Path: utzoo!attcan!uunet!lll-winken!lll-tis!helios.ee.lbl.gov!pasteur!ucbvax!amdcad!tim
From: tim@amdcad.AMD.COM (Tim Olson)
Newsgroups: comp.arch
Subject: Re: Branch Delay Annullment
Message-ID: <22082@amdcad.AMD.COM>
Date: 16 Jun 88 03:06:35 GMT
References: <22065@amdcad.AMD.COM> <1081@nud.UUCP>
Reply-To: tim@amdcad.UUCP (Tim Olson)
Organization: Advanced Micro Devices
Lines: 23

In article <1081@nud.UUCP> tom@nud.UUCP (Tom Armistead) writes:
|     The number of loop instructions is equivalent in either the 
| annul-taken form or the always executed form (for this example
| anyway).  The only slight difference is that no "cleanup" instruction
| was required in the always executed form.  There might be some instances
| where annul-taken is better but I don't know of any specific ones. 

I agree -- delayed branches are a win for the cases in which you can
find a delay instruction which is beneficial in both cases, or is
beneficial in the most-taken direction and is "innocuous" in the other. 
The case I was discussing was when one of these could not be found.  In
that case, an "annul-untaken" jump would allow you to always fill the
branch delay slot in a loop.

This is admittedly infrequent (we didn't add annulling branches in the
29000, because we felt it wouldn't buy much, and could cause
restartability problems when a trap occured between a jump and its
annulled delay slot), I was just wondering what was the intended use of
the "annul-taken" form.  It would seem to benefit "if" statements the
most.

	-- Tim Olson
	Advanced Micro Devices
	(tim@delirun.amd.com)