[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: FPGApple + ARM?
sicklittlemonkey wrote:
I've heard the comparison before, probably because of the simplicity of
both compared to x86 lineage processors. I mean this in terms of the
orthogonality of the instruction set (somewhat, in the 6502, and
beautifully so in the ARM) but more noticeably in the goal of mimizing
clock cycles per instruction.
It probably doesn't bear closer scrutiny though. ;-)
I have on occasion noted the similarities between the 6502 architecture
and a RISC architecture, if you think of page 0 as registers. But the
comparison is pretty loose.
Generally, when one says that "processor B is a lot like processor A"
they mean that someone who has already mastered processor A will be
quickly up to speed on processor B. In this case, familiarity with
the 6502 does not provide much guidance in the use of the ARM.
The ARM instruction set is a work of art. Every instruction is
conditional. I can recall 'borrowing' a small crucial routine written
by Acorn and squeezing one more conditional cycle out of it. It felt
great to beat the makers. ;-)
The predicate bit is a nice approach to conditionals without
branching.
Orthogonality in instruction set design is nice for programmers, but
may be a problem in implementation. For a fast, clean implementation
you want an architecture which permits an "orthogonal" implementation.
To the extent that it occurs, instruction set orthogonality is a
by-product, not a design goal.
The central insights of RISC architecture are: for the best
cost-performance, the implementation must be simple and fast,
the common instructions are the important ones for performance,
and doing something once at compile time is better than doing
it repeatedly at execution time.
-michael
Music synthesis for 8-bit Apple II's!
Home page: http://members.aol.com/MJMahon/
"The wastebasket is our most important design
tool--and it is seriously underused."