#NES/FC/FDS/Dendy
1 messages · Page 12 of 1
Why would you do that? The old Genesis core is still available as an alt core. Also if you really wanted to go that route, it would be easier to just rename your genesis roms to .32x and load them in the 32x core
https://www.youtube.com/watch?v=j0wM5jNrsjI
Seems like it just crashes to a black screen
Just crashes the game but still works!
it should mess with the MMC1 register and then execute STP if the system works right
which stops the cpu
if you dont implement stp, it will try to keep going and be weird
Can the revision of the CPU/PPU or what values there are in unitinilized RAM have any effect on this, or will it always be the same regardless of those?
That video is on real hardware using an original cart, so that checks out
All you guys talking about mappers while I’m busy playing games 
googling 5 screw metroid n the early days they used 5 screws and then switched to the 3 screw. Some games a more desired as the 5 screw because they were usually caught in the transition when 3 screws came out, so some games there’s less 5 screws that were made. So they are more rarer in production. Also before the famicon to nes converter came to America from 3rd parties, a few of the games used the Japanese pcb boards in the 5 screw and that was a easy way to play imported games on an American nes. Basically the imported games had different pins than America games. There was a converter built into the 5 screw American games that used the pcb boards from japan and you would swap board out to play other imports that may not be here fucking video games are awesome
There seems to be only one ROM revision for US Metroid anyway (not counting Virtual Console)
Doesn't necessarilly mean the mapper never had revisions, though
That's always fun to deal with :P
Yep, just thought people here might find it interesting.
also only macos
Interestingly enough, MetalNES is by the creator of Nesticle
But yeah it is way too slow to be playable
it's kind of a research tool, like visualnes
which is similar
made by sour several years ago
I do like the latest issue that’s open for MetalNES
Thought it sounded familiar, sadly the github hasn't been touched in about 3 years. I may still try to get it running on my Mac for giggles.
“Are you still amongst the living?”
It doesn’t compile on mac, there’s a bug iirc
I thought it was macos only
It’s MacOS only yeah, but it doesn’t even compile on Mac
I see
That’s what I heard from 100th Coin
The best NES emulator
I mean, it's only had 7 commits, and the first three were just the readme file. The rest were between Feb to March of 2022. I remember reading about this when it was first revealed but yeah, looks like the dev started it and more or less immediately lost interest.
Is something broken wark? 
@fierce estuary it is just for understanding something, for this game God Slayer - Haruka Tenkuu no Sonata (Japan) from a cold load of the core and load it, SNK logo is shows.
If I reset the core or reload the rom, the logo is not shown.
if you reset the ram it should show the logo
ok !
try setting it to 00 and see if it fixes the problem
00, FF or random the SNK logo is still not shown
hmm, is it MMC3 failing to reset some state?
this behavior is also present on last release so it is not a regression
I think real console also does this
with reset button
but I thought it was a mark in ram
on my agenda is to clean up the reset state of mappers, mmc5 definitely doesn't clean up properly either
it's technically not wrong, because real mappers tend to start in a dirty/random state
but for emulator consistent behavior, it's desirable
Ok ! maybe on save file or something !
it may be in the save ram not as a save but as scratch space and it's saved along with the games
AV Fami does the same, no SNK logo on reset
thanks for the confirmation !!!
The JP Super Dodgeball does something similar after starting a game. The game plays a clip of Kunio saying his catchphrase (namen nayo) and resets to the main menu instead of the title screen.
If you reload the rom, it goes also to main menu ?
MMC1 without battery so it is not store on save file for this game
yes. No save.
I think the US version actually does the same thing, but the voice was removed.
but if you reset at the title screen before going to the menu, the game returns to the title screen without playing the voice
Thank you
Started a playthrough of Rockman 4 Minus Infinity on the core, first time I've played it since probably 2022 or something. And I saw something that surprised me, in a good way.
So I have a R4:MI reproduction cartridge made with real MMC5 hardware and I remember distinctly playing it back in 2013 or so and encountering a very weird visual glitch on one of the bosses in the Cossack levels. I tried it on every NES emulator at the time and none of them reproduced the glitch. Then years later I tried it on the NT Mini and it wasn't there either.
But I was playing today and it's here! and it's exactly like my unmodded Twin Famicom
excellent stuff
really glad that even weird little quirks like this are represented now, it's pretty crazy
i've never seen that
i played r4mi on the nes core about a year ago and don't recall that happening, so i guess it's an expected quirk of the ppu, since it's more accurate now?
Yup exactly
It happens on a rev G Twin Famicom running the game on a reproduction cart made out of a real MMC5 board so it’s accurate
that's exciting the NES core passes all but one of the accuracy coin tests
i noticed with SNAC enabled it only passes 124/128
using the october 5th build
I thought it passed all of them now.
Well they're using an older build right
my bad i got the latest unstable build and see that now
it still fails with SNAC enabled for whatever reason
125/128 tests passed
the controller strobing and clocking and DMA + $4016 read gives a FAIL 1
im not stressed about it or anything 😆
one test fixed 😎
I think Kitrinx said something about SNAC failing because it's two controller ports and one controller isn't plugged in? Something like that.
Also, you have to change your timing setting to NES instead of Famicom.
thank you. i wasnt using the right search queries to look that up!
Accurate SNAC wen
If you think about it, that makes perfect sense because a real NES doesn’t pass all tests either. So you’re just setting it at NES mode instead of accuracy mode.
Kitrinx wins again! 
Now I'm waiting for a Kitrinx Megadrive core.
Imagine the possibilities
Playing Awesome Possum at its most accurate
I heard a rumor that @fierce estuary is working on Virtual Boy next
Paprium core soon?🤔
Well she does like modern games that use special chips to run
Out of spite she might hardcode a specific extension for it, so that people think it something new and undiscovered
Personally, I’d love to see someone work on an FM Towns core. That’s my white whale, the console I never got to play or own.
I'm going to file this under "Things I don't understand probably never should"
so I made a simple TAS-is-active slash input display, what do you think?
The dots in the upper left corner? Seems to work good.
yeah the dots
(it's one too short because I can't count)
you can turn it on without speedrun too, but then the red dot wont be there
Are the two groups for each of the controller ports?
yes
I like it. Simple and to the point without being overly distracting.
I figure it addresses a lot of speedrunner stuff because if you can see the input, it is obvious if they are cheating whether it be up + down at the same time or TAS of any means or whatever
Yeah, as real hardware continues to age, something like MiSTer can be a solid option for that community.
I made the module variable too so any core can drop it in if they want it
would it be hard to make the position of 2p input variable in its location (ie. bein on the right side)?
Looks nice and simple, makes it obvious
yes
I could stack them vertically though
Might be better should it ever be part of one of the cores which utilize more buttons
but looking good simple. Strange that the TAS seems to hold the last inputs when a stage finishes
it's not strange, it only updates when it polls
the game probably doesnt poll for input during that
I see
this is what it looks like when a real person uses it
order is done by polling sequence?
in this case I just fed it in the NES order yeah
I was getting into that as it cut off 
They just make me think of the CRAM dots on the Genesis
It's simple and gets the point across though, but for a human looking at it I admit the dpad being just a straight line is a little hard to understand as you are looking at it
surely there could be an obs plugin which just parses these pixels for a different overlay
Also the module being variable makes me wonder how it would work for systems with analog inputs, or special controllers like a paddle controller
Not that the NES had that.... AFAIK at least
dont use it for those systems
But vcs5200 had that anlog thing, if only I could remember the leading dev...
NES has a rotary controller for arkanoid
it has lightgun
but like.. dont use it for that
that's not what it's for
Nothing wrong with analog values as bits
Was real neat watching a TAS Solar Jetman run on my MiSTer with this, though most of my r08 files desynced pretty early 🙁
Never managed to get past world 3/4 back in the day in that game
solar jetman has to start from reset with offset 0
I must have an older version of the core then, I see the Load TAS option but I don't see the Speedrun options
Where do I get the latest version of the NES core with TAS? I can only find the older versions or ones without the TAS feature
I have no idea
???
I think Kit is working on it - it doesn’t exist yet
Well, it does, but only in her heart ||and on her hard drive||
Ah, sorry. I thought that version had already been released. Also just realized the one I had was older than the pinned one, I thought I was already on that one.
Maybe it has, I dunno
Oh that does look better.
@mint lichen that's forced on for tas, think that'll be okay?
