Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.
Hey Thanks man,
My doubt is whether a system can be designed where the data returns itself when a request is placed without the source having to retreive its location and then get the data.
without receiving the location how can we the system know the address of the data to be retreived... as far as i know there is no way this can be done...
In the present systems it is not possible as the systems are hardwired to the traditional procedure of address and data retreival.
Only a redesign with some AI can do this I guess.
Hi Anand please read the below paper which will help u in understanding the purpose/concept of what Iam talking about.
Please let me know if this is possible or not. I even tried contacting some professors regarding this but hesitated so that I can confirm its possibility first before proceeding.....
Added after 1 minutes:
This paper intends to focus on changing the present methods of accessing data in a
memory
location. An algorithm for this will be discussed in the prceeding abstract. There
is
presently no technical proofs attached here to prove my ideas as the present
technology
doesn't support any tools/software which enable this to be tested. To test this it
would/may
require a new design based on this concept.
As years pass enchancements occur for "faster" processors or systems. we are
concentrating only on the improvement of present technology (ex:Chip fabrication
130nm->90nm->65nm->45nm->35nm chips)
but not on other techniques/methods/innovations. For more better technological
improvements it is
neccessary to "invent" new methods. In such attempt I wanted to submit my idea
(more-like science-fiction or dumb than to be practical/possible) on accessing data
in chips/memory.
In present systems most of the routines/programmes access data
from a
location and such commands are the most common/must actions that we see and occur.
For Ex: if we write
a small 8085 progamme for addition.....
EX:1-
MVI B, 06 //Load Register B with the Hex value 06
MOV A, B //Move the value in B to the Accumulator or register A
MVI C, 07 //Load the Register C with the second number 07
ADD C //Add the content of the Accumulator to the Register C
STA 8200 //Store the output at a memory location e.g. 8200
HLT //Stop the program execution
In the code above you will observe that all the operations require to access data
which is stored in a
particular location which has an address attached to it. To get the data, the source
should
know the address location where the data is present, then retrieve the data from the
location.
Let us know correlate this with a natural example. Supposedly
there are a
million people(data) in a room, there is a Chairman(source) and a
host(medium,processor etc.,)
Let us also assume that all the million dont know their names. There is a
ceromony(code to be
executed etc.,) and the chairman would like to award(access) a person(data) named
"Mike". So
the Chairman places a request with the host for the person named "Mike" so that he
comes to
the stage, to be awarded.
Now the problem is, as the million people don't know their
names, the
host has to get the location/seating no.(Address location) of "Mike" so as to bring
him to the
stage(ultimatly to the source).
This process would have been much easier if all knew their
names, the host
would have called out "Mike, Please come on to the stage". And the step of reteiving
his
location and him would have been unnecessary, saving time.
To elucidate it, if there was a design where in the data(Mike) comes to the
source(Chairman)
when a request(call by the chairman) for the data is placed avoiding the
conventional process of retrieving address
and the data in it, I really hope can save a cycle or two in the Instruction
cycle.
If we consider an example where a list of objects need to searched for an object.
Lets say we have a list
list("Mike" "Arnie" "Chiru" "Nag" "Bal" "Venky"). Now frequently when writing
programs we need to see if
an object (lets say "chiru") is present in the list. To do this If I were to write a
code I would
compare each and every object in the list with the object we want to compare
against. Which takes time (though very less).
If my concept as explained above with example were to work then Instead of looping
through the entire list the object would have
returned itself as soon as it knew(Some design is necessary to make this happen)
that I was checking for this object.
Which I
in my view would have taken less time than the conventional method of looping every
object in the list.
Hi,
What you have explained can be done, if each memory location itself is an intelligent entity as in the case of Mr.Mike. In the conventional computer architecture, CPU is the master and the memory locations are devices simply to store data, as decided by the CPU. Only way for the CPU to store/retrieve data in such a system is by assigning an address.
On the contrary, in a multiprocessor system connected on a bus, if the master CPU simply sends down a command with an embedded address, the addressed CPU can supply back requested data, not just one byte, while other CPU's will not reply. Remember in your case too, you have implicitely sent down an address by calling out Mike's name and he should have been programmed to go to the chairman when his name was called out, otherwise he would have given back only a natural stare at him.
But, you are right in that, it is high time that we thought of better Computer architectures, without simply relying on technology improvement in a narrow direction.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
