Posts for JMC47


Experienced Forum User
Joined: 5/8/2014
Posts: 125
I did some basic input recordings for a project of my own on F-Zero GX and suffered no desyncs whatsoever. I wasn't TASing, but rather just doing some replay/recording stuff. I think it's rock solid, but someone should definitely try it with savestates.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
We messed up the original merge due to a last minute change. Here's the actual fix after we broke everything for a day! https://dolphin-emu.org/download/dev/f00ad2e65c9136d16a93aeec7c225a271d5adf95/
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Before the build is merged; I'd like to have the timings at least "approved" by the people here as better. I know that's a lot to ask, especially with the busy schedules people have, but I don't want a repeat of last time where we think it's fixed, and then we find caveats to that.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
ItsPersonnal: Could you make another side-by-side video showing the new CAV build? I'd like it a lot if we could feature the old video (The one that shows it wrong) and the new video in the progress report. It's really thanks to the diligence of users here that features like this continually get worked on. My White Wii has very bad loading times on it (confirmed via cleanrip) so I'd rather have footage from someone who's already tested and verified the loadtimes. I'm also not a metroid prime speedrunner that knows where the load spots are.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
ItsPersonnal: I got too far ahead of myself last time the load-times were improved; so I know the feeling. Will be absolutely thrilled if this makes things better for the TAS. Regardless, JosJuice did an excellent job whipping up a quick patch!
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Yeah, Hopefully this is closer. I don't expect it to be perfect. We're trying to emulate a top of the line Wii according to the speeds.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Also speedrunners have been using Wiis to get the faster reads anyway. I'd honestly like itspersonnel (sorry if I misspelled your name, lazy right now and don't feel like scrolling back pages) to compare and see if it works in various parts of the game. The best way to fix a bug is to find the people that care about them the most; they'll do the most thorough job looking for problems.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
The graphical puny glitch is fixed in the latest builds of Dolphin, and probably fixed in that build with EFB2RAM, I think?
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Did more studying on disc read speeds. They make no sense. A Wii disc on the very inside of the disc reads at about 2800KB/s, where as a GC game reads at little over half that. Wii disc maxes out at about 3000ish; will test on more discs later. Unless my USB/SD card is the limiting factor at that point. GameCube games max out before that, so I don't think it's my SD card there. CLEANRIP RESULTS: So, I got two freshly polished games at my local shop (Non-Duallayer for the wii game) and dumped them to see how it would work. Okay, so, Wii Discs end up 2800KB/s - 3000KB/s from start to finish on the disc on my white wii. That means the Loadtimes on Metroid Prime Trilogy should be really, really close to console. If anyone has that and wants to confirm, that's be nice, maybe a tiny bit faster or slower depending on our seek accuracy This GameCube game started out at about 1900KB/s and steadily rose to 2500KB/s (2527 to be exact.)
Experienced Forum User
Joined: 5/8/2014
Posts: 125
I know that's weird, but that's the behavior I get in Cleanrip...
