Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!ames!ucbcad!ucbvax!decvax!decwrl!nsc!roger
From: roger@nsc.nsc.com (Roger Thompson)
Newsgroups: comp.arch,comp.sys.nsc.32k
Subject: Re: Branch prediction in the 532
Message-ID: <4233@nsc.nsc.com>
Date: Thu, 23-Apr-87 01:24:19 EST
Article-I.D.: nsc.4233
Posted: Thu Apr 23 01:24:19 1987
Date-Received: Sat, 25-Apr-87 01:07:16 EST
References: <324@dumbo.UUCP>
Organization: National Semiconductor, Sunnyvale
Lines: 27
Xref: mnetor comp.arch:1065 comp.sys.nsc.32k:110

In article <324@dumbo.UUCP>, hansen@mips.UUCP (Craig Hansen) writes:
  
> 
> 	The selection is based on the branch condition and
> 	direction; forward branches are predicted as taken,
> 	backward branches as non-taken.
> 
YES the brochure has misplaced a "non" in the above paragraph.
Reverse the usage.

> Well, I don't see how the branch prediction can depend on the branch
> condition, since waiting for the condition corresponds to no branch
> prediction at all.  (Certainly novel and non-obvious...)

The condition factor that is used is that which is specified by the
opcode, not the actual result from the operation being tested.  The
concept is that the predictor has an algorithm that has multiple
inputs.  Under certain conditions, the predictor determines one answer or
the other.  The prediction should not be confussed with the result.
Our prediction algorithm is accurate to 70 - 80%.

> 
> But when I was a young sprout, they always told me that backward branches
> often represented loops and were most frequently taken. Forward branches
> often represent conditional expressions, and are more ambivalent. 
> 
Agreed.