johnparker007

If you are ever curious about having a dabble (since you mention proficiency in c/c++), the full live source for MAME is here: https://github.com/mamedev/mame If you wanna see how a driver works that isn't too crazy (Scorpion 4 is currently 50,000+ lines, so I may refactor that one!), I've very recently been looking at Maygay M1A/B platform (which needs some work, it has various issues), it's a bit more sane in terms of source size, and the game configs are mercifully already split out to a separate file from the machine logic: https://github.com/mamedev/mame/blob/master/src/mame/maygay/maygay1b.cpp No pressure though, it is a steep learning curve if you're not familiar with chip pin outs and clocks etc! ...but always good to see a fellow coder around these parts

Maybe we could lure you into some emulation programming on MAME...?

I've had a very small taste of the addiction when I was young and relatively skint - I was going down to the local arcade ('Smithies' in Cleethorpes), and contrary to my usual super-frugal approach of going to the seafront with little to no money, then slowly boosting myself up to a few quid so I could play video games, I was doing the danger dance of just taking 'big bucks' for me at the time of like a tenner down there. Then if the machine seemed overdue to payout, but wasn't playing ball according to my expectations, it was like my brain would enter into a grim determination mode, where it was a case of pull a big win/jackpot or die trying (i.e: go skint). Just mechanically slinging my much-needed food money in there (I was on the dole). I remember at the time being probably pretty lucky ironically from also doing a few drugs (not while I was playing!), and recognised that 'end of the night' fiending behaviour. Looking back it still holds true, even after having had access to more and 'better' drugs; there comes a point where you are just fiending, and not getting the 'original pleasure' which motivated you to take the specific drug in the first place. As an example (as I realise I might be more of a fiend than most, so this isn't 100% relatable)... if you haven't taken alcohol for a week, then you get into some drinking (and your brain is wired/tuned up for alcohol)... you'll get a sense of almost euphoria, possibilities seem endless if you just keep doing what you're doing (drinking). Cue 8 hours later, and you're 'chasing the high', earlier a couple of pints and you felt on top of the world, but now an additional two pints is just keeping the wolves from the door (disclaimer I've had various drink/drug issues lol ). That brief dalliance with fruit machine gambling I had in my youth felt very similar - I had transitioned from that phase where was exciting, to the phase of grim determination, where I just had to keep it going because my subconscious brain was willing me too, and I was just along for the ride. I think the parallels between gambling and drug addiction (speaking from experience on the latter) are really similar - your brain gets desensitised to that specific spectrum of stimulus, yet seeks it out, leading to increasing 'dosages' (i.e: stakes in gambling). Unfortunately I'm a bit drunk now, so clearly I haven't learnt! But yes, gambling is a bad path, just like booze and drugs - very similar scenarios ensue. Also - to be clear, I'm not comparing my experience to folks who've spent their entire livelihoods on gambling, just chipping in with my thoughts

Apologies, I've kinda morphed the thread The flyer scan images at the top were showing a model that can remove print artifacts from high resolution source flyer scans (moire/rosettes). The later Andy Capp ones are showing a new model for upscaling from low resolution sources, such as photos of the glasses, but with a focus on preserving text and 'shapes' which is useful for the kind of art on some fruit machines. So they're not really competing, they're working on different aspects of the source images. All very handy though for future 4K+ layouts from 'challenging' resources, since full vector redraws are extremely labour intensive!

This is unfortunately not an open source free model - I'm just testing some commercial stuff to see improvements for fruit machine images, since open source will catch up in time anyway.

Another day, another new AI model - just having a brief play, this one is quite good at preserving text and shapes, maybe best I've seen so far for upscaling these kind of low res images...

Just tried that - again, the sound sample stuff seems a bit iffy. I'll stick on this M1A/B platform for a bit, and work on the Oki MSM6376 - maybe make a new chip driver or summats, get the datasheet, all that jazz. Might not be able to get progress, but it seems worth a shot, since the machines are 'sort of' working!

