[prev in list] [next in list] [prev in thread] [next in thread] 

List:       classiccmp
Subject:    RE: Variable length (in bits) instructions (was Re: ENIAC ....)
From:       "Chuck Guzis" <cclist () sydex ! com>
Date:       2011-01-31 23:29:52
Message-ID: 4D46D570.28555.1196710 () cclist ! sydex ! com
[Download RAW message or body]

On 31 Jan 2011 at 14:30, Rich Alderson wrote:

> From: Chuck Guzis
> Sent: Monday, January 31, 2011 2:13 PM
> 
> > When it comes to variable-length instructions, I'm surprised that no
> > one mentioned the IBM 1401, particularly in its ability to "chain"
> > instructions.
> 
> The instructions on the 1401 are 1 character long, that is, 6 bits.
> 
> Or have I misunderstood the discussion so far.  I thought we were
> talking about different operations being effected by different lengths
> of bits as the op codes.[1]

Are we talking about instructions themselves, or their operands?  In 
either case, the 1401 qualifies--it's a variable word-length machine 
and instruction length is dependent on the presence or absence of 
operands.  

So, there is only one opcode for the "ADD" instruction, but I can 
write:

	A 123 456

Causing the (variable-length) number whose low-order digit is 
addressed by 123 to be added to the (variable-length) number whose 
low-order digit is addressed by 456.  I can also write:

	A 123

Same opcode--but this indicates that the number addressed by 123 is 
added to itself.  Or I can write:

	A

Which adds two numbers preceding (in memory) the pair last addressed 
by an A or other instruction.

Similarly, the "unconditional branch" and the "conditional branch" 
instructions use a common opcode, but the conditional branch appends 
a digit that specifies the condition.

The "trick" behind all of this is that the instruction execution 
logic locates the opcode by scanning until it finds a character with 
the word mark set.  While it's scanning, operands are shifted through 
the operand regsters.

So, I submit that the 1401 is a variable length machine (it's 
decimal, by the way) both in the sense of data that's being 
manipulated and the instructions themselves.

Have I missed a third category of "variable length"?

--Chuck

[prev in list] [next in list] [prev in thread] [next in thread]