@cursive sable there's your core
Thanks, sorry, didn't realize it was not released yet.
it's a feature im working on, i've thrown a few builds in here to test other things that had a wip version of it incedentally included
Red dot means TAS? If it's forced on seems good
yes
red dot means tas
you can turn it on manually too for just your own proof of inputs but there's no red dot
unfortunately it doesn't catch snac atm
but I can add that
Yeah, makes sense, simple solution
stacked is nicer
I wanted to take up the absolute minimum number of pixels possible of screen real estate
but would it be theoretically possible (without much work) to have a one pixel outline per input (or essentially a black line on top and bottom)?
I use the Retrospy script for my controller display, works pretty well
Odd, when I try offset 1 it works for me just like in your video, at offset 0 different scenes play and it crashes
Ah, ok
Trying to get the Battletoads GEG TAS to work, it's only 50 secs. It's synced to the wrong warp but it goes to a cutscene instead of the ending text. Not sure what's missing for it to finish right. Offset 0 worked to get there. I tried all the RAM settings and it was the same.
dunno, could be a mapper issue or core issue
or implementation issue
or I just dont know how to use the movie
the warp one works
Just the regular warps one works? Taking the intended warps?
This is the GEG one, looks identical to the TAS but warps to the wrong screen, should go to the ending sequence in this pic
yeah
it getss stuck on the dialog
this is why I wanted tas, to find these things
feel free to test with the latest core and that pack I put, make sure mem clear is set to bizhawk, mode set to polling, and offset set to 1 for most of them
some games require a specific revision (eg Castlevania needs prg0)
Yeah I was thinking maybe 4 dots in a cross pattern with nothing in the middle? So you can tell left right up down
that's hard to draw
Appreciate the input display being implemented for TASing stuff
the point isnt really for novelty, it's so people can see the inputs and verify a real human is doing legal things
so long as there's a legend/key I'm sure people will figure it out
it resolves "l and r at the same time cheating" issues too
yep
Got Kirby Game End to work, but it doesn't work every time. I remember from watching RTA runs hearing that even if you hit the correct frame for the glitch it can fail, so that might be accurate? Not sure https://streamable.com/avuiot
Nice, I was able to get the Solstice 100% run to work with this, was desyncing before
I hate to ask this of you, but is it possible to get a 60fps video of the orange pacman pixel? we're trying to do some research on the behavior of scanline 0 🙂
oh
this is important
for games that requireme memory init, make sure you load the tas first, and the game second
because the ram init right now only happens when you load the game
Oh yeah, I had to bring up the TAS playback feature in the moderation server of SMB2J as a precaution but knowing you’ve added the input display mitigates a possibility of cheating with MiSTer
Obviously people can edit the source code, but any mitigation is great
Would this work if I had a different game loaded, load a TAS, and then load the game the TAS is for? Also, does this mean for these I should not reset the core after loading both the game and tas?
yes
the tas stays loaded and restarts whenever you load a game
some games require a manual reset, supposedly
that's fine
not all games need memory cleared
@glacial turtle you love spreadsheets right? a spreadsheet of which of those verified tasses break would actually be useful
https://github.com/alyosha-tas/NES_replay_files they come from here
So a testing sheet for all these? I can knock one up tomorrow (late here right now).
I have this one recorded from the other day. Let me know if it's good enough.
that is perfect, thank you so much
No problem, happy to help out.
there
orange dots perfect
hopefully I didn't break anything with these new ppu tweaks, but it seems okay
new update 🙂
https://github.com/100thCoin/AccuracyCoin/commit/f675f78adb61d23f1157901d5eb5cfc2fae6502d
at this pace it might have 200 tests rather sooner than later 
That's not necessarily bad. I helps with fast regression testing.
was there even a core update or did mister just pass the new test without changing anything?
MiSTer core is 130/129
didn't mean to imply that, i just think it's a bit funny resembling a moving goalpost as more tests are made and adaptations to the core being made
I used the last RBF shared by Kitrinx
So like, is this a war of attrition between the speedcoin guy and Kitrinx?
or maybe it's a game for the two
"bet you can't get this to pass!"
"oh sure, i will!"
Awwww Kitrinx has a new friend
As it's an 8 bit system, can it have more than 256 tests?
the most recent test is testing for the pac-man orange pixel, which I discovered
it's actually a complicated set of scanline 0 behavior
I was watching the MLiG video on the MiSTer from 2022 and it's incredible that things are still being developed and refined
It's a very exciting community to be a part of
Meanwhile my wife goes to a book club. What a nerd.
Is the orange dot fix included in the stable release which is being uploaded right now?
that video was pretty good, they should do an updated version
I dunno
I think partly, it'll show rgb behavior though
Sounds like we’re due a new stable core 
I see that your latest PR not included. Thus I guess it is not.
the latest pr fixes the behavior to be properly composite behavior, but it's been there for a while
it's pretty minor changes
wont impact much
mostly tas alignment stuff
So this is newer than the soon to be uploaded new stable release.
yes
Thank you
Oh, it exists - like two days ago
I'll get the testing sheet made up in a bit, need to get some dinner prep done first
LOLOL this was good
some stuff that's important to note about real hardware TAS things:
- They were made on an emulator. Some were made on BizHawk which is pretty accurate, some where made on FCEUX which is not quite so great.
- Several only work on specific alignments of the PPU/CPU, like battletoads GEG is apparently very alignment dependent
- The game has to be loaded last to make sure the memory is cleared to bizhawk
- Most of them should use an offset of 1, but sometimes they use 0 if they are meant to start from a soft reset
- There are different revisions of games and it matters. Castlevania for example uses Revision 0
- The current maximum size is 250ish kb for a TAS video, anything larger will fail
thatss a great post, thanks for explaining it
"hardware verified" actually means they managed to get the tas to make it through one time, with some alignment

I actually use way more tests than just AccuracyCoin
note: they require some understanding of what they are testing to use often
For the snac and test ko on AccuracyCoins I wonder it could be link to this issue ticket
https://github.com/MiSTer-devel/NES_MiSTer/issues/381
I care so little about hacky 2p snac
I'm halfway done with fire emblem chapter 12 of 25 I think.
Fire emblem 2 is gonna be a joy I think.
btw has anyone else actually verified this?
https://docs.google.com/spreadsheets/d/12lRBd49ZHRxUOizItK4Aa0Rq41mRjV2yT-w-p0FDerk/edit?usp=sharing
What else do you need to track on this sheet? Do you also want the bk2 files on there?
no
ideally pass/fail, if fail HOW if failed eg Immediately, Desyncs late, random desync, the mapper, rom revision used if there are multiple (most are going to be rev 0), and the settings used to pass
for example castlevania III will be a random desync, the behavior varies from run to run
submappers can be in the form of 5.2
like mapper 5 submapper 2
So whatv extra collumns would be useful, one for mapper?
Also, some of these files are quite large (bad apple is 11.5 meg) is there a file size limit?
I just told you
#1091056042667937944 message
to be clear the goal in this is to isolate which games don't work for me to try to see commonalities so I can determine if it's an issue with like, mapper timing, address timing, apu timing, alignment, or something else
the most important thing is when/how it fails and what mapper it uses
OK, have updated with column with the files sizes in there, and a Mapper, and Fail Reason columns. Hopefully enough to get people started. Anyone who is able to help test please just request edit access to the sheet
I can do some fancier formatting and tracking once we get some more info in there
That would probably be the most ridiculus one to test, but yeah, way too big, I just went and deleted any r08 file that was 250k or larger from my set on my mister
I kind of have doubts about that, what do you think he means by "flip the controllers in the setting"? swap? Maybe someone can verify 1p and 2p sometime, if not I'll try to.
swap controller 1 and 2 probably
I would have definitely noticed a problem before, maybe its something new
Huh, weird the V2 Ballon Fight TAS keeps pausing and mostly ignoring the bonus levels, I assumed it just desynced, but it's still working normally on the normal levels
I guess it's just like that
there's several snac-related bug reports
but I have no way to verify them
Ok i'll try to check them too
It's either to save time or minimize inputs, I would assume. It probably counts down if you collect balloons
Good to see White Hat logging on the sheet. :-).
Anyone else able to help test TAS files, please just request edit access on the sheet.
https://docs.google.com/spreadsheets/d/12lRBd49ZHRxUOizItK4Aa0Rq41mRjV2yT-w-p0FDerk/edit?gid=0#gid=0
What are the best ways to find out what Mapper a NES rom is using, and is there an easy way to do it on a Mac?
Having that as part of the core somehow would be super useful
What are these r08 files and where would I find them on the microSD?
