Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
Wobmiar wrote:
Out of curiosity, do someone know if Dolphin devs intend to make GBA encoding a thing?
That is something that is intended to be added. It's really just whenever someone has the time to go hook everything up. mGBA has worked fine before with encoding (see upstream and BizHawk) so it's just a matter of time before someone does the thing.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
ALAKTORN wrote:
Was watching the run and had a few questions:
Would it save time to enter the Lavender Town Poké thing and use an Escape Rope after learning Fly? I assume that if it does it’s so little it’s not worth using it over some other location.
In the Safari Zone did you exit the house before using the Escape Rope because it wouldn’t let you inside? Is that some special house or what?
Also noticed at Sabrina you had to be on a teleport tile to Escape Rope but after Blaine you could just go… how does that work?
1. You have to actually heal at the Center for the Escape Rope location to update so no.
2. Safari Zone let's you rope, inside the house doesn't, ask Gamefreak why they designed it that way.
3. Escape Roping on the Teleport tile isn't required for roping out of the gym, but it's faster as it skips the Escape Rope animation.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
MESHUGGAH wrote:
Wow! That was way fast to create a 3:17:50.07 [709348 frames] TAS. Obligatory good job!
According to the submission text it's 9 edit: and 6 seconds BIOS time)
1. It did not take that long since a script applies all the inputs, once it got going it only took around 5 hours and that was partially because my pc is garbage.
2. I said it's 6 seconds if you disregard the BIOS time, implying the BIOS takes 3 seconds (9-6=3).
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
I took a peek at this desync. I logged out LY values during joypad polling, starting right before the code screen before level 3 all the way to the end: https://www.dropbox.com/s/4gjnk805vvqqgft/lyDebug.txt?dl=0
The game polls very erratically. Sometimes it's very early in a frame (LY 91), sometimes it's very late in a frame (LY 8F), sometimes it's in the middle of a frame, sometimes it's right after VBlank is finished, sometimes it's in the middle of VBlank, it's all over the place. I would definitely suspect input timings/rounding screwing us over here.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
I've redone the TAS on the JPN version: http://tasvideos.org/userfiles/info/73397997229424541
I'm not yet entirely convinced this should be the preferred version for this submission, so this shouldn't replace the movie file, at least yet.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
Jigwally wrote:
Well I'm referring to a 100 minute improvement, presumably more than just lag lol
Cut out 60% of the lag and that's enough for said improvement.
EDIT: Also on a peek at that version shows the lag isn't so much different, although the character just seems to move faster overall. Still doesn't sit right to use that version anyways.
EDIT2: Checked every level on the linked TAS and all have an equal number of moves compared to this submission. It has no gameplay improvements in it.
1) There is a lot of visual artifacts that appear in that version which do not show up on international releases. I don't think I would prefer this version if this is the cost to pay for assumingly less lag.
2) Don't think that counts as a gameplay improvement.
Although yeah, if the second point holds up that is indeed an improvement this is missing.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
MESHUGGAH wrote:
edit: so what's up with the improvement called as "questionable"? According to the observation from the submission notes, using the probably next version of BizHawk that emulates better than the current one is okay, as long as that version going to be released where the TAS will sync.
The "questionable" part is due to the fact the TAS is much longer, which is (probably) explainable by more accurate emulation coming from 2.2.1 BizHawk (latest old TAS syncs on) to 2.6.3 dev BizHawk. I only used a fairly small sample size for determining the moonwalking was a lag reduction trick, but perhaps I'm wrong and it adds more lag over a long period of time? Really, only 100% way to confirm this is an improvement is to do the moonwalking on 2.2.1 or do the hold direction for animation on 2.6.3 dev (which granted, the latter is probably very simple with a lua script, but eh, judge can torture themselves with that if they want to go that far).
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
c-square wrote:
But if these runs are being demoted out of moons and being considered for vault, then they should be subject to the same criteria as any other submission for vault. If they pass, they can enter vault, if they don't they should be unpublished.
Runs are not unpublished ever, regardless of any sort of reason (and just impossible short of outright hacking them out of the db). There's actually was a big discussion regarding that, which led to massive logistical issues being revealed. And historically, unvaultable movies put into Moons would not be put into the Vault and just stayed in Moons, perhaps being obsoleted by another branch/older valid movie if such exists. If an unvaultable movie is put into the Vault from Moons demotion, then that's just a mistake and does not set any sort of actual precedent.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
http://tasvideos.org/userfiles/info/74260865280592490
Did an oops and submitted movie is missing the cycle count, here is the same movie with the cycle count saved.
EDIT: Also improved accuracy of the mapper which resulted in more lag appearing. Movie file link is corrected to reflect this.
EDIT2: Some more accuracy improvements, cycle count slightly different, link updated.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
There's something I am wondering about the "syncs to console" part of this idea. How does this work wrt GB/C? GB/C TASes are currently only syncable to console if you go resync them with a GBA platform*.
This is mainly problematic due to this part of the rules regarding platform choice for GB/C TASes:
The default options that appear when starting a recording should often reflect the device the game was specifically released for, and newer device modes may cause glitches, so think twice before changing them.
If glitches that are caused by a newer mode are obvious to unassisted eye/ear in normal viewing conditions, or hinder gameplay, that mode is not allowed.
This part of the rules if I recall correctly was meant as for certain problematic games, a console verification would have to be resynced to GBA mode while the publication would have to use GB mode, and GBA mode would not be able to be published. I do wonder how this ends up working out with this idea.
