#Computer Cores
29996 messages · Page 30 of 30 (latest)
The Macintosh version? 👀
I tested it last night. It runs in MacOS 7.5.5 if you mute sound in the control panel, otherwise it crashes sometimes when sound/music is played.
Yes.
you can also run it on this https://github.com/bbond007/MiSTer_BasiliskII
not that easy as I peviously thought lol, will see if I can make it work tomorrow.
Had some issues building the driver, also not entirely sure I did the kernel compilation ok enough
The stability of W95 JP over standard EN one intrigues me, need to test some more apps on it. I usually test apps on NT 4 Workstation , as I found it the most stable of all (no wonder lol), but the initial boot times (plus size) are making it less desirable for a 0MHz vhd.
Still toying with the idea of having a retro-dev/apps oriented vhd, so far only tested them, and the most functional/close to this idea vhd i done is this, but can't really publish it 😄 :
But that vhd (apart from virii sources) also contains most needed assemblers/compilers/hex editors/etc for a mid 90's DOS programmer
I'm having a heck of a time getting AmiTCP to work with MiSTer using serial.device. Anyone have any configuration tips?
why not use MiamiDX?
Is MiamiDX a front end for AmiTCP basically?
No, totally different TCP/IP stack. More modern...
and there is also Roadshow which is still in active development and is the fastest
speed wise AmiTCP < MiamiDX < Roadshow, but maybe makes less difference over serial.
AmiTCP evolved into Genesis stack, but by then most people used MiamiDX
Got the Roadshow demo up in like 10 minutes lol... I am thinking this is the way to go....
The Demo only going to work for about 10 minutes though 🙂
MiamiDX has a fantastic UI
but roadshow is worth the price too
Has anyone gotten sensible soccer working, with soundblaster sound, on the 486 core?
Ive set it up, but, no music plays. (It works fine under dosbox-x and the beeper works fine in the 486 core. Just not adlib)
Commodore Plus/4
@warm rover what version do you use
just test and v1.1 only support adlib or roland for music it seem ...
What version of dos did you use?
(Thanks for checking, btw)
yes it detect sound blaster ... and that work with roland
adlib seem to be brocken somewhere because this player work before
Having a problem with the most recent amigaVision; it boots fine, but none of the buttons seem to respond in the main menu. I've tried re-mapping the controller, no luck.
Nevermind, just used the button/key swap option to nake a on the pad be the enter button.
While taking a short break from ascii/ansi games, discovered a nice fighting game, some Taiwanese DOS clone of SF, called Tough Guy from Panda Entertainment. Seems to play nice on ao486, supports SB and MIDI, controls are definable either keyboard or joystick
A 0MHz pack for Pink Gear Collection is going up:
Best way to describe this game is steampunk automata meets psychological horror. This one features a really solid fan translation, give it a look!
Lode Runner, VIC-20
Wow, the beauty of CRT displays, the glow, the curvature... it's all pure magic
How could we lose all that?
The unrelenting march of progress?
Energy efficiency, technological simplification, and the pursuit of “display-that-is-less-than-200-pounds-and-still-legible-from-5-feet-away”.
I don't miss the geometry issues and small overscan on consumer TVs.
four-sight is going up:
This is a surrealist adventure game by Synergy, best known for Gadget, Alice: An Interactive Museum, and Yellow Brick Road.
I just want to say
is there an easy way to convert all the bin and cue into a single ISO file for FM Towns?
I don’t believe the FM-TOWNS core supports disc images yet. If it supported .CHD, though, you could compress it to that
it doesn't, and cd-rom option only lists .iso
It’s not worth it, then. I don’t think the core can actually load discs, even though there’s an option for it.
Has anyone been able to get 65535 colors working under Windows 3.1 on ao486? If so, which drivers did you use?
Is there an Amiga vision manuals pack that can be downloaded from somewhere?
oookay
installed recent Amigavision, only to find out there's no option to handle Amiga CD32 games
so how do I go with that?
okay, maybe I figured it out
Yea they're a totally separate download, be sure to read the readme file included with these to make sure it's all setup right
alright!
I got the best* version of Wing Commander to go
-# massively subjective, but this is essentially a nicer amiga version
now I am interested in seeing how both Wing Commander 1 and 2 perform on ao486
both of these games have severe speed issues and don't have a frame limiter (..not by default unless you mod WCAT to WC1)
Iirc, for the 0mhz dos collection or the top300, the system is set to 30mhz for WC1
okay, when I set the cpu preset to PS/2-20mhz the game runs at correct speeds and joystick isn't freaking out
#games-and-patches message
the game according to the dev was rated for 16mhz machines
Wasn't the first Wing Commander released in 1990? Seems odd to still write a game expecting a specific CPU clock by then, that's well past the 8088 IBM era
it did
and it sure did run without a framelimiter with 5 fps on average
now I wonder how WCAT works on ao486 which does mod the game to have a game-wide sync-to-blank-based speed limiting
hooo this is interesting, does it mean that if the cpu clock is set to 90mhz, the game has more frame/is smoother?
yes
man, I can hear that image
at least the game's logic doesn't flip out when dosbox's cpu cycles are set to auto
gonna see if I am able to make a working vhd with secret missions and WCAT baked in, WCAT can launch Secret Missions 2 directly
it works! with no cpu tweaking!
okay, well, it's most stable when you put it in 486DX-33Mhz at least
but it makes me feel bad because the game got harder with better performance
and it spells doom for me in Secret Missions 1 and 2
I do have a ready .vhd if anyone wants it
Upload it to the place
done
look for Wing Commander (AllTinker Overhaul Mod), The 0MHz DOS Collection in the archival place
Abadía del Crimen, Amstrad CPC
Pro-tip, add <setname> to your .mgl file so that it has its own unique .CFG file, eg:
<mistergamedescription>
<rbf>_computer/ao486</rbf>
<setname samedir="1">0mhz_wcat</setname>
<file delay="0" type="s" index="2" path="media/wing commander 1.wcat/wing commander 1.wcat.vhd"/>
<reset delay="1"/>
</mistergamedescription>
Then AO486.CFG file in your archive will be named 0mhz_wcat.CFG and it won't override or clash with the default AO486 settings
okay, I am missing something, because when I load that it does nothing and it's all black, gonna poke through the cfg
nope, I put the 0mhz_wcat.cfg in there and it does nothing
Was that cfg in upper or lower case? it needs to be .CFG uppercase
although this is fat, so maybe it doesn't matter 🤔
I'm not sure why that doesn't work for you. I did the same for my sideline pack and I've seen other 0mhz packs do the same thing as well. Its basically as described here: https://mister-devel.github.io/MkDocs_MiSTer/advanced/mgl/#mgl-format
Only thing I can think of is to resave the settings
I am gonna sleep it over and try again
I figured it out, sorry it was a typo on my part. It should've been this
<mistergamedescription>
<rbf>_computer/ao486</rbf>
<setname same_dir="1">0mhz_wcat</setname>
<file delay="0" type="s" index="2" path="media/wing commander 1.wcat/wing commander 1.wcat.vhd"/>
<reset delay="1"/>
</mistergamedescription>
I missed a _ in same_dir
Hmm, my whole ao486 setup seems borked now. I start booting up a game then it crashes to this screen
Hmm, maybe not completely busted. Some games are working, one that worked last week doesn't. Odd
I borked my setup a while ago when moving to use the same_dir="1" attribute from using symlinks, but I didn't specify the disk in the mgl file. Loading the disk again from the menu after getting the error, saving the settings, then re-loading the mgl from the core menu fixed it in my case. AFAICT the path might have affected a CFG file.
I just rebooted my mister and now its fine 🙃
I must have done something to my setup, the output keeps going out of range on my monitor which didn't happen before. I'm unemployed from tomorrow so I'll have to dig into it then
thanks! I've updated the archived file with my config now.
kinda wanna do The Last Dynasty by Cocktel Vision, which is a weird offbeat space sim and adventure game
but I have no idea how to autoexecute the game after loading Win 3.11, drag it to autostart, maybe?
Create a shortcut in Startup group menu (File - New - browse to your game or simply drag the icon in the startup group iirc)
yeah, that works!
serves me right, setting up the game I've previously installed on dosbox..
looks like I am going to cannibalize a win3.1 pre-install
There's dos-win31-vhd-templates-misterfpga-ao486 that could possibly be found in an archive on the internet
that's splendid, thanks
is there a joystick driver I can use?
the ibmjoy.zip that I've used to install on dosbox is not giving me a good range
Yeah I'm not really sure, maybe something from here?
https://vogonsdrivers.com/index.php?catid=67
mmhhhmmm
I might have to release the game with borked joystick support
the weird thing is that I have that really tight range on windows
when I test the joystick in dos mode using calibration tools it's fine
by the by, win3.1 games load fine if you set this in auto exec
WIN GAME.EXE
and it's live!
to give you an idea what kind of tone this 2 million production has:
it's by developers of Gobliiins and Urban Runner
I am getting a kick out of this
let's see if I can cook up one more
and how terribly it runs
you know, growing up, I don't think I ever had to edit XMS memory
unless my dad literally did it for me while I wasn't looking
well, CyberMage runs
..it sure can run
would you believe it's actually looking like it almost reaches its supposed speeds?
oh, it is literally this game's 31st anniversary, so what the hell, it's on
Wing Commander 1-2 had a Windows version called Kilrathi Saga which was expensive for a while and did do frame limiting I believe
it actually did but the games run too fast
the only way I can reach comfortable speeds is to frame limiting them to 15 fps in dgvoodoo2
..now I wanna see how Terminal Velocity runs
damn, I got beaten
anything past 1993 with heavy 3D stuff is a major crapshoot valley with occasional weird exceptions like Dark Forces which runs at very good speeds, with the magic of multithreading
ahha!
there's another childhood DOS game I can try and share!
alright, got that sorted out
but I wonder if I even can install Gravis Ultrasound in AO486
oh, it's not merged
best not to test the luck
You can set it load the gus core in the MGL
I don't think it will ever be merged because not enough space on the DE10
ouch
I could be wrong but that's the impression I got
But like not a big deal imo if you just use the MGL to load the GUS core anyways
anyhoo, here's another release from me, a very odd childhood game called Radix: Beyond the Void
a game that dares to asks life's most important questions
What if Doom.. was Descent?
-# or is it "What if Descent was Doom?"
runs pretty okay on the core, I've bundled both 1.1 registered and 2.0 rerelease, the latter of which applies automatic graphic detail scaling (as well as enhanced WASD controls that don't keep propelling the ship forward)
I loved Radix back in the day
I am looking for keycaps for USB mechanical keyboards that has the same keys/symbols of the C64 keyboard so it matches the positional keys expected by the C64 core. Does any one know if such a thing exist?
This PR would also solve my issue. No idea why sorg is making a fuss about this awesome feature: https://github.com/MiSTer-devel/C64_MiSTer/pull/130
https://www.cbmstuff.com/ would be the place for that but they seem to have a pause on orders atm
I can't even see their products. Do they have those keycaps?
Can someone here who's on good terms with sorg please talk to him about this feature? I too was surprised by his reaction to it tbh
like I said, its usually the place but they have orders on pause
Kit is sorg's BFF AFAIK
Taking order is one thing. Showing the products is another. But thanks, I will keep an eye!
A bit more info on the caps https://breadbox64.com/new-keycaps/
I am assuming those keycaps are compatible with modern USB keyboards, right?
Actually I'm not 100% sure if they do sorry 🤔
Actually they were meant to go with the Blingboard64 which uses cherry mx style switches
Not as nice but there are some vendors around who sell stickers
I will take blank keys + stickers.
Looks like a pretty cheap way to do it: https://www.amazon.com/Commodore-Keyboard-Stickers-Shortcut-Self-Adheisive/dp/B0DRT3DWPW
There's tonnes of hits on google for C64 key stickers
Commodore 64 Self-Adhesive Black Keyboard Stickers Transform your keyboard with Commodore 64's customizable stickers, crafted to meet your specific needs and preferences in just minutes. Ideal for both beginners and professionals, these stickers are a must-have for anyone using Commodore 64 softw...
There is a patch made on vogons, using the code from Secret Missions 2 expansion, so now dos version is frame limited. A must use.