You can download them from here:
https://github.com/alyosha-tas/NES_replay_files
copy folder over to your NES games folder on the SD
To download them all easily click on the green Code button and select Download ZIP
I just load them in mesen
Is there any key for what the flashing dots mean?
maybe this helps:
if (custom1)
new_color = 6'h31;
else if (custom2)
new_color = 6'h34;
else if (w2000)
new_color = 6'h16;
else if (w2001)
new_color = 6'h03;
else if (r2002)
new_color = 6'h17;
else if (w2003)
new_color = 6'h15;
else if (r2004)
new_color = 6'h09;
else if (w2004)
new_color = 6'h28;
else if (w2005)
new_color = 6'h1A;
else if (w2006)
new_color = 6'h12;
else if (r2007)
new_color = 6'h21;
else if (w2007)
new_color = 6'h06;
oh you mean the controller inputs or the debug dots?
moondandy
do you even NES bro?
The flashing green TAS dots, and I had a red one to the far left when I did a quick check earlier
Red means it's playing back a recorded r08 file, like a camcorder having a red recording light. The green dots are the controller buttons/dpad
I quickly loaded up Bugs Bunny and the TAS file and flashing squares but it didn't do anything, it is right in the file size limit though
I am curious about something though, can these dots only be displayed in the 4:3-ish area of the NES screen? Or would it be possible to display them further to the left in the blank area of a 16:9 screen?
Someone could just crop them out that way, so that wouldn't help identify a TAS was playing
I put them in an area where it would be really weird to crop them
for that exact reason
it should be BizHawk memory, polling, offset 1
load the tas, then load the rom
For me it never works if I load TAS then ROM, have to do it the other way
Yep, that's it, I was opening the menu with the controller
the menu button wont cancel it, but a key combo will
I'll have to recheck some of the Fails then, got a number of Passes tho
Double Dragon 2, 2 player one is good. That's a fun one.
Ninja Gaiden 2 failed but I may need a dif rom
neither ninja gaiden worked for me, nor neither tiny toons
top gun fails fairly fast
archon worked
time lord also seemed to make it
tiger heli too
wizards and warriors, both runs, succeeded (prg0)
super mario bros 2 fails for all version of the rom (prg0 is the intended one)
ironsword succeded
Time Lord failed in stage 2 for me, but I might not have been actually changing the RAM since I was loading backwards
castlevania both runs succeeded, castlevania 3 fails in random ways
I need to re check time lord then
some of them are so boring I tend to tune out
like battle chess
i've tried that several times and I think it succeed but I zone out
I figured if you are submitting a run or something you would be required to have them not be off-screen
Anyone tried the Silver Surfer TAS? It kept desyncing oddly. I think it seeds some RNG values that determine if an enemy spawns or not, where you spawn at the start of a level, or where enemies fire projectiles based on unitilized memory. So far everything I tried caused it to de-sync in one of three places because of these, sometimes even de-syncing differently if I hit reset without even changing any settings, except for one.
Setting RAM initialization to BizHawk appears to work, the Speedrun is still going but has gotten past all those problematic parts and is still going far past them right now.
I got quite far in it
then i think my build finished and I was too eager to try the new one
I've only neen able to get past the first level with RAM initilization set to BizHawk, so far it's still going
bizhawk intializes to 00 00 00 00 FF FF FF FF 00 00 00 00 FF FF FF FF etc
it was once believed frontloaders defaulted to that, which is wrong
but it became the standard for speedruns
That might be it. I tried all 00 and FF and both desynced at different places each, but at the same place for each every time
Setting it to No seem to change behavior at where it desyncs
yeh, well bizhawk is the magic value it wants
Yeah, looks like it's about to complete the TAS with that value
Would most TASes have been made on BizHawk? I know nowadays it's a popular emulator for speedrunning
fceux or bizhawk but both use that initiaization value
it's a good emulator, pretty accurate, top 5
Yeah, it's one of my top software emulators for accuracy too
What's everyone's thoughts on the below?
#1091056042667937944 message
#1091056042667937944 message
Watching the TAS of Adventures of Tom Sawyer and... just what even is this game? Why is Tom Sawyer in the sky fighting Japanese thunder gods and then hopping on a cloud like Goku to fight a Zeppelin? 🤣
Thanks y'all
Are there any reasons to not default to "random" reset behavior? For Metroid and others? When would I need the other ones?
Trying out the Wizards and Warriors all levels TAS. I know this may sound dumb but it caught me so I may as well clarify. For you first time users like myself, you have to load the game FIRST, THEN load the r08 file. Even I can do mistakes sometimes lol
Gotta be some desync here because I tried both the rev1 and original versions of the game and even I can play better than this TAS is doing
Most of the time it seems like it’s just hanging out at the bottom waiting to be killed 😣😆
The SMB maximum score one so far has been a TRIP. It’s unreal seeing the moves!
So far, sufficed to say it works perfect
Ninja Gaiden is a fail. Kept letting an enemy hit them at the end of the first level and then paused the game and stayed there
Same with the pacifist version except this time it loses sync where you climb the Coca Cola looking signs
Monopoly V2 works.
BTW, I did request file access but putting them here for now
The Monopoly 7 CPU doesn’t work. After the auction part, it loses sync and that flashing bar disappears
Monopoly 4 loses sync after a few turns and the bar disappears here too.
it takes longer
also some games actually break with certain values, I think 0 is safer
like that game we were investigating a week or so ago where one of the attract screens was corrupt, it can crash if the ram has a value
you should do it the opposite actually so ram clears
so far so good, no new bug reports since release. You all did a great job testing, thank you
@glacial turtle can I edit this sheet?
Yeah just request access
But would that game crash on the real NES as well?
possibly but probably not, because the ram is likely to decay to something tolerable for it in most systems, but if you switched carts it might
An FYI, I didn't mark the ones that were around 250-290KB as Fail for file size as those could do with a check, so probably quick to check that handful and mark the ones that fail
oh you have multiple pages
oops
well
I threw random data on the first page, can you fix it? 🙂
Thanks Kitrinx for all those changes 🙂
Do you even spreadsheet bro?
no
that's why I need you to
I just smack my code with a plastic hammer like a child
OK, have added new columns to the sheet (the actual NES tab with data on, not the info tab, for anyone who doesn't spreadsheet)
please add the rom revision somewhere
that's an important detail
like castlevania has two revisions and only one will work with the TAS (by design)
oh you have it it's just not populated
the stuff I put on the first page had some revision info
Got Game Revision on there as a column.
and some mapper info
I will be traveling to a gaming tournament next week, first time taking my MiSTer instead of NES & SNES consoles + flashcarts.
Since it can be an adventure trying to eject and reinsert the SD card without dropping it inside the case, I was thinking that I would use a USB flash drive for loading the randomizer seeds that I will be playing over the course of the week. Is there a simple method to get the NES core to find games sitting on a flash drive? I'm planning to use a 32GB drive in exFAT, in case that is relevant.
Afaik just remove the games folder from the sd card and put it on the usb drive.
Path needs to be the same otherwise.
Can we get the non dupes from this set added? I tested a few and they work, Kirby, Karnov, Chip n Dale
Kitrinx, have a look at the sheet now, I entered your data in and added a splash of colour
thank you 🙂
No worries
If someone figures out the exact KB where going over is a Fail I can adjust the conditional formatting on the file size, right now anything 250 and over is red, and I put 230-250 as orange as I imagine that is in the danger zone. Those can easily be adjusted
Desync means an issue somewhere on the core on the TAS sheet file or it could be something else ?
it doesn't imply a cause
just that at some point the script fails and breaks, like jumps in a pit or gets stuck
I see !
reasons a TAS can fail are:
- Requires a different cpu/ppu alignment than the core has
- Electrical difference in controller behavior from a real system (like too short a poll to trigger or something)
- I just implemented TAS wrong maybe
- Possibly memory differences though this is unlikely
- Offset frame differences though this is less likely too since it's adjustable
- Error in the core either in PPU, CPU, or Mapper code
most of them I suspect are the first thing
I did discover a kind-of-but-not really error in the core where I wasn't clearing the controller latch on cold resets, but that's not significant to anything but TAS which needs consistency
So with the ones failing, do you think some of them could become passes if something is tweaked in the core implementation?
do multiple tas software emulators exist which are compatible with the files without any issues/converting?
