Ok guys, so I fixed QuickNES to halt on bad opcodes and disabled the Up+B bad inputs from the bot. So far I made some good progress. However today, just for the fun of it, I tried what would happen if I allow the bot to make glitchy inputs. And to my surprise, I have just observed the first 'glitch' that works in all emulators. See Frame 13685, where our hero glitches through the floor, saving a considerable amount of frames:
User movie #637851337516198347
The bot has also detected other 'tricks' on the patched QuickNES that so far haven't worked in NesHawk. Which brings me to the following dilemma:
A) Keep working on the 'glitchless' version + this glitch I just found.
B) DO WHAT I MUST, and try to find all glitches that work on NesHawk.
If I follow (A), then I'll be done with this damn game in 2-3 weeks. If I follow (B), it will be a rabbit hole: I will need to connect my bot to NesHawk or an emulator that behaves similarly.
The question is: @Alyosha do you think we can use your C++ version of NesHawk for this task?
Ahh finally the treatment this game deserved. GMP had hoarded a bunch of route and execution improvements to the (now obsoleted) movie since last year and refused to publish it until there was an emulation platform that allowed for this game to show it's true colors. Well, today is the day where libTAS+PCem finally overtook JPC-rr for us, and boy it was worth the wait.
Remark: The movie time is actually around 14:10, if you discount the booting times, which makes it faster than my (and +GMP +Challenger) submission
Movie highlights:
0:38 - Can't touch this
1:50 - What are walls?
3:35 - No, really, what are walls?
4:30 - W-A-L-L-S? Nope, never heard of him
5:15 - Bye, bonehead!
5:35 - ktnxbye
5:45 - No walls for you!
6:00 - Parsing error: 'walls' command does not exist
6:45 - Walls are just, like, a social construct, man
7:05 - sllaw
7:20 - Doors ain't a thing either
7:38 - The pain section -- hours lost routing this part
8:20 - Who turned off the light?
9:40 - Just a s(k)ip
10:00 - swordturn
10:30 - the legendary birdskip
11:00 - whoopsie
11:45 - walls. not even once.
12:20 - no, really, how many wall skips does this movie have?
12:35 - another wall bites the dust
13:10 - Great bird avoidance
13:45 - Hey, you dropped your wallet! -really? *BAM*
14:10 -Buggy supersaiyan
14:45 - Kamehameha
This is an obvious YES for me.
Great research. I feared that could have been the case, what a bummer :(
I'll have to patch QuickNES to halt on this opcode and re-run the bot. The result won't be as spectacular, but hey, that's life.
Can we delay this movie until then?
Thanks Challenger and Alyosha for your kind comments and feedback!
You guys raise some valid points about reproducibility and accuracy. Unfortunately, I don't have a good answer for them, given the tools I've used. I tried the trick on FCEUX and it did produce the characteristic flicker you see in this movie (I assume PPU corruption) and didn't crash, but it didn't produce the skip either. I would have to try again with more accurate timing synchronization. On the other hand, I'm thinking of replacing QuickNES by Mesen (another C++-based emu) for the botbut then again, there are no guarantees of cross-compatibility and it might be painfully slow. Perhaps I can also use your C++ version of NesHawk since the routing is mostly done?
Besides that, there are so many things I want to do with the bot like solving other simpler NES games I really like (cough cough Castlevania cough Ninja Gaiden) and even SNES and Genesis (Flashback), that I don't feel like spending too much more time on analyzing what exactly is going on with this movie. I'll be fine with whatever decision the judges make.
Regardless, this is indeed a hairy topic, and perhaps this movie can serve as catalyst for a wider discussion.
So basically the bot found that to break this port of the game is to "Press Up+B in certain places to get weird, unintended behavior to happen" and then you had to figure out the RIGHT "weird, unintended behavior" to generate that didn't softlock or crash the game..
Exactly. Sometimes the bot used variants to UB, such as DUB and DUBA. Don't ask me why.
GMP wrote:
I don't understand anything, all I can see is that the movement is well optimised, and then the prince crouches in random places for no reason, and then the level just ends, lol. It's great you showcased this beautiful port that is extremely faithful to the original and has amazing music. Yes vote.
Hahaha I can sense the irony. Thanks for the vote :)
Patashu wrote:
All the Up+B glitches had me cackling. Very impressive that your bot organically found and exploited these with no prior knowledge.
I know, right? It's amazing.
FractalFusion wrote:
So normally I don't make comparison videos of TASes that have vastly different routes. But I was watching the encode in the submission, and after 2 or 3 minutes, I lost all sense of how this game was supposed to function and my brain started to melt. (See, I don't know this game except for the few DOS/NES/SNES TASes I watched a long time ago.) So I figured I would put the previous submission in as a frame of reference. Hopefully it helps at least a few others out there who don't know the game.