I believe WCAT is more extensive than that
Looking for a few testers who will be spending time with computer cores this weekend or over the next week. Ideally ao486 or minimig, but Apple IIgs, X68000, PC98 and Macintosh Plus work too. Drop me a line please, DMs are open 🙏
OS worth trying: Dynix. Some old library system
I think you're confusing Dynix OS (a BSD variant) with the Dynix library catalog software, which ran on top of PickOS and some other platforms
I honestly heard about PickOS and assumed Dynix was an implementation
If you want to play around with Pick OS, you totally can: #1047332497492553799 message and #1047332497492553799 message
I use arch btw
||Arch deez nuts||
working on the games for the ao486 made me appreciate the fact I am careful with low-spec optimizations for games instead of just messing the cycles in dosbox
Bro come at me, I’m the Linux King
Out of curiosity
What Linux distro would work on ao486?
..and would I be able to run Doom on that?
- many. (eg. #1047332497492553799 message)
- yup (eg. #1047332497492553799 message)
Didn't they JUST remove support for 486s with no FPUs in the latest kernel?
They sure did
yeah, they did
I find it insane that they also removed support for GTX 1000 era GPUs, like, you are JUST now removing support for CPUs from 1989... but also removing support for GPUs from 2012?
still, 5.1x is ok for whatever needs on 486, (see Gentoo or 486Linux)
Yeah, I doubt many people using a 486 with no FPU for whatever reason are installing the latest linux kernel on it
MiSTerFPGA comunity 😄
So technically I could run relatively recent arch build before the 486 abandonment but good luck with the desktop environment
yeap, CLI is your trusty friend
or go 2.x kernel era and it should work ok-ish, depending on WM chosen
Debian Etch with AmiWM is pretty decent for a full GUI, though the best in terms of speed is TinyWM from SkatOS floppy
https://ocawesome101.github.io/486-linux.html (I know I linked it before, but relevant on kernel recompilation to run on ao486)
RenderOrgan is going up:
This one is probably Synergy's most obscure adventure game, which is about repairing a giant, soul-purifying surrealist organ which floats in the sky. As such, it's all about maneuvering this strange machine and learning its mechanics, which are all controlled by different devices around its interior.
When they remove these things from the kernel, doesnt it just mean you can still add it as drivers or something? Or do you need to make a custom kernel to get that support? Or just use one of the old ones I guess
Depends on the level of refactoring. For cpu support I would imagine there are a lot of ifdefs they are going to rip out so it would be harder to support long term as the kernel changes.
I don’t know how they modularize it though
I wonder - what is bottlenecking performance in Crusader: No Remorse? The game is rated for 486DX2 33MHz but it just struggles in ao486
would running the music and sound through Gravis Ultrasound boost the perf or that doesn't matter for FPGA?
the sequel however runs so much better so maybe it is meant for Pentium overhead and they optimized the engine
I can't see it helping but its worth trying to see
yeah but that doesn't beat 640x480 SVGA display
"Key Console EnhancementsAdded Maneuvers: The console versions added a forward roll, a movement ability that was originally only introduced in the PC sequel, Crusader: No Regret.Engine Improvements: The PlayStation version introduced smooth scrolling terrain and transparent walls, which helped keep the character visible behind environmental objects—a feature lacking in the static-screen PC original.Improved FMVs: While the gameplay resolution was lowered (from 640x480 to 320x240), the full-motion video (FMV) cutscenes were actually of a higher playback quality on consoles compared to many contemporary PC setups.Audio Upgrades: The sound system was updated to 16-bit to improve the delivery of the electronic soundtrack and sound effects.New Content: The console ports included new weapons (like the JL-2 Purifier mass driver) and several additional gruesome death animations for enemies."
Still looking for a few testers for https://github.com/treyturner/MiSTer-deskflow-dist. Required setup: a computer/laptop with mouse, keyboard, and dedicated display in the same room and on the same network as a MiSTer with a dedicated display. Existing Deskflow users would be ideal.
And if we have two mister on the same room (though one is behind wifi, but i can plug ethernet if needed), does it work ? 😄
Ok i'm on it...
No, you need a Windows/Mac/Linux computer to run the server. The MiSTer port is client-only.
I would guess its the DX instructions? we only have SX on the core...
i.e. no hardware FPU
ohhh
little progress
looks good
**MAYBE **(a huge one) using Q87 FPU emulator could improve performance. Slim chances, but you can try that (https://misterfpga.org/viewtopic.php?t=6639)
pretty sure this has been attempted and that it was too slow.
I think Quake was made to run with FPU emulation and it would run, but at 1fps or less
Some (very few like FX Fighter, Blue Ice and some other) games run ok with it, but yeah, Quake/Tomb Raider/WipeOut/etc are just 1-2 FPS
and in this case it sounds like its not something that stops the game from running, but I doubt FPU emulation would improve the games performance
Agree, that's why I said there are slim chances. Q87 is not an universal panacea 🙂
I considered whether it might be possible to somehow take the ARM FPU on the MiSTer and have if use some API layer to function as the FPU for the AO486 core but it would probably have latency issues and, most things that need FPU also need 66mhz or more. So its not much gained from it. Still, it would be nice to have native FPU.
If it was FPU with a decent speed, it would probably also make Duke Nukem 3D run better. I think it uses FPU for its slopes, if FPU is available
Reasoning for recommending Wesley to try a test with Q87 is that in some rare instances the program COULD run on SX (but very slowly as it's doing the math on CPU) like Fractint or Arachne DOS web browser (and some others i think) but get a huge bonus of performance with fpu emulator
#1047332497492553799 message -- there's a long conversation/tests/findings when I played around with Q87
This is a project that I’ve been working on for the past eight months now, and I’m finally just about done! It’s a fully-functional Apple Lisa computer inside of an FPGA, complete with a bunch of nice modernizations like USB peripherals, HDMI audio and video output, onboard hard disk (and maybe eventually floppy disk) emulation, and USB t...
👀
this is after it modified the autoexec and showed me the registration form
The "installation" is just copying the files in C:\Q87 and adds the path to autoexec.bat.
Next (assuming you have already functioning EMM386/QEMM, this is important ! ), launch the Q87.exe then the game and see if there's any difference
Shareware version will only last 20 mins, there are means to patch it (if interested, DM me, for obvious reasons I cannot discuss it here. It takes like 4-5 mins at best) or just get a patched one from various 0MHZ games that are using it (FX Fighter, Blue Ice comes to mind)
alright, I just need to mention that MT-32 Dark Forces fresh from 0MHz pack does not work (dos4gw error 1313: can't resolve external references)
okay, just launched Q87.EXE and then the game, no difference, it's still slow
gonna poke a bit more
yeah.. was afraid so 🙁 ... maybe UNIVBE stuff might enhance it ?
have a read here https://misterfpga.org/viewtopic.php?t=2780
I dunno if this game uses VESA 2.0 though, don't seem to find it on any list, so might be as well a goose chase
you need a certain config for univbe to work, as I said, read carefully the forum post I linked, especially this post: https://misterfpga.org/viewtopic.php?p=46566#p46566
And again, this game might not support it at all, I have no clue on this game/never played it
Got two packs by thera34 going up.
First is Tough Guy, a Taiwanese fighting game by the developers of Sango Fighter 1 & 2:
This one's a one-on-one fighting game, and probably one of the best original Street Fighter II-likes I've seen on the PC.
Next up is White Album, developed by Leaf Soft in 1998:
Don't know too much about this one - it's a Japanese-language visual novel, so fair warning that there's a pretty significant language barrier and I haven't been able to fully test it.
However, it's supposed to be pretty good, and is a quite solid showcase of what the core can do when it comes to high-color graphics. The game got an English language remaster several years back, but it apparently cut content from this original version.
Also going up today is a fixed version of the game I.M. Meen:
This version fixes a bug which would previously cause the game to crash when attacking enemies.
??? Was that bug in the original? That would have practically made the game unplayable
Not a bug with the game - it was a bug with the previous 0MHz pack. Turns out, it was a CONFIG.SYS issue.
ah
Finally for tonight, here's Museum of Anything Goes, released in 1995 by Wayzata Technologies:
This one is like if you rolled up the entire 1995 multimedia aesthetic into one game - it's weird as hell in the best possible way, and I can't recommend it highly enough.
So... Hacker News linked to an Apple Lisa on FPGA. It's open source--could it be converted to Mister?
I think the general impression is that it’ll take some work, but it could be ported. There were some discussions between the developer & the MiSTer community at one point to that effect.
I only know of that game becaus e of Vinesauce. It felt to me like the devs just wanted to take advantage of how much random animations and sounds they could fit on a CD and just.... made anything to fill it up with whatever. Like a kid alone in a candy store XD Except here is was going fro the space contraints of floppies to CDs in the 90s.
So I'm wondering if it would be possible to get SNAC going to the DB9 port (maybe even re-use a Genesis SNAC adapter) on the Minimig core so that one can connect... TWO mice, for two player Lemmings
@void belfry I am looking through history and I don’t see any mention of coherent os…have you tried it?
Think I did at some point but can't really remember what went wrong
Maybe should retry it, been some changes since then (3-4 years ago think it was when i tried if my memory doesn't fail me lol)
The question is which player gets stuck with the crusty old amiga ball mouse...
I use a Sordan "tank mouse" and MicroTOM with my real A1200 so I suppose you could use one of those...
In theory, the pinout is all that is different from a bog standard PC serial mouse... maybe could even use both
Amiga can't work with a PS2 or serial mouse without active converter... its not serial
Heh, you are indeed correct. I didn't know it was a different protocol
The cursed solution: https://aminet.net/package/driver/input/SerMouse221
the converters are pretty cheap, but watch out about the "USB" ones because some only work with USB mice that fall-back to serial mode with passive converter.
software solution for serial mouse won't work with games unless they are system friendly
like this adpter for example is really not going to work with many mice https://www.ebay.com/itm/114491223022
then the next problem you run into is sensitivity when you plug your 1600dpi laser gaming mouse in 🙂
Just received this along with some keycaps
I was happily changing the keycaps when I got to this key:
The fucking core doesn't give me & when I press SHIFT + 6
So basically sorg mapping doesn't make fucking sense
He could at least merge that fucking PR since that fixes everything
Damn it
Thanks for letting me vent. I feel much better now! 😌
The stickers are great quality though
I wish I could at least use them 😒
Omg what is this
I guess it might be worth commenting in that PR (or maybe a separate issue) saying that the mapping is wrong?
I will but I need to calm down first otherwise I might hurt sorg’s feelings 
haha, I get you. His responses to that PR is probably up there in my top 5 not great open source interactions. Its pretty close to some of the stuff I saw with the old RPM maintainer and Linus Torvalds when he's getting spicy
𝑀 𝓊 𝓁 𝓉 𝒾 𝓂 𝑒 𝒹 𝒾 𝒶
https://buyee.jp/mercari/item/2JPBfSSsAg6P8DkkufYt9S @lunar trellis @granite umbra a FMR-50 computer seems to be auctioned on JPN Mercari, may be time for someone to purchase it and dump the FMR-50 bios
This could be it!
It's been a while since any updates dropped for the fmr-50, I wonder if it's still being worked on
Last update was mid January
As well it can be found on FromJapan auctioning site at very reasonable prices, put some examples a while ago too
But as @bitter yoke said, last update was in January, plus puu-san is also on FM-Towns/X68k/etc and working at his own pace,so it might take a while until will have to play around with a functioning core
Where is the website to those updates?
The one and only 🙂
but aren’t there two different cores for Fujitsu
One called TOWNS for the Marty and then a separate one for the FMR-50 specifically
FM Towns is a different computer than FMR-50 (even if they share some similarities at some points). Marty is the console version of an FM-Towns so to speak
Straight out of my old 272 megs Seagate HDD that was in my first PC.
Not that it wasn't sourced on any abandonware sites, but for pure nostalgy of passing some bits from 90's BBS/early internet sites (don't really recall where I got it back then lol) to my old HDD then on an image I made of it at some point to preserve my old stuff until now when it lives again on MiSTer 🙂
Noted to add 😄
I played a lot of angband, but I am not sure I would be able to get back into it now
Rogue 2 on the other hand throws this error at any first fight
hmm, hit it so hard it divided by 0?