Don't some of them fail because they were originally made with much older emulators?
well obviously. All of these passed at least once on real hardware, so there is some path to doing so
Ah, sorry, am forgetting these are all hardware verified
but I might have to implement multiple alignments, which I plan to do anyway
and/or fix various things
oh, that answers my question
Whitehat, I've updated the sheet with some new columns for info, for the ones you have checked be good to fill those out e.g. Mapper, Game revision
Cool, will do. Can we get these added? I think they are a little older but still console verified in r08 format, some might still be dupes since they are named backwards, author first, but there are a lot not in the other set
Wondering if it does make sense to add TAS support at all to the core given so many possibilities for failure?
it does to me. I want it as a tool to 1) check for regressions 2) test for accuracy and 3) just for fun
they are ultimately deterministic and there is a path to success for all of them, as real hardware can do it, so eventually they all should pass with the right settings
I am willing to bet before i'm done i'll have found at least a few subtle bugs that only trigger under very obscure timing circumstances
probably in mappers
light right now for instance i'm noticing that Mapper 4 games (MMC3) seem to be having more problems than the rest
so it's possible there is a timing related thing with the MMC3 IRQ? who knows. but that's the kind of thing I want it for
If that's going to make the core "perfecter", then I'm with it! 
the MMC3 IRQ is very alignment dependant
Up to Kitrinx what gets added. Not sure if we want the ones there already to be knocked off before adding more, but will defer to her
Do we need a column for Timing Mode?
technically since Metroid uses frame timing I think, but probably not
"frame" means it wont change the inputs more than once a frame, this is for games that do multiple polls on the same frame to deal with DMC DMA bug, but in theory Poll is the more accurate and accuracy demanding method
that accounts for every single controller poll
maybe a second page. Alys are cleaner so im sticking to them first, then we can work towards the less clean ones. It reduces variables.
The ones in the other pack use Frame in a bunch
for finding patterns the consistency is valuable
Cool, just making sure those ones are known
the failure that really has me scratching my head is donkey kong
Yeah, just jump over the barrel already
it uses frame rules so it might be heavily alignment dependant
Running Lode Runner righy now, weirdly on the end level screen is says Stage X Left 9, on every screen it says Left 9. Is this a known issue with the game?
Sounds wrong. I've been checking the TASvideos page for games then looking at the video if something seems weird. Also has info on which ROM they use
Can I ask a stupid question. Why is TAS support being added?
On a slightly unrelated note: Lode Runner and Choplifter on the SG-1000 (SMS core) are really close to their Apple ][ origins. I like them a lot more than the seemingly beefed up arcade and NES ports
Good shout, seems this is the same in the video
#1091056042667937944 message
Ah okay that makes sense
most things make sense if you torture them enough in your head
I worked in children's cartoons for nearly a decade lol. I know that sentiment all too well lmao
in this case though, the core passes every test I know of, so I need more tools to verify that everything is as it should be
I was running that one yesterday too. Was crazy watching the whole game beat fast like that, I had it back in the day but had no idea how the game worked. Also impressive to see that they managed to fit the entire game and all 50 levels on 24K of memory
Yeah, it is pretty wild. It was so fast at points I wasn't entirely sure how they were even doing it.
It did look like it was running in fast forward, yeah. Then you have ones like Donkey Kong (though that one desyncs early) or Silver Surfer where the character is rapidly flipping in opposite directions at absurd speed
Interestingly I fired up Mario Bros with the version I have on my SD, and it is running along fine as a 2 player with Mario and Luigi, and I checked the video for it and that is a single player run through
Actually, looks like they have come undone now 19 levels in and are just standing there and dying
A 1 player run doing well in 2 player for a while? That's a bit weird, though I suppose if mario never dies it can stay synced for a while
Elite was another NES game that I find impressive from a technical standpoint, though I also find it funny that the NES version is like 3-4x as large as the original version on computers
If someone has the correct Mario Bros rom handy that video is less than. 8 mins long so a quick one to check
Oh, Mario Bros, not Super Mario Bros, right
I tried that one and it eventually desynced, but that was on an older version of the core, should try it again
Joust is another impressive 2 player one BTW that seemed to work fine
If you are testing it would be great to get those logged on the sheet, just request edit access
I'm not sure I know what I'm doing, one of them was acting very weird, I got it to work but I am not sure if it was just chance or I got the settings right
BTW, just tried Mario Bros and it worked fine. The TAS goes up to level 19 and then ends, hence them just standing there.
You can tell the TAS is over when the input display disappears. The game has no ending so it's not like the recording can go on forever
@fierce estuary Does it make sense to have the input display order flipped? To me it makes more sense L R U D S St B A
that's just the order NES controllers feed buttons to the system
I can change it later
when I add decoding for snac
Oh, I assumed it was so it would be harder to try to crop out someone using up/down or left/right at the same time without it being obvious
no, that wasn't a consideration
if someone cuts off half that display it should be sus anyway
Gotcha, at least to me it would be easier to read the other way
Is there a way to tell which https://tasvideos.org/ files/videos for the different systems are verified on real hardware?
I think Where's Waldo is another one that needed a specific RAM pattern. keeps desyncing differently based on the pattern I set, but none of them finish the TAS
Yeah, Waldo has some ridiculous RNG, so it probably needs something very specific
Ah, nice, has a toggle
I think it's an mmc3 alignment thing
Aligned mmc3 wen
the way MMC3 works is that it counts the number of times A12 changes to determine when it triggers it's IRQ. A12 is controlled by the PPU and every cycle has the potential to change it. the PPU ticks 3 times for every time the CPU ticks, which means the particular cycle the PPU is on matters to when the CPU will encounter the IRQ. The IRQ being a CPU cycle earlier or later is pretty normal on a real system, but for a TAS changes the outcomes significantly.
on a real system the PPU and CPU alignment (ie how many master cycles have elapsed before the ppu and cpu start ticking) varies. On cold boot, the time it takes to get power to each chip varies, and at reset, the cpu clock divider is unchanged but the ppu clock divider resets.
additionally, there is an odd/even frame toggle in the APU which can impact an APU IRQ and DMA behavior. This technically starts undefined and is unaffected by reset
to a speedrunner, these things can impact RNG, to a regular player, these things really dont matter, to a TAS, they break everything
So to find the cause you'd have to let the tas run with different alignments, see which one sticks and test whether it sticks deterministic?
I'd guess the tas would use the most beneficial alignment so maybe it's a known eu setting
on a real system it's not super easy to tell what the alignment is
chances are the default alignment on mister isnt the same as the default alignment on fceux or bizhawk or whatever
Right but would the tas work every time on real HW without x number of resets. That would not make sense
no, they dont
some of them require resetting a bunch
some dont care about alignment
And you can't query something on a real system (or write some code) to get the alignment? I'm making it sound easy but I'm just thinking aloud
not under normal circumstances, and the alignment is ephemeral
in this case we're talking about real retail games, so there's no opportunity to query anything without impacting the outcome
Schroedinger’s alignment
more or less
flash carts have some hacks to do it I think but using a flash cart also impacts the outcome
Debug headers?
tas videos generally require a real cart in a real system
what?
How can you play a tas file on a real system then?
I mean yeah you can hook up a logic analyzer I guess but.. that might be out of scope of most people
you plug a device into the controller port
I see. Right. I've only 'hacked' (with permission) some of our routers for pen testing and we always check for debug headers and cli access but NES probably has neither 🤷♂️
Anyone tried the hardware verified FDS files?
https://tasvideos.org/Movies-fds-verified-Obs
Are there any in the r08 format? I didn't think there were
Ah, good point. These are fm2
I have some doubts about how the bios will line up on mister atm
but I haven't tried
there's a script to convert fm2 to r08 btw but it requires some settings
Is there a list of how No-Intro maps their versions to GoodRoms set so I can put correct version in the spreadsheet? I'm having trouble finding it in a search
or are y'all looking at iNES header?
Not sure, I've only been using MESEN to see the mapper. I have a folder of Revisions in the HTGDB rom set I have, so I've been checking there for different versions
The NES had a quantium CPU?
There are devices like TASBot that basically do the inputs for you on a real console, it's been shown off at GDQ a lot doing some insane things like reprogramming games with ACE in realtime to make the console stream a skype call, or keep adding new content to OOT as it's being played and stream the twitch chat to it
you can load it in mesen or something to see the header number
if you see Prg 0 or Rev 0 it's revision 0, sometimes revision 0 has no revision listed
Prg 1 (or 2 or 3 etc) or Rev 1 will pretty much always be included
Update AccurayCoins test rom
https://github.com/100thCoin/AccuracyCoin/commit/e1cfebcf995ee545aa4352b9040f32f225dd4adf
seems minor
yes it is
Core is still hyper-accurate, right?
nope, ultra-accurate 
Maybe... Too accurate
No, no, no. It's omega-accurate!
we'll know when it hits 100% accuracy because a time portal will open and send it back to 1982 so nintendo can "invent" it
As long as it's not too Raph
I love the NES core... It's so bad
Some real shifts put in on testing the tas files, only a handful left. Top work!
