06615: scud: Scud Racer regression, now just a black screen on launch.

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
06615	Graphics	Major	Always	Jun 11, 2017, 23:42	Dec 15, 2018, 18:00

Tester	john_iv	View Status	Public	Platform	MAME (Self-compiled)
Assigned To		Resolution	Open	OS	Windows 10 (64-bit)
Status [?]	Confirmed			Driver	sega/model3.cpp
Version	0.186	Fixed in Version		Build	64-bit
		Fixed in Git Commit		Github Pull Request #

Summary	06615: scud: Scud Racer regression, now just a black screen on launch.
Description	As before, I use the 'scud' set to benchmark. Phil fixed it a few revs. back when it was crashing. Now however it just displays a black screen when attempting to play (fwiw, the bench results are off too if run as 'mame64.exe scud -bench 90'). http://mametesters.org/view.php?id=6492
Steps To Reproduce	1. mame64 scud Expected: Launch scud and display demo. Actual: Black screen.
Additional Information
Github Commit

Flags
Regression Version	0.185
Affected Sets / Systems	scud

Attached Files

No.13983 Phil Bennett Developer Jul 17, 2017, 00:50	PPC execution is stuck on this instruction: FFF00174 mftb r5,tbl This issue is a side-effect of this change: "Convert Sega Model 1 Sound Board and Z80-based Digital Sound Board to use proper UART-based serial interfaces" https://github.com/mamedev/mame/commit/8275551541111567becbee88ff485c1314c85916 The UART introduced a high-frequency clock (500KHz), which has caused the scheduling quanta to be greatly reduced. The mftb instruction requires 100 cycles to execute but due to the way it is handled in the PPC DRC, there are never enough cycles in a scheduling period for it to execute and thus advance to the next instruction. Moving generate_update_cycles() to the end of the instruction sequence of mftb resolves the issue but I'm not familiar enough with PowerPC or MAME's DRC to be confident that this is the correct fix.
No.13984 Phil Bennett Developer Jul 17, 2017, 01:12	Yeah, that's not correct. Perhaps all that is required is to change the allow_exception parameter of the early call to generate_update_cycles() to false to allow the instruction to run regardless of the cycles remaining?
No.14444 Osso Moderator Nov 19, 2017, 08:30	With today's updates it gets to car and course selection, but then it hangs when it should go to the start line.
No.14939 Robbbert Senior Tester Apr 10, 2018, 07:07	Currently it shows the settings, then the message "CANCELLED NETWORK BOARD NOT PRESENT", then it goes back to the settings screen in a loop.
No.15274 Hydreigon Tester Jul 14, 2018, 18:05	Go into test and change the link ID to single. Preliminary comm board emulation has been done so there is no "Cancelled" message and instead goes into the usual network check. Other than that, the regression is still the same.
No.15914 Hydreigon Tester Dec 15, 2018, 18:00 edited on: Dec 15, 2018, 18:01	An commit to attempt to fix this bug, https://git.redump.net/mame/commit/src/devices/cpu/powerpc/ppcdrc.cpp?id=0f1d2011fe284045627647fdb201107dd5d6bce0 , broke another PPC related game, Fighting Bujutsu. See here: https://mametesters.org/view.php?id=7174 Currently, Scud Race runs somewhat fine in its attract mode but hangs right before going ingame (black screen or lock at the "Rolling Start" sequence).