[sdiy] uC languages

[Magnus Danielson]

From: patchell <patchell at silcom.com>
Subject: Re: [sdiy] uC languages
Date: Sun, 30 Dec 2001 08:09:07 -0800

> 
> 
> John L Marshall wrote:
> 
> > Wasn't the PDP-8 the ultimate RISC? Eight instructions? Only one branching
> > instruction.
> > In the early days, Bad Bill of Redmond, Washington wrote all his
> > applications on a PDP-10 (2060).
> >
> 
>     Generally, in order to be considered "RISC",  each instruction needs to
> execute in one machine cycle.  Although, even for a RISC, there are a few that
> won't, but most do.

Um, well... that is the ideal. The main point about that definition is
to rule out micro-coded instructions. What realy motivates the RISC
architecture is the possibility to have less complex and more homogene
instructions so that leaner hardware for a simple implementation can
be achieved. Also, this helps when wanting to do more complex things
like pipelining and register-renaming. It is not uncommon to see that
things like multiplication and division take more time than addition,
even if the multiplication and division is done in dedicated hardware.

Just unified instruction length is one such goal. Many "RISC" CPUs
have usually two instruction lenght. The idea in the "MIPS" school was
that the compiler should take some of the responsibility. For
instance, the carry generation has to be done with extra instructions,
and the idea was that this is done only when needed.

It's sad to see that such a broken architecture as the IA-64 is going
to replace good architectures like the Alpha. Sigh....

Cheers,
Magnus