How many marijuanna's did you take?
just two at some parties and apparently is occasional 😄 Booze is implicit
Anyway, it's basically a "management" game, you also need to write songs/play gigs and tours, go to TV programs , promote and so on, while ofc you are the Rockstar and have to also party as heck
Rock Star Ate My Hamster is a management strategy computer game developed by Codemasters in 1988 and originally released on their full-price Gold label for the Amstrad CPC, ZX Spectrum, Commodore 64, Amiga and Atari ST. The game was written by Colin Jones, later to become known as author/publisher Colin Bradshaw-Jones.
The name of the game was i...
Both
yup, same idea but with ansi 😄
Rodeo Poker is a nice twist of Tetris, you need to form different winning combos per line from the falling cards
Possibly the most complex one so far at least in terms of graphics would be Reaping the Dungeon (still have some more titles to test)
I have to wonder how many of these people would have expected their game to be run in the year 2026
interesting, there are references to a free version of that one called dungeon rogue, but I think you need to email him asking for it
Reminded me of the old Drug Wars game, played the crap out of that as a kid
Oh wow, I forgot about Drug Wars... What core(s) can that be played on?
Also how about Dope Wars?
Oh they're pretty much the same game
Funny at the time I only ever knew about Dope Wars and didn't know until years later about Drug Wars
Should just work on dos, I remember it being a bbs door game but those can just be run separately like any dos app
I think I meant dope wars now that you say that
Maybe drug wars was the stand alone and dope wars was the bbs version
coworker of mine had Dope Wars port on android for years. really nice port, he hired real artist etc.
maybe its still on the store
It was fun to play that on the PalmPilot
New games for ZX Spectrum keep coming
CatGun: Victory Mission G is a promising manga-style PC shooter that was sadly left unfinished. Two early demos exist with playable levels, and there are signs of a more complete version still missing. Can it be found?
https://www.gamesthatwerent.com/2026/05/catgun-victory-mission-g/
Post thanks to Toppy!
Subscribe to our channel
https://www...
Holy crap, that looks and sounds INCREDIBLE!
Like a DOS/Amiga fever dream come true...
Update (screenshot) for CPC+
https://misterfpga.org/viewtopic.php?p=107182#p107182
Hey this looks rad is it C64?
no, ao486
This makes it all the more surprising that they didn’t release any PC stuff under the Vision Factory label.
I feel like Apprentice would’ve done well with a PC port.
How could a C64 game look like that? I mean...it's like 50x what a C64 can do in every possible way 😄
Just glanced at a screenshot in the computer chat after seeing the rock star ate my hamster box 😭 I've got c64 on my mind I guess
A few too many colors in the ui ha
That game does look cool though
It is gonna be part of an ascii/ansi DOS game pack that I am currently working on (albeit with some delays as RL stuff intervenes, or some other rabbit's hole I get sucked into)
Oh that's awesome
Little sneak peak, working on this for PicoMEM, PicoGUS, and PICOIDE but Mister computer cores benefit from all this work if there is an exodos target (eXoDOS 6, eXoWin, eXoAppleIIGS (future)
Working on a couple of utilities prepping for the PicoIDE, one is a VHD maker utility that will auto build VHDs of any type, FAT16/FAT32, auto build the VHD with FreeDOS or MS-DOS 7.1 with utilities, a dos frontend (MyMenu) if a game pack is built using the exodosconverter etc. Expanding on the work from years ago with the Exodos game converter for the MiSTer FPGA I've updated it for eXoDOS version 6 and automated the vhd creation, batch scripts that will mount the CDs automatically when you launch the game selected, Audio selection (Sound Blaster, GUS, etc, same setup as exodos with it's front end just native to dos. There is a GUI verison for windows as well. Still lots of testing for when I get my PicoIDE but getting very close. Everything seems to work with Mister and it's auto cdmounting scripts, sound selection, rebooting to mount a different VHD if needed. Selfishly this just shortens my build process for game packs and makes it easy to play them when you have access to something that can mount VHDs and CDs through automated dos tools. Should be ready by launch of the PicoIDE. Hoping someone else hasn't already done this.
Features:
VHDMount / VHD workflow capabilities
Auto-create ao486-ready/picoIDE VHD packs during conversions
Build one isolated VHD output pack per conversion run
Support single-game autoboot mode and multi-game MyMenu mode
Stage and inject game/menu/support payloads into VHD images
Keep CD/floppy/bootdisk media external beside the VHD (for MiSTeR auto-mount behavior)
Preserve launcher/media command behavior used by generated game scripts
Support FAT16 template-based VHD creation
Support FAT32 template-based VHD creation
Prefer 450M-DOS71 as the default base when FAT32 pathing is needed
Auto-size VHD targets from payload from game selection + FAT/cluster overhead + growth buffer
Auto-expand templates when required space exceeds template capacity
Retry with larger growth targets if copy-time disk-full occurs
Use a FAT32 rebuild expansion path when direct resize paths are unreliable
Linux-side VHD read/write handling is supported through image tooling (mount-like operations without manual mountpoint management)
ExoDOSConverter capabilities
Convert multiple source sets: eXoDOS, eXoWin3x, eXoAppleIIGS
Output to multiple targets: MiSTeR FPGA , Native PC/PicoIDE, Batocera, Recalbox, Retropie, Retrobat, Emuelec, OpenDingux variants,
Provide both desktop GUI and native Linux terminal TUI workflows
Share one backend engine across GUI and TUI for feature parity
Persist user settings/config between runs
Validate collection paths and normalize Linux/Windows-style path inputs
Handle very large game lists with filtering and selection workflows
Load and save custom game selection lists
Generate converted game payloads plus metadata/art/manual outputs
Apply genre mapping and optional genre-folder organization
Apply platform-specific launch/config rewrites and controller mapping options
Support optional download-on-demand for missing game packages
Use per-build output scoping so each run writes into a new child build folder
Include MiSTeR-specific packaging + ao486 VHD generation in the same conversion pipeline
Auto batch creation mounting discs with PIEDCTL (Classic PC) and IMGSET (MiSTer)
Improve Linux extraction robustness for case/structure mismatches to avoid conversion crashes
Just added IMG support as well to auto create system disks floppies for various dos formats and add just the core dos utilities to it.
MiSTer benefits from all of this as that work is already done, so you can just simply make collections or single game vhds with coresponding cd/floppy if they are needed and drop them on the Mister, if a collection is selected then it will add either TDL or MyMenu depending on what the user wants. Bringing all the python code up to spec with eXoDOS 6 latest and there is an Arch Linux TUI or the same setup for a Windows box.
Wall of text, sorry. Will opensource all of this, the vhd/img maker alone has been so nice to have to quickly create compatible disks from the different OS drama, between msdos33/331/5/622/7/Freedos
you will just have to supply the install disks for the DOS version you want in the folder like from winworld and it will do the rest, it has an auto pull of the latest version of FreeDOS and that is the default
fixed a lot of bad code and bad ideas from years ago to make it feel more like 0mhz setups
you can mount the VHDs and IMGs directly from the interface in either Linux or Windows and it will auto bring it up in the event you want to drag stuff into it or change anything before moving to the mister. If any other ideas are out there let me know while I am motivated. 😄 I also have Genres, years, and categories like (CGA+PCJr) so it will quickly filter down the game list from the daunting 7000+ to a smaller list of games if targeting a certain set of games, as long as it is in the exodos metadata. Ok wall of text over
Wont be releasing my own packs but trying to make it super easy for others to mess and create there own without knowing anything
I want to test the Goldorak fangame!
Great! When the toll will be available? Is it possible to use it in MacOS?
I don't have a mac to even begin to build that version, but once I am through testing I will open it up on github. vhdmaker has a command line version that is used by the exodosconverter to keep them seperate or if someone just wants something easy to create bootable vhd or img of any size with whatever os system type on it. I expect you could pass in the code to something like opencode/codex and have it build a MacOS version. If python GUIs work on mac then it might just work out of the box
does the current exodosconverter function on macos with python? I expect it will at least need all the path improvements I did as I current one out there looks for E:\exodos or whatever
looks like it will work but others will need to test as all my systems are linux these days: https://github.com/Voljega/ExoDOSConverter#linuxmacos-installation-and-execution-
50X? ? Looks like 80 column mode with custom character font. C=64 can easily do that in 40 columns. 1 MHz 6510 and a 4.77 MHz 8088 have nearly identical raw performance.
Yeah, that's more ZX81 level of performance
hi there i have a strange issue that happend on amigavision cores
i tried this morning to paly at the AGS core but there're some INI errors shown up
it says :
ini error vrr min framerate unknown option
vrr max framerate unknown option
i tried to change the VRR settings on my INI file but nothing happends...the thing is that only comes up when i try to use 500HD,600 HD or AGS core.
i'm lost i dunno what to do
please if anyone knows how to solve this
thanks a lot to everyone
You need to comment out the two lines on your MiSTer.ini file.
We don't need them anymore.
The Jeff Green Duology is going up, comprising Midnight Stranger (1994) and MODE (1996):
These are a pair of games featuring a unique "mood bar" mechanic, which allows you to give variable emotional responses to any character you talk to. Best way to describe them is social interaction simulators - philosophical, personal, and often quite spicy at times, there’s very little else quite like these two point & clicks.
Ah! Sorry! I though he meant https://www.youtube.com/watch?v=fiHsAwwhHJ4
CatGun: Victory Mission G is a promising manga-style PC shooter that was sadly left unfinished. Two early demos exist with playable levels, and there are signs of a more complete version still missing. Can it be found?
https://www.gamesthatwerent.com/2026/05/catgun-victory-mission-g/
Post thanks to Toppy!
Subscribe to our channel
https://www...
Yeah, that is very graphically impressive! a bit busy for my taste like CD32 version of Zool on Amiga...
Luc Verhulst did some pretty great work. Outside of this, he also did the art for several CD-i titles - including Lucky Luke, The Apprentice, and Dimo's Quest, alongside others.
Goosebumps: Escape from HorrorLand is going up:
This is a multi-genre adventure game, and the first game released by Dreamworks Interactive. It features a neat panoramic navigation system, tons of great visuals that are pretty impressive for the core, and quite a few surprises in the gameplay department. Give it a look, especially if you're a fan of the show or original books!
The Psychotron, an incredibly cheesy point & click spy thriller from 1994 is going up:
Interestingly, this game does something I've never seen in a point-and-click before - it actually has a multiplayer mode, where players are solving different parts of the game's puzzles in turns to try and reach the ending first. It's optional, and you can just play it like a traditional single-player adventure game if you want to, but it's a unique twist on the formula.
Is that Tony Soprano & friends down there at the bottom??
I think they call him like Spinozzi or something like that, but there are indeed random gangsters
Huge update for the WIP Amstrad CPC 6128+ / GX4000 core: https://misterfpga.org/viewtopic.php?p=107212#p107212
Is this purely python? LMK if you'd like some help with setting up environments or builds using pixi, so you can at least do package checks for validity with other platforms.
yes python on both, will ping you when i get closer
sounds good, just when you ping me remind me what project you're working on lol
Another linux distro for @void belfry to try out https://github.com/SharktasticA/SHORK-486
A Linux distribution for 486 and Pentium (P5) vintage PCs - SharktasticA/SHORK-486
That looks pretty cool, and apparently is based on floppinux/gray386linux/486linux of ocawesome, which already tested pretty good on ao486
Will try to build it after work, need to get back to WSL9x as well
I'm too impatient lol :)) WIP (now it compiles openssl ❤️ )
will see in few hrs (when I get back to MiSTer) if it works 🙂
mkey, it doesn't have pppd, so back to drawing board, a.k.a compiling that, adding to rootfs, re-build
ok, recompiled kernel with ppp support, compiled and added pppd binaries, removed ethernet/pci/scsi/other stuff, added SB16 support and rebuild the whole image, will see in ~3hrs if it boots on ao486, preliminary qemu/86box tests seemed ok so far
IF it works properly, gonna need more apps (MC, BitchX, maybe SynchronetBBS terminal with ncurses and all/ others) . Was thinking to try to compile them directly on ao486 to see if it can be done (as i done on older Debian/NetBSD), but don't think I will have enough space on this image lol (gcc is already installed)
And it frigging works and it is the fastest linux booting on ao486 😄 PPP works as well lol
Thanks @hollow estuary , you just discovered the 19th entry on my list (will update it shortly) 🍻
wow, nice
#1047332497492553799 message --- now officially to character limit😄
am I reading that correctly…it is running the latest kernel?