Would be useful for when "No" is set for the "Memory" to mark that on the sheet
But not like in touching your kids way, more like an LL Cool J way
One of those needs to be removed, kyman, it's a bk2 file
About the only ones I haven't tried at this point are the ones for romhacks/homebrew
Everything is pretty much tested now
There were a few homebrew where I couldn't find the exact revision so they Failed, but I'm pretty positive they would pass
Only ones out of those I tested I could not get to work were Metroid, Donkey Kong, and both Tiny Tunes IIRC
oh, and Where's Waldo
Though sIlver surfer is very tempermental
Thanks, I didn't spot that, have deleted it
So only two left:
Famidash_2025_April_Fools.r08
Nova_World_6.r08
I left notes on those
Great to see all the Mappers added. Would be good to have all the Memory ones that were "No" rather than "Bizhawk" etc. marked as "No" just to confirm that is what it should be
Nice to see 70% pass though, hopefully enough info now for Kitrinx to dig into why some aren't working and get that rate up
Nova should be a pass, I just don't know what setup it needs, the regular game was good
Woulden't any option work for those that are "No" though?
a lot of games dont care about memory
Ah, so Bizhawk is likely a false positive.
for a number of them anyway
I'm going to be away for a few days from tomorrow, so on my phone but no computer to do any spreadsheet work. Is there anything that needs added before I go, or are we good to not be adding a new tab for more files or anything else large before mid next week?
Huh, I had some different results than on that spreadsheet.
For Silver Surfer I could only get it to work if memory was set to BizHawk and I load the rom after loading the tas, any other memory setting or order of operations caused it to always desync in one of four specific locations in the first level for me
I also played Solar Jetman V2 on default settings multiple times just fine, unless the tas still happens to beat the game while desynced
it's hard to know which it impacts off the cuff but it's a good default
afaik none of these require any memory pattern other than bizhawk, but most don't require any memory pattern at all
Yeah, I have mostly been setting mine to bizhawk when testing. If a game needed a specific memory pattern and it's not bizhawk's, then no woulden't really work unless you get really really lucky with what the ram happens to currently be initilized to
can you check your solar jetman rom for me and see if it's using a nes2.0 header what the mapper is set to?
Just tried Marble Madness, and it completed for me again. Though on the second to last turn it looked weird, almost like it failed and loaded a save state
Yeah, one sec
marble madness seems to exploit glitches
I had a few that would not work with Bizhawk, I had to try others
I don't think there are any glitches in Marble Madness, it just stays ahead of the camera and pauses at the very end. Not sure why mine kept failing the final bounce off the flag
iNES 2.0, Mapper 7, submapper 2
Would it be worth having the core default to Bizhawk rather than No, do you think?
no
smb1 seemed pretty forgiving from what I remember, does any of the settings mess up those runs? Was thinking that might be a good one for people to try first.
Solar Jetman uses glitches too right?
If you can get it to pass, you can overwrite mine
All the Mario 1 stuff I tried worked, hacks included
Let me double-check solar jetman, I last tried it on an older version of the core that didn't even have any settings to change
smb1 is one of the most tolerant games there is, it uses no fancy mappers or system features
if your sprite0 hits work chances are smb1 is running about right
Memory contents can have an effect, because people use to use a game like tennis to change ram, then insert smb without powering off and reset and I think it changed the minus world or something I don't remember.
Woulden't that only work if the game assumes an uninitilized part of RAM will always have the same value, and uses that value?
Would be pretty silly to do that since, well, the value can be anything if it has not been initilized
I think it was only used to seed RNGs speifically because it can be random
Go ask SMB3 why that's a bad idea XD
How can you get to glitch worlds in Super Mario Bros. by using the game Tennis? It's all explained right here.
LINKS
Twitter (updates): https://twitter.com/RetroGameMechEx
Patreon (support): https://www.patreon.com/rgmechex
SubscribeStar (support): https://www.subscribestar.com/rgmechex
Discord (discussion): http://discord.rgmechex.com
there a nesdev forum page about it too, if you want that
Wasn't there a similar trick where people could beat Dragon Quest III extremely quickly?
on emulator it could be very difficult to trigg i think
it would be pretty easy if you can poke the ram values in the emulator i think. The trick was involving continuing and not the minus worlds, I was wrong about that.
I haven't heard of the DQ3 trick but maybe
Should be pretty easy to test on emulator just boot game and change 07FD in ram then continue. Wark (edit: I tried it, it works fine)
what world you get is random based on what you did in tennis
Have you ever seen that trick before Kit?
I see 🙂
same, honestly
Trying Jetman V2 again and now it keeps desyncing for me, I had tried it multiple times on the first version of the core with TAS (the one that had no options) and that one had passed multiple times
He does mention it checks for cold or warm start and clears ram, which might explain why smb runs usually work. He also mentions Tennis only clear some of the ram, and other games might also only clear some ram which could be problematic for tas runs
not specifically for smb but swapping carts its usesd in a lot of hacky things
So those could require the right settings that the emu used for the tas run to work right
Cool I showed you something new 😃
Trying out spookatron, which is like a twin stick shooter homebrew. It allows for two nes controllers, a snes controller or virtual boy controller on real hardware to achieve the desired control scheme.
The only one that seems to work on mister are the two nes controllers.
My question is is there any way to map this games controls to a dual analog controller? Keyboard remapping perhaps?
this spreadsheet paints a pretty clear picture to me
mapper 7.2 (AxROM with bus conflicts) and 4 (MMC3) have issues
mmc3 passes all the tests so likely it's a timing or alignment related issues
not irq
So what's the next step?
are all mappers (supported) tested ? if not there is another source of r08 to test them ?
Just a small selection of Mappers covered by this set of tas files that are verified on real hardware.
there's other r08's but they are older and mostly frame
How does someone physically record one of these tas files and then play it back on real hardware? Is there something you have to buy or build yourself?
I have a last question regarding the new horizontal border stuff: On the analog out, is always the full picture sent to the TV now (with the overscan options like full, borders etc just affectimg the digital out)?
Generally yes, see #1091056042667937944 message
I am a bit surprised it isn't a bigger thing for people making their own Arduino tasbots and running TAS files on actual consoles and recording the videos. I would assume if someone goes to the effort of making a TAS file of their speed run they would want it confirmed on real hardware, and it obviously helps in getting emulators to become more accurate (like Kitrinx is doing now verifying against ones in real hardware). Looking at the site linked, very few systems seem to have files verified on real hardware.
I'm not surprised, it's like three levels of niche, haha. Speedrunning interest into TASing into hardware confirmation, very few people are going to go that deep.
most of them are probably on this this server or the tasbot server
if you're really interested https://discord.gg/GySG2b6
if you remove the fds filter, the amount of hw verified does increase noticibly
I find it odd there doesn't seem to be any PCE videos on real hardware, and basically no Mega Drive for instance. Lots of emulator ones though
I think emulator ones can be fun but they don't really serve my main purpose for them which is as a verification and debugging tool
Have you asked there, or anywhere else, if people doing recordings on real hardware can do some more?
Maybe it is one of these situations where for example, no/not enough Mega Drive ones exist but if someone were to say "we would like to support tas in MiSTer Mega Drive core, but in order to do that we need files verified against actual hardware with a video, so we can be sure the core is accurate enough and then you can not only run your tas files on the core but be confident it is accurate enough for speed running on" then maybe some people would see a reason to
I think all of the runs are originally made frame by frame on emulators and they just play them back on real HW, it would probably be possible for you to record your run on mister and then play it back.
If those non hw verified ones work it would tell you the core works like the emulator at least. like ram starting contents are the same. There is revisions of HW also, I don't know it any would work with one revision and not another but it's possible.
Alyosha seems like the only person consistently doing hardware tests
@amber willow have you ever looked into making tasbots? In my head the connector part would be like your SNAC adaptors in reverse, with an Arduino attached
Yeah I made one out of an arduino. I used it to try and verify snac was working correctly with it, I tried it on real hw too.
I think I turned Kit onto tas stuff too around that time.
Ah nice, did you just do NES or different console connectors?
I only tried nes
Like I played back the same file on emulator, mister through snac and the tas load feature and real hw. Some worked on all 3 platforms
But some didn't and there was so many variables it was hard to figure out what was wrong
So I'm glad Kit is getting this figured out 🙂
Yeah, I can imagine. Why it would be great if there were more (and for other systems) of ones verified on real hardware, so people know they are actually legit speed runs and not just ones done on inaccurate software emulators. Then emulators can be verified against tas files known to be accurate on real hardware and match that behaviour.
So unless I am missing something, if Kitrinx were to go on to add TAS to the 2600/7800 core, if none of the tas files for those are verified against hardware, then you are just comparing the core to how these run on a software emulator, which could well be inaccurate.
Is native composite output (like how the original PPU generates it) technically possible on the NES core? I know it's not implemented, but I'm wondering if that's a technical limitation or if it's just low on the priority list.
Anyone know if the NES was meant to display at 5:4 or 8:7 aspect?
neither
All I know is the dot clock is 1.143
According to pin eight anyway. I just apply that stretch to the 1:1 pixel image
Well I actually let the RetroTINK do the calculations but if I ever want to do it myself, that's the stretch I apply
4:3
It was meant to be viewed on consumer CRT TVs in Japan, which were 4:3, 60Hz, 15kHz sets
Although I think the NES displayed a refresh rate closer to 60.098814 Hz or something like that
But it's also not a basic stretch to 4:3
Yes, due to the variance in overscan on CRTs
I.e. how some people think the 320x240 displays of some consoles is 4:3
Nah it's not the overscan. It's outside of my technical knowledge tbh but it's to do with how the beam is blasted out
it's 4:3 of 283x242
I always thought the NES displayed at 256x240 without counting overscan
and now you know better
Yeah :P
the overscan is part of the picture
the drawn area within the borders is 256x240, with some of that 240 being a little disfunctional
Also was this a response to me or a response to the aspect ratio question?
aspect ratio
kind of sort of
I could make the digital part of the logic that has the johnston counter and stuff, but it ultimately connects to this big serpentine silicon thing, and then it becomes a matter of guessing at the analog voltage of it
you'd get some kind of integer which then you'd have to convert into voltage on a wire somewhere
Is that what leads to the coveted shimmer?
Is creating analog logic even possible on an FPGA?
Was asking because those were the options I had when I was messing with puNES. 1:1, 5:4, 8:7, and 11:8, with the last two being labeled as NTSC and PAL.
you can emulate it in various ways
Like using an external DAC or something?
8/7 is the reduced fraction of 256/224
I'm guessing that's a 1:1 pixel scale, which would look too thin
Haven't used puNES in a while, though
Yeah, I was trying to check them out and 5:4 felt the most accurate, but I have a bad eye for this kinds of things
Well, accurate is probably the wrong word, more like looked the most correct
4:3 is what you want, though
Oddly, not an option
Yeah, IDK about puNES. When I use software emulators, I stick to Mesen for NES
The aspect ratio options seem odd
I usually use it through RetroArch, which has many display options
mesen would be what I would use of the software choices
Or compile the NES core using Verilator or something and run it at 60 seconds per frame
Probably with no video output either :P
I always wondered, any idea of the voice synthesis in those Jaleco/Bandai Famicom games was using an analog chip? Nothing software emulation or FPGA seems to support it
last I looked at that, which was admited me several years ago, nobody had managed to get a dump of the data
Yeah, I have been trying to find out what I can out of curisoty and there have been several threads on nesdev trying to work it out but they mostly peter out, I think the most recent was from 2022 where someone decapped one of the chips
I think I once ran into one of the chip's whitepapers on archive, IIRC it has syllables encoded on it that it combines to make words, but I have no idea if they are stored digitally or somehow analog
I noticed if I change system type to pal it fails some tests in accuracycoin is that something that matters if playing pal games or doesn’t it effect anything?
Generally depends on which tests fail, but unless you are doing glitch ACE-based speedruns chances are you are not going to notice a difference in any test that fails on any even moderately accurate ancient emulator, much less on the NES mister core
Most emulators that can actually run games at all are going to pass the ones that could effect gameplay in a noticeable way
The accuracycoin tests were designed for a specific revision of the NES, and are so strict that even on real hardware some of the tests will fail on a Famicom or PAL console, or even on the NES they were made for if running it off an everdrive in the normal way
Thanks for the info that’s good to know 
NP
those tests aren't made for PAL
Hey I tried it too, smb was boot rom, then loaded Tennis and ran around, then loaded smb and continued and it did work. I started on a world past 8.
But I was thinking about that warm boot cold boot thing. Should loading a rom be like cold boot? How does the ram fill thing work? Just wondering if that could effect the tas thing. Like maybe it would work good the first load only and say if you didn't have a boot rom.
loading a rom is a cold boot
the cpu, apu, and ppu all have specific behavior that's different on cold and warm boots
this is the problem right here with mmc3 games
see how it's writing FF to 4017?
that's wrong
How does smb have the ram values from tennis if it's a cold boot?
4017 is controller port 2 as you know
you can cold boot a console without clearing memory too
sram will hold it's values for a few seconds
maybe make a build that fills ram with whatever pattern you pick on each load. it might help
I didn't test your tas build, I just tried tennis on one of the release builds
ok I might need to update, I thought you were still working on it. Good job getting it in 🙂
the latest core has a few new features in advanced and different overscan options
as well as a lot of accuracy stuff
Ok setting Ram Clear to something other than No seems to make the smb trick not work so crisis adverted.
No tas stuff in release though, you had me fooled there.
https://www.romhacking.net/hacks/7899/
Any chance at getting this to work on the core? Just boots to a grey screen
#1091056042667937944 message
It works when the header is fixed
does the 100thCoin test ROM check for the TMNT randomness issue?
what is the tmnt randomness issue?
I saw in one of Kitrinx last videos some options under Speedrun > alignment. So maybe she's in the process of publishing a solution soon. But I don't think that build has been posted yet.
it's impossible for a test to test for variable alignments because it can only have one alignment while running
but yes, im adding that
Variable alignment is not misalignment. There are a finite amount of 'correct' alignments but as I (hope) understand correctly some are more beneficial for speedruns. Example: frame lag that doesn't happen on certain alignments. Do correct if this info is false
Jotego already has an implementation of the Jaleco ones in fpga, but as Kitrinx mentioned no dumps have been made. https://github.com/jotego/jt7759
it's kind of right
there's a few different ways a system can align
the first is how many ppu cycles have happened in a cpu cycle as the system boots. Like dot 0, 1, and 2 will all happen inside a single cpu cycle, so is it dot 0 1 or 2 on the cpu's phi1?
second is what point in the ppu cycle the M2 rises and falls. There's 4 possible master cycles it can land on per PPU cycle
the third thing is the apu's odd-or-even flag, which determines which alternating cpu cycle DMA's and APU IRQ's occur on, this also boots in a random state
though less impactful, it's also noted that components like the address decoder that triggers the chip selects can have variable timing depending on temperature and things, and that can impact the way the intra-ppu cycles line up with m2 reads and writes
HI Everyone it's the first time that i write something in there so ....hi just wanna talk about compability 'cause i'd heard that the NES core is now 100 % accurate ...though i notice that a game that i own and love who goes by the name of CAPTAIN SAVER has glitches and cannot be played in a proper way ...i dunno the cause so that's the reason i end up here on behalf of @little schooner who told me that on some of his video...btw i'm french so be gentle with my english .....merci a tout le monde !!!!!!!!!!!!!!
Did you download and install the latest version of the test core? The links are usually located in the pins of this channel https://cdn.discordapp.com/attachments/1091056042667937944/1427131027901513748/NES.rbf?ex=68f25c17&is=68f10a97&hm=e82c14786abc88723a9c3509daf134802c30d93bca739b76e483280ebc3a1287&
i've just tried as we speak but no changes on famicom nor NES behavior ... it maybe use some special mappers ? who knows
Can you describe the glitches with more detail? I played the game for about 10 minutes now and it seems like it is running okay. But I am not familiar with the game.
ow i'm surprised ...maybe i got a bad rom 'cause in my case, as son as the intro begins glitches appaers on the guy's face (splited in two)
as youy can see if you have played at this game ....that 's weird
Looks completely different here. Try finding a new rom. (Edited for easier to understand English 😂 )
😆 sounds legit, i will try to find me another one or if you could share yours that 'd be nice 👍
That we can't do.
oh sorry ! finally it took me a couple of tries but that's ok now !! thanks for take the time to awnser 👍
Odd, are these chips hard to dump or something? I know many have looked into them and even decapped them, but nobody has been able to dump them? Is it true that they have some analog components then? I would assume that makes dumping rather impossible and you would have to "rip" the analog audio off if them somehow then in that case?
Captain Saver is also known as Power Blade 2 in US if you want an english version
maybe someone translated it to french, You'll have to look
Also are you sure that's for the same chip? That link says it's an implementation of the NEC 7759, but from my understanding those games used an NEC D7756C and NEC D7755C. Also there was a Bandai game that used a Mitsubishi M50805
Could this be a header issue
Their problem was yes, he found a working one I think
Did you read the architecture section of the readme in the link? All those NEC D775x are the same except for type of ROM. 7759 uses external ROMs. Most of the other part numbers are just different sizes of internal ROMs. Dumping the internal ones used by Jaleco is difficult. My understanding is the mask ROMs do not show up in decapping pictures the same way most are done. There might not be a way to read them at all.
what games was he talking about here? with the undumped roms
here was someone trying to get the rom from one of the games https://forums.nesdev.org/viewtopic.php?t=24220
I had made a list when I was looking into them, but I am not sure how extensive it is:
https://pastebin.com/7DYd109N
Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.
tell me more about this?
where does that game come from?
