Telephony switches - is there anyone out there collecting these?

dwight elvey dkelvey at hotmail.com
Sat Jun 30 18:25:52 CDT 2007




>From: woodelf <bfranchuk at jetnet.ab.ca>
>
>dwight elvey wrote:
>>I just saw a web site with truly reduced instruction set. It had one
>>instruction. Subtract and branch < 0.
>>One can do a lot with a simple model.
>
>One or TWO's compliment?
>Decimal or binary?
>What if I want Subtract and Jump if < 0?
>I've seen the same site. :)
>
>The Ultimate RISC page is here.
>http://www.cs.uiowa.edu/~jones/arch/risc/
>The minimal CISC  page is there.
>http://www.cs.uiowa.edu/~jones/arch/cisc/
>

Hi
I don't think it matters what form of number represented.
As for jump or branch, I like branch better for portablility.
Still, jump is the easiest to implement in hardware. Sequencing
could be done with an LFSR instead of a incrementor.
This is even easier to implement. Of course, one would
most likely need a compiler to figure addresses out but branch
would just be an xor pattern. The ultimate in simple.
One could most likely get even simpler with a bitwise
nand of the data and a branch if zero ( or jump if you like ).
One can create add and subtract with just nand and invert
( invert is just nand( n, 1's ). This would be binary to makes
sense.
I think one might like to add some form of subroutine call.
The way I saw it done in the Nicolet would be fine. The first
location of the call would receive the return address. One
could make it so that all branches were like calls, it would
just be that the return address was ignored if it was not
a call.
Still, only one instruction but multiple ways to use it. :)
Dwight

_________________________________________________________________
Picture this – share your photos and you could win big!  
http://www.GETREALPhotoContest.com?ocid=TXT_TAGHM&loc=us




More information about the cctalk mailing list