EXACTLY !!!
when your fpga core is more secure than the mister on which it is running 😄
I changed some things from the default kernel config author made *by adding ppp/SB16, removing scsi/paralel port/other unused crap), other than that it just worked
nice. Did it come with X?
even pppd is newer 😄
yes, it has some X (TinyX and TWM) implementation but :
might have been missing smth on kernel config
haha, either way, thats a fun new one…now you can answer the question of “can the ao486 core build the ao486 core” 😄
Wow, that's something! I have always wanted to go back to DOS boot times on my GNU/Linux systems, and indeed I have succeeded, but that's on computers orders of magnitude what ao486 core implements!
yup, it boots in seconds, absolutely amazing
now added make as well so will try to compile ncurses directly on it, though I might have done a bad mistake having the VHD sized to only 500 megs (only 85 megs free)
Yup. been there when building disk images for many systems 😄
I normally follow your OS-testing saga on the fence, silently, but now I wonder: Did you get major problems with ao486 memory/CPU? Things to report in order to improve the core, etc
For DOS it is, as of today's unstable, very compatible with anything one can reasonably expect to run a 486 computer
But for other operating systems, things may be different
more with the IDE and floppy controller, not many unix/linux variants can work because of that
Ah, that's unexpected. Have you reported those? Humply is always improving the core
Yes, many of my sshots on different OS tests here contain different IDE related errors
Started to configure/compile ncurses directly on core, will see how much time it takes to do it.
Ofc, when I will have a full list of things to add, will do it via WSL2 and either re-build the VHD or simply add them via ftp (which would make more sense as I already got ppp working, have wget and other stuff ready as well)
Another major point would be to make the GUI work, probably missed FB/VESA support stuff
Will do some more digging, as I am not some Linux Guru/King, as @copper slate is with his Arch (which btw can be used as support platform to build SHORK-486 images) 😛 /jk
Sry ASCII/ANSI DOS games pack, you'll have to wait some more days, Linux rabbit holes take priority (they have sudo after all)
on the cpu/mem/cache side, the only notable OS with issues is OS/2 starting from 2.1 which needs L1 cache turned off (https://misterfpga.org/viewtopic.php?p=45942#p45942)
at some point will have to do a SHORK-486 floppy as well https://github.com/SharktasticA/SHORK-DISKETTE for completeness
Sry for rambling a lot, but never been so hyped lol 🙂 This (from the very first tests) might prove to be the best Linux (modern as well) that runs on ao486. With some custom stuff added, it could be the very ao486_Linux we can all enjoy (sure, probably not so much on the GUI side, but otherwise... sky is the limit)
I'm enjoying it, but just quiet and reading. Will be interesting to see what they do with Kernel 7.1 or are they just stuck now on 7.0.x
they are removing 486 support right? I thought I read that
I honnestly thought they did already lol
i see..
anyway, as @hollow ice said, running a newer kernel than the MiSTer ARM is hillarious AF:
/root# uname -a Linux MiSTer 5.15.1-MiSTer #1 SMP Wed Apr 2 20:01:54 CST 2025 armv7l GNU/Linux
Speaking of support removed, IPX one was dropped since kernel 5.15 and as well on pppd 2.5.0 🙁 it can be patched back though, at least on 5.15.x kernel (we still use pppd 2.4.8, so adding IPX support is no issue)
I use arch btw
McDonalds dosen't count :p
ncurses needs sort which is missing, coreutils is 14 megs just the tar.gz srcs, which when unpacked will transform in more megs than I provisioned :)) Clearly need a larger VHD lol
not bad for good ol' ao486, eh ?
You technically have 3 other hard drive slots
yeah, I can do some storage vhds and mount them, no probs, but for today just want to compile smth complex enough
Just create a massive one for Src and builds so you can keep the main images smallish
Then noted for later to test floppy support, fb/vesa stuff and add other needed packages from the get-go (aka compile stuff on wsl and move to shork)
this will be difficult 😄
and yes, I could have made a bigger vhd since I am babbling about , but where's the fun without some self induced pain ?
And I will for future tests, but still, 500 megs for full default install is not enough, leaves just 85 megs free to play around anyway.
Will test smaller builds as well, but for today I'll just go with the current
picture instead of 1k words 🙂
so now, anytime you absolutely need to ssh into something from the mister, use the ao486 core instead? 😉
absolutely 😄
Now that I am procrastinating with compiling stuff, we need a proper name for a future ao486 SHORK-based "distro", eh ? Any thoughts ?
MiSTer-SHORK-du-du-du-du-dee-du
First I thought to 1337 Linux, but it is taken iirc :))) I am so not good on picking names
plus no ref to MiSter/ao486
Meanwhile cross-checking with notes on Slack 12/Debian 4 stuff that I installed to have a better picture on packages/dependencies I need/wanna add
nano is a must 😄
If you get x, try zed
Oh, for X, there will be a whole chapter of apps to be tested 😄 first ever will be xeyes ofc
You got dosbox working before, right?
Wonder how well shork would work on shork
ahahaha, inception all the way 😄
That would be insane and I love the idea !
meanwhile ao486 is still crunching the config for coreutils, I might be able to pull it off afterwards with just selective make on sort, even in the tiny space I have left on VHD
Ultimate stress test (providing it passes the more "simpler" ones) would be to do a kernel recompiling on ao486, and for sure I'll be doing it (thus locking away my MiSTer for a few days/weeks)
Anybody knows if @lost nest is progressing on the Thomson MO/TO core ? I just can't wait for TO8 on Mister - nostalgia on such a practical and faithful machine...
I had been implementing the thomson ef graphics chip from the data sheet for the PET core….but if grab is doing an implementation I would defer to his
speaking of, need to make it work also the other way around, meaning sshd listening on ao486 side for easier remote management (same I did with Slackware/Debian Etch)
would solve the scaling issues
As I suspected, dropbear ssh server is not built by default, so need to compile it as well
In any case, will build also a larger vhd (800 megs), add sshd, ncurses, BitchX, mc, syncterm, coreutils, etc and their dependencies
For now just brainstorming with myself (and bit with Claude to summarize/mini-scripting), but once I am done with all tests/apps added/functional (to whatever extent is possible) GUI/etc, besides providing a ready-to-use vhd, also thinking on forking SHORK-486 on github with all these changes/additions, so anyone can build his/her own image
But first things first, adding stuff to make it more functional
Small updates:
- built an 800 megs image
- added pppd/make/sort/some other coreutils
Now adding/testing ncurses/glib2/mc/bitchx/sshd/syncterm/perl/etc, so hopefully by evening will have a more complete VHD to play around.
Sharp X1 POC (not running fdc or tape)
https://misterfpga.org/viewtopic.php?p=107335#p107335
Well, there are still many problems. Now I'm trying to debug the TO8/9+ floppy controller "THMFC1". It's a custom chip with barely any documentation, I have to dig into reverse-engineered emulator source code.
Got a rare mid-week game going up, and this one is a personal favorite.
Beyond the Wall of Stars is going up, from 1992:
This is a game written and created by R.A. Montgomery, best known for writing most of the Choose Your Own Adventure books.
I’ve been experimenting with this driver on and off. So far, I haven’t been able to get it to work in ao486 - if someone is able to run it in high color mode (65535+ colors) on the core, please message me and let me know
I haven't posted about this project I'm working on in the discord, but I've been hard at work (along with CLAUDE) building a filesystem manipulation tool called "Rusty Backup". My goal is for this to become the swiss army knife for all things hard disk/image related. I am finalizing basic Amiga and MSX filesystem support in the next hour or so.
Have a look at the project repo https://github.com/danifunker/rusty-backup
If you're a core developer and want this app to support your filesystem, please DM me so I can prioritize those filesystems to support your project. Since I'm also working on some of those Mac cores, it of course it fully supports HFS and HFS+, along with apple .DSK images.
@void belfry which filesystem are you using for your image? I think you can inject files directly into ext2/3/4 using rusty backup. It's been a while since I've supported that
SharpX1 can load FDC
https://misterfpga.org/viewtopic.php?p=107407#p107407
Toy Top
EXT4.
For the momment I am compiling stuff in a staging folder, move them to root and simply rebuild the image (without busybox/kernel as those are already done).
After coffee will just setup a simple web/ftp local server and will wget the binaries/libs on the core via PPP link from my WSL2 instance
BitchX functioning now, connected to ##MiSTerFPGA channel, so ncurses also work ok
Next on my list will be glib2 and mc
I cant seem to load any .d64 files on latest C64 core. and in the bottom right #8 18 is flashing. Any suggestions?
its on the unstable that .d64 doesnt load. On latest official core it loads
got Dropbear ssh server working, moving on 😄
Haha
Wow all kinds of fun stuff happening today
That looks like broken ansi?
no, that is ascii, nothing ansi there
this one was: #1047332497492553799 message - 2nd sshot
I mean all the 1;34m
yup, solved those ny baking TERM=xterm-256color
Oh. I replied to the wrong image. Sorry, my thumbs are too big for the phone interface 😀
that's why i only use phone just for sms and calls 😄 (and those 2FA ofc)
ASCII stupid question, get a stupid ANSI I guess
go go go 😎
Small update for today on the SHORK-486 adventure:
- got missing coreutils correctly, along with ncurses
- added and solved BitchX garbled stuff by setting TERM variable to xterm
- added bash and set TERM to xterm-256color
- got dropbear SSH server working ok
- parked for now MC and glib2 dependencies, didn't yet got a fully functional form (issues with term/functional keys), will have another look after other planned additions
- got a functional .config for kernel 7.0.6 tailored for ao486 (author already jumped now to 7.0.8, probably will not make much difference but will have to check)
- modified config/build scripts and added some diff-patches, but yet to iron out some bugs/issues before even attemtpting to update this fork
Great !
This is all open source software? It should be fine to direct people to a link? @copper slate
Panic on hard drive access?
Still debugging the MMU, now Amiga almost boots to the workbench, crashing when SHARE: is being mounted due to wrong mapping
Yeah, it will be a github fork in the end, so anyone can build it's own vhd with enough tools to make it fully usable. For now didn't publish anything than a sshot on my fork, as I need to clear A LOT of crap (and only for the work i done so far)
I am working under WSL 2 with Debian, so Fedora/Arch might need also have a separate look (and not really sure I'm up for the task atm, will see how it goes)
Got the last commit to boot to Workbench with mmu.library and 68030.library, so that is a big step forward. Now debugging the NetBSD kernel panic
Nice!
if anyone wants to take it for a spin, or tey to boot linux or something
I have some debian image to test, will give it a spin tomorrow morning first thing after coffee
working here
Any idea why it reports so slow?
Not optimized enough yet?
The caches are forced off while I was debugging the MMU
Makes sense
Question - if I have games that are aesthetically interesting, but not really playable in English, would there still be interest if I made 0MHz packs for them?
I’ve got two in Japanese and one in French that fit that category set up at present
Is Knights of Xentar among them ? 😄
Is on my naugthy to-do lcheck-list, but been side-tracked by other stuff lol
The three I have set up are Eriko Tamura - OZ, Noah’s Ark (1995), and Big Bug Bang: The Sequel to Commander Blood.
I’d also like to see if I can get Putlestory working. The CPU is unfortunately too slow for Incarnatia, Cookie’s Bustle, and Ultra Resort Keroncuel in testing.
we need a NSFW section of 0Mhz... just sayin'
Does Leisure Suit Larry count?
pfff.. maybe just as history 😄 True Love FTW !!!
I have a vague memory of obtaining a copy of the original LSL1 decades ago and being thwarted by the age verification, which was basically just "trivia questions that are easy for Boomers".
you need the new unstable mister main for the new core
With the ZX Spectrum core, how do I access my saved games? As an example, once I'd completed the first two quests in Valhalla I saved my game, but when it came to loading it in again the system just tells me to "press play"! What do I need to do, or do saved games not work? Thanks
Are you playing it from a tape image, or a .Z80 snapshot?
