RetroArch Android releases (v1.0.0.2)

Anyone else notice a performance drop in the MAME 2003 core from 0.9.9.6 to 1.0.0.0? Mortal Kombat (1) has pretty much 0 slowdown on my old original Galaxy Nexus with the older version, but it’s a lot choppier with 1.0.0.0, with all the same common settings set the same in both versions.

Luckily both versions can coexist as installs at the same time.

Have you forgotten to enable ‘DCS Speedhack’ inside Core Options?

Previously we would honor the ‘feature low audio latency’ option - now we assume that it just tells a bunch of BS and we just turn low latency audio on anyway. That might be the problem.

Looks like I will have to add an option/workaround for that as well. Looks like we can’t trust on certain devices that it tells the truth, and on other devices we have to honor it.

But no way in hell am I going to put ‘device-specific’ hardcoded hacks in there.[/quote]

I have performed more tests and I think there is a problem with the touchscreen overlay. When it is active, the sound is choppy and the image is slow, no matter if I am touching the screen or not. When it is disabled, the sound is good and the emulation is fast.[/quote]

Might just be touchscreen overlay overhead then. We thought that having separate buttons was going to be faster vs. one big texture before.

Anything new with the problem of the Custom binding of keys ?

Thanks,

Regit

First of all, thank you so much for all your work so far.

I’m an old 0.9.9.6 user and the retroarch cores used to run flawlessly on my Nexus 4. I saw the 1.0 update when it was re-released on the play store and decided to remove the older version and simply install the new one from the store. Back then I tried several games in several cores and all ran perfect, on the PCSX rearmed i ended up trying tekken 3, FF7, Chrono Cross, etc and all ran without issues.

With the 1.0 version however, the only core I tested this with was the PCSX ReARMed core and Final Fantasy VI.

Sadly it seems to be crashing now once I’m through the intro videos for some reason, something that did not occur back in the 0.9.9.6 version. I will do some testing with another game and see how it goes.

I have the appropriate bios on it’s folder (which I have selected on the folder options) but it doesn’t seem to be “fetching” the bios from the folder as I get the “no bios found message”. This issue also occurred back in the 0.9.9.6 version but since I never really had crash issues back then I simply ignored it. Not sure if that’s the cause now, but it certainly wasn’t causing issues previously.

Were there any considerable changes to Retroarch to cause these crash issues? :
Is there a way to temporarily revert back to 0.9.9.6?

Thanks in advance

It’s enabled (even though I don’t think MK1 used DCS)…settings that are common between both emulator versions are set the same, including the refresh rate.

With the FPS output enabled, RetroArch 1.0.0 shows that MK1 is churning along at 40-41 fps after the opening title screen, in-game, while r22/0.9.9.6 has 0 audio crackling and the game is running full speed.

I’m having audio crackling and video stuttering issues with the PCSX core as well. The core does appear to work better when the overlay is disabled, but I can’t test it. My wireless 360 controller isn’t being detected and custom bindings aren’t working.

Also is there any harm with RetroArch incorrectly identifying my tablet? My Tegra Note 7 was detected as a Nexus 7.

While using the latest version on my Shield, I noticed 2 issues:

  1. Fast forward doesn’t seem to work anymore? I used to pull up an overlay with the fast forward option but now it doesn’t function and even crashed the app when I was tapping on the fast forward icon.
  2. If you hit the home button in the middle of a game then return back to retroarch via its icon and select Resume Content, the controls will be messed up. I only tried this with pcsx, not sure if it happens with other cores. If you return to the app via the multitask window instead, controls map out correctly again.

I can confirm that I have this problem too. And in more than one emulator(core) if I remember correctly.

Hey. First I want to say that 1.0.0 is a HUGE improvement over the previous release!

However, there are a few issues that I see. I only have a JXD S7800B, so I don’t know if these problems are specific to the device or not.

