Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
narutosasukefx1998 wrote:
where is randil
I'm still here! I haven't forgotten about this, but progress has been slow lately. I'm still working on this though. I'll try to have something new to post about in a few weeks or so.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Ah, yes, you're right about MPCfold. :) Here's an updated version of the script:
http://pastebin.com/EaK6cJxt
I also added the variable "arrfile" so you can change the name of the array file if you find it necessary.
I'm using version 2.0. I got it from here:
http://www.unfungames.com/mariopaint/
The things I did just now in order to be able to play your song was (all this was done in the Prefs-folder):
1. Create a file SMB201]MarioPaint.txt and paste the song data here.
2. Create an array file called SMB2]MarioPaint.txt that only contained the text string SMB201.
3. Append the text string SMB2]MarioPaint* to MarioPaintArrList.
Note that the file names have to have this exact format - for example, at first, I accidentaly named the song file SMB201]MarioPaint.txt.txt (i.e 2 txt at the end), and this crashed the program. Perhaps you did something similar?
All this is done automatically by the program, so it is weird that it doesn't work. Please try again with version 2.0, and if you still have problems you can upload the files and I can take a closer look.
EDIT: Also, Super Mario World (I think this is my best one yet):
Link to video
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
The first version of this new script is done.
http://pastebin.com/9tE9J30f
New features:
*Automatic bpm calculator. This works well in some songs, but it assigns too high a bpm for most songs. You can test it out for fun, but you probably want to manually set it if you want a correct bpm.
*Works on any MIDI file with any number of channels, and multiple notes per channel at any time.
If you find a MIDI file it doesn't work for (or works horrible for), please let me know.
Patashu: Yeah, it worked quite nicely with your songs, though I had to manually set the bpm. Here's the text data for "Patashu in the stars" if you want to hear it (you do know how to play this I hope?):
http://pastebin.com/MGVErPBi
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
I'm currently writing a more general version of this script that will (or at least should) work on any MIDI file. I will continue to test it, but if all goes well this script will replace the one you're using. Its code will be much more commented and won't call any home made external functions (such as this num2str2 that I've made). I'll let you know when this script is finished.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Yeah, perhaps I was a little short in my answer.
juef wrote:
1) What exactly is MarioPaintArrList.txt? I have no such file, yet it seems the script wants to read from it. I have simply commented all related lines for now.
This is a text file that contains all all available playlists, called "arrays" in MPC. It should be the Prefs folder in your Mario Paint Composer directory, otherwise my script should create it. An MPC song file can only have 384 beats in total, so most songs have to be split into multiple files. To solve this, you can put multiple songs in a so called "array list" that simply litsts all the songs you want to be played, like a playlist. In order for MPC to know what array lists that shows up when you click on "load array", you have to put the name of the array in the file MarioPaintArrList.txt. What my script does is simply to add the name of the array to the end of this text file so that it shows up when you look at what arrays you have available.
So, you shouldn't comment out this section, since it's necessary for MPC to know what songs it can load.
juef wrote:
2) Sometimes, F was accessed in an out of bounds location (2nd dimension), which of course halted the script. I'm guessing this is because of the number of instruments in the track?
Yes, that would make sense. As of right now, the script is only intended for MIDIs that come from NSF songs. It could work for any MIDI file in theory, but I would need to tweak the script some - for example, right now it can't play more than one note at any time from one channel - this is no problem for NSF MIDIs but can be a problem for MIDIs in general. Of course, you can get a MIDI to work if you're lucky, but the script isn't intended for that (at least not right now)
Perhaps the next version of the script will be more general in nature. :)
juef wrote:
3) How does one set up MPC so that it realizes we've added a song to it? Editing the songlist file seems to do it, but very often, loading a file halts with a 'Script error; continue?' problem.
If I understand this question correctly, I answered it in point 1 - the game knows what songs are available from what is in MarioPaintArrList.txt.
