Posts for Alyosha


Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
This has been fixed in the dev build.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Not uninitialized RAM. Paperboy runs a loop from power on that basically measures how long it takes to get an NMI (or some other ppu condition I forget exactly.) FCEUX adds a frame at the start of every movie which is basically a dead frame. I don’t know why exactly since this is not console behavior. This is also why you have to typically delete one frame from a FCEUX movie if you convert it to BizHawk and try to sync it up. So FCEUX gets a very incorrect count hence strange house positions. This Probably effects at least some other games that we just aren’t aware of yet. Edit: if someone out there has a NES and paperboy, it would be cool to just repeatedly power up the game and see how many different House variations you can actually get .
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
There are 2 examples I know of: [2599] A2600 Seaquest by morningpee in 01:39.80 [749] NES Paperboy by Randil in 11:30.78 These are emulator bugs. Last I checked the FCEUX one is still present. [2821] A2600 River Raid by Lord Tom in 1:22:47.90 is also basically one big emulation error but it at least looks passable. I’ll try to remember some more examples that I have happened upon in the course of working on BizHawk. EDIT: [3407] A7800 Scrapyard Dog by Noxxa in 11:04.14 contains obvious and fairly annoying graphical errors. I basically wrote A7800Hawk to fix this because it bothered me. [2585] A2600 Dodge 'Em by yep2yel & morningpee in 00:52.93 and [2226] A2600 Superman "pause glitch" by jlun2 in 00:11.42 need updated movie files and encodes but the runs themselves produce correct results. [814] SNES Star Fox "maximum score" by YtterbiJum in 19:47.37 the publication even says the game is not emulated correctly.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Warp wrote:
I think the idea was to unpublish the few TASes that might still exist that are clearly and blatantly against the rules (such as not completing the game, or is known to be using faulty emulation). I have no idea if there exist any such TASes that are in a published (non-obsolete) state.
There are 2 examples I know of: [2599] A2600 Seaquest by morningpee in 01:39.80 [749] NES Paperboy by Randil in 11:30.78 These are emulator bugs. Last I checked the FCEUX one is still present.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
^ noted, I’ll add it to my TODO list, it will be a couple weeks though.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
I think this is an interesting thread. My fear is that a lot of the work already done will likely get lost and there will be a gap in institutional knowledge . At the same time I can see that tasvideos is poorly equipped to deal with such fluid environments. Maybe a separate website would be better suited to focusing all the work and getting it out for general distribution? Like TASLabs instead of TASVideos. At least you could forgoe all the formalities and document and categorize things as you please. Oh well just a thought I had while reading through the thread.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
This is an amazing advancement in TASing. I would not have guessed at the start of the year that a Shovel Knight TAS would be possible right now . Great TAS and great work to keylie and anyone else associated with libTAS for making it happen. Yes vote.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Oops, I guess I shouldn’t have submitted it then, and it should have been rejected. My bad.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
No, the inputs are unique, I still had to find them. Otherwise it would work the same in both cases. But my point is that the uniqueness is not due to me being somehow better. The emulator was better.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
No. It’s the same glitch it both movies. The setup is only different due to emulation differences.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
That was a fun watch, yes vote . How long did it take to make out of curiousity?
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
It’s an IRQ timing related glitch. So emulated timings dictate how / when the glitch is possible. In other words, I didn’t ‘improve the gameplay’, I just found the glitch on a different emulator which is much more accurate, and it happened to be shorter .
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
feos wrote:
You mean it's impossible to improve the old SML2 run using VBA v24?
The button presses in the current run certainly wouldn’t work in VBA v24. Just Like the button presses in the original run doesn’t work in GBHawk / Gambatte. I don’t know if it’s impossible, since it’s just random button mashing until it works, but I didn’t know that the sequence of inputs that worked in GBHawk would be shorter ahead of time .
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
MOD EDIT: Posts originally moved from the discussion thread of #6023: TiKevin83's GBC Pokémon: Red/Green/Blue/Yellow Version in 1:36:42.94. -Mothrayas
</hr> Sorry for the late reply but I found this conversation interesting enough to post to. I’m glad the movie file got updated. This is state of the art stuff, and making sure things are correct to the furthest extent possible I think is pretty important. Just as food for thought, I’d like to point out that [3746] GB Super Mario Land 2: 6 Golden Coins "game end glitch" by MUGG, Masterjun & Alyosha in 00:46.08 is a ‘resync’ in exactly the same sense that this movie is. It is pure happenstance that it ended up being 9 frames shorter so that it superficially met the requirement of being an ‘improvement.’ But really it’s just slight timing differences in emulation. Also, the behavior that was missed in the original Pokémon run is very significant. Personally I think that not having it emulated puts it in ‘technical foul’ territory. Sure nobody knew at the time , but luckily Pokemon people had good R&D to move things forward .
