Sub-1 route?
6 years ago
Kentucky, USA

there was a new route that is possible for sub-1 0-exit runs, sethbling and furious used jailbreaks and modified roms to do this to reset the ram if they need too, so it doesn't count Sethbling: 53.1 Seconds Furious: 55.6 Seconds Maybe they will get verified if they don't use modified roms

Edited by the author 6 years ago
Denmark
CarlEmilEx
He/Him, She/Her, They/Them
6 years ago

Their runs are technically legal, but they're going to make rules against it soon.

Kentucky, USA

Maybe it will count if they did on a unmodified rom of the game

North Carolina, USA

It was already done on an unmodified rom of the game. It requires some setup before the speedrun that isn't quite in the spirit of the category, so it won't be on these leaderboards, and there is no small fix that is going to make it magically accepted.

Staffordshire, England

Yeah, Seth had the idea with the intention of saying "technically this doesn't break rules, but it should" so there's discussion about how to adjust the rules to prevent manipulation like this

Germany

http://tasvideos.org/forum/viewtopic.php?p=444295#444295 If I interpret this post by Masterjun right, this method allows you to beat any game in 0 seconds.

British Indian Ocean Territory

I don't think I understand what Masterjun is saying there, but as far as I can tell there is no way to warp directly from the title screen directly to the credits using this technique. A 0-second time should still be impossible for this "category" if I understand things properly.

When doing ACE (such as credits warp) there are generally 2 pieces. The first is setting precise values in RAM which will be read as the desired code (the payload). The second is crashing the game in such a way that the game jumps to the written RAM and executes it as code, and then returns to normal operation (the exploit).

Using dirty RAM bypasses the first step. In the case of the speedrun this saves time because much less shell positioning is required (a small amount is still required to get the game to follow the correct codepath to get to the prewritten RAM). But it doesn't allow you to intantaneously execute the preinitialized RAM; you still have to crash the game to get execution to jump to the code you wrote. Incidentally, in the current TAS, the payload is loaded via controller registers during the crash itself, and so using dirty RAM for a TAS would only save frames.

As far as I know there is no way (in SMW) to crash the game without entering a level. The fastest way to do it is to eat the first chuck in YI2; even Masterjun's (admittedly old) fastest crash TAS () does this. In a game like SMB3 it would be possible to crash the game in a matter of frames via a buffer overflow, but knowledgeable people have told me that is completely impossible for SMW and that we can't crash the game from the title screen. And if the game executes normally without crashing, execution should not get anywhere near the uninitialized RAM we write. It does read, and even execute (a small portion of the game's code is actually written in RAM rather than ROM for optimization reasons), some RAM addresses, but they are all initialized at startup and so can't be used for this kind of ACE.

Everything I just said is conditional on using dirty RAM only. If instead, we alowed for corrupted SRAM, there is a known save-corruption glitch which would allow for an effectively instantaneous credits warp. This was demonstrated fairly recently in this video: (and can be used to much greater effect such as Seth's "jailbreak" mod). For that reason a credits warp "speedrun" with corrupted SRAM really would be trivial in SMW. But the runs that are being discussed here don't require this glitch (though they do induce it as a completely unnecessary side effect) and could be done on a cart with no initial save data. One thing I am not completely sure of is whether it is possible to load a save file listed as clean on a cart which also has corrupted SRAM and have this warp to credits instantly. I wouldn't be surprised if this is possible but the runs in question did not do this.

So, starting with uncorrupted save data, I don't see any way to warp to the credits faster than we can eat the first chuck in YI2, or about 34s by RTA timing. That's way faster than the runs in question, but not 0s by any measure. That said, obviously Masterjun knows more about this than probably anyone else in the world, so if I'm wrong about anything here he or anyone else is free to correct me.

Edited by the author 6 years ago
333Rich333 and Sten like this
British Indian Ocean Territory

Also, I'll just mention while I'm here that, as best I can tell, no one (including Seth and furious) who seriously runs 0 exit thinks this should be a valid strategy, at least in any meaningful SMW category. The current discussions in the community are not over the validity of the runs.

Rather the discussion is over what the appropriate rule is to rule out this sort of run without invalidating any future legitimate runs. It isn't so simple as "you can't touch uninitialized RAM" because some legitimate routes could use uninitialized RAM (which actually is not very random even with a hard power-off). Nor can one easily say "you need to do a full power cycle" because in some cases even 30s would not be enough to accomplish this and requiring a 30s reset between runs would make the category very bad to run (and invalidate probably every current PB on the leaderboard). I don't know what the right rule is, and so far there has not been much discussion, and we may not even have all the technical research done to state the "right" rule, but whatever it is, these runs won't be considered valid 0 exit runs.

Edited by the author 6 years ago
333Rich333, authorblues and 2 others like this
Kentucky, USA

The power on ace method I think needs a emulator that pushes a button multiple times in a frame so you make it tool-assisted playback push every button 6000 times in frame after start (it worked in smb3)

Kentucky, USA

lock request?