Yeah that kinda gels with what I've been seeing with hacking around with 'Eastenders - The Queen Victoria'. I think there may be bugs not only in the core emulation but in the sound chip emulation. For instance on Eastenders QV, when the numbers count up from the reels onto the word 'Eastenders' you have to fill up, the ascending 'ding' samples fire slowly, then it plays a couple of wrong samples, then finally the next credit is available to play. It has a Oki M6376, apparently this is a ok substitute for the Oki MSM6376... but yeah, it might be what is causing at least some problems. When the chip receives a request to read NAR, if I always return 1, then it break some sounds, but then all the things that did go slow, such as the Eastenders letters, the actual feature board, and the bank counting up 10p at a time... then go either really fast (feature board), or about the right speed (Eastenders/Bank 10p anim). So I do think it may be the sound chip causing these speed issues... I've definitely seen something like this before on other fruit platforms in MAME, where it's clearly waiting too long for a sample 'callback'... so I think there's some low hanging fruit around those areas to get started with Gladiator - I believe that needs an alternate Reel MCU... there's a part-done implementation commented out in MAME. If it works in that really old MFME I have, then there's good potential of porting Chris's rev engineered version over into MAME's framework.

Ok I tried Eastenders Queen Vic instead (also Maygay M1 A/B hardware) and that one does boot and play. However, something funky is happening so the machine seems to keep briefly freezing as if it's waiting for the end of every sound sample it plays before it can continue... so I'll have a look into that If you lemme know the names of a couple of those M1A/B hardware games that play but don't have sound and they seem like they should, I'll also have a look at those and see if I can spot a problem in the code - ta

Ta I'll have a look, while I'm looking at sound stuff. Just tried Eastenders (m1eastnd) and it doesn't even boot, gets stuck on 'RAM CLEAR' even after repeated forced resetting...

Ok the Scorpion 4 Volume control emulation is implemented (in 'our' fork, it'll make it to the main branch 'at some point' ).

OK I've added that to my sig - will do a post for it in the future once it has a comprehensive set of test/settings access procedures for the various platforms [ Fruit Machine Settings/Tests Guide ] https://tinyurl.com/yuebw8b5

Thanks, I didn't know you could do a doors closed test with a DIP switch! As I mentioned to Mort, I think I may start a community public doc to compile a 'cheat sheet' for some of this stuff...

Aaah right, I was a little tired when I got my Scorpion 4 'console logs' of the volume working in MAME last night... maybe they do save to NVRAM from the test mode as well! I'm currently working on using MFME 20.1 as the 'ground truth' since I don't have access to any machines. Thanks very much for this info I may start to compile this info about the steps to get into Test/Settings menus into a public community doc if there's general settings that work per platform (and perhaps manufacturer within the platform groupings). For instance I noticed on Scorpion 4: for Bell Fruit's DOND, the volume range in the test menu is what we would expect; 0-100 inclusive. However in Mazooma's Cash On The Lash, while the test menu looks very similar, for some reason they have changed that range to 1-99 inclusive! I'm not sure why that is, but it does show the rebuild manufacturers like to mess with stuff a bit... (oh and yeah, I'm hoping to be able to mirror what MFME v20.1 does with these things in the MAME drivers, it's a nice change from Oasis until we have the MAME classic layout export stuff working Later planning to move onto fixing stuff where the machines simply don't work in MAME if I can work out the issues, I think there's a lot that should be just about in my wheelhouse these days).

Hey all Ok so I can 'test' changes to a machine's volume in the test mode on Scorpion 4, I've been doing: Open back door Await full reboot Press Test button and then I can test volume changes like so: However... I assume there must also be a 'settings menu', so I can change the volume, and then have that volume saved in the machines NVRAM as the one it will use on subsequent boots for standard sound effects/music. The reason is that to ensure my volume emulation code for Scorpion 4 is working properly/complete, I believe that if we set the volume to very low/zero... then reboot the machine a few times, to trigger the On/Off alarm (or some other alarm like coin alarm/hopper alarm etc), that alarm should actually play quite loud, regardless of if the machine's saved volume setting is very low. So I'm wanting to see if that just works with my existing emulation code, or if I need to find an address that is mapped to a temporary 'Volume Override' command. Thanks for any help

Not quite finished, but there was some code in Chris's source that was not implemented in MAME - I've implemented it, and now I have a working representation of the digital volume pot that is part of the Scorp 4 hardware, that is adjusting with the sound test Showing test value 69 of course So tomorrow, I'll continue, and work on implementing that so it scales the output waveform in some accurate way to cleanly represent what's happening in the MAME source (perhaps see how other drivers do it), and that'll be another minor audio bug fixed

Got some serious 'old school' MFME on the go over here, running in Windows XP! It's the trusty Cash On The Lash test layout... as there are some bugs in MAME that don't happen on this old version of MFME... so I think the strategy for some of those types of bugs will be trying to spot the internal difference between MFME and MAME emulation states/ram/irqs/read/writes etc as the bug happens, may help with developing fixes for MAME... that's the idea anyway

Ok - so to give a bit of variety to my FME dev stuff, I've bit the bullet and got the stuff sorted out so I can build MAME locally. First fix is to the Scorpion 2 broken volume, I made the edit I mentioned in the BFM Showtime Spectacular thread, and with a bit of debug, we can see the Volume is now emulated correctly I'll figure out a way now so I can have these commits in Github, without needing to wait for the MAME dev team to merge them, which has often been rather problematic. At some point in the future when convenient for them, they could grab all of the fixes and merge them in to the official MAME. Once that's done will update with a link to the repo here (and prob in my sig). Edit: the base repo on github for now that will contain the bugfixes/features is: https://github.com/johnparker007/mame

This test is unfortunately not using free software... it's just a proof of concept I was playing with, to verify that these patterns can be removed using an AI model. If you have some decent resolution scanned flyers that you want processed that have this specific issue of rosette/moire patterning, you could send me a batch and I'll process them here and send them back... best I can do for the mo! I was getting these from @spa's drive, he has a decent collection there. At some point it would be possible to train/find a model/weights and use them with a ComfyUI setup + NVidia GPU (so then it would be free software), but I've got too much on with Oasis to get sucked down that rabbit hole, which will cost me loads of time One day perhaps I'll sort something

A few more tests:

I've been having a play with another AI image processing angle that might be useful for creating good background images from scanned flyers; attempting to remove the printing artifacts like the rosette pattern/moire: Original on Left / Processed on Right:

Quick summary of Dif by our robot friend Desert Island Fruits is a community-driven website focused on the emulation and appreciation of vintage fruit machines. It caters to enthusiasts of fruit machine emulation, particularly those using the Multiple Fruit Machine Emulator (MFME). The site offers forums, layout reviews, video tutorials, and a gallery of user-created layouts for various machines. Members actively contribute by sharing ROMs, layouts, and technical advice on recreating or modifying old fruit machines digitally. New members can access a "Newbies Help Area" for guidance on setting up MFME, downloading files, and contributing through donations or by uploading resources. While some downloads are available freely, there are limits on downloads for basic members, which can be increased with contributions or active community involvement. The community aspect is a significant part of the platform, where users share nostalgic stories about their favorite fruit machines, discuss cabinet building projects, and even honor the memory of Chris Wren ("Wizard"), who was instrumental in developing MFME. Overall, Desert Island Fruits serves as a hub for enthusiasts to relive and preserve the legacy of classic fruit machines through emulation. Totally gives itself away with its weird AI phrasing, though one day I'm sure that'll be solved so it won't do a final summary sentence starting with the word 'Overall' which makes it sound utterly mental by British writing standards. Nice to see a shout out to Chris, though I'm sure he wouldn't be best pleased with occupying the same sentence as the American spelling of 'favorite' and 'honor'

I've tried it on some 'retro' asm (the latest Claude 3.5 / o1 mini preview which are SOTA coding models) and it is ok but it does hallucinate. A lot... a very long way off a skilled (human) disassembler right now, but maybe one day I could see some usefulness

A lot of work has been going on with Oasis... but none of it lends itself to screenshots! But it's been a while, and I realised this hasn't been seen yet, so this is an example of a 'View'. ...you are not seeing double On the left is the 'Base' view, and on the right (for this early testing) is the MAME view. Later these will also be the 'perspective-corrected' views, that are part of the process of making the 2d MFME layouts into 3d Arcade Sim machines... but just dealing with the fundamentals for the moment... The commits can be seen live here, though they're all fairly technical at the moment: https://github.com/johnparker007/Oasis/commits/main/

Thanks for having a look, much appreciated! I set up this mini-side project, as there were a few folks who had high res (much higher than the MFME Layout) resources just locally on their home drives, and these machines are only getting rarer to see in order to take a photos of, so those high res photos are a super valuable resource for future layouts