stingray being busy - Page 2

extralife · 22 September 2008, 16:16

Hey Sting, I am patient, just interested thats all m8

extralife · 30 December 2008, 14:04

Just thought I would bumb my own thread to see, if u are still up for this challenge stingray

maybe this one fix is a bit more of a challenge than was expected

Anyways hope u haven't giving up

happy new years amigasceners

StingRay · 07 January 2013, 14:34

Finally I have good news, I have found and fixed the problem with the glenz vector part, it works fine on my 060 now. A bug in the the demo I patched yesterday gave me the much needed clue as to what the problem could be. Hardwired fully working WHDLoad patch should be available in the near future.

BippyM · 07 January 2013, 15:23

Out of curiosity what was the problem??

skan · 07 January 2013, 19:33

Quote:

Originally Posted by StingRay

Finally I have good news, I have found and fixed the problem with the glenz vector part, it works fine on my 060 now. A bug in the the demo I patched yesterday gave me the much needed clue as to what the problem could be. Hardwired fully working WHDLoad patch should be available in the near future.

Raytraced or it didn't happen!

pmc · 07 January 2013, 20:15

Quote:

Originally Posted by skan

Raytraced or it didn't happen!

StingRay · 07 January 2013, 21:19

Quote:

Originally Posted by bippym

Out of curiosity what was the problem??

A wrong instruction to enable blitting with the copper (bset works with bytes only when the destination is a memory location and besides reads the memory before the bit is set which is also not exactly what you want when dealing with a write only register). No problem on OCS/68000, it apparently worked by accident on such configs but on AGA/680x0 it didn't.

Quote:

Originally Posted by skan

Raytraced or it didn't happen!

Toni Wilen · 08 January 2013, 08:51

Bset and friends are interesting, these kinds of demo bugs also caused problems in emulation.

They work differently for some reason, when it does read-access, write-only custom register see zeros-only write = nothing happens when destination is DMACON/INTENA etc. Normally CPU read-access to write-only register writes whatever CPU's previous fetch was. (and messes up contents of above registers)

I am quite sure it is 68000-only "feature" (possibly 68010 too) and most likely related to bit number calculation which is unique to bset and friends.

StingRay · 08 January 2013, 16:21

What's also interesting is that on 68000/OCS COPCON ($dff02e) apparently allows to write to the blitter registers as soon as it's not zero anymore, i.e. it's not necessary to explicitly set bit #1 (CDANG), any will do (Hardwired bug, bset #1,$dff02e). On AGA/680x0 this doesn't work and seems to completely freeze the copper.

Toni Wilen · 08 January 2013, 16:32

It needs bit #1, I have confirmed it

bset #1,addr works because custom registers always see word access and when writing, original byte is in both low and high byte of word (= $0202).

I think data lines 0-7 and 8-15 having same value when writing byte is another 68000/010 only feature.

StingRay · 08 January 2013, 16:44

Thanks, my other guess was that it was because bset reads the memory before setting the bit.

In any case, this took me way too much time to find even though it is rather obvious.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Animated Busy Pointer Image	Leandro Jardim	request.Other	1	23 May 2012 02:41
Wow emuunlim have been busy	Paul	Amiga websites reviews	1	05 January 2002 02:27
Server busy problem	Paul	project.EAB	3	03 January 2002 17:03

22 September 2008, 16:16	#21
extralife Registered User Join Date: Jan 2008 Location: denmark Posts: 268	Hey Sting, I am patient, just interested thats all m8

30 December 2008, 14:04	#22
extralife Registered User Join Date: Jan 2008 Location: denmark Posts: 268	Just thought I would bumb my own thread to see, if u are still up for this challenge stingray maybe this one fix is a bit more of a challenge than was expected Anyways hope u haven't giving up happy new years amigasceners

07 January 2013, 14:34	#23
StingRay move.l #$c0ff33,throat Join Date: Dec 2005 Location: Berlin/Joymoney Posts: 6,863	Finally I have good news, I have found and fixed the problem with the glenz vector part, it works fine on my 060 now. A bug in the the demo I patched yesterday gave me the much needed clue as to what the problem could be. Hardwired fully working WHDLoad patch should be available in the near future.

07 January 2013, 15:23	#24
BippyM Global Moderator Join Date: Nov 2001 Location: Derby, UK Age: 48 Posts: 9,355	Out of curiosity what was the problem??

08 January 2013, 08:51	#28
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,502	Bset and friends are interesting, these kinds of demo bugs also caused problems in emulation. They work differently for some reason, when it does read-access, write-only custom register see zeros-only write = nothing happens when destination is DMACON/INTENA etc. Normally CPU read-access to write-only register writes whatever CPU's previous fetch was. (and messes up contents of above registers) I am quite sure it is 68000-only "feature" (possibly 68010 too) and most likely related to bit number calculation which is unique to bset and friends.

08 January 2013, 16:21	#29
StingRay move.l #$c0ff33,throat Join Date: Dec 2005 Location: Berlin/Joymoney Posts: 6,863	What's also interesting is that on 68000/OCS COPCON ($dff02e) apparently allows to write to the blitter registers as soon as it's not zero anymore, i.e. it's not necessary to explicitly set bit #1 (CDANG), any will do (Hardwired bug, bset #1,$dff02e). On AGA/680x0 this doesn't work and seems to completely freeze the copper.

08 January 2013, 16:32	#30
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,502	It needs bit #1, I have confirmed it bset #1,addr works because custom registers always see word access and when writing, original byte is in both low and high byte of word (= $0202). I think data lines 0-7 and 8-15 having same value when writing byte is another 68000/010 only feature.

08 January 2013, 16:44	#31
StingRay move.l #$c0ff33,throat Join Date: Dec 2005 Location: Berlin/Joymoney Posts: 6,863	Thanks, my other guess was that it was because bset reads the memory before setting the bit. In any case, this took me way too much time to find even though it is rather obvious.

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