Experienced Forum User
Joined: 5/8/2014
Posts: 125
I've been experimenting with Cleanrip on Dolphin (it actually runs and works!) and I've found out that Dolphin runs at about 3000KB/s constantly. Metroid Prime should be about 1500 - 1700KB/s on the inner rim and 2600KB/s to 2700KB/s on the outer rim. Older builds of Dolphin went extremely fast before eventually failing, so this is a lot better than before, but it leaves a lot to be desired. Efforts will continue to make it more accurate in the future.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
I totally believe that Dolphin's load times could be wrong thinking about it. It sucks, but if things are as far off as said, something isn't being emulated close enough.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
I went through and compared my GameCube to Dolphin in the opening areas for load times. Now, I don't have much in the terms of speedrunning knowledge for this game, so I just kind of played through, and tried to measure a few of the doors, and couldn't get a notable difference. Considering that load times vary per console, version of the game, etc, there will likely be cases where Dolphin is faster. Heck, Dolphin is just simulating a disc drive, and the way that Metroid Prime loads things may hit something that we don't emulate well enough. If you guys are dead set on making the load times more accurate for this game and this game only, I'm sure you could mess with some of the numbers to get it reasonably close.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
If it doesn't work, don't worry about it. I was hoping it'd be merged before all of your TAS work started, but there were a few delays with the pull request and a few things that had to be researched and checked. I just hope the old bounding box emulation doesn't cause you any heartaches.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Not to be off topic or anything, but I don't think this is appropriate to be argued in the Wind Waker TAS thread. I think every game can be enjoyed regardless of its popularity. Just because game X didn't sell as well as game Y doesn't mean the TAS is any less important or exciting... I'm not a huge sonic fan, but I absolutely love the speedruns, they bring out the best (and sometimes beyond) in those games and I always look forward to them. The hyperoptimization is ridiculous! On the other hand, Zelda is a series near and dear to my heart, and I'm excited to see the ways in which the games can be broken and how far they can be pushed compared to my original experience. To say decide what is and what isn't a major title is kind of silly. Any title, big or small, has the potential to make a great TAS.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
I know this probably is a bit too late, but a way better form of bounding box emulation was merged into Dolphin a day or two ago. It notably affects the Paper Mario Wii and GC games. It should just be graphical issues in Thousand Year Door (if any remained), but it was known to cause a crash in the Wii game with the old emulation.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Maybe you could report the pixel shader error, assuming you have a modern graphics card, that shouldn't be happening. Actually, If you send me the bad PS file, I can send it to someone who would understand what's going on.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
You have to manually combine audio and video, but it works in my experience. Just open something like Virtual Dub at the very least and combine it, or if you have more advanced programs, Adobe Premiere and Sony Vegas work too. Everything syncs up as it should in most titles.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
The Punies should work relatively well in master, there's a Pull Request to completely fix everything (bounding box related) as well. It should be noted that bounding box isn't actually for collision detection stuff, but rather some graphical effects.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
I wonder what's causing all the desyncs; it bothers me, especially since stuff like Netplay syncs up. I'd guess more variables and such aren't being savestated properly, but I have no idea what.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Removed this post, because the fix in question turned out to break something else. I'm sure something will get revived to fix the music eventually, though. If you want to play with the music in this game, you can use this build, but it won't be seeing light in master. https://dl.dolphin-emu.org/prs/pr-994-dolphin-latest-x64.7z
Experienced Forum User
Joined: 5/8/2014
Posts: 125
I think the complaint is legitimate though; the load times affect the game, but emulating disc media load speeds isn't exactly easy. Disc speeds are "fairly accurate" in Dolphin right now for most games, but if it's not up to someone's standards for a TAS, so be it. It only means there is room for improvement in the future.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Yoshifan generously provided 112 replays from console, and I'm happy to say the ones I've gone through (about 60ish) sync up perfectly between Dolphin and Console.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
Well, I think the general discussion of loading times is very relevant to Metroid Prime. It affects the game and how it plays. Dolphin added support for more accurate loading times more or less because it is known to affect the playability (and thus glitches) of certain games. One of the things hoped to be accomplished by the feature was that Metroid Prime's speedrun route would be possible on emulator (at least, from me, that was something I wanted to make sure worked) but, I don't think frame perfect load times are plausible. If Metroid Prime's load times are that far off still (I did some remedial testing on my white Wii, with Revision 0 Metroid Prime, and it was very, very close playing through the frigate) then we can't really put the issue completely to rest. Yes, Dolphin more accurately emulates disc load times, but that doesn't mean it emulates them perfectly. And I think if the emulation quality is below the standards of some TASers, then they shouldn't be bullied over their decision to not want to work with it. And, like any open source project, the code is available to view. If anyone spots a potential problem, or thinks they can implement it more accurately, they can try their hand at it. Slowing down the loads just for Metroid Prime isn't a solution (for master, if you want to do a special version for a TAS, I think that's understandable) because other games seem to line up very, very well to console (Luigi's Mansion, Wind Waker were among tested games) and sometimes were slightly slower even. I hope this more or less explains the discussion going on here. I'll be doing more hardware testing as I go to try and verify if there are any (easy to spot, I'm not a capable coder) problems in the current disc read speed setup.
Experienced Forum User
Joined: 5/8/2014
Posts: 125
well, if the disc read speeds can be made more accurate and can look at the code and get better results, I definitely think they should. The problem is that changing it for one game affects it for every game, so if any single little variable is wrong, it won't seem right for all titles.