Debian AGA console display issues

mark_k · 22 February 2014, 19:47

Quick time test booting Debian HDF; WinUAE 2.7.0 223 secs vs 2.7.1b8 179 secs. Not bad, about 20% faster.

There are a couple of problems with the Debian AGA display (both console and X):

- Since WinUAE 2.4.0 the rightmost ~24 pixels are blanked, maybe fewer in later versions (~20???). With WinUAE 2.3.3 only about 3-4 pixels are blanked. See screenshot, rightmost pixels of letter d and all of e and f are blanked. I mentioned this before, but thought I'd re-test with the latest 2.7.1 beta (no change).

- During the boot sequence (using the Debian HDF I mentioned here, type boot-hda1 at the prompt with A4000 config) the screen blanks and reappears several times. I assume that's normal, it happens with ECS too. However with AGA when the screen reappears it's squashed/distorted, see pic. Happens with 2.3.3, 2.4.0, 2.6.0 and later (2.5.x just gives an always-blank screen). With 2.3.3 and 2.4.0 the distorted display looks a bit different. Log output near that time:

Code:

BEAMCON0=5BB3 VTOTAL=01C0  HTOTAL=0071
  HSSTOP=001A HBSTRT=0001  HBSTOP=0023
  VSSTOP=0004 VBSTRT=01C0  VBSTOP=0027
  HSSTRT=000C VSSTRT=0001 HCENTER=0045
  HSYNCSTART=00E3 HSYNCEND=0041
BEAMCON0=5BB3 VTOTAL=01C0  HTOTAL=0071
  HSSTOP=001A HBSTRT=0001  HBSTOP=0023
  VSSTOP=0004 VBSTRT=01C0  VBSTOP=0030
  HSSTRT=000C VSSTRT=0001 HCENTER=0045
  HSYNCSTART=00E3 HSYNCEND=0041
BEAMCON0=5BB3 VTOTAL=01C0  HTOTAL=0071
  HSSTOP=001A HBSTRT=0001  HBSTOP=0023
  VSSTOP=0004 VBSTRT=01C0  VBSTOP=0030
  HSSTRT=000C VSSTRT=000C HCENTER=0045
  HSYNCSTART=00E3 HSYNCEND=0041
BEAMCON0=5BB3 VTOTAL=01C0  HTOTAL=0071
  HSSTOP=001A HBSTRT=0001  HBSTOP=0023
  VSSTOP=000E VBSTRT=01C0  VBSTOP=0030
  HSSTRT=000C VSSTRT=000C HCENTER=0045
  HSYNCSTART=00E3 HSYNCEND=0041
warning: 5BB6 written to BEAMCON0 PC=000EF14A

If I open the settings window and change the line mode or interlaced line mode settings, the display corrects itself.

Toni Wilen · 23 February 2014, 10:30

I found one more bug in programmed mode horizontal pulse to display positioning calculations. Should be correct now, without breaking other special monitor modes..

mark_k · 23 February 2014, 11:34

That mostly fixed the right side blanking, however there are still a few (2 or 3?) pixels missing.

The display reappearing distorted after blanking problem is still there. My guess is, WinUAE detects a programmed mode change part-way through Debian writing the registers, so it picks up invalid/bogus settings. And for some reason WinUAE doesn't update the display when the last programmed-mode register is written. And manually changing line mode causes WinUAE to re-evaluate the programmed mode registers which fixes the display.

Toni Wilen · 23 February 2014, 12:03

I don't get any flickering or corruption.

Missing pixels are probably bug in console display mode setup and it may not work with all monitors (without also having clipped pixels). Not going to bother with it.

EDIT: programmed mode recalculation may now detect changes that it missed previously. Still can't duplicate any glitches (Different config?)

mark_k · 23 February 2014, 14:24

Did you change something about IDE emulation in the latest winuae.exe (dated 2014-02-23 11:49)? Booting my Debian HDF there are now messages like:
hda: irq timeout: status=0xd0 { Busy }

ide0: reset: success
and this in the log:
IDE0 initialize drive parameters, CYL=31001,SPT=63,HEAD=4
IDE0 recalibrate

Anyway, with the latest winuae.exe the display distortion is fixed.

If you want to try reproducing the issue with the previous winuae.exe, use the Debian HDF I linked to above and the config attached. The display blanking and reappearing first happens near the end of the boot process. The Login: prompt is printed, then the screen blanks and reappears (distorted) several times. I think xdm (the X display manager) starting is responsible for that.