*(Ok technically SGB TASes have been able to be verified but even then your "good" choices are either lsnes (not a very good choice in the first place but way better than BizHawk BSNES) or the BSNESv115+ BizHawk core (actually the most accurate SGB rerecording emulator at this point), and even then some resyncs might be practically impossible since the SGB filters out L+R/U+D and some inputs are not going to be possible due to frame timing going off SNES VBlank)
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
So this is a rather interesting case. While it might seem like this submission can be Vaulted due to trying to aim for full completion, it ends up falling flat on that for two reasons:
1. Memory manipulation is not used in this TAS. Going off RTA leaderboards, memory manipulation appears to save over 2 hours of time. It saves so much time that the RTA record using memory manipulation is around 1.5 hours faster than this TAS, which even if you account for RTA using a PS2, that's still much faster than this TAS. Thus, this TAS does not set a record for full completion.
2. This TAS aims for 200%. This is full completion if only 1 player is used. However, if 2 players are used, this isn't actually the case. An extra trophy can be obtained with 2 players, making the maximum percentage actually 201%. For the purposes of the Vault, 201% should be the aim for full completion, not 200%.
So, since this run is ineligible for the Vault, it could only potentially go into Moons. And so far, feedback has been very mixed. So I need more feedback on this submission, as so far I'm not seeing Moons level feedback.
Also personally, I don't think this is Moons either. Huge chunks of this run are just boring load screens and general waiting around, along with this just being 5 hours of minigames the gametm. It gets boring real fast.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
MESHUGGAH wrote:
There is a change in 2.6.2 BizHawk which made my Lua scripts broken from 2.3.2.
edit: the change probably introduced earlier but I've only trying it now. also changed version of where my scripts working.
The problem is the following:
NLua.Exceptions.LuaScriptException: [string "main"]:4: attempt to call field 'registerafter' (a nil value)
Example lua which is broken, copy pasted from Wiki: LuaScripting/Registers
local function Fn()
--Insert somewhat mediocre code here
end
emu.registerafter(Fn)
Used NESHawk
That function has not existed in BizHawk for a long while, being removed in 1.2.0 in favor of event.frameafter() (which at some point became event.onframeend()).
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
c-square wrote:
I can attest that developing and tuning the lua scripts to tas it optimally was certainly not a trivial task.
Movie Rules wrote:
If a run consists of executing a trivial strategy where the only challenge is having to do it over the course of several rounds, it will still be judged as trivial.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
Alyosha wrote:
Thank you for the detailed report. I am in NES development mode for now, so my recommendation is to use Gambatte in the dev build for the time being until I can look further into this (which may not be for a month or so): https://ci.appveyor.com/project/zeromus/bizhawk-udexo/build/artifacts
Gambatte-Speedrun r717 works so a dev build is not needed, a recent BizHawk release (2.4.1+ should match r717) should work fine.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
SMWAgent09AF wrote:
A "true" credits warp can only currently be done on 3DS for now.
Judging by the video, I would presume this means the "credits warp" for now hits an invalid opcode. The 3DSVC is just Nintendo's shitty emulator, which ignores these opcodes rather than hardlock as original hardware did.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
MESHUGGAH wrote:
I think this is OK, but what would be the motivation for a resyncer?
Believe me, there's enough people who hate noticeably inaccurate emulation enough to go resync the runs to newer emulation.
Also, console verification can be a factor for relevant platforms, although just the hate against noticeably inaccurate emulation would be the main factor probably.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
Requesting delay until BizHawk 2.6.3/2.7 is released, the current movie desyncs on console due to an inaccuracy fixed in dev builds.
Resync to latest dev build: http://tasvideos.org/userfiles/info/72994296903763990
Console verification using this resync:
Link to video
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
ThunderAxe31 wrote:
What do those corruption glitches used by RTA consist in?
From what I understand, it refers to the same glitches used in the published Any% TAS, which allows for a glitch shop which can corrupt various places in memory, which can also be used to wrong warp all over the place. Wrong warping with the glitch would be the primary time save, since you can just warp to anywhere you need to go recruit the Digimon to get to 100 Prosperity Points.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
TiKevin83 wrote:
the current published movie runs at the wrong framerate due to being on VBA and doesn't account for either the GB or GBC bootROM.
Not only that, but VBA is horrible at timing what it actually has. It just uses VBlanks for timing, which is a really bad idea for timing GB/C due to the LCD being able to be disabled and VBlanks do not occur when the LCD is disabled.
(As an extreme example of why this timing is very bad, let's say a movie has 10 minutes worth of VBlanks, but also 10 minutes worth for the LCD being disabled. VBA will ignore time the LCD is disabled and report a time of 10 minutes instead of 20 minutes).
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
This is an interesting submission.
Optimization seems good, the only immediately questionable part is not resetting to skip the credits, although it doesn't retract this from being a speed record in comparison to RTA's NMG runs and isn't particularly bad in the grand scope of the entire run.
The main problem I'm seeing with this is simply the "no memory corruption" part of the branch. This restriction ends up extending the run by around an hour, and there is already an RTA run with the memory corruption glitches this TAS avoids, so in no way can this TAS be accepted to the Vault under full completion (it does not beat RTA records).
So, for acceptance, this TAS must be able to go into Moons. And I'll be honest, I'm not exactly seeing that here. There have been no replies so far and just 2 No votes.
Of course, this is an RPG, so I can perhaps expect somewhat a lackluster response. I would like some more feedback before making a decision on this movie.
Emulator Coder, Experienced Forum User, Judge, Published Author, Experienced player
(610)
Joined: 2/26/2020
Posts: 699
Location: California
Seems like this TAS does not like AMD GPUs, forcing my Intel GPU to be used seems to make the TAS not desync on that spot.
I'll update this post when the TAS finishes playing back on my side.
I can confirm sync:
