Mail Archives: djgpp/1998/07/04/23:20:09
Arthur wrote:
>
> >> You say "only." I know of no other processor other than the x86 (and
> >> compatibles) that have such an illogical and unfriendly instruction set.
> >
> >That's because you don't know much about RISC processors.
>
> That's a bit harsh. I do actually know a bit about RISC processors, and of
> the processors I know the x86 is the hardest to program.
I suppose it depends on what you do. Writing assembler math routines
gets rather unpleasant on machines with only shifts and adds.
> The 8086 was not a RISC processor, and yet (some) 8086 programs will run on
> my P200MMX. To be able to do this there must be some backwards compatability
> between processors. The x86 family was not originally intended to be a RISC
> based chip. The Pentium (and others) have therefore a mish-mash of new
> instructions and old instructions with the original structure of the 8086,
> so how can it be truly a RISC? It seems to me (in comparison with something
> like the Archemedes, which _is_ a RISC computer) that the x86 is a sort of
> hybrid chip - sort of quasi-RISC. In this case what is the advantage of the
> chip being a RISC?
I believe that recent chips have a RISC-ish core, with the most
frequently used instructions (mov, jmp, etc) being done in hardware,
while the stranger ones (daa, bsf, etc) are done in microcode. There
aren't really that many new instructions since the 8086, and those that
are new are decidedly more CISC-ish. (Can you say MMX? ;-)
When the instructions are simple and few, more silicon can be used for
important things like registers, and it's easier to execute in parallel.
This is now totally off-topic. :)
--
Nate Eldredge
nate AT cartsys DOT com
- Raw text -