  1. I can use the MAME 2003 menus via the R2 button, but neither input changes nor dip switch settings save anywhere.

  2. In MAME 2010, I can also enter the MAME menu, but can’t do anything with them because no button seems to select anything. In MAME 2003, the B button selects. I don’t know whether it saves anything, as I am unable to change anything to save it.

  3. VBA-Next crashes in the title screen of a game I wrote. The game tuns fine on hardware and VBA-M (though slowly). If I press a button to start the game quickly, it won’t crash, so it’s definitely something in the title screen VBA-Next doesn’t like. I know it’s no big deal if one programmer’s PD ROM doesn’t work right, but I’m sure this can’t be the only thing out there with this crash. The advantage is I still have the source code to my game and will help track down the problem.

  4. There is a bug in the GBC compatible channels in VBA-Next. I’m going to try to figure out exactly what’s going on when I have time, but it may be treating Sound 3 like it’s on an actual GBC, when Sound 3 on a GBA has improved functionality, Again, it is correct in VBA-M, but slow.

  5. and 4) also existed in the previous release.

Keep up the good work! I played Championship Wrestler (a personal favorite) today, and the only complaint I had was that it’s much harder to “waggle” a D-pad than it is an arcade joystick! :wink:

Regarding the input issues, they should be mostly–if not entirely–resolved in the latest git code, hopefully, so lordashram’s next build might be a good one to test to see if anyone’s reported issues are fixed or not. I’m not sure if his latest build includes the fixes or not, but it might worth the time to check.

@Bargaintuan VBA-Next is essentially VBA-M with some targeted speedhacks and some tile-based rendering changes, so it’s not surprising that it has some problems that aren’t evident in VBA-M. That said, if you can find out what’s breaking, it may be worthwhile to fix the issue(s), assuming the fix doesn’t drag the speed down too much.

Squarepusher, turns out the speed problem I was having in v1.0.0 vs v0.9.9.6 was entirely because of the new multiple image overlay setup.

On a whim, I disabled overlays entirely (usually even when using a gamepad, I have the opacity set very low), and MK1/etc ran at the full speed I was used to.

I also enabled+loaded the included single-image ‘720-med’ overlay, and it was still not affecting speed at all…only switching to any of the other overlays did I get any extra slowdown.

So, that was the big mystery :slight_smile:

I understand what VBA-Next is, which is why I ran VBA-M to see how far the issues went. The issues are obviously in the optimizations/hacks that allow VBA-Next to run full-speed on my S7800B where VBA-M doesn’t.

I wrote a music tracker for the GBA, which is what I used to produce music for all my GBA games. I should be able to narrow the sound problem down using that. It’s pretty lightweight, so VBA-M should run it at full speed and then I can compare the sounds the two cores make from the same inputs. I suspect it’s a Sound 3 problem, but with four channels going, playing things I wrote a decade ago, it’s hard to tell what’s going on.

As far as the crash goes, I’ve found the bit of source for what it’s running at that moment. It’s setting the transformation matrix for background layer 2, setting the coordinates for that layer, checking for button presses, updating the music player, generating a random number (I use this method to seed the generator, since the GBA doesn’t have a clock and players won’t press START on the exact frame twice), checking to make sure we’re out of V-Blank, and halting the CPU until next V-Blank via a software interrupt (BIOS call).

Most of this stuff happens all the time in my games, so it’s hard to imagine why the emulator picks this particular time to crash, especially since it doesn’t crash immediately and instead runs a few frames into the title sequence before it dumps you to the Retroarch menu. My first guess is that it breaks when the registers BG2X and BG2Y go negative.

@Bargaintuan Sounds like you have it narrowed down pretty well. At this point, I would recommend opening an issue at the github repo for VBA-Next where Squarepusher can take a look. He probably won’t get to it immediately, as getting the 1.0 bugfix out the door for all platforms (and releasing 1.0 for Xbox/360) looks to be his highest priority at the moment, but that should all be finished soon.

I haven’t really narrowed it down that much. I know where it crashes in my code, but it’s doing several things at that point. I actually alternate two different music tracks on the title screen, so by playing and dying, I was able to check for the crash with a different track. It crashed in the exact same spot, which means it’s probably not anything in the tracker. My best guess is still the background layer registers, but it’s only a guess.

I found a new problem, though. In some MAME ROMs under 0.78, none of the controls seem to work on my S7800B. I can’t even insert coins. The ROMs I’ve found this on are starwars and esb, which are two awesome games. Worse still, it seems no vector games work in MAME 0.139.

I haven’t really narrowed it down that much. I know where it crashes in my code, but it’s doing several things at that point. I actually alternate two different music tracks on the title screen, so by playing and dying, I was able to check for the crash with a different track. It crashed in the exact same spot, which means it’s probably not anything in the tracker. My best guess is still the background layer registers, but it’s only a guess.

I found a new problem, though. In some MAME ROMs under 0.78, none of the controls seem to work on my S7800B. I can’t even insert coins. The ROMs I’ve found this on are starwars and esb, which are two awesome games. Worse still, it seems no vector games work in MAME 0.139.[/quote]

For MAME 0.78 - those games would need to be button mapped. This hasn’t happened for those two games since I never bothered to play them. Maybe you can remap the buttons by bringing up the MAME UI and changing them there. Not sure if you can do that right now with the Android version.

Re: MAME 0.139 - I believe we implemented a patch for that in MAME 2013/2014 (ie. 0.151/0.152) - it might be that we have neglected to backport that yet to MAME 2010 (0.139).

Anyway - for both of these two issues - go to our Github and make issues about them in the respective repos.

I haven’t really narrowed it down that much. I know where it crashes in my code, but it’s doing several things at that point. I actually alternate two different music tracks on the title screen, so by playing and dying, I was able to check for the crash with a different track. It crashed in the exact same spot, which means it’s probably not anything in the tracker. My best guess is still the background layer registers, but it’s only a guess.

I found a new problem, though. In some MAME ROMs under 0.78, none of the controls seem to work on my S7800B. I can’t even insert coins. The ROMs I’ve found this on are starwars and esb, which are two awesome games. Worse still, it seems no vector games work in MAME 0.139.[/quote]

For MAME 0.78 - those games would need to be button mapped. This hasn’t happened for those two games since I never bothered to play them. Maybe you can remap the buttons by bringing up the MAME UI and changing them there. Not sure if you can do that right now with the Android version.

Re: MAME 0.139 - I believe we implemented a patch for that in MAME 2013/2014 (ie. 0.151/0.152) - it might be that we have neglected to backport that yet to MAME 2010 (0.139).

Anyway - for both of these two issues - go to our Github and make issues about them in the respective repos.[/quote]

The main problem with button mapping is that changes don’t seem to be saved, at least on my device. I can’t reach the MAME UI on starwars or esb using the device’s controls, but if I had to, I could use and overlay or even connect a keyboard to get stuff mapped, if I didn’t have to do it every time I ran the game.

I’m going to try and gather more data on the VBA-Next issues. Like I said, I have a music tracker I can use to find exactly what the sound issue is. For the crash, there are a few more things I’d like to try before submitting a report. I want to give as much info as I can find, and I’m digging through stuff I haven’t looked at in years.

Anyway, despite the few issues that I’m sure you’ll work out, it really is a great improvement over the previous version.

[/quote]

maybe you don’t have the good roms and the pb is in your own , as for me these games work perfectly in 0.139 ,like other vector games .

http://imgur.com/FB3YC2J http://imgur.com/eKzLX3c

Nice Update , thanks, my controlers are working better than evee! o But Parasite Eve II still not works on PCSX reARMed, aways freeze on the same part, and I don’t know how to delete cores .-.

By the Way, thanks for the nice update. ‘-’

I’ve been playing a lot with the windows version, but just tested this on my Galaxy S III.

NEC CD in bin+cue and ogg music are working now. The onscreen controls seem a lot more responsive as well. Menus are nicer too.

Awesome work.