To give an example: Let's say you've made a song that is split up into text files metroid01]MarioPaint.txt, metroid02]MarioPaint.txt and metroid03]MarioPaint.txt. What you do(or rather what the script does) in order for MPC to know that you want to play these after each other is to make an array file, let's call it metroid]MarioPaint.txt. This file will contain the text
metroid01
metorid02
metroid03
i.e. all songs to be played. The Matlab script adds this the name of this array file to the list of available arrays automatically, and this is done between lines 124 and 134 in the code (the "arrlist"-part). I hope this made sense!
juef wrote:
4) Long tracks seem to be split in many files. I'm guessing that is because of a limitation of the MPC format?
Yeah, see above. In my first version of this script I didn't know about this, so all songs I made had to be really short. I'm glad I found out about this feature!
juef wrote:
Thank you for the script, details and help! :)
No problem, good luck with this. In hindsight I should probably have commented the code better, but hopefully you can make some sense out of it.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
I'm glad you like it! Here's a quick breakdown of how these are made:
1. The NSF file is converted to MIDI using the program NSF2MIDI (I use the english version 005a). This can be downloaded from Zophar.net. This is the first step where I rely on an external program to correctly convert the NSF music. I can only hope that nothing gets lost in translation, though I haven't been let down yet. I suppose this is where it gets decided if expansion chips sound gets included or not. I think not, because the Solstice main theme (which uses one, right?) doesn't sound quite right.
2. I'm a Matlab programmer, so the program that converts the MIDI file to MPC text file is a matlab script. The main Matlab toolbox I use is a MIDI toolbox that can be downloaded from
http://www.kenschutte.com/midi
3. Using functions from the toolbox mentioned above, Matlab converts the notes to a matrix where each row is a note being played, and the columns are the note number, start time, end time, sound channel, and some more (for us unnecessary) info. So this matrix contains all information about the song.
4. The last step is quite simple - since I know all the information about the notes, I tell the script the number of beats per minute of the song and the script assigns each note to a beat in the MPC song. This is the main script that I wrote. I don't know how familiar you are with Matlab, but I've uploaded the code here (maybe you can get a rought idea anyway):
http://pastebin.com/yMAAcALA
You should be able to just replace the infile argument to any MIDI you've converted from NSF.
I can select additional parameters to the program, such as bpm, what sound channels I want to use, and what instrument to use for each sound channel. The number of bpm is decided by a parameter I call beattime. A value of 1 means that one MPC beat is 1 frame, so the MPC song will then have 3600 bpm, and can thus play any song as mentioned above (at least in terms of timing, the notes themselves might be out of the range). Usually, a value of 6 is enough for most songs, which gives a bpm of 360. Too high a value will cause too many notes to be assigned to one beat, so you have to trial and error a little until you find the right bmp.
Also, here's Kid Icarus:
Link to video
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
John11 wrote:
That's really cool and it's one of my favorite songs ever too. Could you try to do the intro song to Solstice? It bet it's too complex though.
I just tried this song, and although it did work, it didn't sound very good. The notes are all over the octaves, so the Mario Paint version of it only slightly resembles the original song (since i have to increase or increase the octave in order to play them).
Interestingly enough, I had to use a bpm (beats per minute) of 3600 in order to play this song correctly - this is the highest possible bpm for NES songs, since an NES sound channel can't play more than one note per frame = 3600 notes per minute. The Dr Wily song I posted above had a bpm of 360.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
In Mario Paint Composer you can create music just like in the SNES game Mario Paint, but with some new features such as having sharps and flats. If you like the SNES version, you should definitely check it out.
All songs you make are saved as text files, so this renders itself quite nice for converting other formats into this (simple) format.
I made a script that converts NSF music (music from an NES game) into a Mario Paint song, by first converting it into a MIDI file. Of course there are limitations such as the narrow note range of Mario Paint, but most songs can be played quite nicely. I made a version of Dr Wily stage 2 from Mega Man 3 here:
Link to video
For reference, here's the original song:
http://www.youtube.com/watch?v=GsqMYXqGAVw#t=19s
If anyone's interested I can give details on how these were made and share my scripts. I also take requests! :)
I hope to see other people's Mario Paint Composer work here too!
