05025: neogeo.c: Neo Diagnostics show Z80 issue

ID	Category [?]	Severity [?]	Reproducibility	Date Submitted	Last Update
05025	Misc.	Minor	Always	Sep 27, 2012, 06:28	Mar 14, 2014, 04:38

Tester	Reyn	View Status	Public	Platform	MAME (Official Binary)
Assigned To	hap	Resolution	Fixed	OS	Windows Vista/7 (64-bit)
Status [?]	Resolved			Driver	neogeo/neogeo.cpp
Version	0.147	Fixed in Version	0.153	Build	Normal
		Fixed in Git Commit		Github Pull Request #

Summary	05025: neogeo.c: Neo Diagnostics show Z80 issue
Description	Issue is in neogeo.c, can't be selecetd from Driver menu. The Neo Diagnostics v.018b from smkdan show that there is a problem with Z80 <-> 68K communication -> See attached image. In neogeo.c sourcecode beginning from line 590 this is listed: /* this is a guess */ audio_cpu_clear_nmi_w(space, 0, 0); Maybe Neo Diagnostics can help to identify the issue.
Steps To Reproduce	Go to http://smkdan.eludevisibility.org/neo/diag/ and download Neo Diagnostics v.018b. Use set ' irrmaze ' for testing, replace ' 236-bios.sp1 ' and ' 236-m1.m1 ' with files from ' 018b.zip ' Fire up mame.
Additional Information	Also the following beep error code appears: ' Z80<->68k COMM. ERROR (DATA) 001100 '
Github Commit

Flags	Verified with Original, Noted in Source
Regression Version
Affected Sets / Systems	neogeo.c

Attached Files	neo_error.png (59,417 bytes) Sep 27, 2012, 06:28 Uploaded by Reyn
Attached Files	mt05025.diff (2,742 bytes) Nov 3, 2012, 16:59 Uploaded by Reyn File for comment 09074 [Show Content]

No.08976 hap Developer Sep 27, 2012, 14:00	Thanks. Do you happen to know if the sourcecode is available for this test suite?
No.08977 Lord Nightmare Developer Sep 27, 2012, 14:39	http://www.neo-geo.com/forums/showthread.php?236157-Neo-Diagnostics-ROMs as for source code, you'd probably need to ask smkdan for it privately
No.08991 Reyn Tester Sep 30, 2012, 07:53	Some usefull? info: http://wiki.neogeodev.org/index.php?title=68k/Z80_communication&t=20120814005148
No.09024 smkdan Tester Oct 12, 2012, 11:49	Hi, I was shown this page and thought I'd share some issues I found with MAME driver. I only needed minor edits to neogeo.c to get my debug build of MAME working as my systems do so I'll just list the changes instead of digging up the source. This is based on tests with my own MVS/AES systems. The cart S1 and M1 ROM are supposed to be enabled (or disabled) together. MAME has the enable code with the BIOS instead but games still work since the first thing the BIOS tells the Z80 to do is a soft reset anyway. There is also no Z80 reset on switch. Z80 switches are done by jumping to a loop in RAM on command by 68k so there's no need for the reset. Try watch/breakpoints on the last few bytes of Z80 RAM. The BIOS overlay is an actual swap so when enabled, reading $000000-$00007f is supposed to read the BIOS and not the game P1 area. Z80: out ($00) zeroes the input port. The pictured test just writes to port $00, then checks if $0c is actually zero to see if a certain trace is damaged. out ($08) enables NMI and out ($18) disables it. There's a comment on this in the source but it's not hooked up to anything. It's disabled on reset with real h/w. I think there's some other stuff I'm forgetting so I'll post it up later when I check my notes. I remember the sound test being mute in MAME when it's supposed to beep some FM channels.
No.09054 Reyn Tester Oct 27, 2012, 08:01	@smkdan: Could you please take a look at this: http://mametesters.org/view.php?id=5041 Does the mentioned bug also occur in your edited neogeo.c MAME debug build? Thanks
No.09074 Reyn Tester Nov 3, 2012, 16:59	Attached a small diff which fixes the following: S1 rom enabled on system start / reset ( M1 already was ) Removed Z80 reset on audio bank switch Hooked up NMI $18 -> Disabled on machine reset (is this done right?) Someone needs to figure out how to hook up NMI $08; the current function used is 'audio_cpu_assert_nmi' -> line 354 NMI $08 can be found at 'Audio CPU port handlers' Can someone please check this -> See attached MT05025.diff Thx
No.10359 hap Developer Mar 13, 2014, 23:29	Z80 comms should be ok now if you boot the program with button D held down. There's a remaining problem though: if i boot with A+B+C+D pressed, it will run the tests instead of bringing up a menu and will specify that the Z80 failed
No.10360 AWJ Developer Mar 14, 2014, 04:38 edited on: Mar 14, 2014, 04:40	If you use irrmaze as the "donor set" to run the diagnostic ROM in MAME, then you can't press A+B+C+D because those buttons don't exist; that game has a trackball where P1's joystick and buttons are on a regular neogeo. For testing purposes, you have to modify the GAME line for irrmaze in neogeo.inc to make it use "neogeo" instead of "irrmaze" for inputs. If you modify the irrmaze inputs, run the diagnostic ROM and hold D on startup, you get "SM1 slot switch ignored". This is as expected, since irrmaze is a modified MV1B board, which has no sm1 ("audiobios") ROM. If you hold B and D on startup like the instructions for the diagnostic ROM say to for MV1B/MV1C, you get "all tests passed". However, if you soft reset (press and hold 1+3 or just press F3) and run the test again, then you get "YM2610 unexpected IRQ" on the second and all subsequent runs. Interestingly, if you modify neogeo.inc and add sm1.sm1 to the irrmaze set, or if you use a different game as the "donor", then you can run the Z80 test without holding B (only holding D) and it'll pass every time. So I wonder if the "YM2610 unexpected IRQ" is actually a remaining bug in our emulation, or if it's a bug in the diagnostic ROM in the codepath it executes for MV1B/MV1C... Anyway, I'm marking this resolved because the NMI issue is definitely fixed now. We can open a new bug if smkdan proves that the failure when re-running the test on SM1-less boards isn't a bug in his code.