suggest an old cpu architecture that doesn't fully suck
category: general [glöplog]
Quote:
I think RISC vs. CISC was mostly a marketing stunt and not so much an actual single difference in architecture (or overall benefits of one vs the other). With current architecture, the distinction has become quite meaningless (if it ever held much meaning). :)Load-store and register-memory are mostly used in the RISC vs. CISC realm nowadays, which the CPUs we're talking about predate, so it might've been incorrect from my part to start the thread with a false dichotomy.
Quote:
Close, Virtual Boy was a V810, the precursor to the V850 (which then evolved into the RH850 series, used in automotive ECUs).Quote:other contenders: SuperFX, TMS320, V810/V850/RH850
V850... that rings a bell for sure. Didn't Nintendo use a V850 core for the Virtual Boy?
MIPS is fun to write for. It's load-store, but also really comfortable, minimal and makes sense in the same way as 68k.
When it comes to general-purpose processors, assembly flavours A vs B were never much of an issue, imho - they were just what came with the machine at hand, and learning a new flavour was a matter of a few hours. It's all assembly, in the end.
Then... i came across DSP assembly that looks like
Then... i came across DSP assembly that looks like
Code:
for a single instruction word, which was harder to get into, and not much fun. Former workmate loves this stuff, though. =)f8=f1*f4, f12 = f8 + f12, f1 = dm(%5,m4), f4 = pm(i9,m14);That’s a perfect summary of that moment when you feel like an assembly master... and then DSP shows up and says: ‘Hold my coffee.’ 😄
If you’ve ever coded in BASCOM for AVR, this might feel oddly familiar. Think of it as retro-hardware minimalism meets high-level comfort — just enough abstraction to keep your brain cool, but close enough to the metal to still smell the silicon.
If you’ve ever coded in BASCOM for AVR, this might feel oddly familiar. Think of it as retro-hardware minimalism meets high-level comfort — just enough abstraction to keep your brain cool, but close enough to the metal to still smell the silicon.
I don’t know about the rest of you, but I can’t listen to this. Too cringe. It’s like Sheldon Cooper trying to be Jim Morrison.
I wasn't expecting to get attracted to 6502 and MIPS after chatting a while in here. I got PS1 to start with MIPS I, so that's good, and I can't even begin to count all the 6502 boxes out there, so that's great.
I'd like to mention one weird thing, in case anyone wants to go down a rabbit hole as well.
Motorola made a RISC CPU in the mid 80's called the 88000. It's nothing like the 68000, fully different architecture, pretty much everything you'd expect from a RISC, but it's interesting to see how Motorola went about the design: A very trimmed down instruction set (has about the same number of instructions as THUMB16), many general purpose registers, separate control registers, FPU, MMU, one ALU instruction per cycle, you get the drill.
Reasons it failed I think because the chip was huge to the point they had to split the FPU and MMU into a separate chip, which ramped up cost, and even then they started to hit yield issues and unreliability. About the only mildly successful consumer thing that used it was the Omron Luna-88K, which is pretty much a playground for the CPU, it barely had anything interesting in it. 88k, 16MB of RAM, your typical expansion/IO and a 2048x1024 frame buffer. Nothing more, nothing less.
I'd like to mention one weird thing, in case anyone wants to go down a rabbit hole as well.
Motorola made a RISC CPU in the mid 80's called the 88000. It's nothing like the 68000, fully different architecture, pretty much everything you'd expect from a RISC, but it's interesting to see how Motorola went about the design: A very trimmed down instruction set (has about the same number of instructions as THUMB16), many general purpose registers, separate control registers, FPU, MMU, one ALU instruction per cycle, you get the drill.
Reasons it failed I think because the chip was huge to the point they had to split the FPU and MMU into a separate chip, which ramped up cost, and even then they started to hit yield issues and unreliability. About the only mildly successful consumer thing that used it was the Omron Luna-88K, which is pretty much a playground for the CPU, it barely had anything interesting in it. 88k, 16MB of RAM, your typical expansion/IO and a 2048x1024 frame buffer. Nothing more, nothing less.
Quote:
When it comes to general-purpose processors, assembly flavours A vs B were never much of an issue, imho - they were just what came with the machine at hand, and learning a new flavour was a matter of a few hours. It's all assembly, in the end.
Absolutely not an issue, and actually the fun part of writing for a particular system comes with the design of the system as a whole, not just the CPU. Without any peripherals to interface with the CPU is useless, and the human feedback you get as reward from your work always comes from peripherals. Video, sound, or whatever. Still, I like diving into the different assembly language flavors because it's still a big part of the puzzle, and I just have fun with it. This is purely recreational :)
As for 6502, with its one accumulator + 2 index registers... check out this weirdo variant that breathed life into Furby. ABI is like 6502, but without the Y register. Only A and X. Ugh. =D
It’s a pedestrian traffic light as opposed to the original traffic light CPU! No orange/yellow.
