Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 (Tek) 9/28/84 based on 9/17/84; site tekcrl.UUCP
Path: utzoo!linus!decvax!decwrl!pyramid!hplabs!tektronix!tekcrl!patc
From: patc@tekcrl.UUCP (Pat Caudill)
Newsgroups: net.arch
Subject: Addressing modes (conditional branches)
Message-ID: <546@tekcrl.UUCP>
Date: Sun, 23-Feb-86 11:53:38 EST
Article-I.D.: tekcrl.546
Posted: Sun Feb 23 11:53:38 1986
Date-Received: Wed, 26-Feb-86 07:07:11 EST
Distribution: net
Organization: Tektronix, Beaverton OR
Lines: 45

. .. . ... . (line eater kibbles)

	The use of a stored branch as the way to effect a
conditional jump is not new. In "First Draft of a Report on the
EDVAC", John von Neumann, Sec 11.3 discribes the conditional
'branching' facilities of the EDVAC.
	Some background, the EDVAC's ALU (arithmetic organ) had
three registers. I and J were inputs and O was output. Numbers were
loaded to I with the previous value in I shifting to J. When an
operation was performed the result appeared in O. There were operations
which allowed I or J to directly go to O, and O to I. The conditional
operation transfered the contents of I or J unchanged to O depending
on the previous value in O. To conditionally branch to one of two
locations the inputs would be loaded with two jump instructions
the correct one taken as the result and stored in memory then
jumped to. (often the clever programmer would arrange the store
address to be the next location.
	For example:

   if ( a < b ) then goto L1 else goto L2 ;

 would code up a:
    (my own notation, von Neumann's is very cryptic, and I dont have the
	Greek alphabet on this terminal)

		LD	A
		LD	B
		SUB				; -
		LD	#JMP+L1
		LD	#JMP+L2
		COND				; s
		STO	L3
	L3:	JMP	L3

	Turing for the Pilot-ACE ( I think, look in Early British
Computers from Digital Press for more details)just had an instruction
which set a negative accumulator to zero or 1 depending on its sign. He
suggested for a conditional branch:

	lbl1 * condition + lbl2 * (1-condition) + jmp -> nextInstr

	We now look down on self modifying code, but at that time
it was considered a great step forward. Ahh, those were the good old days.

			Tektronix!tekcrl!patc