68030 vs 68020 line-f behaviour

jotd · 15 July 2017, 15:53

Hi,

Trying to fix a 030-specific bug report on a whdload slave, I found out that the game uses LINEF emulator vector

When 68020, 2+A7 contains the address of the LINE-F instruction
When 68030, 2+A7 contains the address of the LINE-F instruction + 4 !!!!

(tested on WinUAE)

Anyone can comment on that difference? Where is that behaviour documented? (I suppose that it wasn't coded like that in WinUAE just for fun)

Toni Wilen · 15 July 2017, 16:02

Which opcode word?

(Either it is a bug, or MMU instruction which was "executed" normally and following F-line caused the exception)

jotd · 15 July 2017, 17:53

F005 (followed by 43FA 0A1C)

So on 68020, taking the "code" of the line-F gives "5", whereas on 68030 it takes 0A1C (and "skips" the LEA opcode)

could be a WinUAE-only bug: http://mantis.whdload.de/view.php?id=3171

Toni Wilen · 15 July 2017, 20:10

It is PMOVE but it does not appear to be valid PMOVE. It probably causes F-line on real 68030.

68030 MMU instructions does not do much validation, needs to be fixed..

jotd · 15 July 2017, 20:23

ok thanks for confirming.

Toni Wilen · 15 July 2017, 20:35

It is not yet 100% confirmed, that needs 1-2 days more

Toni Wilen · 16 July 2017, 19:25

Confirmed. Game uses MMU instructions that have invalid mode/reg pair and also other invalid bits set that 68030 won't accept and will generate F-line.

I tested all real 68030 MMU instruction bit combinations. Emulation should be correct now.

(68040+ don't have this problem, they don't have multiple different modes and "unused" opcode bits)

EDIT: Thanks for reporting it instead of quietly adding some workaround!

jotd · 16 July 2017, 20:29

You're welcome Toni! (I doubt I could've workarounded this anyway)

15 July 2017, 15:53	#1
jotd This cat is no more Join Date: Dec 2004 Location: FRANCE Age: 52 Posts: 8,303	68030 vs 68020 line-f behaviour Hi, Trying to fix a 030-specific bug report on a whdload slave, I found out that the game uses LINEF emulator vector When 68020, 2+A7 contains the address of the LINE-F instruction When 68030, 2+A7 contains the address of the LINE-F instruction + 4 !!!! (tested on WinUAE) Anyone can comment on that difference? Where is that behaviour documented? (I suppose that it wasn't coded like that in WinUAE just for fun)

16 July 2017, 19:25	#7
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	Confirmed. Game uses MMU instructions that have invalid mode/reg pair and also other invalid bits set that 68030 won't accept and will generate F-line. I tested all real 68030 MMU instruction bit combinations. Emulation should be correct now. (68040+ don't have this problem, they don't have multiple different modes and "unused" opcode bits) EDIT: Thanks for reporting it instead of quietly adding some workaround! Last edited by Toni Wilen; 16 July 2017 at 19:52.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Understanding the (in)famous copper's lost cycle (PAL all line, NTSC short line).	ross	Coders. Asm / Hardware	10	13 March 2017 17:50
Overclocking 68020?	Marchie	support.Hardware	8	11 October 2016 13:33
68020 33 MHz	Leandro Jardim	support.WinUAE	2	02 January 2012 19:21
Questions about 68020 CE	Maren	support.WinUAE	11	09 December 2009 21:01
Apollo 1220 - instructions and 68020 to 68030 upgrade	fc.studio	support.Hardware	11	10 January 2008 20:30

15 July 2017, 16:02	#2
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	Which opcode word? (Either it is a bug, or MMU instruction which was "executed" normally and following F-line caused the exception)

15 July 2017, 17:53	#3
jotd This cat is no more Join Date: Dec 2004 Location: FRANCE Age: 52 Posts: 8,303	F005 (followed by 43FA 0A1C) So on 68020, taking the "code" of the line-F gives "5", whereas on 68030 it takes 0A1C (and "skips" the LEA opcode) could be a WinUAE-only bug: http://mantis.whdload.de/view.php?id=3171

15 July 2017, 20:10	#4
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	It is PMOVE but it does not appear to be valid PMOVE. It probably causes F-line on real 68030. 68030 MMU instructions does not do much validation, needs to be fixed..

15 July 2017, 20:23	#5
jotd This cat is no more Join Date: Dec 2004 Location: FRANCE Age: 52 Posts: 8,303	ok thanks for confirming.

15 July 2017, 20:35	#6
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	It is not yet 100% confirmed, that needs 1-2 days more

16 July 2017, 20:29	#8
jotd This cat is no more Join Date: Dec 2004 Location: FRANCE Age: 52 Posts: 8,303	You're welcome Toni! (I doubt I could've workarounded this anyway)

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)