Anyone else who never heard of John V Atanasoff?

[wilfred]

/.

[helius]

"Stored Program" is a technical term that refers to an architecture that has been named after John von Neumann.
Its distinguishing feature is that a program is loaded into the same memory as the data operated on, and can be addressed the same way. This is important for interactive computing and assemblers or compilers.

Neither the ABC (Atanasoff-Berry Computer) or the ENIAC was a stored-program computer, they were programmed with cables. The first functional stored-program computer was the Manchester Baby.

[helius]

This is a critical primary source. The ABC would have been forgotten entirely, but there was a controversy around the validity of a patent filed by Eckert and Mauchly and acquired by Remington Rand, granted in 1964 and which would have been in force until 1978: Atanasoff gave testimony about his prior invention of the ABC and how he described details of the machine to Mauchly in a collegial way.
http://en.wikipedia.org/wiki/Honeywell,_Inc._v._Sperry_Rand_Corp.
The existence of the Bletchley Park Colossus was also antecedent to the ENIAC, but it was never revealed until the 1980s.

[helius]

The problem with a tagline like "Father of the computer" is vagueness. What do you call a computer?

During the Manhattan Project, calculations were carried out by hundreds of women sitting at mechanical tabulator machines. The machines weren't computers, but the operators were. The use of the word "computer" to describe these people predates its use to describe machines. The dream of "automatic calculators" that could evaluate complex formulae faster than a human operator was around since the 1830s; Babbage's "engines" were computers of that sort. The dream next awoke in the 1930s, when several engineers, working independently, created electromechanical machines that were designed to solve specific problems. Zuse and Aiken read instructions from paper tape to perform math on accumulators; Stibitz read complex numbers from a telegraph and did complex arithmetic; Atanasoff solved linear equations stored on a drum.

Despite these forerunners, the ENIAC was the first computer that could be programmed for more or less any purpose. Maybe "programming" deserves scare quotes: it was not a series of instructions, but a maze of patch cables connecting different components. The "programmers" needed to understand the entire machine in minute detail, connecting gates and registers in a manner that could solve a problem. They were really doing hardware design without a HDL, since everything happened in parallel.
http://eniacinaction.com/

After ENIAC was built, Eckert and Mauchly applied the concept of an instruction sequence (as used by Aiken) to be stored in electronic memory, to their work on the EDVAC. Later on, their names were deleted from the Draft Report On The EDVAC, written by von Neumann. That is how the stored program concept came to be called the von Neumann Architecture.

[T3sl4co1l]

There's a certain (if rarely used) tradition that the person doing the activity is the -er, and the machine used to facilitate it is the -or.  Hence, computer/computor (I have some 12AX7s which proudly claim "COMPUTOR TUBE" on them), welder/weldor, etc.

Although maybe I have that reversed, or it's situation dependent (I seem to recall a "weldor" is a person)... or it's not a "tradition" at all but more of a corruption...

Tim

[helius]

I confess I'm not familiar with that convention. I kind of thought that the "-or" suffix just makes you sound more Latinate.

On the subject of inventions and patent trolls, here's an article about the first microprocessor.
http://www.righto.com/2015/05/the-texas-instruments-tmx-1795-first.html

[magetoo]

"Stored Program" is a technical term that refers to an architecture that has been named after John von Neumann.
Its distinguishing feature is that a program is loaded into the same memory as the data operated on, and can be addressed the same way. This is important for interactive computing and assemblers or compilers.

I'd say that the distinguishing feature is that the program is loaded into memory at all.  You are of course right about the distinction between having the program in memory on one hand, and having the program be a part of hardware on the other.

But consider that there are also Harvard architecture computers, who have different memories for programs and data.  They are also stored program architectures, since the program is stored in memory, but they are not Von Neumann architecture computers, since program and data are separate.

(And if we want to get philosophical, is a micro running a program out of ROM still a stored program computer?)


Apparently there are two schools of thought on what constitutes a stored program computer; I had no idea until I checked the Wikipedia article.  (And now I'm lost in Wiki again.)

[helius]

But consider that there are also Harvard architecture computers, who have different memories for programs and data.  They are also stored program architectures, since the program is stored in memory, but they are not Von Neumann architecture computers, since program and data are separate.

I take an operational approach. What can you *do* with the program stored in memory? And how does it get there in the first place? If you can swap an array from "data memory" to disk and back through another path into "program memory", then it works pretty much the same as a von Neumann machine. If you can't, then your computer basically cannot ever be self-hosted. In the old days this didn't matter much: programs were loaded from punch cards or paper tapes and moved by hand between the editing punch, the collator, the ASR33, etc. In a batch system, Harvard architecture works the same way (biggest limitation is that you separate your "data" punch cards into another deck!). In a timesharing system with interaction between processes, that's no longer true.

(And if we want to get philosophical, is a micro running a program out of ROM still a stored program computer?)

While that is not a self-hosted computer, you normally can load data out of the same ROM.
For more philosophizing, consider a tagged architecture (like a B6500) where instruction words are marked with a tag that prevents them being read as a number.

Apparently there are two schools of thought on what constitutes a stored program computer; I had no idea until I checked the Wikipedia article.  (And now I'm lost in Wiki again.)

It depends on who is looking at it. From the processor's point of view, there is a huge difference between data and code, because only code can be executed. It looks different from higher (interpreter, compiler) or lower (microsequencer) levels.

[Mechanical Menace]

It depends on who is looking at it. From the processor's point of view, there is a huge difference between data and code, because only code can be executed. It looks different from higher (interpreter, compiler) or lower (microsequencer) levels.

I don't know, yes code can be executed but it's still a type of data. A special subset but you can manipulate it in the same ways and that can lead to interesting applications. Self modifying code isn't always bad, and can still have practical as well as academic purposes.

[T3sl4co1l]

(And if we want to get philosophical, is a micro running a program out of ROM still a stored program computer?)

Mask ROM or Flash ROM?  Most MCUs with the latter have self-erase-rewrite capability.

Tim

[magetoo]

(And if we want to get philosophical, is a micro running a program out of ROM still a stored program computer?)

Mask ROM or Flash ROM?  Most MCUs with the latter have self-erase-rewrite capability.

Tim

Real ROM, of course, the kind you have to write using hammer and chisel.  None of this modern "floating charges" nonsense, to borrow an expression.  :-)

For more philosophizing, consider a tagged architecture (like a B6500) where instruction words are marked with a tag that prevents them being read as a number.

Not familiar with it, but it feels like it would be analogous with your example of the Harvard architecture where bits can be moved from one memory to the other.  (Assuming those tags can be set/cleared.)  Like that, except now the data and code memories also happen to share the physical implementation.

And modern architectures have tagged pages (the "no execute" bit), where data can not be executed.  It seems like things aren't necessarily one or the other with a clear dividing line down the middle..

[T3sl4co1l]

Real ROM, of course, the kind you have to write using hammer and chisel.  None of this modern "floating charges" nonsense, to borrow an expression.  :-)

You tell me how many strikes to hit the chisel with, and I'll make you a robotic programming attachment; it can even be ICSP capable!

Tim