Putlestory is going up:
I'm not exaggerating when I say this might be one of the most beautifully animated adventure games ever made. Here are some screenshots, for reference:
There is a bit of a language barrier, as the game features a good amount of Japanese text, but after some testing I found it didn't significantly impact playability.
Valhalla I'm loading from a tape image, then using the built-in save and load commands.
Did you try pressing F1? That starts & stops playback of the virtual tape for .TAP files
Judging by the photo the last person who tried lost their hand 😮
Wish there was a way to save just a delta file when writing to a floppy disk so I don't change anyhting on my original disk images
Same thing happened to people who pirated NES games in the 90s and early 00s
Mmm I'm still really excited for this 386 core whenever it releases
Is sysinfo getting wrong MHz readings? 80.60? The results would indicate something like 68030@33 MHz.
CheckMMU program doesn't start and testing some games and program I have crashes or memory issue messages on my side
It's slowing on the memory accesses, if I turn off the TC.Enable in VATestProgram, it will show 90.70 MHz
Still, it's a bunch of debug builds, and I'm doing all sorts of errands in between...
I barely got it to boot to Workbench without removing setpatch or the mmu libraries, so it's still a work in progress
I'm trying to figure out why the kernel is panicking on NetBSD, so definitely there are bugs.
Virtual Murder: The Complete Collection is going up:
This comprises the four games in the Virtual Murder/Murder Mystery series by Shannon Gilligan:
While I'm working on stuff here, does anyone have any requests for 0MHz packs that I could put together?
Tex Murphy?
I believe Mean Streets, Martian Memorandum, Under a Killing Moon, and The Pandora Directive have packs already
Ah, sorry I misunderstood. I read “put together” as compilation
I played a lot of countdown more for the “impressive” sound and graphics on my pc.
I know it is related to the Tex Murphy games but not sure how.
I believe it was also made by the same developer, Access Software. They released it the year before Mean Streets, and I believe it uses the same engine as that game
Ah, that's what I need to do, right, I'll give it a go then. Thanks Elphive!
Ocean software games, Codemasters games, Bitmap Bros games, Sensible Software games, Team 17 games ?? Some have wildly varying genres, but, some, eg, Sensible, just made a handful of types
Most of the packs I've seen, have been fairly US centric.
Oh, Horrorsoft games.
I've done a bunch myself, too. Actually, I've done the horrorsoft games. But, I went over a bunch of 0mhz games, and made better versions. Eg, Monkey Island 1, I included Soundblaster, Adlib, Beeper, MT-32, and CD-Audio along with the talkie versions. I should really upload them, somewhere, I guess.
I made a bunch of smaller (3mb-20mb) blank vhd files, if anyone wants them. Just to save having to make them over and over.
Nice - can definitely use those. Thanks!
I only made the smaller ones, because, most of the 0mhz games default to 8mb. Where, a bunch of the smaller games work fine, in a 3mb image. (etc)
Next up is tackling Gravis Ultrasound. Undecided if I should make a second .vhd, and mount it as E: and point the ultrasnd dir to there, or not. Save around 11mb, over about 70 games (potentially) Or, is it not worth the hassle...
Oh nice. My rusty-backup tool is able to handle A LOT of file level backing storage things, including VHD and CHD files.
I'm extending my app so it can be CLI accessible, so maybe you can stamp out files from fresh, and being scriptable things like that.
For Gravis? Well, it just needs the env. entry set, and the files to exist. A single set is all is required. The existing tools should let me mount a second vhd from within DOS, after the C: and D: drives have already been assigned. The only "problem" is, is it worth the effort to save potentially, 700MB maximum.
No problem!! 😛 The other thing I'm doing, is, for games that provide it, add a config/setting/menu entry for joysticks. Since, today, I imagine, the vast majority of MiSTer users will have one.
None of the other collections, eXoDOS etc, are doing that, since they're abstracting controls, and using key-to-joystick type programs.
Where, in MiSTer world, real joystick is required 😛 Well ... "real"
Same with stuff like Monkey Island. Including the CD on the menu, gives not just CD audio, which is different to MT-32/adlib, but, if you load the actual CD based menu for the game, you get an exclusive, otherwise unknown, cd audio track. (And, as a byproduct, can easily pick multiple languages)
Oh I'm just talking about "in general" plus if you want something added to the app you know the developer is on this discord 😉
Maybe you want to pump a bunch of things out like 0mhz
Yeah, played it a bit yesterday. It's a great sequel.
Caldor's core
Cool. I would love to see it expanded into its own thing one day.
It is quite slow for the moment
Today someone (who is not on Mastodon) released a collection of more than 570 distinct operating systems, pre-installed with VM configurations for the 250+ different platforms, going back all the way to 1948.
Now, I have to admit I'm posting this without trying it myself, as I'm running low on disk space on this machine.
Because the full download is 121GB (174GB unzipped!). There is also a lighter version at 14GB that will download stuff on demand.
824
829
Update core CPC+
https://misterfpga.org/viewtopic.php?p=107585#p107585
WIP Nabu MiSTer CORE
https://misterfpga.org/viewtopic.php?p=107591#p107591
Nabu is an interesting computer, especially for its network boot, iirc it was also done via serial but different~~ baud speed~~ standard (rs-422)
SAM Coupé, Street Fighter
This might be over achievable clock speeds, but it reminds me how much I'd like to see a webTV core. Given the webTV Redialed project is a thing, it would be neat to have a reimplemented version of the hardware itself
Updates on Nabu: https://misterfpga.org/viewtopic.php?p=107672#p107672 😍
very nice 🙂 I am working on a better turbo, but... its likely weeks a way. Quite a slow process it seems, because every part likely stops the core from booting at all and its very difficult to debug.
We can use boot rom from ao486 on z386 and also vhd of 486 boot
Fujitsu FM Towns Marty hype
That’s better than I would’ve expected, given 386 instruction speeds. How far can you overclock it?
I also remember someone talking about a 286 implementation a while back - was that ever finished?
I would definitely be curious about that as well for Tandy VIS/Modular Windows.
Seems like if the clock speed was a little faster it would be more comparable to the 486 core
At least for Doom benchmark
Interesting DOS32a revisit, especially for LFB: https://github.com/calculatetech/dos32a / https://www.vogons.org/viewtopic.php?t=111282 About
A patched and optimized version of DOS/32A for real DOS environments. Original 9.1.x versions had changes that served dosbox use cases but broke linear framebuffer write combining. This release aims to fix that. "I discovered the 9.1.x series removed HDPMI32 compatibility for some DOSBox related reason." "I recall having issues with Terminal Velocity and 9.1.2, but that game is working fine now. Descent works great, much smoother than DOS/4GW. "
Not sure it will have any impact on ao486 but still interesting. I wasn't aware of the compatibility removal in 9.1.x
at a wedding but will be interesting to see if it has any effect on benchmarks for fdoom, VESA games, decscent, duke... "People already use DOS32A successfully on ao486 as a DOS4GW replacement for games like Duke3D, etc.
There have been various memory/cache/VGA-related quirks reported on the core over time, and LFB performance has come up in discussions around Doom ports and other VESA games.
Try the new 26.0 build (or whatever the latest stable is) on games that had issues or felt slow in LFB/VESA modes.
Compare it directly against older DOS32A versions or DOS4GW.
Pay attention to games that rely on VBE 2.0 LFB (FastDoom's VBD mode, certain 3D games, etc.)."
The fastest real world 68882 FPU is clocked with 50MHz.
What is DBurst, Cache related? I just ran xSysInfo on my A1200 with TF1230 and motherboard soldered FPU and I have DBurst on for me.
My Gary is also showing as 'FAT Gary'
Thanks for this great work
Edit: I got the wrong demo! Now corrected - For a practical test, I am sure this demo needs an FPU. It's also in AmigaVision, (which should run without any tweaks on your Core): https://www.pouet.net/prod.php?which=52968
wow! anybody tried it yet? I'm also curious to see the if the FPGA utilizatiin is lower than ao486 (leaving room for adding more fearures)
It uses same boot rom as ao486 and we can use same vhd
Doom, Prehistorik 2 and Nicky Boom run but Rusty and Tyron hang during the load
is the speed configurable? Lion King apparently ran too fast on ao486
no speed configurable for the moment
Still debugging, so it's like 1 step forward, 2 steps back
Cool no worries. Keep plugging away, amazing job your doing. This is most likely going to be the last big addition to Amiga on this generation of FPGA 👍
Debian almost boots now on Amiga core 🙂 Before it stopped and went to guru error right when uncompressing the kernel
Thanks for the core dump
Rusty loads for me for some reason, along with Descent, Doom engine games including Heretic/Hexen,Terminal Velocity, Raptor, Wacky Wheels, Prince of Persia, and Rastan. Fewer games didn't load, including Comanche, Jazz Jackrabbit, Build engine games (Duke 3d/Shadow Warrior) and Tyrian as you mentioned. I used an XMS only configuration with EMM386 or JEMM with EMS. I also used DOS32A v26 with most titles using DPMI, although the sb.exe utility crashes trying to restub/bind an executable.
As for things that don't work (yet?), the HX extender's HDPMI32.EXE hangs, as does a core reset. MIDI support isn't present. Joystick input seems to do nothing, either as a joystick or remapped to keys. I couldn't get a CD-ROM drive recognized either, I'd love to test Sideline when it does work. Windows 95 crashes as well, but it's still early yet and it's more surprising how much is working in the first release.
hopefully can get around to messing with it today
@terse totem I know you are working on the ne2000 as a target, but the 2065 uses the AMD Am7990 which has an implementation in the sparcststion core.
I'm aiming to make the HPS side generic enough, so that it could be used for other chipsets, and other cores...
yeah, I had looked into it a little when I was talking SLIP with Thera, but since you were already experimenting I didn't dig too far. ne200 adds a lot for a lot of cores as well.
ne2k was the lowest common denominator decision
yeah, I had mainly looked at the amd lance chip because it worked with the existing sparcstation core, had the 2065 support in amiga, and was used in some 3com ethernet cards on x86
it may have had some uses in some of those weird unix installs, but I can't remember
on the Amiga side, I looked at the drivers and tools that could help me with debugging, and there were a couple of the cards based on the rtl chipset that were ne2k compatible, so that was handy from that perspective
Either way, I hope we can convince srg to accept it.
hehe, I think he's aware of my attempt at 030
030 is one thing, but the Linux changes are a bit more invasive.
My plan was only going to ever work on wired Ethernet
So, not really a super intuitive experience
I think the last attempts didn't even needed a tap kernel module, just a bit of changes to main
and from the menu, I had a switch to activate/deactivate ethernet
so that was all C code in there
Hmm, how were you routing it? I had planned to build a kernel module to manage the tap copying into a buffer in ddr ram to avoid using main, but it is easy to say stuff like that 😀
But the tap can’t really deal with wifi
bridging via raw socket on the eth interface
copying packets (not the cleanest solution, but avoids extra plumbing)
and then using shared memory to communicate between the FPGA and HPS
Yeah, that is pretty much the same I had planned to do, except in a kernel module to keep Ethernet out of the main loop
I just needed something for debugging, and it was a pretty practical approach where I could easily confirm things sent and received
Well and using tap. I never did come up with a solution for wifi, but one of the new bridging protocols like Geneve may solve it?
little steps
https://eab.abime.net/showthread.php?t=122494 --- new browser, with modern features is being developed for Amiga 🤘
Just found out this morning from 10 Minutes Amiga Retro Cast last video
yes maybe it could help any accuracy issues remains
Data burst, it allows for transfers of more data at once to the ram or the peripheral
How does one clear the selected hard disk in the MacPlus core again?
probably del like with ao486
actually the MacPlus core periodically "clears" the HDF on its own in my experience.
ahh, I thought the FPU part was working already. I merged some of your branches into my CDTV/CD32 core stuff on top of also merging the IPF/external floppy stuff from Rob Smith.
