ooh, a Transputer link - nice idea!
1 Like
Ed - just thinking of a suitably retro and interesting example.
You could make an array of Picos with ~12Mbaud links (baudrate subject to confirmation).
As you have control over GPIO pinning - you can create north, south, east and west links on convenient pins.
A 16 x16 array of RP2040 chips is going to be about 300mm square!
2 Likes
When I read about PIO it immediately reminded me of the channel I/O of mainframes and also about the https://en.wikipedia.org/wiki/Communications_Processor_Module (warning: I wrote most of this article) being part of many Motorola/Freescale processors. It was present in PowerPC designs like MPC860 that were at the core of many Cisco routers and their interface cards.
So offloading I/O this way is not very novel concept as suggested in the HackSpace magazine article quoted above?
Interesting about the CPM. Some other microcontroller has PRU which is again a bit of configurable or programmable resource near the pins. But I think I’ve read that the PIOs are more handy - this might be a matter of opinion! But I think in the field of MCUs, with these two exceptions, this kind of offloading is novel - it’s just that phenomenon we see over and over again, that innovations from the 50s and 60s eventually reappear in integrated form. Novel in one sense, not in another.
But also, perhaps there are distinctions to be drawn between low-level accelerators such as the PIO, which remove the need for the CPU to do bit-banging, from mid-level accelerators which remove the need for the CPU to move data from A to B, from high-level accelerators which remove the need for the CPU to know how to drive a device (or how to understand a filesystem, if we go as far as CBM’s smart floppy drives.)
1 Like
Having discovered CPM (I was wondering what “PowerQUICC” meant in Cisco routers’ CLI output) I came to conclusion that one of the factors why Cisco router were so successful was that interface cards were independent computers handling I/O independently on their own. I am still using MPC8272-based router at home - that thing must have been designed ca. 2002 so I think that slowly qualifies for retrocomputing subject in itself, except that I gather from the NXP website that I can still buy new ones, but at 72 euro per piece, are they still “MCUs”?
It doesn’t surprise me, and it doesn’t surprise me who did it, either!
1 Like
Sure are a lot of little unix like OS’s popping up now days. Sadly they all seem to follow
the UNIX /MINIX trend of not having a C compiler in source code, or having one at all
making bootstrapping interesting.
I wish I knew about this around 1984 when I was trying get a S100 system working, Ben.
All these OS experiments needing a nanny (other computer + SDK) for their life cycle are boring.
E.g.: Why isn’t an ESP32 (½…8½ Mbytes RAM) big enough to run its own C-compiler? We could do that in CP/M days on 8-bitters.
Because most people do not feel like squeezing a toolchain into a limited environment, these days/anymore?
1 Like
I’m sure there’s room for a constructive thread about any bootstrappable OSs that we discover. I agree that it’s a desirable feature.
I’m not sure about the value of moaning about how something doesn’t exist, or hasn’t yet been found. Or how something which does exist lacks this or that aspect: somebody took the trouble to build it and to share their work, and they took the tradeoffs which seemed appropriate to them. Encourage such efforts!
2 Likes
This is probably an unnecessary diversion for me to bring it up, but Unix from Bell Labs always came with the sources for its C compiler, as has the Berkeley Software Distribution, and they have always been self-hosting. BSD 1.x and 2.x were patches to Bell Labs Unix, while BSD 3.0 and beyond were complete distributions.
It was only with the commercial distributions of Unix that the compiler became a crown jewel, not to be shared with the populace.
2 Likes
I started MINIX1-ing when I saw a patchset plus documentation including BCC16 and BCC32 to turn MINIX1 into a 32bit OS. I wouldn’t have bought both (MINIX1 plus that patchset) otherwise.
It was fun and after a while Linux-0.11 showed up and having MINIX on an other partition helped a lot in its earliest days. At work (SunOS4 era) I never asked for the C compiler’s source. It existed as binary and worked. I didn’t need more those days. Previously on AtariST I got used to have at least 3 of the 5 C compilers as sources too (if my lossy memory compactor doesn’t trick me too much): Sozobon-C, LCC, GCC. That always felt like an insurance…
…but having the sources sometimes isn’t enough: I stopped playing with RetroBSD on PIC32 when I no longer could build its toolchain on Debian and building the Propeller’s PropGCC on an uptodate system is something from pain to nearly impossible too today… :-/ …ok… if all else fails, there always is a VM with an ancient Debian, but FUN is spelled differently.
Sorry for the nostalgia attack blurbarber… :-(
1 Like