22 Jun
2004
22 Jun
'04
10:45 a.m.
> > How much on other kinds of processors?
(RISC)
> That should be covered in another class: computer architecture.
I'm not able to completely separate them.
I can't do a good job of teaching assembly language without
dealing with architecture, and
I can't do a good job of teaching architecture without
some dealing with assembly language.
Therefore, the Assembly Language class includes some
review of computer architecture, and some lightweight
coverage of processor architecture.
For example,
The easiest way that I know of to explain RISC is after
they've managed to create a few simple programs, to then
ask some variants of:
would you be willing to give up the DIV instruction,
and replace it with SUB in a loop, which we will inaccurately
guess would take twice as long, if doing so permitted a 300%
improvement in the speed of MOV?
[class says YES]
What if the improvement were 30%?
[mixed
responses]
Let's look at how often each
instruction gets used, and multiply that out to see when it would
pay for itself. Hmmm. Whose program shall we get those statistics
from?
What if we really gutted a lot of less commonly used instructions
that could be replaced by combinations of other instructions,
but used that to have a core of instructions that are really fast?
OK, now let's look at some real world cases."
That is the minimum coverage for RISC - just explaining what it is
and why. It could easily be expanded rather significantly.
--
Grumpy Ol' Fred cisin(a)xenosoft.com