[karmakaze]

This sounds like a much better spec. Using the 65c816 keeps the 8-bit vibe but doesn't force you to play the bank switching game to access more than 64K. DVI instead of VGA itself is also enough to give it a much better chance of success.

[cmrdporcupine]

It's true you don't have to bank switch... but... the 816 doesn't have 32 bit or even 24-bit registers, so working with memory regions over 64k (such as the video framebuffer) is surprisingly awkward. You end up fiddling with the bank registers a lot, and doing a lot of futzing around. Having screen resolutions that require a framebuffer >64k isn't a good match for it, for example. And the program counter is a 16-bit register, too, so you can't have programs over 64k without incrementing the program bank register at the end of the 64k region, etc. Also the stack and direct page can't be located out of the bottom 64k, etc.

So while changing a bank register is easier than doing bank switching with a 6502, it's actually not that different. And using the 16-bit registers on the 816 require a mode change, so working back and forth between 8 and 16 bit values is also awkward.

The 816 on paper sounds good at first. In practice, after working with it for a while (I wrote an emulator for an earlier version of her machine and then started working on software) I found I didn't enjoy the 816 as much as I thought I would and understand better now why Commodore and Atari made the jump to the 68000 instead of to the 816 or other modifications to the 6502.

[hth313]

You can work with 24 bit pseudo registers in the direct page using the [dd],y addressing mode, basically scaling up the very useful (dd),y on the 6502 to be 24+16 bits instead of 16+8.

I am currently working on a compiler for the 65816 and has taken the approach of avoiding mode shifts as much as possible. I also lock the direct page and bank register to a fixed place while running normally. This simplifies it a lot.

Segmenting your functions/code block on 64K is not a huge limitation. How big functions do you write? LOL

The 65816 does require more discipline when programming than the 6502.

Yes, they all went 68000 which is no wonder as it was a step up further. The 16 revolution never really happened, we went from 8 to 32 bits computers.

Commodore did go back to modify the 6502 later, that was the Commodore 65. That the C65 project went as far as it did is beyond me. I would have stopped it immediately if I was shown it at the time, both from marketing and technical aspects. Today, I suppose it makes some minor sense to revive it as a curiosity, just for the sake of it.

I wonder how you can find the 65816 odd, when the Mega 65 comes with three altered variants of the 6502. Also considering that some well known 6502 programming idioms no longer works as before, as soon as you move the zero page or the stack. Having studied them all, I find the 65816 puts those 6502 variants to shame.

[karmakaze]

There was sort-of a 16-bit revolution--it wasn't connected to the 8-bit one. The 68k systems were 16-bit hardware with all new/different operating systems. It was much later we got 68020 32-bit hardware. The IBM PC/AT was 16-bit hardware following the 8-bit 8088 PC. That got us to DOS/16M and Windows/286 and OS/2 v1.3 (good times). On the fun side we got the 16-bit wave of consoles. It was all very semi-business though--I really missed the 8-bit days all through this period. The best times were probably on the Amiga in EU.

[cmrdporcupine]

Yeah the Mega65 is also going to have these oddities. From reading though it does sound like it offers a nifty 32-bit single register from merging A,X,Y,Z. And I believe it also has relocatable direct page. So some tricks borrowed from the 6809.

When I was playing with the 816 I did use the technique you describe using the direct page. But I recall there being limits on where I could use that.

The 816 is a bunch of compromises. And some awkward ones because the opcode space was all eaten up.

[hth313]

I looked at the Mega65 with its three (not quite) 6502 CPU variants and I just found it so weird. Basically no 16 bit operations, instead a couple of 32 bit operations. Z register is no longer 0 and you have to set it non-zero to change bank configuration. Five instructions (nine bytes) to change bank and I do not understand how you can save the previous bank setting in case you want to restore it.

If you move the stack or zero page, the well known trick of using absolute plus index register no longer works the same as on the 6502 (breaking existing code). Sure, you could hardcode a different page, but the point of moving the stack is (probably) so you could do multi-tasking. But if processes are given different stack pages, they can no longer share code (if using such well known tricks in the code). Moving zero page is in 256 steps, so you cannot use it as a movable stack frame (the 65816 can set it to any location in the first 64K).

