07026: vr: vr gameplay calculations and physics issues

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
07026	Gameplay	Major	Always	Jun 19, 2018, 13:34	Oct 27, 2019, 18:30

Tester	wuemura	View Status	Public	Platform	MAME (Official Binary)
Assigned To	galibert	Resolution	Fixed	OS	Windows 10 (64-bit)
Status [?]	Resolved			Driver	model1.cpp
Version	0.198	Fixed in Version	0.215	Build	64-bit
		Fixed in Git Commit		Github Pull Request #

Summary	07026: vr: vr gameplay calculations and physics issues
Description	For some reason the computer is doing a lot of mistakes during attract mode and gameplay in comparison with MAME 0.196: - In the first minutes of attract mode, the car get out the racing track at the first corner (verified with original) - During attract mode, It has trouble to pass other cars (verified with original) - Traffic cars at the wrong position on the racing track (verified with original) - Near the end of the attract mode, the arcade show that the main racing car pass the second check point of the track, the emulation it doesn't even get close (MAME 0.198). MAME 0.196 near the second check point two white racing cars get in the way. - During gameplay at the "Big Forest" track, near the corners if there are two or more computer cars, they always crash at it. - The racing track got very unstable, it shakes too much in 0.198, 0.196 is much more stable.
Steps To Reproduce	Watch attract mode or play the game at the Big Forest track.
Additional Information	Related? http://mametesters.org/view.php?id=7025
Github Commit

Flags	Verified with Original
Regression Version	0.197
Affected Sets / Systems	vr

Attached Files	vr-side-side-01.png (428,378 bytes) Jun 19, 2018, 13:34 Uploaded by wuemura

No.15182 wuemura Viewer Jun 19, 2018, 13:52 edited on: Jun 19, 2018, 15:05	Arcade and MAME 0.198 Traffic cars always crash at the the corner and get in the player way, notice the crash and tire screeching sound. The traffic cars don't crash at the corner in MAME 0.196
No.15187 NekoEd Senior Tester Jun 20, 2018, 05:55	I think you've covered this pretty well. Given it locks up at trying to display polygons for me, something is definitely wrong.
No.15199 Haze Senior Tester Jun 21, 2018, 13:03	fwiw I've had Virtua Fighter 2 fail to display polygons once or twice too when just left in attract after startup, so might be something uninitialized in the TGP core as that is a random behavior and probably not related to this bug. the car movement issues are clear to see from the moment the player car incorrectly cuts across the grass at the 10 second mark in the video anyway. maybe it would possible to add some logging code to intercept TGP commands and show their results, to find out which ones are giving significantly different values with the new core vs old core? I imagine OG had a plan for it anyway tho.
No.15203 NekoEd Senior Tester Jun 22, 2018, 06:21	The no polygons turned out to be outdated TGP ROMs, so my error there, sorry.
No.15206 Fujix Administrator Jun 22, 2018, 10:12 edited on: Jun 22, 2018, 10:16	You can LIKE a note now! (testing yet) It shows only the number of likes, does't reveal who liked it. It is not displayed even for admins. Enjoy liking!
No.16101 Hydreigon Tester Feb 3, 2019, 17:33 edited on: Feb 3, 2019, 17:50	I'm going to return to this bug and confirm that this is related to 06861 but not a duplicate. This behavior currently happens despite the updated TGP roms but only in Virtua Racing and Formula. I'm starting to think this is a TGP/CPU issue causing both a bad camera tracking (06861) and the car accelerating faster than usual in attract. Interestingly when running in a linked setup, of two players at least, the slave machine won't have that TGP issue as much compared to the master machine in attract mode. Idk what would happen with more than two players.
No.17018 Tafoid Administrator Oct 3, 2019, 22:33	Your reporting about issues related to timing in networked MAME builds should not have been piggy backed onto this bug report. I'm tempted to delete them as they confuse the overall issue first posted with no comparisons done with other versions or arcade behavior - it makes it a difficult sell.
No.17028 Hydreigon Tester Oct 5, 2019, 00:29	You can delete both images if you want. And no, it's not a timing issue that the car accelerates. Both instances of MAME itself synchronize fine. Decided to test a 4p setup for curiosity's sake and the results were pretty much the same. Gaibert recently did a commit that improved the calculation issues but it isn't fully fixed yet: https://git.redump.net/mame/commit/?id=b51ae75c3ae7b5bb297b27a31b00de603b14220f
No.17137 hydef Tester Oct 27, 2019, 18:30	I wanted to register to thank for the work put into fixing this. When I first played it and did a couple of games it was great, I thought it was completely fixed. The issue I had was at usual point on the track (the turn before the final turn on Big Forest), something would slow your car for no reason. It happens far less frequently now, though. And it's impossible to reproduce every time. And once earlier I was driving along the bridge on the first track and something briefly slowed me like as if I'd driven off track.