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Thank you all for your responses. In the end, I used the trace logger file to dissect the RNG - I wrote another program that translates each row of the trace logger to plain language, so I could more easily follow what is happening. P.JBoy's list was very helpful for this, so thank you very much for that.
I have a question though: For the STA command, how do I know if it stores an 8 bit or a 16 bit number? For example, if the trace logger says
STA $0000
and A contains the hexadecimal number $012C, will it place the value $01 on address $0000 and the value $2C on address $0001? Or does it only move the 8 bit number $01?
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Well, I must admit that I haven't had much time lately to look into this. I hope to have some time to look into this weekend. I would like to understand better how the RNG works, maybe then I can better predict and plan how to alter RNG dependent rooms.
I'll keep you posted on my progress.
EDIT: Also, 2000 posts, yay! :)
EDIT2: I've had some time to work on this now, and I have some news: I've discovered several more frame rules in this game - there are frame rules for screen transitions and when enemies start to move once you've entered a room. This has an impact on rooms where enemy movements are crucial. Those 5 frames I gained turned into 3, then to 9, and then down 4, so the gain keeps oscillating because of these frames rules. Right now I've just reached room 9, so I'm pretty close to the boss. I think I might be able to save some frames in room 8, though - I lost 5 frames in this room alone, which stings a lot, so I will try to win them back.
If anyone wants to take a look at how the run looks now, I've uploaded it here:
http://dehacked.2y.net/microstorage.php/info/1814839486/goof_troop_new.smv
Note that I might abandon this and go with my previous WIP if this doesn't end up well (for example, if I get horrible luck on the boss no matter how hard I try to remedy it).
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Nice improvement, I'm glad this game got some more attention. I vote yes, of course, good job as usual. Perhaps we will see a frame war on this game eventually? :) (I think it lends itself quite good for competition)
It's been a long time since I TASed this game, so I'm afraid I don't remember all the different tricks and strategies. Health management is crucial though, so you need to plan in advance where to take damage.
Taking damage from an enemy as soon as you've made a jump will give you a speed boost, so that's something to consider.
Good luck with this! Maybe Aglar has something more to fill in with?
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Hi, I'm trying to understand how RNG works in the SNES game Goof Troop, and I figured that I should be able to decipher how the RNG works by using the trace logger.
Here's what I know about about RNG: The current RNG value is stored as a 16 bit word at RAM address 7E1A58. The RNG changes when you pick up a pot, but I don't know exactly how. I made a trace log for a few frames during which I picked up a pot and thus changed the RNG.
So somewhere in this trace log file there should be information on how the RNG changed, but I have a hard time figuring out where. Does anyone here have experience with the trace logger? Is there a guide or manual somwhere? I didn't find one when googling. What should I look for in the trace log file if I want to find out how 7E1A58 changed during these frames? (I searched for the string 7E1A58 with no luck, all addresses here seem to start with 8)
Also, while I'm at it, can someone point me to where I can read up on what all these 3 letter commands stand for? I know a few (lda=load, bne=branch on not equal), but not all of them. The closest thing I've found was this:
http://en.wikipedia.org/wiki/List_of_CIL_instructions
but only a few are represented here.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Thanks, it was 5 frames faster, though it did affect the RNG (because of the bee in this room). I will try to remedy this, but if it fails I will need to wait 4 frames to get the same RNG as in the current WIP, thus only saving 1 frame by this.
EDIT: Argh, I'm not having much luck. The changed RNG gave me bad pirate movements in room 4. I have managed to get 3 different RNG settings (by making the pirates in room 3 move differently), but all those settings ended up in bad enemy movements for room 4. I also tried editing in the improvement with the same RNG (i.e. wait 4 frames thus only saving 1 frame), but because of a 2 frame rule, the RNG did not stay the same... I will try to fix this, but right now it looks like it might be hard.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Blechy wrote:
Uh oh, found an improvement in early stage 1. In room 6 pass 1 on your map, goofy can actually make it out the right exit faster than Max can.