I know there is a lot of interest in the Mega65, but I found it had so many poorly made design decisions that made me want to stay far away.

So far I find the 65816 to a well designed 16 bits ISA (I am also familiar with PDP-11, MSP430, MELPS7700 and Motorola 6812) that gives a lot of 6502 vibes, but is just a lot better. There are a couple of odd corners, probably due to the opcode space as you mention, but I would not call it a bunch of compromises.

[cmrdporcupine]

To be clear, the value (to me) of the MEGA65 is not that its CPU is superior. It's that the developers of the project are spending a lot of time on the hard part: cases, keyboards, software, and tooling.

They have GEOS running on it, they have backwards compat with the C64, they have games modified to use its extra capabilities, they have injected molded plastic cases, they have keyboards and keycaps manufactured, they have a version of the Kernal and BASIC ported and expanded, they have a user manual, etc.

And it's a) open source and b) open.

And b) is key because they're also making it possible for their hardware to be used with other FPGA bitstreams, including ones developed for the MiST/MiSTer. They demoed their hardware being used with a GBA core running a GBA game.

Further, they seem to have a CC65 backend -- I don't know of what quality -- but it seems C based software is running on it.

That makes it an interesting and more complete platform and consumer product.

And if I want a better 6502 variant on it, I could do that myself and use one of the various 65xx variant cores out there, or write my own.

EDIT: And, yeah, from my perspective from just casually reading their docs is that the CPU in the MEGA65 is an 8-bit CPU with bank switching and a couple weird 32-bit ops. It's not a 16-bit machine despite.

But the thing about the 816 is that it's only _sort of_ a machine with a 24-bit address bus. It effectively just has a bank switching mechanism, but not external. Mensch really just bolted some stuff onto the front of the 65C02 and widened the accumulator and registers by 8 bits. But the 16-bit registers are a pain. I just wish he'd added some new registers (and a way of combining them) instead of having the mode switch.

[derefr]

> Using the 65c816 keeps the 8-bit vibe

Given that a 65c816 is what’s in the SNES, I’d say it has a distinctly 16-but vibe.

Don’t get me wrong; 16-bit is my favorite! Far too few retro projects targeting early-90s era tech, or its portable early-2000s mirror.

(Imagine if you will: a plug-and-play with extended GBA-clone hardware, with a tile-and-sprite PPU that could push 1080p@60 with tons of layered graphics colour-math + windowing + DMA effects, but which explicitly didn’t offer a bitmap mode, let alone 3D acceleration; and tack on both Flash storage and Ethernet APIs, exposed through simple MMIO hardware registers to the games, with all the complexity shunted to a Wii-like IO coprocessor running its own RTOS. It’d be the perfect modern platform for retro-modern 2D JRPGs to be released on. A WonderSwan+++, per se. A PlayStation Zero.)

[TeaDude]

Don't forget the hardware divider and the DS's sound capabilities!

I don't think retro 3D is a bad idea though. Just do what the DS does (keeping it separate from 2D in the process) and I think it would probably be the greatest console ever made.

[derefr]

It’s not that 3D is un-retro; it’s just a tempting crutch. Give people “free” 3D, even a little, and they’ll use it to the exclusion of interesting 2D effects. You’ll get FF7 instead of Terranigma / Symphony of the Night — still retro, but 1. the wrong era, and 2. not taking any advantage of the distinctive cool stuff the hardware can do, instead looking like every other low-poly 3D game.

Same with bitmap modes: people will use them as a crutch by combining them with software rendering.

(Maybe this could be fixed by putting a hard cap on the 3D frame rate. Maybe even make it so low that all 3D must be precomputed, baked down into a tile map and used through sprites! Under that restriction, it could look arbitrarily good, but the results would still look more SMRPG than Goldeneye.)

And same for sufficiently-good PCM sound. The ideal early-90s sound chip would be something like 128-voice polyphonic MIDI with a huge wavetable (but banked — no playing PCM by walking wavetable samples), passed through an arbitrary FM synth module, with the whole thing being bytecode that can do loops and branching.