I found it in japan
make sure the header is set correctly 4E 45 53 1A 08 10 00 38 00 00 00 00 00 00 00 01
Seems like it was indeed a header issue, I tried it and it worked fine for me too, then I looked up an older set I had on an old harddrive and the header was different, when I tried that one I had the exact same issue with half the face being cut off in the intro
could it at least display the alignment info? not that it can test it if it’s set, but the user would have to note that the alignment never changes. or maybe asking the user to hit reset or power off, save data could note to continue checking for different alignments and “pass” when it detects changes.
not very well
Looks like Cyber Akuma got all the relevant ones. https://www.nesdev.org/wiki/List_of_games_with_expansion_audio#NEC_µPD7755C_(Jaleco)
The following is a list of games that use the Famicom's expansion audio.
hi everyone i wanna thank you all for helping me so far .... indeed i had some issues with a bad rom i had to search a couple of Captain Saver on the net a finally i found a good one .... i had some issues with my former rom, i posted some photos on my previous messages with @supple jasper who helped me to solve this out ...so thanks to everyone the problem is solved
just a bad dump i had i guess "merci a tous a la prochaine 👍 see you next time !!!!
I hope you enjoy playing your game!
problem solved just a bad dump !!!! merci 👍
Anyone having problems with TLOZ Perils of Darkness romhack? It freezes as soon as I enter the house to get the sword...
40th anniversary to NES
Yep. Same problem for me. Just tried it.
What is the behavior of this hack on original hardware? It's possible that it's a hack that only works in emulators or uses a weird custom mapper.
I was able to confirm but in weirder ways. If the game loads with no save data and you go in the house you can pick up the sword and the game locks up on exit. If you load up the game with existing save data the game locks up while entering the house.
But we need a confirmation from real hardware that the hack isn't bugged or that the hack mapper isn't incorrect. Mesen reports Mapper 1 horizontal
Seems to work on AV Fami. I tried with one ROM and it did crash on AV Fami but worked with another. The one that worked on console crashed on MiSTer, so idk what the deal is.
Are we all using the v2.0 versions of the hack?
I got it off romhacking, I think it's 2.0
an mmc1 game?
can someone try it on an older version of the core to see if it's regressive behavior?
Trying on the core before all the new changes and it seems to work
I see
it gets stuck in a loop reading 2002 waiting for something
i'd have to assume waiting for NMI
no, it's getting the vbl flag
29 40 F0 F9 AD 02 20
it's looking for sprite0 hit
looks like it's being caused by oam corruption triggering a little overzealously
it probably happens on real hardware sometimes too on certain alignments and systems
It actually did crash for me on AV Fami my first try, I thought I needed a different ROM and then that one worked. But now both are working.
so it's technically accurate behavior, it just happens a little too consistently since we have a single alignment
trying to find a common reason for these failing is frustrating. I suspect it is APU related but dmc doesnt see off no matter where I catch it https://docs.google.com/spreadsheets/d/12lRBd49ZHRxUOizItK4Aa0Rq41mRjV2yT-w-p0FDerk/edit?gid=0#gid=0
I noticed a lot of patches expect 1.0 headers for some reason.
I also noticed the exact same behavior on TLOZ Redux romhack, and it does mention the MMC1 mapper on its changelog.
but the Zelda hack crashed for me too after patching the 1.0 headered rom, so that's not it
I originally downloaded the roms already patched, but I patched an original 1.0 headeared rom and still happens.
it's causing oam corruption
it's not a bug exactly but it's undesireable in this case
im thinking about how to deal with it
is there a reference to know what some tests mean if they fail?
in particular the APU Register Activation and Address $2004 ones
probably no games really care about those though
Is anyone else having an issue with the nes core locking up the mister? I had a compile I build just before the latest release which was working fine, the ran update all and now the latest core shows "kbd mode: Joystick 1" then just hangs up the mister at that point. Have to unplug and replug power. Oddly the nes core I compiled the other day now isn't working either. No other core are affe ted, not sure what to do
Something like this always happens on the days I have an hour or two to myself 😂
Thought it'd be an idea to delete the nes.cfg but.. there isn't one
Nevermind, the cfg was there but didn't show up when I ls'd the directory. Anyway, nuking that fixed it
it depends what the error code is
that one can mean they aren't getting activated when they should be or are getting activated when they shouldnt be, but the most likely reason is because they chip select is based on the external and not the internal bus value for the top bits of the APU address
ie, DMA is not able to correctly trigger it
realistically most retail games wont care about that, but who knows when it might trigger by accident
weird, the config bits did change a little but im not sure why they'd lock stuff up
I'm not sure either since the core I built was only a day previous to the release and worked fine but running the release broke that one too some how 🤷♂️
someone asked before about detecting ppu/cpu alignments
Kitrinx I have a question. Is it normal some VS games converted to nes mapper don't used mapper 99 ?
yes
only some vs games use 99 I think
I really need to add that mapper so people stop talking about it
once I do it'll never come up in conversation again because nobody really plays them
maybe a little
ok. I have tested some on the new core and see some can render correctly and after checking mapper I see not all used the mapper 99.
From my experience with VS SMB, it’s incredibly easy to “crack” Mapper 99 games and convert them to a different mapper, either CNROM with PRG-RAM or MMC1 if you wanna be truly accurate and don’t want to force inaccuracies with NES 2.0
Plus, for Vs. Excitebike you’re better off playing the FDS version
it's not a hard mapper
we're just being weirdly apathetic about it for no particular reason
well
I dont think it's a hard mapper
it just requires knowing how to work with the rest of the system too since it's not really a mapper but a system type
I have placeholders all over for it in the code
like, it doesn't skip dots so the timing is subtly different
it happens on a Chinese "Lava FC" that uses a real PPU and CPU. From your deacription, I suppose the mod might be causing the APU misalignment
CPU is rev G but the PPU is a RP2C02G-0 which I think can fail some of the tests
ah maybe I can try that ROM
Isn’t it basically just CNROM with PRG-RAM and no bus conflicts?
Minus obviously the Vs. System specific features such as DIP switches and coin insertions
dips coins and most importantly the higher color, scrambled rgb palettes
ah gotcha
Yea, and I know VS Excitebike is one of those
And I would be one of those people grateful if you implemented it but probably only talk about it once or twice 😂
That reminds me, would I be able to run the NES Arcade games on this core as is or would they have to be the patched ones like I was using on the Everdrive?
i've been busy working on multiple ppu cpu alignments
So I noticed these weird artifacts in fire emblem. Is it supposed to look like that? 
Inbuilt grass generator
I got an AV famicom and all tests of accuracycoin have been passed. AV famicom is detected as a NES by this test.
But 5 tests fail in the Famicom Twin. I don't know why, because the system works well.
A lot of Robbys in that image 
Hardware revisions and the test is geared to a very model of the NES/Famicom
I think it’s the most popular one
Congrats, looks like your Famicom Twin isn’t a real Famicom!!!!
Ouch !
MiSTer more accurate than Famicom Twin
😂
getting the timing right for every possible phase is very difficult
phase3 fails on real hardware a lot too though, supposedly
also can someone try this test on real hardware with several different resets (for phase variations) to see if it randomly f's up? supposedly palette reads are unreliable on real hardware
are you writing these test roms too, Kit?
no
I just get this screen on rgb mod av famicom and everdrive n8 pro. In game combo doesn’t work so I can’t reset it but fired it up several times.
Both if that's what you want https://streamable.com/iale8p
Unhappy with merely being more accurate than original hardware, Kitrinx continued to improve the accuracy of the MiSTer NES core
I am soo excited about the current speed of NES development. NES rocks!
And Kitrinx of course 
but where does it end?
in a black hole and the self destruction of the universe by the paradox of the accuracy...
I don't know what alignments do. I guess it's like in shin megami tensei games and Kit is trying different story paths. 
Shin Megami Tensei - The alternate alignment
well, depending on the boot it could be Shin Tensei Megami
there is 131 tests now
from the 100coin discord heres the new ones
Does the core pass them all?
good question id like to know too, will give it a test when i'm home from work 🙂
“Sprites on scanline 0” should be an easy pass, that’s what Kitrinx was working on recently
Nice 😁
MiSTer wins again! 