Thanks. I should be able to edit this in without any problem. I'll let you know how it goes.
Blechy wrote:
Edit: Watched your videos, that boss is insane. We were thinking that double kills might still improve it further, but considering that everything gives you something to throw, might not improve by much. Again I think it's just a case of how far can you push the insane amount of luck in the game.
Yeah, looking back on it, it's pretty quick. I will continue to try to improve it further though before starting on level 2.
EDIT: Can you give details on how Goofy reaches the exit faster? I tried using the same strategy but exit the room with Goofy instead of Max, and it was 11 frames slower. I also tried picking up the pot with Max and immediately running to the exit with Goofy, but this was 8 frames slower.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Blechy wrote:
Cyg & I noticed in our runs that some pots reliably will never drop fruit. I was messing around just now and had a breakthrough as to why this is: it seems pots picked up via hook will never drop an item.
Yes, this is the case. Also, pots picked up with the hook won't affect the RNG (picking them up normally does) - this is useful in a TAS, but just a "fun to know" fact for a console speedrun.
As for my progress, I haven't had much time to TAS in the last few days, but I hope to have some time this weekend. I'll get back to you on the progress.
EDIT: Orange Claw Hammer wrote:
- The blue guy in the room before the golden key can either stay put long enough for Goofy to grapple twice, or he can even walk downward toward Max and Goofy can grapple him once. Either way, it is definitely possible for him to be closer to Max when Max throws the barrel upward at him than is the case in the WIP (I think even pretty much adjacent to Max).
I just managed to beat this room in this way (get the pirate to walk down, so Max could kill him faster), but the barrel hit the pirate on the same frame, so no potential frames to be saved here it seems. I don't see how the pirate could be closer to Max than he is in this case. If anyone wants to take a look: Movie file.
EDIT2: I played around trying to kill two enemies with one bomb, and although it worked quite well, I didn't manage to improve my best time, even though it came close. However, I did manage to improve my best time by 60frames = 1 second, using a different starting RNG. Now the boss fight is 82 frames faster than the published run. Take a look: Movie.
Also:
Orange Claw Hammer wrote:
Also, one question: is it possible for you to find out how the final in-game time works, the one displayed after all the credits have rolled? We're not sure if it takes into account hidden demicals cut off in the individual stage times or if it simply adds the integer figures displayed after every stage is completed and spits that out.
Only the integer figures shown after each level are added. Interesting!
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Orange Claw Hammer wrote:
- The blue guy in the room before the golden key can either stay put long enough for Goofy to grapple twice, or he can even walk downward toward Max and Goofy can grapple him once. Either way, it is definitely possible for him to be closer to Max when Max throws the barrel upward at him than is the case in the WIP (I think even pretty much adjacent to Max).
This was an oversight on my part, I will look into it.
Orange Claw Hammer wrote:
- I noticed you got a couple bombs in the boss fight this time around, yet still no double kills! Perhaps it wasn't possible with the exact pattern you got here, but it should still be obviously faster to have a fight with double kills going on than all single kills.
I have not yet looked into this in detail, but I have full intention to do so (think of the current WIP as version 1). Given that I do get some bombs, this might be possible (I hope so, it would look really cool).
Orange Claw Hammer wrote:
I realize luck isn't very readily mainipulable in this game as some others, so if it simply wasn't possible for you to implement these strategies then I understand.
Yeah, this is indeed a problem. For example, I've managed to get around 10 different boss fights (=10 different RNG settings) by manipulating the RNG in the room before the boss. It's better than nothing, but putting it in the perspective that there are 65535 RNG values, it's not that much.
Orange Claw Hammer wrote:
Also, one question: is it possible for you to find out how the final in-game time works, the one displayed after all the credits have rolled? We're not sure if it takes into account hidden demicals cut off in the individual stage times or if it simply adds the integer figures displayed after every stage is completed and spits that out.