Wow!! Thanks so much for the comparison video, it really helps to see the difference the new tricks made to the route!
Can I request a download link to the patch that doesnt require an account on RomHacking.net to download? I am not a fan of websites that force you to create an account just to have access to certain files.
I have a SNDDRVRS directory that's absent from the HDD image. Any reason why it's not present?
Additionally, I have a MIDI.DRV and a DIGI.DRV that's also absent from the image.
Yes, SNDDRVRS contained additional drivers provided by Broderbund for that era's sound cards. Since we use only PC-Speaker, we have no use for this additional folder.
MIDI and DIGI.DRV are the specific drivers for Sound Blaster and MIDI music copied from the drivers folder by the game's setup. Again, since we use PC Speaker, these files are also not needed.
Hope that answers your question!
I'm curious why PC Speaker audio is on for this—is it actually a benefit time-wise instead of having say Soundblaster sound?
Time-wise there is no real benefit for either, because we skip the end level jingle anyways, but in JPC-RR soundblaster/adlib and tend to cause jittering in the final encode and also various other emulation issues, speaking from our experience with PoP1. Since this game is more memory intensive as well, we didn't want to take any kind of risk and hence went with pc speaker.
It's even worse. Using SB + JPCRR crashes the game when you hit a guard, no matter how you configure it :(
Sorry, dumb question.
Why wasn't the equally bad SNES port used?
The SNES port is, from what I recall, considered the worst version of the game, being glitchier, having several missing elements, and most notably completely lacking the Escher-esque final level.
Edit: Oh, I forgot this version was never actually released. So I don't know if anything besides the final level that was missing from the SNES version is actually in this one.
This version was chosen for being very close to the DOS version. Many of the mechanics/tricks that work here, also work on DOS and vice versa (with some notable exceptions mentioned in the description). Memory analysis also confirms that game-specific variables behave largely equally in both versions.
Level layout is also equal on both versions. Again, with some exceptions: e.g., the beach level has one tile less in Genesis than in DOS. Also there are no health potions in the last screen of Jaffar's labyrinth.
Therefore, this version was the best proxy to the DOS version when applying the bot (applying the bot to the DOS version is just not feasible with the current state of DOS emulation, and we don't have its source code as we did with PoP1).
Hi all,
This is to let you know I'm working on a bot-driven TAS for the genesis Prince of Persia 2 (1996) cart. This is an almost complete game has never been released and it only surfaced/leaked relatively recently. I cannot post the link to the information page because it also contains a link to the rom and goes against forum rules, but a quick google search would suffice but I quote:
The unreleased Mega Drive port of Prince of Persia 2. It was dumped by Jurai and released by drx on August 11, 2006. [...] This game was canned for unknown reasons.
The game is almost complete because there is a bug that prevents you to get onto the horse that takes you to the last few levels of the game. To overcome this, a fix patch has been released: http://www.romhacking.net/hacks/3755/
As of now, I've linked the BlastEm! emulator to my route generating bot, Jaffar2 (https://github.com/SergioMartin86/jaffar2) and I'm working with the PoP speedrunning community to bring the full TAS to life. I estimate to to take 6-12 more months to complete.
The purpose of making the TAS is to satisfy our own curiosity and lust for TAS, but also to eventually port the movements to the DOS version of PoP2. Early tests indicate that the ports are mostly similar and tricks work equally in both ports. We work on genesis because a bot-driven TAS for DOS games is still very much in the horizon.
THAT being said, my question would be: would this genesis version be acceptable for publication here once it's done? I know it's very far fetched from the ground rules (it's a non-published game and it needs a patch), but then again I know judges have some degree of freedom to take discretional decisions.
Don't worry, it's ok if it doesn't quality. We won't can the project if it isn't and we'll probably come back later on with the DOS version of the TAS. But for now I'm curious to know whether this particular port could work it out here.
Why was this created/chosen?
How was this created?
Also, I'm trying to import this image, however JPC-RR says it doesn't recognize it.
Edit:
Okay, I see the file just needs to be copied directly to the drive library.
Hi Nach,
For most of JPC-RR's configuration, I followed this page verbatim:
http://tasvideos.org/EmulatorResources/JPC.html
The FreeDOS image is a build shared by CoolHandMike in that page. It used to point to the google drive link but it seems that someone replaced it with a link to a static file in this server. FreeDOS is the DOS of choice since it's GPL and is, allegedly, perfectly compatible to the licensed MS-DOS.
