didn't add correctly. All
> > operations of the CPU can be reduced to addition, complimentation, and
> > shifting. Addition is the only purely mathematical thing that it does.
> > There is no excuse for making such a mistake. As it turned out the chips
> > were replaced; there is justice after all!
> 
>         Can you name the instruction that was bugged, and use it in a short
> assembler program? From your statement, I am not even sure you know what
> an instruction IS. The error took place in the least significant digits
> (you know what that term means?) in a 128 bit operation. All the rest of
> the instructions were fine.
> 
>         Bart Lidofsky

I'm sorry I snapped at you Bart, and I can't name the instruction, but I
have studied assembler for 360 series computers years back. Now you I
suppose would know that errors in the LSBs can easily become disasterous
when the computation is in an iterative form. And, such forms are quite
common in the world of number crunching. Maybe it was limited to one ADD
instruction, but that doesn't make for a decent computer. You know what
they say at Silicon Graphics, Intel outside. 

Vincent
-- 

vincent@dmv.com

http://home.dmv.com/~vincent/