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
This is some cool stuff, great work getting this to the level where it is accepted by the site . I’ll be really looking forward to what crazy tases come out of it.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Unfortunately it’s not easy to compare because Gambatte doesn’t use a big enough integer size to compare cycle count in a trace log. It’s fairly likely that they are both the same just with different frame lengths. I know that this is a big source of confusion between the cores but I’m not really willing to try to adapt GBHawk to have fixed length frames like Gambatte , as I don’t consider it worth the added hit to complexity. Also be aware that accurate gameboy emulation is an ongoing area of research. There are a couple of pending accuracy improvements to Gambatte and likely GBHawk when I have a chance to catch up. If you want to really know which is correct, probably TIKevin83 is the only one who can really know for sure since he has the console verification setup, but he is busy with Pokemon . @ezgames69: sorry I missed your post earlier but it seems like the same thing as jlun2 mentioned Edit: @jlun2: oh, and I’m not really able to check movie files and such anymore unless it’s from strictly free to distribute games (ex. Streemerz) so sorry I won’t be able to actually compare your movies.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
A bit late but Cadash is gone!
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
@jlun2: frame count is inconsistent between the cores , so don’t worry about that too much. As for input lag frames I’ll look at it when I get a chance since I forget what I did. I’ll be back to working on BizHawk / GBHawk again in a couple weeks I think, hopefully not too much has piled up and my code isn’t all foreign to me now
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Cool! But I won’t be at my computer to implement it for several months, so if someone wants to implement it go ahead.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Great work Challenger! Easy yes vote. I tried to tas this myself a while back and this game is not so easy to deal with. The high level of optimization and understanding of the game is obvious here. Also that boss health glitch is silly.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Augusto wrote:
Hey the Atari 7800 Sirius graphics is breaked in current BizHawk.
Fixed now. (There wasn't enough emulated cart RAM)
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Sonia wrote:
I've been wanting to report those for a while, but only got around to it now. Akumajou Special - Boku Dracula-kun (NES) The bottom of the screen shakes constantly at the dancing girls minigame. ... Lagrange Point (NES) Similar to Dracula-kun, the bottom of the battle screen shakes constantly. ... Megami Tensei II - Digital Devil Story (NES) In the battle screen, there's a graphic glitch below "AUTO". ... Necros no Yousai (PCE) Game crashes during the intro, making it unplayable. __________ None of the first 3 bugs exist in Mesen, and Necros no Yousai is playable in mednafen.
Ok I fixed the screen shaking stuff. Turns out VRC IRQ timing was slightly off (being triggered one cycle too early and being clocked at the same time it's enabled.) Actually the first one is more an issue with CPU emulation, it really should be re-cast to check for IRQ's at the correct time, but that's not happening any time soon. Please test any other VRC games that you are interested in to make sure there are no regressions.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
There's nothing I can do about the beeping if it's related to the skip bios option. It's a crude hack that may or may not work and have various side effects. I added the remaining memory domain for SGM. It's now split into High / low regions. EDIT: Tomarc: here is how the game picks a starting position:
8377: ED 5F        LD A, R                    AF:0093 BC:00FF DE:000A HL:024C IX:A3B7 IY:A289 SP:73B9 Cy:41437612 CNp-H-zSe
8379: E6 03        AND 03h                    AF:2621 BC:00FF DE:000A HL:024C IX:A3B7 IY:A289 SP:73B9 Cy:41437621 Cnp-h5zse
837B: FE 03        CP 03h                     AF:0210 BC:00FF DE:000A HL:024C IX:A3B7 IY:A289 SP:73B9 Cy:41437628 cnp-H-zse
837D: 20 01        JR NZ, +01h                AF:0293 BC:00FF DE:000A HL:024C IX:A3B7 IY:A289 SP:73B9 Cy:41437635 CNp-H-zSe
8380: 32 53 72     LD (7253h), A              AF:0293 BC:00FF DE:000A HL:024C IX:A3B7 IY:A289 SP:73B9 Cy:41437647 CNp-H-zSe
That R register is an internal counter that increments with each instruction fetch. I don't really see any way for this to be a value different then what it currently is. Extensive testing on ZXSpectrum seems to indicate that it is behaving correctly as well. So, I don't know how it could take any of the 3 values at power on.
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
The super game module is only relevent for games that are made for it (ex Zaxxon Super Game.) It is a fan made module, so no official coleco games will make use of it. But I did forget to add a memory domain for it when it's activated, I just did so now for the RAM. Also I fixed the super action controller crash. I don't hear any beeping in looping though. Can you make a recording of what you hear that doesn't sound right? I'll have to take another look at Tomarc since I totally forgot what happened with that one (even after re-reading the github issue.)
Alyosha
He/Him
Editor, Experienced Forum User, Published Author, Expert player (3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Hurray for accuracy! I consider this a pretty big achievement. Great work to TIKevin83 and the pokemon community for putting in the work to make this happen, It's very impressive to behold. A few side notes: I left in the debug code in GBHawk that will tell you when the relevent glitch that effected this run pops up. It's actually not too uncommon. It actually happens twice right at the start of the run on the 'Game Freak' screen, but this is too early to effect the run in any meaningful way. Also this test rom: halt_if will only pass if the glitch is implemented. But this test was all the way back in 2015! I guess no one noticed. I do think that whatever rule is keeping this from being published should be revised to allow it. The existing run is demonstratably wrong, and this one is demonstratably correct. Why would you want to keep the wrong one? Yes vote!