06870: mario and clones: Missing music/sound effects

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
06870	Sound	Major	Always	Feb 8, 2018, 09:45	Feb 13, 2018, 03:30

Tester	kunikuni	View Status	Public	Platform	MAME (Official Binary)
Assigned To	AJR	Resolution	Fixed	OS	Windows Vista/7/8 (64-bit)
Status [?]	Resolved			Driver	nintendo/mario.cpp
Version	0.194	Fixed in Version	0.195	Build	64-bit
		Fixed in Git Commit		Github Pull Request #

Summary	06870: mario and clones: Missing music/sound effects
Description	Missing music/sound effects in Mario Bros and clones
Steps To Reproduce
Additional Information
Github Commit

Flags
Regression Version	0.194
Affected Sets / Systems	mario and clones

Attached Files

No.14735 Tafoid Administrator Feb 8, 2018, 11:20	Appears due to MCS48 changes early in the day on January 23, 2018.
No.14742 AJR Developer Feb 13, 2018, 03:30	The regression undoubtedly occurred with a98f32babe5fb37d41aca5527e4329f2d28e9360, which made the MCS-48 program counter wrap at 11 bits (as consistently documented). Before that commit, the mario sound program could execute the entire 2048 bytes of the "internal ROM" before getting to the beginning of MB1 where it needs to be. However, the "internal ROM" in all sets (excluding a few heavily modified bootlegs) is just zero-filled by the loading routine, meaning all NOPs. The actual hardware almost certainly couldn't work that way. (The schematics might well be wrong, though.) The workaround I've devised is to hack SEL MB1 followed by JMP $000 into the internal ROM area. This gets the program counter to the beginning of MB1, but uses far fewer instructions to do so.

No.14735

Tafoid

Administrator

Feb 8, 2018, 11:20

Appears due to MCS48 changes early in the day on January 23, 2018.

No.14742

AJR

Developer

Feb 13, 2018, 03:30

The regression undoubtedly occurred with a98f32babe5fb37d41aca5527e4329f2d28e9360, which made the MCS-48 program counter wrap at 11 bits (as consistently documented). Before that commit, the mario sound program could execute the entire 2048 bytes of the "internal ROM" before getting to the beginning of MB1 where it needs to be. However, the "internal ROM" in all sets (excluding a few heavily modified bootlegs) is just zero-filled by the loading routine, meaning all NOPs. The actual hardware almost certainly couldn't work that way. (The schematics might well be wrong, though.)

The workaround I've devised is to hack SEL MB1 followed by JMP $000 into the internal ROM area. This gets the program counter to the beginning of MB1, but uses far fewer instructions to do so.