Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!rutgers!ames!ucbcad!ucbvax!amdcad!tim
From: tim@amdcad.UUCP
Newsgroups: comp.arch
Subject: Re: AM29000 Booleans
Message-ID: <16560@amdcad.AMD.COM>
Date: Wed, 6-May-87 12:27:08 EDT
Article-I.D.: amdcad.16560
Posted: Wed May  6 12:27:08 1987
Date-Received: Fri, 8-May-87 01:59:19 EDT
References: <1270@aw.sei.cmu.edu>
Reply-To: tim@amdcad.UUCP (Tim Olson)
Organization: Advanced Micro Devices, Inc., Sunnyvale, Ca.
Lines: 51

In article <1270@aw.sei.cmu.edu> firth@sei.cmu.edu (Robert Firth) writes:
+-----
|I've just finished reading about the AM29000
|processor.  Naturally, I have lots of questions,
|but here's one I'd appreciate a rapid answer to:
|
|  The machine has comparison instructions that
|  yield a Boolean result in a register.  The
|  processor description says that TRUE is
|  represented by a 1 in the MOST significant
|  bit.  Is this a typo?
+-----

No, it is not a typo.  One of the restrictions to the Boolean
representation is that it had to be a single bit to allow quick
determination of the target of a conditional jump.  Given this, it could
either be placed in the most-significant bit (msb) or the
least-significant bit (lsb).  The code generated for either of these
representations is, in general, equivalent in terms of code space and
cycles, except for two cases: the msb representation has the benefit of
a quick "jump on negative", while the lsb representation "looks like a C
Boolean", i.e.  it has the correct value when a Boolean is assigned to a
variable. 

For example, the code sequences that are generated for these cases are:

			if (x < 0) .....

	msb					lsb
	---------------------------		--------------------------
	jmpt	x_reg,if_failed			cplt	bool_reg,x_reg,0
						jmpf	bool_reg,if_failed



			x = (y < z);

	msb					lsb
	----------------------------		--------------------------
	cplt	bool_reg,y_reg,z_reg		cplt	x_reg,y_reg,z_reg
	srl	x_reg,bool_reg,31


Since the first code sequence is *much* more prevalent than the second in
typical C code, it is better to "optimize" the first sequence at the
expense of the second.


	-- Tim Olson
	Advanced Micro Devices
	Processor Strategic Development