XPost: comp.lang.misc, sci.physics.relativity
From: already5chosen@yahoo.com
On Mon, 8 Dec 2025 04:25:32 +0100
Janis Papanagnou wrote:
> On 2025-12-07 15:26, Michael S wrote:
> > On Sun, 7 Dec 2025 11:42:40 +0100
> > Janis Papanagnou wrote:
> >
> >> On 2025-12-07 10:22, Thomas Heger wrote:
> >>> Am Donnerstag000004, 04.12.2025 um 09:57 schrieb Janis
> >>> Papanagnou:
> >>>> [...]
> >>>> That's the date of the Z1, isn't it? - The Z3 came later, 1941.
> >>>
> >>> Sure, but the first computer using 'von Neumann architecture' was
> >>> actually the Z1 of 1937.
> >>
> >> Well, yes. At least mostly. That's why I've written upthread that
> >> the concepts from the earlier Z1 were reused in Z3.
> >>
> >
> > I'd say, no. Neither Z1 nor Z3 are von Neumann architecture
> > computers.
>
> Right. - Considering all properties, von Neumann's computers had a
> von Neumann's architecture.
>
> > And it has nothing to them being either electronic or mechanical.
>
> (This was just one common example by some to attest its innovation
> and being "the first".)
>
> > The key element (==distinguishing feature) of von Neumann
> > architecture, at least in modern (say, of last 60-65 years) meaning
> > of the term is that program store and data memory reside in the
> > same space.
>
> Yes. But is that crucial for a programmable computer? Is that the
> functionally necessary or important element? - I'd clearly say no!
>
Is it not particularly important for programmable computer.
When I design MCU-based systems, my MCUs are physically able to do von
Neumann (i.e. to run program from RAM) but I don't utilize this
property, treating them as if they were Harvard.
The biggest and fastest growing computing business of recent years
is based on programmable computers (GPGPUs, TPUs, NPUs) that can not
modify theier own programs.
However, computers with which we interact most, so called
general-purpose computers, from smartphones to servers, are very
heavily dependent on being von Neumann.
>
> > Which leads to possibility of self-modifying code.
>
> And that specifically is neither a necessity for a "[universally]
> programmable computer" - IMO the historic noteworthy key property! -
> nor an example how systems sensibly should be (or are) programmed.
> We avoid in practice exactly that property (modulo virus-developers,
> maybe, and similar corner cases).
>
You are thinking about SMC at small. Think about it at larger scale.
Level one: OS loads application. It's rare that it just blindly
copies the image from disk. More often it modifies it to fit at
particular address. Esp. so today, with ASLR considered must.
Level two: JIT. For good or for bad, a cornerstone of modern web.
> > Which led to Von
> > Neumann's claim that index register is unnecessary for array
> > processing. Which is undeniable mathematical truth and serious
> > engineering mistake at the same time.
>
> Yes.
>
>
> An inherent logical problem lies also in the argumentation chain
> we commonly see...
> "Contemporary computers are "basically" all characterized by
> von Neumann's architectures."
> "Von Neumann's computers are defined by ...property list..."
> "Von Neumann was the inventor of [contemporary] computers."
> (I assume you notice the dodge.)
>
> I really don't want to engage in such discussions[*] but I think we
> should at least understand the mechanics behind that. The interests
> and the rhetoric/argumentation moves used to establish such agendas.
>
I don't find it interesting.
> Janis
>
> [*] We know how the inventions have been and (partly) still are
> attributed, we see the various areas, and the actors' agendas and
> interests, and we observe that also on various levels (countries,
> competing companies, partners, plain sponges, gender status, etc.);
> there's countless examples of historic mis-attributions.
>
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
