Altered Beast behaves strangely in FBA.

[modeler]

Moved from the issues list from RetroArch.

There seems to be an issue with one of the Sega System 16 games "Altered Beast", where some enemies appear without heads and do not disappear when killed. This is present in RetroArch 0.9.8-beta2 on both Windows and Linux. It also occurs when using the libretro build of iMAME4all, which is odd because the Android version is fine. Other System 16 games such as Golden Axe, Shinobi, Out Run and such appear to work fine.

This is visible immediately after the game has been started, with the first wave of enemies on the first level. I tried the same ROM in the original Final Burn Alpha emulator version 0.2.97.28, and the issue does not occur. Discussed further here:

http://forum.themaister.net/viewtopic.php?id=229

Here is a screenshot:

--- Want to back this issue? **[Post a bounty on it!](https://www.bountysource.com/issues/6798917-altered-beast-behaves-strangely-in-fba?utm_campaign=plugin&utm_content=tracker%2F820398&utm_medium=issues&utm_source=github)** We accept bounties via [Bountysource](https://www.bountysource.com/?utm_campaign=plugin&utm_content=tracker%2F820398&utm_medium=issues&utm_source=github).

[modeler]

This bug is also present in the Android build (0.9.8.1).

[inactive123]

Confirmed, but no idea as of yet what is causing it. Will try to look into it.

It's not an endian issue at least since it happens on PC as well.

[peteuk]

I have also found the same bug in the Android and PS3 versions.

[modeler]

It's a weird one is it appears in multiple cores such as mame078-libretro, which points to an issue with RetroArch itself but I'll never know how the devs would go about debugging something like this.

[inactive123]

If it happens in MAME 0.78 as well - then I'm leaning more towards it being some kind of emulation bug - especially since MAME and FBA share a lot of sourcecode (in fact, most of the drivers are just re-wired MAME drivers).

[modeler]

That would make sense, but it doesn't appear in the Final Burn Alpha emulator itself (tested version 0.2.97.28 on Windows). It doesn't appear in the iOS or Android versions of iMAME4all either (I don't recall seeing this bug in any version of MAME over the last ten years or so). It only seems to affect the libretro builds unfortunately. If I can help with any further testing, please let me know!

[sergiobenrocha2]

"A little warning for self-compilers and port maintainers. Later versions of GCC subtly mis-compile the M68000 core. An example is heads missing from the zombies in Altered Beast"

http://www.barryharris.me.uk/article_view.php?id=198 http://www.fbalpha.com/view/217/

I compiled with gcc 4.4.7, 4.6.4, 4.7.3 and 4.8.2, same issue. With clang I have the sames issues too.

[barbudreadmon]

Fixed in my libretro port of current fba stand-alone : https://github.com/barbudreadmon/fba-libretro

[inactive123]

barbudreadmon - Thanks for doing that.

Do you want us to fork off your repo or should we move the repo to the libretro organization? I would give you admin access over it too then so you still hold control over it and can push to it yourself.

[barbudreadmon]

Move it

[inactive123]

OK, transfer the repo over to me (twinaphex) - you can do this from the Settings page of the Github repo. Once transferred I will transfer it over to the libretro repo and give you admin access over it as well.

[barbudreadmon]

As mentioned on forum, the new core seems slower than the old one (noticed it on rpi2 with cps3 games, they don't run at full speed anymore), so it could be a good thing to keep both repo separated.

[barbudreadmon]

You didn't receive the transfer ?

[inactive123]

The transfer request seems to have expired, can you try sending it again?

And yeah we will also keep the older version for those systems that need the extra performance.

[barbudreadmon]

Just sent it again. As mentioned on forum, i found out where the issue with speed comes from, but my knowledge of C/C++ inner workings is still pretty poor, so perhaps someone could help.

[sergiobenrocha2]

@barbudreadmon could you point out the fix? Is it related to Musashi 3.32 update?

Edit: for reference: https://forums.libretro.com/t/i-wrote-a-new-libretro-port-of-fba-based-on-current-fba-stand-alone/2103/22

[sergiobenrocha2]

Missing effects in Super Hang-On too (super charge fire)

[barbudreadmon]

Yes, it is related to fbalpha2012 using the old Musashi.

[barbudreadmon]

Btw, if you update musashi, m68k games will probably end up running slower, which would void the purpose of using fbalpha2012.

[ghost]

Hi @barbudreadmon I was just wondering if you could share the fix for Altered Beast, I compiled fbalpha 2012 and the latest one from 2017 and they both present issues on PS3 build. Monsters without head, GFX issues, etc. Btw, wanted to ask, have you found a way to make megadrive games work? I find it weird that they do not work on PS3 build.

Thanks in advance for any info! SeeYa! :)

[barbudreadmon]

Sorry i have no info about this, all i know is that updating to musashi 3.32 solved issues with altered beast on x86/arm platforms. Megadrive games also work on those platform. If ps3 still have issues with those games, i would say it is likely a big endian issue in their driver part, not the cpu part (since they are using same cpus, neogeo/cps probably wouldn't work if it was an issue with the cpu part). I suppose you could get some help upstream if you ask "gamez fan" (he works on a xbox360 port of fba, which run on big endian too)

[ghost]

a new 3.4 core update is out never seen this till now i was trying to figure out what was going on. in mame 2003. There isint many changes going on in the updates that will effect speeds as far as i can see. The only real loss you would have is if your using this core instead of the assembler core.

https://github.com/kstenerud/Musashi