I figured I would try to just add an FPU option to the 020. I finally got it working and can run it in SysInfo 4.4, but AIBB 6.5 crashes when FPU is enabled. SysInfo does not crash and the speed shows the same, but, it still shows as an 030 CPU instead of 020 and while I tried to disable the MMU stuff it is still listed in SysInfo as having a disabled MMU. Maybe because 030 is supposed to have MMU?
I figured both things could be gated behind... well, gates and then the features could be added later as they worked. I guess I will try to show if my AI seems to figure out anything useful. I have it working on AO486 FPU... using simulators. Pretty sure once everything is ready as HDL the AI figures out that there is not enough room for FPU on that core.
But not sure if there is any HDL FPU for x86 cores, at least not 486 CPU FPUs. So... if it ends up working it would be interesting to play around with I guess.
i think it'd be fair to fork ao486 for an fpu version, there's plenty of stuff you could remove to make room
or ifdef it anyway
The 68882 FPU has a bit of an integration that needs to be fixed. The MMU is usually on the full 68030, and it's disabled in 68ec030
My commute to/from work takes up about 7 hours these days, so I mainly fix things on weekends
7 hours
😮
There are still some bugs that need to be fixed before I stamp it as 1.0, I only got the AmigaOS to boot up on that thing like a week ago, so it should be considered fast and dirty
Dirty with regressions
I got some tokens to burn this week with the AI, so I think I will have it look at the FPU some more. But FPU seems pretty advanced. I have had sims running for the AO486 FPU for... maybe a week or so now. But it has made some huge iteration logs about things that has been implemented, what works and such. All without using the MiSTer itself, Which is nice since I am working on SuperCPU and Amiga on the MiSTer testing things.
But I am getting a timing... thing that is about 3.5ns. Seems it still works, but that is while disabling most the MMU stuff, trying to get it ignored by the build process.
And... I need to figure out why the CPU is still seen as an 030. So probably need to get it into a larger research process for this
I'm on the $100/mo Codex, and $20/mo Claude, 5.5 is so much better at not hallucinating bitfields
IIRC, the 030 detection is based on the CACR probing
also if you just want to run quake you dont need a whole fpu
quake being opensource you can point your AI at it to see whats used
I did setup a "Codex" skill that uses 5.5 as a helper
Something like trying to enable data cache, and if the readback is 0, it gets detected as 020
So my Claude Code gets help from it
Claude is only good at creating reports with a shitload of emojis
Should DCache be possible to control from software btw? It seems ICache can be, and SysInfo seems to expect it to be possible to control the DCache
Otherwise useless, keeps chasing its own tail when you try to debug stuff
I have had pretty good results so far I think. Its only... I think two days ago I began using 5.5 at all. I have gotten working SuperCPU that runs Doom and such, CD32 and CDTV on the Amiga core
But in some cases I had to stop Claude when it was just doing really really small iterations, testing one thing, rebuilding a full core, then testing another thing
I was starting last year with Claude, and it was doing a good job creating the base, but then it would just pollute it with things that don't exist, and that required a manual intervention
Well, at least two days ago that I began 5.5 as a skill for Claude Code and had it do anything with these MiSTer projects. I am only on the Plus plan with ChatGPT and I am on x20 with Claude
Then it would at random go for git checkout, and if you didn't have your stuff committed...
I did go down to x5 again though for next month. I might upgrade again but... I accomplished most thing I hoped for with the MiSTer already at least
You can have one check the other's work
I am forcing Claude Code to max out at 300k tokens to avoid it getting stupid. It does support 1M tokens, but I have it create handoff files instead of compacting
Yeah, that is what I am using the Codex skill for, then Claude now sometimes have Codex check on its work or problems
So it just runs the Codex app as a CLI app with specific instructions for one run
Use ISSP for on-system debug, it helps a lot
ISSP?
Mini USB next to HDMI,I think
ahh, yeah. I have that. Does not seem to be used much, but its connected
Do you have some skill files on how the AI should use it?
Because mine found that it could use it, but mainly builds cores with tracers for UART debugging
You can get readouts of the values of registers and stuff from the running system
and on the C64 it makes PRGs and CRTs that helps debugging and an OSD debugging overlay
But can you get those readouts without SignalTap?
Ask the ai 😉
Yeah... maybe its because I used Opus, but it so far failed to get signaltap to work. It wanted me to set it up and I tried and it just did not work
alanswx got it working without needed to use the UI I think
What are you using Win or Linux?
Win and WSL
I have Quartus on both though
and I guess I could switch to Linux. I have Kubuntu installed
In Linux you just need to add a few lines in /etc/ and it's good to go
But I have this setup where I run all the AI in XTerm so that I can host it on my PC and access all the sessions remotely
You can set up RDP, and have it going in VS Code/OSS Code
It could probably run on Linux as well without much issue. But I have several projects running on Windows and skills and such based on Windows, using WinUAE and Vice emulators
Yeah, but this way I can access it in a browser, even on my phone
Might be a problem if I used RDP from my work PC to my home computer
But its not the main issue, because I am pretty sure I could get that whole AI control center setup working on Linux as well. It is just a webserver thing that I run from my computer and tunnel through my website
Pretty sure its a minor thing to get it running on Linux. It would probably run better there. But so many other things would have to be Linux specific as well then. I made some skills that use local AI for checking screenshots
It was mainly because of annoying API errors, but seems like a nice way to save tokens
Qwen3.5 0.7B can actually describe images very well for such a small model
Don't mix work stuff with home stuff
That is why I made the website thing, so it stays its own thing
But now I am considering making a similar thing for work instead of having to run a bunch of terminals and then having to restart them every time I restart the computer
But yeah, I guess I will try to have some sessions that look more into SignalTap.
And simulators also really help, can often save a lot of time when running simulators to do some specific tests for things before doing builds
But I am a pretty heavy multitasker. I have a lot of hobby projects, but I have even more non-hobby projects at work. Multitasking helps keeping my brain active or something. Without that adrenalin I can become slow. Works better for me to keep a high pace.
the bigger problem is there's basically no fpu software
unless you're big into spreadsheets
my understanding is that software had to be compiled with or without support for it. Since 486SX's were pretty common, nothing required them that normal people would use
later things made for pentium required them, because pentium had it standard
eg quake
yeah, its pretty useless for the 020 and... well with AO486 I think maybe Duke3D might benefit a bit from FPU and maybe a few other games as well. But... I think if it actually ends up working, it will be too big for the core and to do any kind of test some things would have to be cut from the AO486 core to make such a test and get it built into the core.
It was mainly because I thought the Amiga FPU part was working with the 030 that I wanted to test it. There does seem to only be apps that use FPU on the Amiga mainly and otherwise if I remember correctly, you would need both FPU and MMU? But it was things that I heard or read somewhere that I did not look much into. When it comes to games the current MiSTer already covers most of them... and even more so with CD32 and CDTV support. But if FPU would fit with all the other things added on top as well, like floppy and FLUX image support then might as well I guess.
But there are a few games where its a problem if D-Cache is enabled and if D-Cache is supposed to be possible for software to disable, then I guess this D-Cache problem might only be about the software not being able to disable it
I tried making a forced I-Cache on or off as a way to get original 020 speeds and found that some games would just not work with forced I Cache on or off.
btw you could do what I have done, but in reverse. Have Codex / GPT 5.5 use Claude Code as a skill it can consult with
I compiled a list (it is not in any way complete) of games and apps that would require FPU on ao486 when I was testing the Q87 FPU emulator: https://misterfpga.org/viewtopic.php?t=6639
In the apps category, the CAD ones are a clear winner even with the Q87
As for games...so-so, many of them were slow even on a real DX/4 (Quake, Tomb Raider, Wipeout etc), some earlier like FX Fighter are more than ok even with Q87
OS-es like B-TRON (B-Right/V) are also expecting FPU, could not find any way to get sources/recompile/see even if it possible to work on non-FPU 486 (like I did for some Linux distros), but I guess that is pretty niche
duke3d uses an FPU (if present) for its sloped surfaces
Looks like the NABU core is progressing https://misterfpga.org/viewtopic.php?p=107826#p107826
Those games you mentioned are all made for pentiums
They won't rub well on 486 architecture even with fpu
right? I thought so... and Duke3D already runs pretty well but has a few slowdowns. So maybe FPU could help it run a bit better. But I am not sure how well its progressing. Its doing a lot of tests and they are all simulated so far. FPUs seem to be quite complicated. I used to figure it was "just some math stuff" it did, but... well, math can be complicated after all.
yeah, overall there would be very little benefit from having FPU on this core. I do not think anything can be removed from the core to make FPU fit, that would actually be worth removing to get FPU. Because pretty much anything else that is already in the core seems like its more useful than FPU. But... more toys to play with if an AO486 core with FPU could be made 🙂
I guess it would make Windows 98 more stable (possibly)? Or at least it would remove the need to disable the FPU check.
the only stuff realistic to remove is sound hardware
everything else I think is fairly mandatory
Yeah, I cannot think of anything. But if I do need the space, I guess... hmm, I might as well try to get an overview of what features the core has and get an estimate on how many resources each feature uses... and I guess it should only focus on features that would make sense to remove of course.
Does the core have FPU support maybe? That could be removed 😄
maybe in the near future ai will be good enough to make a cleaner 486 implementation or optimize the cpu
I dont see any human writing it soon
maybe if we get ahold of the microcode like the 386
We do not have specs for x86 CPUs? Oh right, they are not open source. I guess that makes sense. I think 8086 was released? Or recreated.
I mean we have some
obviously
but we dont have real microcode for most
286 and 386 I think
What about those PC emulators? Box86 and the original one that Box86 was forked from. That must have something to go by? But whether its actually more optimized... AO486 started as a Sorg project, right?
not exactly
the 486 chip was programatically converted from box86
I think it was then adapted slowly into more of a mister core, but it was loaded with bugs
The AO486 one? I knew it was auto converted and then the stuff that came from it was cleaned up over the years
But I did not know the source was box86
robert actually fixed it up a ton
needless to say the 486 cpu core is long of tooth and could certainly be better if redone
Yeah, I was so exited when they began working on it and making it a lot better. Freeing up a whole lot of FPGA logic... and then... was it robert that added the L1 and L2 cache?
That made so much difference, going from Doom not being playable at all to being very playable and then even Duke3D being playable
But I still see new changes made to the AO486 core, making it more and more compatible. I think there were some Win95 issues and compatibility stuff that got fixed, making it more stable to try to use Win9x
Which probably helps when people try all those different Linux distros on the core 🙂
the only one I can think of that might run well is https://nerdlypleasures.blogspot.com/2015/08/retro-city-rampage-486-review-of-new-ms.html
it can use the q87 emulator on ao486 but is pretty slow
SuperCPU Doom... but still at 0 fps:
I think it has a 16 color mode. Not sure how to use that though
I had Perplexity look into the AO486 core and there is a lot of audio stuff that could be removed. But it estimates it could free up to 14k ALMs to remove such non-essential features. Then it estimates a minimal FPU to need 5k-9k ALMs and a full FPU to need 12k-18k ALMs. Currently it should have between 7k and 8k ALM free. But that would put it pretty close to the max for the full FPU implementation. Hopefully some simple sound thing could be added with it, but... I guess it would not be that important just to test how well FPU stuff might run. Although audio could help make it clear how badly something might be running.
It might be completely wrong of course. But its estimates anyway.
Ohh... actually with the 14k ALM freed it would be while leaving in OPL2 support, just removing OPL3, midi, C/MS, CDDA, 1 channel IDE so CD and HDD being possible
and it should leave in SB16 as well
Would an overclocked 486 core not outweigh whatever sped up calculations an FPU might give
Well, that's what it is, and the thinking behind all the caching
I think speeding up the 486 core has already been looked into a few times, but I guess there might still be ways to boost it some more
And every 6 months or so, somebody insists that it's not enough (but generally use Pentium-class software as examples)
Oh, my FPU thing is also looking into adding 256kb L2 cache. Which might speed it up... and it might just not work at all
ao486 already has multiple better-optimized caches helping speed it up. FPGAzumspass had sped it up by a factor something like 5 back in 2020 or so
But yeah, I do not expect FPU to be added to the official 486 core, its just for science!
