Designing a .NET Computer | Hacker News Reader

Designing a .NET Computer

daeken.com

30 pointsdaeken15 comments

Open articleView on HN

"Remember the Intel 432" seems in order.

Ah... BTW, also give a look to Lisp machines made by Symbolics, LMI and Texas Instruments.

And, when you feel like it, google for the Smalltalk machines made by Xerox, Tektronix...

Either that or, to quote George Santayana, "Those who cannot remember the past are condemned to repeat it"

"Remember the Intel 432" seems in order.

That was my first thought too!

In a way, Im sad that the 432 never took off. It would be really interesting to work with an object oriented processor architecture with built in garbage collection.

I've done a lot of research into Lisp machines and a little bit into Smalltalk machines (Renraku is a spiritual successor to the Lisp machines of yore), but I'm seeing this is more of a joining of the Jazelle Java acceleration on ARM and the managed OS work in Renraku.

That said, for the nitty gritty details, such things are certainly the place to look. They've solved many of the problems already.

Did you check those monsters from Azul Systems? They are doing some interesting Java stuff on more or less bare hardware

Don't forget the object-oriented Linn Recursiv CPU.

Ohhh I forgot about this one. That was an interesting architecture too, not just because it was innovative and ahead of its time, but also because it was designed by a company that wasn't in the business of processor development!

Yep. First time I read about it, I did a double-take: "wait, _Linn_ designed a CPU?!". Sounds like it might have been an interesting place to work.

Too bad there's next to nothing online about it.

Same here, and I used to be a huge Linn-fan (still proud owner of an LP12). I've even met Mr. Linn himself, Ivor Tiefenbrun. He had that peculiar combination of ego, craziness, genius, and "how-hard-can-it-be" that makes a great old-school inventor.

I found this story about the Rekursiv (note the typical Linn "k") chip:

http://groups.google.com/group/comp.arch/msg/991720740417ec4...

Another system worth looking at is Symbol.

Its compiler was implemented in hardware....

http://portal.acm.org/citation.cfm?id=633642.803976

http://www.computer.org/portal/web/csdl/doi/10.1109/MAHC.200...

Its worth mentioning the Great Microprocessors of the Past and Present website[1], especially the section on Weird and Innovative Chips[2]. Some very interesting designs in there and some interesting history overall.

Completely opposite of a crazy CISC machine, but just as interesting, is the Transputer T-9000[3]. They actually produced processors which could be slotted together to form vast, highly parallel, arrays of simple processing nodes. Think of the TILE64[4], but where the tiles are separate units, allowing you to add more over time.

[1] http://jbayko.sasktelwebsite.net/cpu.html

[2] http://jbayko.sasktelwebsite.net/cpu7.html

[3] http://jbayko.sasktelwebsite.net/cpu7.html#Sec7Part5

[4] http://www.tilera.com/products/processors.php

Less-known than it deserves to be: http://www.azulsystems.com/ has been selling Java computers (with almost 1000 cores) for several years now.

Why?

I mean in some contexts it made sense historically. There were the Lisp machines someone already mentioned. I've used the 8052AH-BASIC: an Intel 8052 microcontroller with a built in BASIC interpreter, and there were a few small chips that booted into a Forth interpreter. But those were controller-oriented microprocessors in a time when compilers (and silicon in general) were expensive, so it made sense to build a chip with its own language, especially when it would find application in a product with a lifetime measured in decades.

But in 2009? Besides raising a host of upgrade problems, why would you do this?

I'm guessing it would run cosmos.

http://www.gocosmos.org/

Well, hopefully there'd be a few competing OSes for it. I'd love to see Singularity on it, and of course I wouldn't mind seeing Renraku (my own managed OS, which inspired this) running on it.

While Cosmos is interesting for the same reasons Singularity is interesting - and I'm very much in favour of any OS which tries to focus on security, verifiability and correctness - from my little knowledge of the system, it doesn't really seem different enough. Theres countless toy, research and hobby OS projects out there, some more ambitious than others, and few truly try to innovate. Now, I'm not very familiar with Cosmos, so its unlikely that I'm giving them enough credit and I apologise for that.

What I want to see is an operating system which forgets all the technologies which were a great idea 20 years ago, but may not be so suited to todays computing world. We have a lot of computing power available to us now, so we can try some truely innovative idea cheaply and easily. I want to see highly extensible systems, applications which are hugely parallel from kernel up, I want to see orthogonal persistence, I want a system which can determine what applications to run for me, so that I dont have to and can simply focus on the task at hand - getting work done. I dont ever want to have to think about filesystems and directory structures again. I want to be able to see the internals of the running system in a REPL, if I so choose. If I'm using, say, an instant messenger and I'm typing text, I want to be able to use vim (or whatever your favourite editor is) to edit that text!

Ok, some of these may not be great ideas or very realistic - the point I'm trying to make is that I want to see operating systems truely try to achieve something different in an end-user (ie not simply for research) OS.

This is the type of system I envision to run on such a computer, were it to exist!

End of Rant ;-)

Reply on news.ycombinator.com