Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site ucbcad.UUCP
Path: utzoo!linus!decvax!tektronix!ucbcad!ucbesvax.turner
From: ucbesvax.turner@ucbcad.UUCP
Newsgroups: net.arch
Subject: Re: Barrel Shifters - (nf)
Message-ID: <239@ucbcad.UUCP>
Date: Sun, 10-Jul-83 22:17:43 EDT
Article-I.D.: ucbcad.239
Posted: Sun Jul 10 22:17:43 1983
Date-Received: Tue, 12-Jul-83 05:48:03 EDT
Sender: notes@ucbcad.UUCP
Organization: UC Berkeley, CAD Group
Lines: 22

#R:turtleva:-20900:ucbesvax:12800004:000:808
ucbesvax!turner    Jul  9 16:56:00 1983

	Indeed, I was quite indignant when I discovered that 68000
shift/rotate timing was (sigh) a distinctly linear function of the
number of shifts.

	I suppose barrel shifters do rather horrify the micron-hackers,
since it's just a big array of simple switches.  How can anything that
clean-looking be useful?  (Oh know, I don't really believe that they
think things like that.  How do you say in netspeak?  ":-)")

	But yeah, 32 bits is a lotta bits.  32-bit machines almost *need*
to be virtual, I think, since there's so many high order bits sitting
around in memory being 0 or 1 most of their lives.  Better bit-field
support in high-level languages and the use of folding algorithms in
loop optimization would both probably help a lot.

	From the Mental Blocks of Michael Turner,
		ucbvax!ucbesvax.turner