Newsgroups: comp.arch
Path: utzoo!henry
From: henry@utzoo.uucp (Henry Spencer)
Subject: Re: A simple question on RISC
Message-ID: <1988Nov15.235728.26718@utzoo.uucp>
Organization: U of Toronto Zoology
References: <6544@xanth.cs.odu.edu> <75577@sun.uucp> <1618@imagine.PAWL.RPI.EDU> <419@augean.OZ> <23541@amdcad.AMD.COM>
Date: Tue, 15 Nov 88 23:57:28 GMT

In article <23541@amdcad.AMD.COM> tim@crackle.amd.com (Tim Olson) writes:
>There are a number of reasons, most of them due to the main problem of
>limited microcode space:
>
>	3) Microcode is not as easily optimized...

Also, optimization of microcode can require more microcode space.  One
reason why the simple instructions tend to be fast is that a lot of work
(microcode space, decode hardware, etc.) is put into optimizing them.
For example, the hardware may break out the major variants of the MOVE
instruction, rather than having the microcode do test-branch-test-branch
to finish the decoding.  This sort of thing can't be done for the whole 
instruction set without exceeding the available resources (space,
development time, staff), and the difference can be major enough to make
the fancy stuff slower than "doing it by hand".
-- 
Sendmail is a bug,             |     Henry Spencer at U of Toronto Zoology
not a feature.                 | uunet!attcan!utzoo!henry henry@zoo.toronto.edu