01650: ridgerac: Unstable Freezes that ends with Crash and needs totally reboot under XP

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
01650	Crash/Freeze	Critical (emulation)	Always	Mar 27, 2008, 13:22	Apr 3, 2008, 23:09

Tester	Mock	View Status	Public	Platform	MAME (Official Binary)
Assigned To	aaron	Resolution	Fixed	OS	Windows XP/Vista 32-bit
Status [?]	Resolved			Driver	namco/namcos22.cpp
Version	0.124	Fixed in Version	0.124u1	Build	Normal
		Fixed in Git Commit		Github Pull Request #

Summary	01650: ridgerac: Unstable Freezes that ends with Crash and needs totally reboot under XP
Description	When ridge racer start or other Namcos22, it runs 90 FPS or drops to 1 FPS where it freezes random sometime the freeze last longer. first time its possible to exit without problem, next time it freezes the GFX driver so i have to force shutdown mame in Vista, XP needs a complete restart. i found out that mame 1.20 doesnt have this problem and run at a constant framerate. I also tried disable cores and tried diffrent mame option, with disabling multithread. I also tried disabling crossfire so that i appears as a normal Videocard. the DOS,UI and other build has the same problem. CPU utilization will max out when the picture freezes, where other games runs with variable usages of the cores.
Steps To Reproduce	just start mame with default parameters from dos or GUI.
Additional Information	System tested that produce the problem Intel QX9650 3.0ghz 12mb cache ATI 3870X2 CrossfireX 2GB X-Fi soundcard Asus Formula SE 2X 1GB Corsair 8500C5
Github Commit

Flags
Regression Version
Affected Sets / Systems	ridgerac

Attached Files

No.00431 Tafoid Administrator Mar 27, 2008, 13:51 edited on: Mar 27, 2008, 13:52	Not sure what the issue is. I just tried on my setup (Athlon XP1800 - Windows 2000) on 0.124 and 0.124a and although it was very slow (about 40%-45% with frameskip 10/10) I was able to finish the race on both and there were no high unexpected slowdowns or crashes. Is it possible with all that power on the system you have insufficient cooling - since this game tends to run a processor hard? I'll wait to see if others can duplicate your problem.
No.00432 Mock Tester Mar 27, 2008, 14:11	as i write, older versions of mame run good enough. I think maybe i could be a problem with Quadcore because my dualcore worked fine last i checked mame. also it could be some ATI driver issue. but again when an older version of mame run same with about 80& Framerate that it must be something changed in the newer versions. What about the things Aaron Giles talked about making some kind of SLI using dualcore tecnoligy. or other optimizations that only work with dual core but not quadcore
No.00433 Mock Tester Mar 27, 2008, 14:15	at first i also thougt about the cooling like TMprotection. but im running with all stock speed, alså tested prime95 and other stabability tests that ran fine. I can also test it on a Intel E6750 Vista if this has the same issue. As i remember i also tried running a benchmark with no video output that never finished. don´t know i that eleminates a video driver issue
No.00434 Fujix Administrator Mar 27, 2008, 15:19	It works fine for me with C2D on XP. I think the problem is too vague. You have to pinpoint the real cause of the problem if you want someone to take measures to the problem. Vista vs XP, quad core vs dual core, dual vs single videocard (not only disabling dual card by the utility), with or without -mt, combinations of other MAME configs, etc.. And to avoid confusions, the current baseline release is not a "DOS" version, but a command-line version without UI. Also the version number is not 1.24 but 0.124. Please keep in mind that all texts are subjects to be searched.
No.00442 couriersud Developer Mar 27, 2008, 23:26	Please try the following: a) Open a console window (cmd.exe) b) set OSDPROCESSORS=2 c) mame ridgerac
No.00443 Mock Tester Mar 27, 2008, 23:43	Thank you very much Courlersud. That fixed the freeze when tried your suggestion, the version i got that dont need that command line is MameFX 0.119U1. what have changed since there is need to set OSDPROCESSORS=2
No.00447 etabeta Developer Mar 28, 2008, 07:51	so in the end it was a quadcore problem. notice that OSDPROCESSOR has been added (iirc) in 0.119u3, when multiple cores have been started to be exploited by MAME. you don't need it (nor it is supported) in 0.119u1 @Fujix: shall we close this, or are quadcore bugs supposed to be valid?
No.00450 Mock Tester Mar 28, 2008, 11:44	its good enough for me to close it now that i solves my crashes. And it runs as before, i tested alot of roms where no problem appeared. I also tried diffrent GUI version that also worked fine.
No.00451 Tafoid Administrator Mar 28, 2008, 13:34	I'll resolve this bug. Some notice of this parameter should probably be put in the official documents soon.
No.00495 aaron Developer Apr 1, 2008, 21:40	Re-opening. This is a legitimate bug. You should not have to force the OSDPROCESSORS environment variable to make it work. More from couriersud: Using e.g. OSDPROCESSORS=3 on a 2 processor machine will freeze sdlmame at a random time at 100% cpu usage. The freeze always is attributable to the scalable lock functions. I replaced the scalable locks with pure (non-blocking) pthread locks and sdlmame scaled up to OSDPROCESSORS=8. Setting the cpu affinity mask to bind threads to a given cpu even yields a comparable performance like the blocking scalable locks. It looks like that the locks would block as soon as the OS places two threads on the same processor. I do not have a solution for windows but just wanted to document what I found out.
No.00506 aaron Developer Apr 3, 2008, 07:48	I switched from using scalable locks to Windows CriticalSections, which should prevent the hard hangs. Give it a try.
No.00537 Mock Tester Apr 3, 2008, 23:09	I tried 0.124u1 and it works perfect and run very fast now. almost 100% in ridgeracer. very good work aaron