This I can definitely look into. I'll get back to you on it.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
I thought I'd make a new post since I have a new WIP:
Movie file
This is up to the first boss, and I've saved 38 frames so far. I had some bad luck towards the end, so in the last room before the boss I took damage instead of zig zagging between the pirates. This is exactly as fast, but doesn't look as impressive. I will try to manipulate luck so I can do the zig zagging, since it looks better, but if I'm unsuccessful I'll keep this as it is.
EDIT: I just gave the boss fight a try, just testing stuff out, and managed to improve the fight by 20 frames! I'll try to improve it further, but still, this is a good start.
Movie file
In game completion time is 1m 38.45s. I will redo the last part of the fight and move Max around more before the level ends.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Personally, I support that this should obsolete the 1 player run - it's faster, there's more happening on screen with two players, and there are some nice new time savers that can be used. Although I like the game, I don't think it should have 3 categories (I definitely think this run should be published).
I'm afraid I didn't spot any potential improvements in the WIP. Never the less, I look forward to the finished product, keep up the good work!
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Thanks, I've added some of that info now, I'll gradually add the rest in the upcoming days.
It seems what you get when you pick up a pot depends on the usual two RAM addresses: 7E1A58 and 7E1A59, the same addresses that determine enemy movement. This means that if one pot held a diamond due to the RNG value, all pots would have held a diamond - all pots share have their item determined by these RAM addresses.
I'll look into this more this weekend, in the worst case I can make a lua script that simply loops through all 65535 possible RNG values to get a table of what item you will get depending on the RNG. It could be interesting to see, as it gives an estimate of the chance to get a certain item from a pot.
EDIT: Digging seems to work in exactly the same way - if the current RNG says that you'll find a banana under a pot, you will also dig up a banana with the shovel wherever you dig.
EDIT2: Ok, big edit: I've looked into the player order more in detail today - I've calculated a rough estimate of how many frames would be saved/lost in each room if the player order was reversed. Short story is that switching Max to player 1 should be about 30 frames slower than having Goofy as player 1, so it's a very close call. I do believe, though, that I was quite thorough when I checked this, so I do think I have enough material to say that using Goofy as player 1 is faster.
Also, maps, with room numbers, of all the 5 levels:
Level 1: http://i41.tinypic.com/2zzhxj7.png
Level 2: http://i44.tinypic.com/2d1u7t3.png
Level 3: http://i39.tinypic.com/9icq4x.png
Level 4: http://i41.tinypic.com/141861i.png
Level 5: http://i42.tinypic.com/23rpp1k.png
Here's a google document spreadsheet with my calculations:
https://docs.google.com/spreadsheet/ccc?key=0Akge7o4BbEEgdEl2UUNCTkdCTjY5SDljeXllX1JOWWc#gid=0
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Here is a map of level 1 where I've added room numbers, as they are named by RAM address 7E0067. I will refer to room numbers using this map:
http://i41.tinypic.com/2zzhxj7.png
Also, the Game Resources page has been updated, so check it out if you're interested.
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
mmbossman wrote:
Awww sad, I was hoping that our run was going to be safe from further optimization. If you keep going Randil, I only have one request: Make sure that Goofy still fellates that enemy in level one. A wholesome Disney character performing an unmentionable act like that is just too awesome to leave out.
Of course, this will definitely be kept for the new run! :) (perhaps I can add more of them? There are some more guys just innocently hanging around on level 1...)
Experienced Forum User, Published Author, Skilled player
(1890)
Joined: 4/20/2005
Posts: 2160
Location: Norrköping, Sweden
Well, taking damage seems to have big support, so I'll go for it.
It seems I need to go back to the drawing board for this run - the player order is something I've dreaded looking into a bit, but it seems I'll have to do so. In the worst case scenario I'll actually have to make 2 parallell TASes to see which one is faster in the end.
There are many tricks and strategies one can use in this game, so I've made a game resources page where I intend to document them:
http://tasvideos.org/GameResources/SNES/GoofTroop.html
As you can see, not much so far, but I'll add in more when I get home from work.
EDIT: Orange Claw Hammer, please do send me all your findings, even if I won't have any immediate use for them, I'm still very interested in hearing them.
