RPCS3
"MEM: Access violation" on [BLES01102] - Ar Tonelico Qoga - RPCS3 v0.0.16-12251-cacb852a
As I think this issue always happened with this game in some way or another, I'm reporting it now:
The game randomly crashes with a memory access violation error when doing almost anything. In my case, I'm just alternating between two different maps on the same "level".
This is the save file so anyone can reproduce it: just walk to the left until the map changes, and then walk to the right until it changes again. Do this a couple of times and sooner or later you'll get the memory access error, it always happens: Savedata.zip
This is the emulator log. Additionally, the emulator kills itself when you try stopping the emulation after getting the error:
Log from v0.0.5-7576 Alpha [2018-12-11]: RPCS3.zip
Log from v0.0.16-12251-cacb852a: https://github.com/RPCS3/rpcs3/issues/5418#issuecomment-841271500
Your Build is months old, you need to update it to the latest version as there is many fixes made in that time
I just downloaded the latest build in the "History" tab of AppVeyor. Sorry I guess.
I doesn't even matter anyway, as the issue always happened. Here's the exact same problem on 0.0.5-7576 Alpha [2018-12-11]: RPCS3.log
Can confirm the bug happens in [BLJS10075] as well in build 7560. The "MEM: Access violation" error happens in the following conditions afaik:
- When map changes
- When item crafting (or イメージ調合) starts or finishes (before or after renaming panel) Log(ASMJIT) Log(SPU LLVM)
- When encounter in the map
Notably, if spu llvm is used instead of asmjit, The error message will be The process has finished and the game will quit immediately when changing maps.
Was that always the case, or was there a build where it worked?
I certainly can't remember a version of the emulator where this issue was not present. It always crashed at some point.
I'm not sure if a bigger problem needs to be raised and attach all these games to it, but the MEM: Access Violation error is rampant in many games.. All versions of Diablo 3 including Ros xpac and the Ultimate Evil edition. C&C Red Alert 3 is another one. There was others, but cannot remember the titles off hand.
This is a pretty generic error and it can have multiple causes. So you can't just bundle that into one issue.
I also have this issue but get around the spots this happens most often in by setting the PPU to Interpreter instead of LLVM Recompiler. This has other side effects like sound stuttering so it's not ideal, but that may help narrow things down?
Yeah, tried all that, and yes, you can proceed a bit further on, but it just freezes up completely soon after, and the performance hit isn't worth it. It will be good to get these Access violations sorted.
Been playing around a bit with it - whenever myself or my follower attempt to hit a boss, boom! immediately F {PPU[0x1000005] Thread (ServerGameManagerWorkerThre) [0x008d9ffc]} MEM: Access violation writing location 0x0. I cannot proceed any further because of it.
So... any love for this thing? As of recent versions, it seems quite more stable at least to me, but still crashes after a little while with the same error...
Well, with PPU Interpreter fast and SPU LLVM I can play over an hour easily without problems. No sound glitches, constant max fps, the only thing is that from time to time some things load a bit longer than expected, but still it's less than 2 seconds.
i9-9900k @ 4.8GHz GTX 1080 16 GB RAM
@vietnam13231 Yeah, well... there's absolutely no way my 4670K can maintain 60 FPS on Interpreter so I really hope this gets properly fixed sometime, even if it gets done as a side effect of fixing other game.
Welp, it was time for me to get into this game and the issue is still exactly the same as it was. Furthermore, Interpreter doesn't fix the issue either, it just delays the inevitable a little longer than LLVM. Is there anyone looking into this?
logs using latest master RPCS3-scene_change.zip RPCS3-synthesis.zip
both stopped at sys_mutex_unlock function according to logs, other mem. access violation crashes also said the same and either x,y,z,w has NaN value on it (which is maybe the cause of the crash, but idk what's that tbh)
Just for the record, this game has had the same random crashes since forever (2017). Slowing the game down indeed helps but only delays the inevitable. Its a well known issue and more or less affects every game with the same engine (such as Atelier Ayesha).
