Emulator Issues #3944
closedDirectX 9 plugin - random crashes
0%
Description
What's the problem?
Using the DX9 plugin I experience random crashes at the start or during games.
The errors are "memory could not be read" or "exception unknown software exception", as you can see in the fotos.
Crashes are completely random: you can start a game 4 times and get the error 2 times, then start 30 times a game and get no error, or get into the game playing a lot of time and then get the error, or get it already at the start of it...
Dolphin version with the problem (as it appears in the title bar, Ex.: "R
4779" or "R 6403M"):
r6802 and later
(optional) Dolphin version that does not have the problem:
r6801
Operating system and version:
32-bit or 64-bit:
Windows XP SP3 32-bit, e8400, hd4850, 2gb RAM - DX9 plugin
Game ID (as it appears in game properties, Ex.: "GZ2P01" or "RSBE01"):
Many games (both GC and Wii)
I made fotos of some, but not all the crashes that I encountered (Mario Kart Wii, Wii Party, Wii Sports, Star Fox Adventures, Resident Evil 0 etc. crashed too randomly, but I did'nt made a foto of them).
Foto 1: Batman the Brave and the Bold (credits intro)
Foto 2: Cursed Montain (Wii warning logo)
Foto 3: Michael Jackson Experience (Ubisoft logo)
Foto 4: New Super Mario Bros (at Time's Up screen, during the game)
Build command-line (not on Windows):
Was the ISO a plain dump from disc, compressed and/or scrubbed?
plain dump
Please provide any additional information below.
I use the deault DX9 configuration, with Copy to textures and aspect ratio: stretch to window.
Problem started with r6802 and then got better or worse revision after revision...r6809 (almost immediately reverted) seemed to fix it, but I cannot be sure because the crashes are very random, so to be certain I 'd need more testing time...
In the next days I'm gonna try accurately r6809 and latest revisions with OpenCL activated...
At least OpenGL plugin doesn't seem to be affected by this problem...
Updated by hatarumoroboshi over 14 years ago
Forgot to say that Bartman and Cursed Mountain isos are compressed, while Michael Jackson and NSMB not.
Updated by hatarumoroboshi over 14 years ago
Bartman is not a new Simpson game...of course I meant Batman:-)
Updated by gabrielmorano over 14 years ago
Did not happened to me EVER... and i tested it a LOT :s
(and i havent had any of this issues since i started using dolphin last year) :s
also i finished batman and didnt had a single issue with that game.
cursed mountain does not work (it never did)
didn't played the other 2 games
try updating your video drivers :S
Updated by hatarumoroboshi over 14 years ago
I don't think I can upgrade my driver because is the last one available (Catalyst 10.12), If you have developed a newer version post the link:-)
I still didn't get deeply into the game, but Cursed Mountain seems to work perfectly for me with JitIL (perfectly with OpenGL:-/)...
Anyway other users have experienced this kind of problems after recent r6802...so saying that you finished Batman maybe a year ago doesn't mean anything...
I'm still testing to figure out where's the problem (If I can find it)...
Updated by gabrielmorano over 14 years ago
yes i noticed cursed mountain was fixed on this previous revisions and it does work, i have played up till the monk asks you to use the sight and had not any error (besides been slow)
nope i finished batman last week on r6850
the only problem i seems to have with the r68.. are fifo problems
Updated by dariosamo over 14 years ago
I can confirm this issue with even more popular games such as Brawl and DKCR. Trying to track the revision which caused it, trying out with 6880 ATM.
Updated by gabrielmorano over 14 years ago
@Darios let me ask you something, do you have ati video card? (cuz this could be an ati compatibility issue)
Updated by Hgomez2020 over 14 years ago
Build r6882 keeps crashing when I start a game using DirectX 9 plugin. Dolphin just crashes all the time with that plugin.
Updated by gnick79 over 14 years ago
- Priority set to High
It seems we lost the stability on the roads.
In order to find a solution under statistical method, I want sure results under high percentage; so "maybe" isn't a good word inside your answer.
Let's care what exact revision breaks... if work about last SSE optimizations, FIFO changes or any other thing.
Spend time with a revision and stay on it even more than 1 day for a "robust" outcome... since this happens randomly.
Remember... quick test = unreliable results.
Updated by gnick79 over 14 years ago
Issue 3969 has been merged into this issue.
Updated by gnick79 over 14 years ago
- Status changed from New to Accepted
OK. First analysis by hataru is right.
[cut]
Problem started with r6802 and then got better or worse revision after revision...r6809 (almost immediately reverted) seemed to fix it, but I cannot be sure because the crashes are very random, so to be certain I 'd need more testing time...
[/cut]
In substance recent changes about SSSE3 instructions cause a lot of randomly crashes.... they were unlucky. In this specific case, maybe xascha and james.jdunne can help us.
Updated by bgtsetso over 14 years ago
Well they are not so completely random, if you check my issue 3969, which got merged into this one, you will see that:
- Resident Evil 1 Wii PAL always crashes right after or during the screen "Accessing. Please do not touch the Power button or reset.", which is in the very start of this game.
- Mario Kart Wii PAL ALMOST always crashes during the screen "Saving. Please do not touch the Power button or Reset." or just before this screen, which is in the very start of this game. Sometimes this issue does not happen, but for most of the time it does.
Updated by gnick79 over 14 years ago
this happens with r6081 also?
If yes, I can split your issue from this without problems.
Also you should find the exact revision that breaks.
Updated by hatarumoroboshi over 14 years ago
For me Mario Kart Wii doesn't crash at saving screen, if it crashed it crashed randomly...from your original issue I don't understand if the game crashes only with DX9 plugin or with OpenGL too...
Updated by gnick79 over 14 years ago
I'm doing massive tests.
You should try to exit emu, reopen it and play a game.. crash is probably.
game tester = SMS :°D poor.. a lot of crashes ahahahah and ....
... surprise surprise :O with r6836 and r6837 neither only one crash.
r6837 has both FIFO and SSE changes, then my last conclusion was wrong.
Updated by hatarumoroboshi over 14 years ago
After lots of DX9 plugin testing I can say that:
r6802 introduced random in-game crashes with DX9 plugin probably fixed by r6809 (still testing this revision), BUT THEN r6811 introduced random crashes at start (from r6811 on if I try to start Wii Play with DX9 plugin 1 time on 15 it crashes)...so now I don't know how these two problems are related, in fact I've always thought that the "crash fix" of r6812-r6813 didn't fix the problem, but isn't it possible that now the in-game random crashing depends only on r6811?
For me now is important to test accurately r6809 and r6811...if r6809 has no-crashes at all and r6811 crashes even in-game, then the problem depends only on r6811; if 6809 doesn't crash and r6811 crash only at start, then the problem is r6811 AND r6812-6813.
Updated by hatarumoroboshi over 14 years ago
I meant Wii Sports, not Wii Play...and this is a photo of a random crash at start (as I said it happens usually one time in a row of 15 tries) introduced by r6811...
Updated by Xsacha over 14 years ago
r6811 has a known crash due to buffer overflow in titlename. Soren fixed this in r6818 (actually the one before it but it introduced an issue fixed in r6818).
Do you guys have crashes in r6818? If so, it is related to SSE changes in r6812 or r6813
Updated by gnick79 over 14 years ago
in r6811 char array size causes crash in DX11 only... array too short. However this was fixed in r6817. Did you have crash experience even in DX9?
Updated by gnick79 over 14 years ago
You guys try new r6896 for progress or not.
Anyway It seems we have more crash variants.
I've seen different behaviors between x86 and x64 versions.
Goodnight, see you tomorrow.
Updated by bgtsetso over 14 years ago
the problem with RE1 and Mario Kart continues....
Updated by pepodmc over 14 years ago
sometimes there was this problem with sonic colours and sonic and the black knight but perhaps, they are fixed know, someone need to try
Updated by hatarumoroboshi over 14 years ago
The Soren fix in r6817-r6818 was for Dx11 not 9...now I'm gonna try r6896 to see first of all if the crashes at start are fixed (easier to find out), then checking if in-game crashes are also fixed too (much more time needed).
Updated by hatarumoroboshi over 14 years ago
With r6897 still DX9 random crashing at start...
Updated by gnick79 over 14 years ago
Uhm. I'm not sure if this change fix something.... you can test it.
as I said before there is more than 1 crash pattern, and this commit doesn't fix for sure my SMS crash at the starting game. This happens 1/2 times and with x64 builds... and this crash pattern has been introduced from last changes by rodolfo.
Updated by gnick79 over 14 years ago
hataru download r6875 and try with this revision. Check is you have, at least, a crash reduction...
Updated by hatarumoroboshi over 14 years ago
r6875 crashes randomly at start as well...
Updated by gnick79 over 14 years ago
@hataru
ok your crashes are different of mine...
check r6810... in this revision you should not have crash anymore.
Instead with r6815/r6818 crash return again....
If yes, we can isolate a group of crash patterns around last SSE changes.
Updated by hatarumoroboshi over 14 years ago
r6810 is definitely the last revision without DX9 crashes at start in Wii Sports (tried more than 300 times in a row)...from r6811 included every revision has this problem (wich I still don't know if it is in some way correlated with the DX9 crashes in-game)
What does r6811 exactly do? Is something important? Isn't it possible to try to completely revert it or would it be a problem?
Updated by gnick79 over 14 years ago
changes related to DX9 in r6811 were reverted in r6896... but you have still crashes, then Xsacha solution against previous james.jdunne's fix should be the problem. Let's ask to him if he has a solution instead to try reverting that revision.
Updated by gnick79 over 14 years ago
other crash pattern:
[cut]
diff between r6874 and r6876 (http://code.google.com/p/dolphin-emu/source/diff?spec=svn6876&r=6876&format=side&path=/trunk/Source/Core/VideoCommon/Src/VertexLoader_Normal.cpp)
Why 64bit builds with r6876 crash always and with r6874 never?
What does the compiler?
[/cut]
tester: Super Mario Sunshine
builds affected: all x64 revisions above r6875 (none/rare for 32bits version)
frequency: almost always.
Since r6876 changes obviously are evolved, but crashes persist.
Let's see if rodolfo has some idea about this strange case.
Updated by hatarumoroboshi over 14 years ago
Looking at the differences it seems to me that r6811 was only partially reverted...anyway I doubt that r6812 and r6813 could introduce crash at start since this kind of issue was not present even in r6802 (it is possible, but how many possibilities are there that a successive different revision reproduces perfectly the same issue?)
What is certain for me at the moment is that r6802 brought random crashes in-game and r6811 random crashes at start...as I said before it would be great to find out certanly if r6809 and 6811 do not have crash in game, because in that case r6811 would be still responsable of crash at start and r6812-r6813 of crash in game...
Updated by gnick79 over 14 years ago
@hataru
try this unofficial revision (6816 less 6811 changes) http://www.mediafire.com/?dngji2u2ps6rqkn
Updated by hatarumoroboshi over 14 years ago
Yep, seems I was right about it, with this unofficial release I have no more random crashes at start (tried over 350 times)...now going to bed, tomorrow I'm gonna try to test if random crashes in-game are fixed as well...
Updated by Xsacha over 14 years ago
hatarumo: Surely the latest rev doesn't include the bad changes from r6811? So you shouldn't experience crash at start in latest rev either. You are using DX9 or DX11?
I'm actually not experiencing any crash whatsoever here in 64-bit or 32-bit. Tried 3 different machines. Restarted games many times, played for numerous hours with and without wiimotes.
r6811 of course crashes at start but it was fixed in later rev.
Updated by gnick79 over 14 years ago
r6901 should be fix a couple of problems in DX9 I think in equivalent way to unofficial revision.... try it please
Updated by hatarumoroboshi over 14 years ago
Yes, r6901 fixes the DX9 random crashing at start for me (tried with r6906)...now let's see if the in-game crashes are fixed as well...
Updated by gnick79 over 14 years ago
please try if you merged issue persists or it's fixed as well...
Updated by bgtsetso over 14 years ago
@ gnick79 - ok I'll do. btw shouldn't you revert r6896 since this was not the problem source?
Updated by bgtsetso over 14 years ago
I tested r6907 and the crashes during startup of RE1 and Mario Kart seems to be fixed :) Still I think that 6896 should be reverted in this case
Updated by skidau over 14 years ago
r6896 does not hurt and can be kept. Can this issue be closed or are there still random crashes occurring only with dx9 (and not ogl nor dx11)?
Updated by hatarumoroboshi over 14 years ago
Still testing, but for now no DX9 crashes at all (at start or in-game)...so for me it can be closed.
Eventually if I'll experience again in-game crashes in the next days I will ask to re-open it...
Thank you all.