Yeah, it was pretty lousy before that
It went from barely being able to run anything, at least at reasonable speeds... to now it can actually run Diablo and Starcraft in Windows 95 (laggy but running)
There is some software that refuses to run without a math coprocessor, but for the things that compensate, I doubt you'll get serious improvements
Yup... and I would not argue that FPU would really be worth removing any of those audio features. Its only for testing FPU. If the FPU works, maybe it can be used on other FPGAs, in other cores or... something
It's up to you how you spend your free time
I have tokens to burn anyway.
Its the last days of my x20 subscription month, I will be going down to x5 again. I have some 6+ projects running, but some of them needs me to do a lot of testing because I have not figured out reasonable ways to get the AI agents to do tests
But my initial goals have pretty much been met. SuperCPU on the C64 core + CD32 and CDTV support for the Amiga core. Getting the SuperCPU to run at 20mhz though... there is a plan but its not going to be done anytime soon it seems.
And several of my other projects have reached their initial goals, projects I have had around for years mostly, but AI has really come far now.
Like this feature. Castle Master, added VR support and made it able to run in a browser. But the nicest feature is portals replacing the doorways, windows and such:
I am thinking I will try to get that feature into ScummVM where I got the code from to make this possible. Have been following that project for years.
And then it could run in ScummVM on MiSTer cores 😄
I even had it support portals within portals. But it does not... at least not yet, support animations through portals.
I am getting a lot of requests for an Amiga console core. So it would pretty much be a Minimig core, but with a new name and the UI being redesigned, making it console focused.
I think PSX and... probably other cores usually expect the boot ROMs to have a specific name and be in a specific folder? So I was thinking that could be done for this version of the core. Make it so all the settings are on one menu page, or maybe with one sub menu to be able to change main and ext ROM if you want to. That way CDTV and CD32 mode could just be one option.
Dual mice support for minimig core, so that you could play settlers and lemmings
an fpu wouldn't really make anything on the core playable that isn't now
quake might run but so poorly its moot
There is a surprising amount of Pentium-class software that runs well on the core - I've had several programs list Pentium 75 and run well.
I tried the first Quake demo on my DX/2 66 when it was released. I can verify that it wasn't quite a slideshow, but it wasn't that far from one either.
oh yes... I thought maybe this had already been implemented. Could it be the AO486 support dual mice? I will put this on my list.
yeah, it would mainly to be a test for the FPU and that it works well enough to handle Quake I guess. I think people already ran quake with FPU emulation on the core and got the lower than 1 FPS but still running Quake gameplay?
something like that
things made for pentium expect more than just the fpu in terms of speed
also the pipelining and other improvements
Yeah, very much so. The main gain from the FPU is to have FPU and be able to test it. Then if we get larger FPGAs that could do more than our DE10-Nano to be commonly available and priced somewhat reasonably, we could maybe use it as a stepping stone towards Pentium? Not sure how different the FPU on a 486 and a Pentium is, but it sounds like they are quite different. Speed is one thing, but I think the Pentium FPU does more instructions / types of math than then 486 one?
I think a lot of MiSTer users like me use the MiSTer just to test strange things, to push some bounderies. Like... you do not really want to play Diablo or Starcraft on this. But you can and its nice to get it to run such games.
Somehow the SuperCPU implementation was working without the original bios. I think it reverse engineered the parts needed for Doom and Wolfenstein 3D. That was probably a lot of time wasted when it should be possible to just use the original bios instead. Might also explain part of why SuperCPU has taken so long to implement despite having an open source emulator to test against and learn from.
But now it finally shows the SuperCPU bios boot text instead of C64 or other alternate kernels boot texts 🙂
Hopefully this will help with the turbo implementation, because SuperCPU did a good job of making sure the parts that needed to stay at 1mhz did so. Sucks that what TC64 and C64U did for their turbo modes seems to be closed source.
duke3d may actually run at a playable framerate given an fpu
its a game that runs badly on a 486sx but well on a 486dx
Come to think of it, I never tried Duke3D with Q87.
I do not expect much improvements honnestly (like for FX Fighter), but will see over the weekend
In the meantime, if anyone around has any knowledge of DOS floppy images of CrossCountry Texas and/or North Dakota , please let me know.
I know these two were distributed among schools on those states, but couldn't yet source any image (if they even exist)
https://en.wikipedia.org/wiki/Crosscountry_(video_game_series)
How well does FX Fighter run currently? Something like 5fps or worse? I played it on my 100mhz Pentium I remember.
It's very much playable, there's a pack on archives with it (search "fx fighter ao486"), I have not played this game other than on ao486, but worked very well
#1047332497492553799 message --- my old tests on this game and Q87
This game + Blue Ice one (and maybe Humpty's Scramble, had to revisit that sometime) work just nice on the core + Q87
i'd think at best it'd be the same speed it already is, worst its even slower
its already falls back to a slower software fpu internally
yeah, most probably
There's an issue opened for this iirc
maybe that is what I read then. I loved playing Settlers and Lemmings that way 🙂 Well, Lemmings did not stay fun for long tbh, it was not that well made for being competitive. But Settlers was awesome.
meh, similar speeds, no gain using Q87 apparently from some quick tests I did
Ohh... iteration 150... the agent says: "The synth-unblock track is fully done (FPU now synthesizable). Which direction for iter-150?" 🙂 I guess it is making progress. It asked to do a Quartus fit but... yeah, I am pretty certain it will be too big. It suggested 3 different sim tests, each being a whole series of tests, regression tests and such. So I asked it to do all 3 of those.
I really hope this works. I plan to then have it try to do its own fit check before using Quartus to see if there should be any chance of it fitting. Otherwise Quartus can take a very long time to try to build and figure out it just cannot fit.
But there is just such a huge chance of this having all kinds of problems before having any kind of practical use... oh well.
Failure to fit isn't always "too big". There could be spare gates but insufficient routing/clocks/etc.
Seems it has gotten the FPU down to about 12k in size by "dedubbing" some of the parts, making different part reuse similar logic. Not sure what it started at though. But it needs some pretty serious trims it seems to get room for all of that. or it needs some of the FPU logic removed. But would be nice to keep all of the FPU to begin with at least, so it might be easier to test what parts are actually used, before removing parts of it.
Hmm, so now I am removing all the audio and the HDMI scaling. This build is mainly to see if this would make it fit. I do have a CRT to test on, but... I think even with the CRT I always used scaling to get that to work.
But apparantly there might still be more of the FPU that can be deduped
It did also add 256kb of L2 cache, and has reduced it to 128, but I think normally the core only has 32kb of L2 cache. But apparantly not much is gained from going down to 32kb of L2 cache. But I think I remember reading someone already tried adding a bunch more L2 cache with little to no effect. I should probably test the L2 cache on its own.
AI is absolutely abysmal at writing efficient hdl, it tries to serialize everything
Well, it uses the emulation code as the basis, so it just converted all of that, seeing each part as a feature I guess. But then it does at least see that it can be "dedubbed" due to that.
It did the build, removed all of those things, but had a timing problem, so it is working on that now.
A timing problem in the FPU
It has done a bunch of sims to check that all the parts should be working, but, hard to know before real tests and be done, and currently the main thing being tested is whether it is capable of even getting it through the Quartus Fit and I guess it could get fitted, just not in a reasonable way. It had a timing lag of 106.6ns in the worst path. But it seems to have found a bottleneck that would help solve the worst timing fixes because they all went through one point or something.
It got past the fitting part this run, but then the next step is expected to take 50 minutes or something. Sucks getting so close to the limits I guess. And I am pretty sure this wont have a screen or sound. Or at least I would need the screen to have a native resolution output. Would HMDI work without scaling? Maybe I could add one of those external scalers?
But... good chance the build might still fail anyway.
It's probably using the division operator
That code will really have to be done at least partially by hand to be any good
probably. But if it can make something that works and just needs some adjustments, that would be a good start. But it would need a larger FPGA to do much good I think. It seems to have gotten it from the 100ns down to -9.65 now, but apparantly that is still a problem. So now its reducing the L2 cache that was expanded while doing this as well, from 256kb til 128kb to see if that can help fix the timings... or something. Although there is also some dedubbing left to do that it knows to be possible still, but it did not expect it to gain much in terms of ALMs.
And I am not sure with the cache, because that is usually BRAM and I think its stored differently than ALMs, the M10K or something. But it might still help on the timings or something
with the current state of AI and the space and timing constraints, the odds of starting from AI and tweaking are very low, versus writing a base and having AI fix
AI seems quite capable of refactoring now. If it is told how to write the code or in what way it should be made different. Especially if you use the top models like Opus 4.6-8 or GPT 5.5
Things that AI would get wrong 10 times in a row a year ago, it is likely to one-shot now. I also used to just use Claude Sonnet models and Haiku for most things, but then I had some tasks were those models had made errors and I then found Opus to just fix it all and make it work in one attempt.
The smaller models could sometimes go down a wrong path and mess up things, sometimes making it faster to start over with an Opus level model and have it do it all right from the start, or at least have a much higher chance of avoiding a really bad foundation. So I rarely even dare to use anything less than Opus if the task has any amount of complexity to it.
Privateer 2 needs square root from the FPU. I was looking at Wing Commander 3 and 4 performance on MiSTer. The main issue is SVGA is unusable.
Has anyone played the new Impossible Mission 3 on the c64 core. Mine just locks up after selecting a difficulty level. It's working fine on the ultimate c64.🤔
I was able to do a AO486 build with FPU, but... then it has no HDMI and.... I want to avoid having to use VGA to test it. But it seems there is room for HDMI currently, so I will try that and hopefully it can be tested.
Probably a good chance that it will just end up with the core having regressed in critical ways, but hopefully the sims has helped avoid that. I will be positively surprised if I can at least do regular AO486 stuff with this build
Are you guys exploring the ability to put more of ao486 onto the arm side in order to free up resources in the core itself?
I looked into using the FPU on the ARM and maybe having some kind of layer that translates those calls to the core, but the AI argued it would add too much latency for it to be practical
Which I think was also the conclusion when I asked about such an option over a year ago
There might of course be other parts that could be moved to the ARM, but not sure which ones
But:
"Fit GREEN at 99% (41,478/41,910 ALMs, 432 ALMs headroom). Let me confirm M10K + actual WSS."
Now that is a tight fit
Maybe not the FPU, but possibly a different component?
wow, 99% is really tight, I'm surprised if it works
Yeah, I have not tried having a core use that much of the ALMs before. But, it also means I probalby cannot add tracers or anything as it is here.
Well, it boots. Now I guess I need some way to test it.
Seems I had to reduce it to 56mhz to get it to boot and... I am trying to load some benchmarks and they seem to crash.
Maybe I should just try running some games, because if nothing runs, I guess its not working
Not sure if this was supposed to be able to f.ex. get the Cache and CPU type right. But it doesnt here I guess.
I was able to run Descent without audio, but setups for games seem to generally crash and freeze the core
I think the crashes come from the audio having been disabled maybe. I can run some games. Like Descent and Wolf3D. Many of the benchmarks and system info tests crash.
But since it is at least booting, I could maybe make a floppy that could be used to test, for the AI to run some automated test. Or maybe it could write some DOS apps that could help do the tests. Last time I tried to get AI to create DOS apps it did not work out very well, but, there is better AIs now and probably more options. And there might already be some better ways to test it, like FPU test software.
Yeah, to my knowledge a separate “fpu” is a lie- the fpu was a full 486 and just disabled the primary 486 on the board.
I think using ARM to software emulate is somewhat defeating the purpose of the mister being a FPGA device. Also I assume it would be a pain to time right since the ARM side does a lot of work in the background for the mister in general, and the more features you have loaded the more timings can be effected. I remember a few years ago this was suggested to get N64 on the Mister back when it was considered "impossible" but the idea was dropped before the core was even made
Sometime ago, somebody tried to do a hybrid cpu for 68k, and apparently the Cyclone V has quite a bit of latency shoveling the data between the HPS and FPGA
Try Privateer2, it uses the FPU and will actually test it. Descent 1 and 2 run without the FPU
yes, there's a forum thread about it
I know, this was a test whether the core actually could run anything at all. But it did not work very well, a lot of things crashed. I was unable to run Duke3D at all, and I could not run the setups for games, not even the games that would run like Descent.