@al0xf AT3 is two years older than Ayesha and was developed before of Gust being adquired by KoeiTecmo, so I think ~~the engine is probably different. Ayesha and AnS having the same engine, for example, is indeed very likely.~~
Atelier Rorona, Totori, Meruru, Ayesha and Ar Tonelico are on PhyreEngine. Atelier Escha & Logy, Shallie, and (in japan only) Sophie is on the "koei tecmo" engine which is used by many other KT games.
Rorona -> Ayesha all have crashes to some degree, though Ar Tonelico seems to be affected the most.
@al0xf You're indeed correct, if what the Wikipedia says is true. They looked quite different for me. AnS is definitely running on some KT engine, but I think it has some MEM: Access Violation problems too in some cases.
It seems that some time ago the error message given by the emulator changed, but exactly the same thing keeps happening, of course. The "new" message reads something like:
F {PPU[0x1000000] Thread (main_thread) [0x00518c04]} VM: Access violation reading location 0x8122cafc (unmapped memory)
I don't know it at least gives some more information about the issue or it's just the same thing worded differently though.
Unfortunately, it is still crashing regularly. RPCS3 Version: 0.0.13-11496-7fc26b1f Alpha
E {PPU[0x1000000] Thread (main_thread) [0x01a81ee4]} SYS: 'sys_fs_stat' failed with 0x80010006 : CELL_ENOENT, “/dev_hdd0/game/BLES01102/USRDIR/Res/sound/se/us/btlvoice.cpk” [3] F {PPU[0x1000000] Thread (main_thread) [0x004ea5a0]} VM: Access violation reading location 0x0 (unmapped memory) [type=u32]
EDIT: OK, so I may have found a setup which works. I'm sure it can be optimized, but I have high FPS and no crashes since an hour now:
EDIT 2: And 12 seconds later... : F {PPU[0x1000000] Thread (main_thread) [0x004eaaf8]} VM: Access violation reading location 0x8122cafc (unmapped memory) [type=u32]
I think the problem may have mutated and sometimes is now is crashing with something like:
E {SPU LLVM} SIG: Thread [SPU LLVM Profiler] is too sleepy. Took 2.700µs to wake it up!
E SIG: Thread [SPU LLVM] is too sleepy. Took 1.770µs to wake it up!
so... progress, I guess?
@RaulDJ Have you gotten that message before or after #9817, there was some progress on that front.
@HenningJW That test was done on the "0.0.14-11772-8be2a55c" and I had not seen that message before, but I only try one version every two or three months or so.
I just tried the "v0.0.15-11838-8e4451d1" version and after 10 seconds I got the
VM: Access violation reading location
again, so I guess not progress at all after all.
:( Was worth a try. I assume you have tried playing around with the driver wake-up delay setting (try 200 and 500)?
If you want to give this problem a good chance to be tackled, you might want to reupload the full log of the latest version, just the message in the log is not enough.
Sadly, none of those two values for the delay did anything. I'm uploading two logs from v0.0.15-11853-ea5e837b: one with an access violation and one without it (it seems to crash without that error when rushing through the menus quick enough). RPCS3.log.gz RPCS3_no_access_violation.log.gz
I'm just about to completely give up on this thing being ever emulated, but I'd like to at least thank you a lot, @HenningJW, for at least trying to help.
I've been playing this game again lately (few sessions for about an hour each), and had no problems at all. No crashes, no graphical issues, 400 fps if not capped... Smooth and normal gameplay. Settings mostly default, only "enable thread scheduler" off and "multithreaded rsx" on. Builds are current, since each time I played I updated the emulator.
@vietnam13231 You said the same thing two years ago, so I guess your 9900K runs on magic. Nor my 4670K managed to do it, neither my current 5900X does it ¯\_(ツ)_/¯.
I had this crash back then. Exact same as in title of this thread. All that changed from that time is SSD in my system and emu version. And weird thing, it forced me to reinstall disk data of the game. It wouldn't let me play otherwise.
Just load a savedata where you're inside an inn, go straight to the "Make image" screen, try to make something and it will crash 99% of the times. The faster you get to that screen, the higher the chance of it happening. I could replicate the issue every single time 2 or 3 years ago and now it's exactly the same thing.