The game does not in any way indicate that this potion does that.
Well, the music plays for a limited time after drinking the potion, so it's somewhat of an indicator ; ) I wonder if it was pointed out in the game's manual, however...
Exactly, there is a mellow music that plays while you are 'under the influence'. Since we disable music in the RTA and TAS runs to save frames in between levels (otherwise, the game won't advance until the music ends), this music is not heard in the encode.
In casual play, you are presented with this unusual 'green' potion at the beginning of the level and can only get to drink it after passing most of the level. So the idea is that you are kept wondering the whole time what is it about. And then you hear the music and should kinda figure out from the music that you can drop safely now.
Funny thing, you don't need the potion at all as you can drop down without dying into the exit room. We do drink it in the TAS because it is the faster route.
Thanks all for your kind comments. I'm really glad you've enjoyed it.
PLANET wrote:
PLEASE (!) consider doing a TAS of Shadow and the Flame (Prince of Persia 2)! That could also be stunning : )
Yes, Pop2 is next on the list. We've found an unreleased Genesis ROM that seems to be 100% compatible with the DOS version, which will make emulation a lot easier to work with. Another goal is to obtain sub 12 in this game, but I need to take a break from it for a while to try and get a new perspective.
By the way, if anyone wants to encode this, you'll need the exact disk images I used. I provided the link for the FreeDOS image, but the original PoP copy can be obtained here: [MOD EDIT: Removed link to copyrighted game material - please read the site rules -Mothrayas]
In any case, I assembled both in one zip for convenience: [MOD EDIT: Removed link]
Using SDLPop would greatly simplify our work, as we already have developed tools to feed it frame-by-frame commands and produce replays that can be played by SDLPop everywhere else. That would relieve us from painstakingly inputting the commands into JPC-RR and also solve the sound emulation/jittering issue.
If we are allowed to submit these movies under Linux/SDLPop, then I'll start working on it immediately. I would note in the submission though, that these runs can be proven to be reproducible in the original DOS game.
I think we have a problem with the game files used in this one. So this run uses a cracked version of the 1.3 PRINCE.EXE file that skips a copy protection sequence. The copy protection is a mini-level that appears after level 1 where you have to look up a letter in the manual and drink the corresponding potion. It looks like this:
Link to video
Our movie rules state:
- Do not use fan translations, cracked game versions, or otherwise hacked ROM images — translators do not want you to use obsolete versions and we prefer non-hacked games. Hacked versions are labeled [h] and [t].
- Exceptions may be made for bad or cracked ROM images only if no good ROM images exist, or are not obtainable.
From what I can tell, this is a very common game and the second point does not apply here.
We have certain standards for game authenticity. If the copy protection was simply a password screen, maybe we could let it slide with a warning to use the untouched version next time, but this is skipping gameplay in a mini-level that can be TASed and optimized. What is the reasoning for RTA runners using this version, and why should we make an exception here?
A similar objection was made in my submission but I guess we should follow the discussion here. Here are my arguments for the acceptance of this version:
- The binaries used in my submission and this one are the gold standard for RTA runs (see https://www.speedrun.com/pop1/guide/r6nek). They can easily be considered the 'official' versions, as they are widely available and agreed upon by the speedrunning community.
- The copy protection screen does not count towards the IGT, therefore they serve no purpose other than wasting time in runs.
- From a spectator point of view, this interruption serves no entertainment purpose, as it requires no skill.
- Once the copy protection screen is passed once, it does not repeat upon restarting the game with CTRL+R. Although this argument may not be relevant here because you measure time since boot up.
If these arguments are not sufficient, I won't complain. But I wanted to put them there in case you hadn't considered them. Thanks for reviewing both entries!
After completing the first level in the DOS version of Prince of Persia, the game asks the player to drink a potion which corresponds to a certain letter found in a particular area of the game's manual. How was this skipped?
We use (for TAS and for RTA speedrunning) a version of the game that has this screen disabled.
Edit by Nach: Please do not post illegal links to games.
User movie #637851337516198347
The bot has also detected other 'tricks' on the patched QuickNES that so far haven't worked in NesHawk. Which brings me to the following dilemma:
A) Keep working on the 'glitchless' version + this glitch I just found.
B) DO WHAT I MUST, and try to find all glitches that work on NesHawk.
If I follow (A), then I'll be done with this damn game in 2-3 weeks. If I follow (B), it will be a rabbit hole: I will need to connect my bot to NesHawk or an emulator that behaves similarly.
The question is: @Alyosha do you think we can use your C++ version of NesHawk for this task?