Transistors instead of trauma — nice to read about computers again, rather than just how AI hurt someone.🚦
Quote:
ABI is like 6502, but without the Y register. Only A and X. Ugh. =D
What a weird thing. If I wanted to have headache for breakfast I'd read dinky microcontroller datasheets all morning. A 6502 minus one index register doesn't sound that bad compared to those 4 bit MCUs that came in Game & Watches though :S
Quote:
It's still useable. In C-64 (and friends) demo code, however, you're always short of registers and processor cycles, and find yourself not only running code in the zeropage to selfmod literal arguments, but also using the stack pointer as a shadow register for X... etc. =)A 6502 minus one index register doesn't sound that bad
Personally, as also a 68K addict, I really enjoyed programming the PS2 and the Inty, the GI CP1600 was really fun and easy to play with.
The Motorola 6809 from 1978 is the most advanced and elegant 8-bit CPU. It supports recursion, dual stacks, and addressing modes straight out of a dream manual. 🖥️
Quote:
In C-64 (and friends) demo code, however, you're always short of registers and processor cycles
I get it. I sometimes end up juggling registers too much in 68k when I'm making games, so imagine. In fairness, making games and demos is not the same because with games you have to be more conservative with your state management and there is more outside information to handle, opposed to a demo where it will always do the same thing, and you can just call a loader between parts and each part can just run as its own independent program and not give a damn. Though demos like EON had to do some memory management to load in and decompress parts from disk in parallel, so it's not always like that.
Well, maybe the better comparison is that you need to do more things in a demo than just drive motors in a furby lol
I think it's funny that no one has commented on x86 yet. I wonder why that is :P
Personally, as a 68K addict, I really enjoyed playing with the PS2 architecture (more than a CPU) and strictly on the CPU, I really liked the Inty General Instrument CP1600, really fun to code.
You can’t always JSR LOADER. Sometimes he’s off optimizing the next interrupt and responds when the stars align. 😄 Krill isn’t a subroutine — more like an overlay. Loads when you least expect, vanishes on RET.
When it comes to x86, we all carry the trauma. No one wants to unpack the segmented memories — it always ends in a calling convention debate that lasts until sunrise. It's the architecture that refuses to die — we honor it in silence.
When it comes to x86, we all carry the trauma. No one wants to unpack the segmented memories — it always ends in a calling convention debate that lasts until sunrise. It's the architecture that refuses to die — we honor it in silence.
Quote:
Personally, as a 68K addict, I really enjoyed playing with the PS2 architecture
PS2 seems nice, but trying to get straight to weeds with MIPS IV, vector units and the other peripherals of the system bare metal might be a bit too extreme. Still, I think it would be fun to start with some C SDK, make some simple demos and then start peeling off the abstractions until I understand what the system does. But for that I'd prefer to start with PS1 first.
About the Inty btw, I never even looked at it, and I constantly forget about its existence hah
Quote:
You can’t always JSR LOADER. Sometimes he’s off optimizing the next interrupt and responds when the stars align.
True, a loader isn't just "jump to the next part plz" without any delays or syncing, but that's what I meant with the EON example. In big demos, for DMA based systems like the Amiga you'll be polling disk, decompressing, jumping from interrupt to interrupt, setting timers and doing other things all the time. For DMA based systems like the Mega Drive where you only have one HINT and one VINT you'll be counting cycles and praying to god that your DMA transfer doesn't leak out of VBlank and that your interrupt handler has the minimum amount of overheard. Trying to rewrite palettes in HBlank is a pain too, sometimes you cannot get CRAM dots out of screen and you're like for fucks sake end my life now
That's a problem in the NEO GEO too. Color effects are always such a royal pain
Quote:
PS2 seems nice, but trying to get straight to weeds with MIPS IV, vector units and the other peripherals of the system bare metal might be a bit too extreme. Still, I think it would be fun to start with some C SDK, make some simple demos and then start peeling off the abstractions until I understand what the system does. But for that I'd prefer to start with PS1 first.
About the Inty btw, I never even looked at it, and I constantly forget about its existence hah
Sorry for the double post, Pouet fooled me :)
Yeah, I liked the PS2 due to his architecture, DMA stuff, calling the VUs, the GS.. more than the MIPS itself (I used assembly mostly for the VUs in fact, only C on the MIPS so it makes my comment irrelevant :D)
The Inty was a real surprise, I used to code on the 2600 so having 1K of RAM was a luxury, but really the CP1610, one of the rare 16bits CPU from the 70s, said to be inspired from the PDP-11 (I did not use the PDP-11 enough to find it was that true, many the octal stuff was the reason :D) gave me nearly the same satisfaction I has when coding on the 68K. So.. unexpected but good!
Less talk, more action.