@half pasture
200 tests soon
And here I am, just playing nes games for fun.
aS LoNg As yOu HaD fUn! 
A very smart man taught me that. 🥺
I thought that in the end it doesn't even matter?
Yeah but she tried so hard and go so far.
Has anyone here played the Metroid hack 'Metroid + Saving' on the MISTer? When I launch the ROM right after loading the core, I don't get any graphical glitches. However, if I launch the ROM second or perform a restart, I get graphical glitches. Here is the link to the hack: https://www.romhacking.net/hacks/1186/
I only get graphical glitches on the title screen and when selecting the save file.
I haven't tested if there are any other graphical glitches yet.
Could this be due to a faulty ROM?
Maybe the patch still has some bugs to be fixed
Oh, I must have overlooked that! Thanks for the hint!
Thank you for the hint! I'll try it out this evening.
Well that other one adds a permanent minimap that is quite intrusive and doesn't really help you to find your way because of its zoom/size. Personally I would rather stick to the snarfblam one which is tested by the times. Especially with the "US" RNG now fixed by Kitrinx. Make sure to set the reset behaviour to random though.
what does it do on real hardware?
I wonder if the last update of AccuracyCoins could help test that failed with SNAC. I don't have SNAC to test on my side.
https://github.com/100thCoin/AccuracyCoin/commit/b538d6728e03cdca1f181f71c99bc57796d9f333
This time I added some checks to make sure there isn't garbage data coming through the controller ports.
I haven’t tried it on real hardware yet. I’ll do that sometime when I get the chance. But I assume similar graphical glitches will occur there as with the NES core.
huh?
Maybe using it to translate?
Maybe he is Chat GPT
Chat gpt is into fpga gaming.
Yes, I use ChatGPT for translating because my English isn’t perfect. Is that a bad thing?
I was burned once for using AI to generate memes 😔
Not at all. We were just confused
Sorry for the confusion!
Not at all, my friend. please don’t think twice about using it - it’s helping 🙂
English can be pretty complicated.
Flipside is you get to talk shit and when somebody complains you can just back paddle and say you didn't mean it since it's not your first language.
Question. I have an expansion for me nes that fits into the bottom which enables really nice sound on some Japanese games. Is the mister nes core able to reproduce this?
I think what you're talking about is expansion audio, and yes. Try the japanese version of castlevania 3
Does the snes also have an expansion that goes in it's butt? The nes has one, the n64 has the dd, the gc the gb player. Nintendo seemed to like the idea.
Wow that's cool.
Well, it is more commonly used for the Satellaview.
iirc, it was originally planned for the CD expansion that never happened.
Don’t forget the GameCube!
EDIT: nvm you did
Everything since then also did... But they were just stands/docks
The GameCube had three
It was possible to use two at once with official plugins
This is like Pokemon just when you think you know all of them they add more
So many that they can't even fit them into one game anymore!
Three holes….
I don't get it
If you don’t get it, you’re too young for this server, man 🤣
giggity
Someone reported on MisterFPGA's reddit about having issues with Klax (USA).
https://www.reddit.com/r/MiSTerFPGA/comments/1oguj3h/klax_usa_on_nes_core_graphics_issue/
there's something about this game
I fixed it once
ah yeah it's the RAMBO-1 mapper
the redheaded step child of MMC3
atari crappy clone of it
it's almost certainly a mapper issue
It seems the JP version was converted to MMC3?
the american one is an unlicenced game
There are two IRQ modes: PPU A12 mode (also known as scanline mode) and CPU cycle mode.
In scanline mode, the counter is clocked using a very similar method to that used by the MMC3 and follows the same restrictions. In comparison to the MMC3, the actual interrupt triggers slightly later. Specifically, it is delayed until M2 falls twice after the PPU A12 rise that would have triggered the MMC3 interrupt.
In CPU cycle mode, the counter is clocked every 4 CPU cycles. The actual interrupt triggers one M2 cycle later than one would naively expect.
Whichever IRQ mode is being used, the counter behaves the following way:
When the IRQ is clocked by scanline or CPU cycle modes:
IF $C001 was written to after previous clock:
reload IRQ counter with IRQ reload value; if non zero, this value is ORed with 1 (see notes).
ELSE IF IRQ counter is 0:
reload IRQ counter with IRQ reload value.
ELSE
Decrement IRQ counter by 1.
If IRQ counter is now 0 AND IRQs are enabled:
trigger IRQ after 4 CPU cycles.
Notes:It's still unknown how the extra kick works. Klax still requires +1 to run perfectly.
Most emulators run Skull & Crossbones with a glitched scanline on the "Continue" screen OR during the game (score bar).
Following these rules in scanline counter mode has demonstrated accurate behavior, but this needs to be refined and merged with the description above:
I know, but it's mostly identical to the US version
One of two Tengen games officially released on Famicom in Japan. Altron published Super Sprint and Hudson published Klax.
Alien Syndrome and Rolling Thunder were also released on Famicom
Not developed by Tengen
ABII is Sunsoft (so is JP Fantasy Zone, but that's a different port, though the NES Tengen version was also developed in Japan).
the RAMBO-1 mapper just needs a little TLC
it's easy to fix I just need to spend time on it after im done with what im working on
Can’t wait for RAMBO-2!
RAMBO-4 is my favourite
is that the one where he fights tommy?
Yup. Tommy was a traitor.
How alliterate
Shouldn't the mapper be called First Blood?
NES Fantasy Zone is an odd case. It was developed by a company named Pixel, who worked on Over Horizon with Hot-B. It does have the same composer. Masaharu Iwata aka Rezon.
Whoa! Wondering if the latest unstable fixes this? https://github.com/MiSTer-devel/NES_MiSTer/issues/282
Is this game have same header as More Glider (game fixed with the last PR)
these flash writes are not persistent, it only allows some features to work that the game may rely on (like in this case: the continue after game over)
if this is supposed to be persistent saving, I dont see a way how to overwrite a PRG-ROM from a core. save states are a good workaround
it is also required to have the battery flag set in the header for writes to work, as specified on the nesdev wiki
interesting to see those other issues on mapper 30, and on 111 as well
if homebrew carts do flash saving a lot, maybe it makes sense to look into a persistent saving routine for those
So it is same as Courier and this games also the save doesn't work
seemed like a poor fix to it
it's not that hard to get mister to save it if saves are implemented
you probably do these things 10x as fast as me 😅
saves are always gonna be somewhere in sdram, more or less
///////////////////////// STATE SAVE/LOAD /////////////////////////////
follow around the save_written
you might have to add a flag to tell it to pull from a different sdram address
cheers. I might take up this challenge.
Yeah, those flashrom save games are a bit of a curveball since they write to the same bit of flash that the game itself is stored on. They could in theory overwrite parts of the game itself if they wanted to.... or have a bug that corrupts the entire flashrom thus wiping the game.
I assume most people probably would not want their .nes file to be able to overwrite or write to themselves though since the game can never then be reset to first boot status by just deleting it's save file.
I was thinking if it would be better to just use a .sav file like normal but have that .sav file contain the entire flashrom that's being saved to itself, not like NES games are large enough that it would cause a space issue, but then you are basically storing the game itself in the sav file so that might have issues with threading on copyright, especially if people share those sav files.
Would it be possible to write anything it saves to some kind of delta file that only saves the changes?
I believe that the save portion could be mapped to a .sav file
That logic should be inside the mapper though
Is there a seperate save portion? I thought they can just save to anywhere on the flashrom if they wanted to, even over the game rom itself?
the game has to follow a sequence before a write to flash works. if you do it accidentally without following the sequence, it just gets ignored, so there is write protection included in this
but yeah I was thinking the same thing. if the game can flash itself basically anywhere, you'd want to pull the PRG-ROM from RAM and use that as a save
How is your your NES rework going Kitrinx? Much left to do in the CPU/PPU?
I mean the save file would be the rom, but oh well, their fault for making a shitty absurd mapper
why absurd?
flash beats sram by a long shot technically speaking but we didnt have that stuff in the 80s/90s
and if you can make a cart with only prg-rom and you can reuse it for saving, sounds efficient. as long as the game does not self destruct 😄
because it's anacronistic and having the ability to corrupt the games code is dumb
Just so understand correctly: this mapper would make missingno in Pokemon red/blue not just corrupt the save file but have a chance to corrupt the actual rom permanently?
Sorry missingno is the only corrupting glitch I at least halfway understand. 😅
any glitch that leads to undefined behavior could potentially. you'd hope they put plenty of guardrails in their code to stop it happening
fair enough