kolla wrote: ↑Mon Oct 12, 2020 6:34 am
rhester72 wrote: ↑Mon Oct 12, 2020 1:03 am
Oh dear word no. The ARM side is nowhere near fast enough to provide satisfaction.
Really, 800MHz Cortex-A9 is not that much different than the 900MHz Cortex-A7 in Raspberry Pi 2, right?
The Pi 2 can't do it, either. A Pi 4 is nice, a 3B+ is adequate, anything less isn't getting it done.
kolla wrote: ↑Mon Oct 12, 2020 6:34 am
It MIGHT be enough for the 020 core itself
There is no "core" when we speak of software emulation of a CPU, it's a matter of translating 68k instructions to ARM instructions. Since the beginning of MiSTer, there has been talks about replacing the CPU core on the FPGA with software emulation on the ARM, so called "hybrid emulation" - are you saying this is not worth it? And if so, based on what experience?
Sorry, we've been calling CPU emulation modules 'cores' LONG before FPGA existed. LOL Same term, different meaning. I was referring to the emulation components of the 68K only. I'm not saying it's not worth it, I'm saying it's downright impossible the way you envision it. See below.
kolla wrote: ↑Mon Oct 12, 2020 6:34 am
but certainly not for the custom chipset or beyond. It wouldn't even handle AGA.
But the custom chipset and AGA is not the topic here - like I wrote earlier, the ENTIRE POINT of using FPGA in the first place, is to put the chipset there, but once you go RTG (P96), RTA (AHI), fastest possible I/O etc. you are no longer using the chipset, so why do anything on the FPGA? You could run FS-UAE for the Workbench/productivity experience, and launch games (or legacy programs) that require the chipset on the FPGA from there.
Of *course* the custom chipset is the topic, you just don't care about it.
Unfortunately, even very base-level AmigaDOS (without Workbench) cares *VERY VERY* much about it, and quite literally can't do without it.
If what you're envisioning is an 'Amiga' without a custom chipset at all, just a UAE-like framebuffer on P96 (uaegfx) strapped to an AHI thunking layer for sound and an '020 CPU in ARM, you're not talking about an Amiga anymore, you're talking about something more of a DraCo. I'm confident it would run the 3 or so apps that ever existed for it quite nicely if someone wanted to invest the thousands of man-hours in creating such utterly purposeless emulation.
And because the custom chipset _is_ a requirement, even if you don't "want" it...you're right back where you started. You're not going to get blazing 68K performance from the ARM side, and you're still stuck with all the 'legacy' design on the FPGA side, so...you don't really get anything.
Then, of course, there's always the question of where the fast RAM lives. It'd make sense to put it on the ARM side, of course - that way the CPU and P96 can get at it readily, so you're golden, right? Oh, wait...P96 needs access to chip RAM to mirror Workbench, so now we have to port all the chip data over to the 68K side in real time for that, and at that point you're stressing the interconnect between the two sides of the DE10 to the point of breaking.
It's a nice fantasy. I cannot see a way of it working in any meaningful way in reality, nor can I see the benefit. Run WinUAE on a processor made in the last 10 years and it will smoke anything you will EVER hope to build on the DE10 Nano, and it's here and available right now.