So I ran a build yesterday that would enable the sound, but I think it would only be stubs of it, so the sound would not be there, but software would see it as if it is. So maybe that would fix the crashes. But now that it is a bootable core, I think I need to get this agent setup with skills and boot floppies or hd images so it can run autotests.
I have had it focus on dedupping the FPU code, figure out what different FPU features that can be made to share similar code to save some ALMs, hopefully that will help make room for some tracers for debugging.
yea, and I think it worked... somewhat... because it was the whole CPU it moved and then... hmm, well, it still had issues and a lot of that was probably because of the latency to the other chips like those audio chips, IDE controller etc. I think the Minimig core uses DDR3 RAM for the chip and fastRAM which also helps make it possible, because the DDR3 RAM is already on the ARM side. Using SD RAM for the CPU, SD RAM being FPGA side, would probably cause latencies that would not be able to be solved with caches and such.
Update core Sharp X1
https://misterfpga.org/viewtopic.php?p=108013#p108013
IIRC it worked but it didn't give a huge boost to the current setup, and as always, simplicity is best
Does this bench use FPU?
The core now has sound, HDMI and FPU it seems. There has been several iterations of dedupping the FPU code
It cannot run 90mhz, but it can run at the 56mhz setting
Dunno if that benchmark uses FPU if available, but you can try a MOD/MP3 player that totally needs it (XTC-Play version 0.97c https://modland.com/pub/software/players/DOS/XTC-Play/ )
Nice. I can try to test that later. I tried two benchmark apps that saw the CPU as a 486SX2 though, so FPU might not be seen by software yet.
I wll make a test with that later
As for games, FX Fighter should be best candidate to test (just get the VHD from archives and remove the Q87 from autoexec.bat).
Or some other apps/games that I mentioned in my Q87 forum post, but on those YMMV
Snood uses a FPU, that's the only other game I can think of but that works fine with Q87
Interesting, didn't know this one needs FPU too... Will test it later too and add it to the forumpost list
Apparantly the core makes itself known as not having FPU, the id thing was found to make it show itself as a DX instead of SX. But it also does not show itself as having L1 cache which skews some of the benchmarks. I think one benchmark showed itself as testing to have more than 600mhz, that might be related to the lack of the L1 being visible.
And that probably becomes even more problematic for the benchmarks with the L2 cache... not sure if the 486 ever had L2 cache. But it could also just be these tools not trying to look for L2 if no L1 is found
But I am not home now, so I set the AI to look into the L1 cache not being visible and whether it could figure out why the CD test failed... hmm, might just be the DOS setup.
The AI mentioned Norton having an SI app, also called it SystemInfo and I do have some SystemInfo benchmark apps, but they seem to crash as it is now. And while the core now has the FPU along with audio and HMDI... somehow... I think its above 95% ALMs, probably still around 99%, so adding tracers might not be possible. I guess if a benchmark floppy could be setup that does everything automatically and then tracers help check that it runs without crashing or detects the causes of the crashes, that might help the next steps.
But it also has some huge timing issues, making it surprising that it runs at all, let alone is capable of running Duke3D with audio and such. Still cannot run setup for any games I have tried so far, but the ones where the games already have the correct settings for audio and music, it seems to work. I think Duke3D probably runs slower than before now since I have to use the 56mhz setting instead of 90mhz. But it probably also did not try to use FPU the last time I tried running it.
IIRC L2 was an option for 486, usually added later on mobos (eg: https://www.vogons.org/viewtopic.php?t=95925)
Well, at least on those that supported it, my old DX4 surely didn't had that added but vaguely remember seen it on bios as present but not populated)
interesting, I hope I can get this to be something useful.
yeap, definitelly 486 had L2 cache support, probably some mobos didn't had it populated/had slots for it. Remembered an old LGR vid on upgrading it on some of his 486 https://www.youtube.com/watch?v=e4rw3d7mu28
Got 256KB in cache chips to install alongside the 486 DX2-66! Installing them in the Woodgrain 486 DOS PC to see what they can do.
● Watch the 486 PC build video here: https://www.youtube.com/watch?v=fbjYkPKRm-8
● Here's the video adding a Sound Blaster Pro 2.0: https://www.youtube.com/watch?v=g15J44xB2zU
● Consider supporting LGR on Pa...
Assembly language, HEX editor, checksums! This video has it all! I received enough feedback from my audience to attempt reverse-engineering the fake cache BIOS of my 486 BEK-tronic BEK-V429S motherboard. Will we be able to patch and revert the change scammers have introduced in the BIOS to always display 256K L2 cache regardless if or what cache...
i remember this with low cost motherboard
website description
So my mobo had some cache after all on it, and my memory on that failed me this time lol
The only thing I have left of that PC is the Seagate 272 megs HDD (still works lol), which at some point I imaged it and restored my old stuff on MiSTer
||like my precious virii sources||
Interesting... hmm. I had a goal to add 256kb L2 cache but I think it might have been reduced to 64kb to have room for the FPU though.
Btw, the SuperCPU for the C64 core seems to have Wolfenstein 3D running at 3-4 fps. I was thinking I had reached a dead end with that project and maybe just hoping I could put it in a state where it might be easy to take the SuperCPU stuff from this for the C64U developer to put into his FPGA code. He already has a 64mhz turbo option that would probably work really well with all of this SuperCPU implementation and the C64U also already has an option to "pretend" to have SuperCPU because that apparantly is enough for some SuperCPU apps.
But I played it a bit and could shoot some enemies and such, I did not do a Doom test. Hopefully it means it is also running at 1+ fps.
don't think that one does, but there should be some in this pack that do: https://www.philscomputerlab.com/dos-benchmark-pack.html
yeah, but the ones that do seem to crash. I tried FX Fighter though and even though the CPU should now show that it has FPU the game still crashes when you go into a fight. Turns out there are some more things missing in the current FPU implementation 🙁
So some things might work, but if both FX Fighter and Quake wont launch... its probably not good
this has been looked up a lot for a bunch of cores, it really doesnt work out
why does this have anything to do with room, a cache should be bram
not at all related to fabric, unless you are implementing it in a super wrong way
true... it might have been related to timing
Seems there should be M10K room for the 256kb afaik
I think you are probably trying to infer ram which is a terrible practice
use the megafunction or one of the many wrappers
The main focus for now is the FPU. There were several "transcendental" functions that were not implemented yet it turns out
When they are implemented, I suspect the core is back to the problem with not having enough space for FPU + audio + HDMI scaler all at once
well yes, we've discussed there
you have almost no hope of making it fit unless you can write code like robert
Well, I certainly can't. Maybe the AI can, but that is also unlikely. But for now the plan is to implement those 5 things, then have Claude and GPT figuring out how to deduplicate the FPU operations and see if maybe there could at least be room enough for the FPU and then I might have to use a CRT to test the core with no audio
and no HDMI
But HDMI is pretty important to allow for the AI to be able to test. I guess I could use some VGA to HDMI thing to make it possible in theory, but I already had AI use OBS to help see the OSD menu and that sucked. Trying to get the AI to control the OSD was pretty bad.
But in this case it would mainly be for the screenshots I guess.
I did not think SuperCPU would work, but it seems it does. Getting more speed for the C64 core though, that seems to be going nowhere. Wolf3D was runnning at 3-4 fps, but then Doom would not run at all. I am still pretty surprised it got as far as getting those ports to run on the C64 core
I did wait with AO486 for last after the Amiga and C64 stuff, because I know its already been worked on so much, and already the timing is quite a bit off and it cannot run 90mhz mode without being unable to boot
and it is already using a lot of the MiSTer ressources
the timing issues is probably also why most the benchmarks just crash.
Some small rural community or town probably has no electricity or water due to the @worthy zealot 's FPU efforts.
I wondered why it was so hot today…..
at least the humidity is low
Yeah, the data center has used up all the water
No worries. Elon is going to put data centers and AI in space
Is that even feasible? How will they cool anything?
thermal radiation (literally glowing the heat away via infrared wavelengths)
but ya. think its like the tunnels system and mars. BS
Yeah, I didn’t think there was enough cooling capacity for data center level heat in thermal radiation. Maybe he will build a space elevator too and just have a long bucket brigade of water going between them
but then the heat is like back on earth contributing to global warming.
Yeah, I am not sure engineers and such agree with Elon on datacenters in space being a great idea. But its Elon, so he might make it happen anyway.
What's the alleged benefits to having them in space? Seems a lil out there but also kinda cool at the same time
I think the argument is that it will be a lot easier to cool servers that are in outer space and that the radiation and such can be shielded away. Also that solar energy could be the main power source since its a lot more effective in space
But then it could also all be taken out by a big enough space rock and I am pretty sure there is also a lot of other potential problems with it
Hmm, I guess outer space might be the wrong word when its still in orbit?
Low earth orbit most likely. We definitely don't need more space junk in orbit.
Sure is a lot up there and I think we have more satellites launched every week or something
Yeah, about 10 per day. I hoped I remembered wrong that it was a daily thing.
SpaceX alone seem to be launching 250 satellites per month, at least in 2025.
There are at least 4 programs that tries to do some space junk cleanup, but its probably still a growing problem
Well, it seems the AO486 FPU might be able to fit while only removing audio, making it testable. The AI suggested removing some of the FPU operations that games might not use to save space, but that seems like it could be a long process.
I guess a system could be setup to build a bunch of cores with different parts of the FPU enabled or disabled and see what different games and apps actually use. Or if there is space for it, maybe some tracers could be added that would log which ones actually get used
But first it needs to work
quake and duke3d are both opensourced so you can just check exactly what they use
dosbox can probably be used to figure out any others
stick a bunch of logging into its fpu code
true... and I could use box86 or something like that. Its what I have done with the other cores, have an open source emulator to compare against
and shouldnt it maybe have some FPU implementation of its own?
I would think it should be straightforward to have a disassembler find all FPU instructions (assuming there's no self-modifying code or elaborate obfuscation method). Compared to optimizing an FPGA core, it should be be a much simpler problem for an AI to work out, no?
does seem like AI is pretty good at dissambling. I did it a bit with Dark Queen of Krynn, but it can be pretty time consuming. Its just that I should have a full FPU on the MiSTer now. Also even if AI can do it, it will probably burn a LOT of tokens dissambling something. With all this core stuff, a lot of the time is spent on building the cores, mister main, running tests and such, so its not really that token heavy
And I am thinking if there already is an open source emulator with an FPU it should make it easy to log such things
I don't mean have the AI chew through each binary but have it help write an "FPU instruction cataloger' to detect any FPU instructions smartly (e.g. "Here's where the program tries to detect an FPU, here's what it does if it finds one, here's what FPU instructions it will then use").
sure, but it would still have to scan through files and such, and I am not sure its simple to find in assembly or what it ends up being. I guess I will ask an AI what way to go. Would be nice if it would be possible to just scan the exe file for FPU calls
yeah wont they have distinctive signatures in assembly/hex?
you can just use like, regex once you know them
If they are distinctive in that way. I just imagine there would be several ways to do it, and maybe several jumps in different areas.
if you can get duke and quake working you can probaby just ride off into the sunset 😛
Well, if Quake runs, it would probably still be a slideshow
With Duke... I am still wondering why it did not crash after I set the CPU type to DX
it might go from seconds-per-frame to frames-per-second
But it sounds like you can just make a FPU instruction scanner
That is my problem with the SuperCPU so far. At least with Doom it has stayed seconds per frame, but Wolf3D did end up being 3-4 fps in one of the builds
But then Doom would not run at all
I think someone ran Quake on AO486 with the 87 emulator, and that it was seconds per frame, but it did run
I suspect it would be the same with this
duke might not trust the cpu's ID and actually try some math to detect fpu
honestly duke is a bit of a weird one, i think before its sourcecode came out people wern't even sure it used an fpu
Hmm, yeah. I guess going over the source code would be a lot simpler than doing that scan
It was already suggested, and the same with Quake, makes it simpler to figure out the exact requirements
But that wont work for FX Fighter
So I guess a scanner would help as well, and maybe the Quake and Duke sourcecode will then make for good ways to test the scanner since I have a result to go by