Toni Wilen · 23 February 2014, 14:38

Quote:

Originally Posted by mark_k

Did you change something about IDE emulation in the latest winuae.exe (dated 2014-02-23 11:49)? Booting my Debian HDF there are now messages like:
hda: irq timeout: status=0xd0 { Busy }

ide0: reset: success
and this in the log:
IDE0 initialize drive parameters, CYL=31001,SPT=63,HEAD=4
IDE0 recalibrate

Anyway, with the latest winuae.exe the display distortion is fixed.

If you want to try reproducing the issue with the previous winuae.exe, use the Debian HDF I linked to above and the config attached. The display blanking and reappearing first happens near the end of the boot process. The Login: prompt is printed, then the screen blanks and reappears (distorted) several times. I think xdm (the X display manager) starting is responsible for that.

I still can't see any blankings or ide problems.

But disable More compatible, it shouldn't be allowed if MMU is enabled.

(Or did you upload multiple images that have identical file name?)

mark_k · 23 February 2014, 14:48

How is the displayable area determined for programmed modes? I checked various ECS modes (the monitor drivers use slightly different settings on AGA machines, maybe something related to fetch mode?).

It seems that HTOTAL + 1 - (HBSTOP - HBSTART) gives the number of displayable colour clocks (so multiply that by 8 to get number of 35ns pixels). At least for all modes I checked except EURO36.

Calculating that for the Debian mode (HTOTAL=0071, HBSTRT=0001, HBSTOP=0023):
71 + 1 - (23 - 1) = $50. $50 × 8 = 640
(At the moment WinUAE shows 638 pixels for that mode I think.)

By "max horizontal overscan" I mean the largest width you can set using Overscan prefs.

EURO72 (max horizontal overscan 680): HTOTAL=0071, HBSTRT=0001, HBSTOP=001E
HTOTAL + 1 - (HBSTOP - HBSTART) = 71 + 1 - (1E - 1) = 72 - 1D = $55. $55 × 8 = 680

EURO36 (max horizontal overscan 720): HTOTAL=00E2, HBSTRT=0008, HBSTOP=002F
HTOTAL + 1 - (HBSTOP - HBSTART) = E2 + 1 - (2F - 8) = $BC. $BC × 4 = 752
For EURO36 the maximum selectable overscan width is 32 fewer than that.

Multiscan (max horizontal overscan 680): HTOTAL=0071, HBSTRT=0001, HBSTOP=001E
HTOTAL + 1 - (HBSTOP - HBSTART) = 71 + 1 - (1E - 1) = $55. $55 × 8 = 680

SUPER72 (max horizontal overscan 936): HTOTAL=0091, HBSTRT=0001, HBSTOP=001E
HTOTAL + 1 - (HBSTOP - HBSTART) = 91 + 1 - (1E - 1) = $75. $75 × 8 = 936

DBLNTSC and DBLPAL (max horizontal overscan 744): HTOTAL=0079, HBSTRT=0001, HBSTOP=001E
HTOTAL + 1 - (HBSTOP - HBSTART) = 79 + 1 - (1E - 1) = $5D. $5D × 8 = 744

Example of different register values used on AGA machines by the DBLNTSC monitor driver:
DBLNTSC: HTOTAL=0081, HBSTRT=0001, HBSTOP=0021
HTOTAL + 1 - (HBSTOP - HBSTART) = 81 + 1 - (21 - 1) = $62. $63 × 8 = 784
Max selectable AGA horizontal overscan is 720; 64 shres pixels (= 8 colour clocks) less. Is the difference related to 4x fetch mode and/or sprites?

mark_k · 23 February 2014, 14:54

The HDF I'm booting from is in the archive Debian_4GB_HD_CDorHDinstall_CL_installed5.tar.xz:
Edit: The old Multiupload links for that archive have expired, I'll re-upload it.
New Multiupload link

Toni Wilen · 23 February 2014, 14:59

Quote:

Originally Posted by mark_k

How is the displayable area determined for programmed modes? I checked various ECS modes (the monitor drivers use slightly different settings on AGA machines, maybe something related to fetch mode?).

It seems that HTOTAL + 1 - (HBSTOP - HBSTART) gives the number of displayable colour clocks (so multiply that by 8 to get number of 35ns pixels). At least for all modes I checked except EURO36.

It isn't that simple. It only tells maximum possible visible area.

Display can't start very early and DDFSTRT/DDFSTOP and DIWSTRT/DIWSTOP also needs to fit perfectly inside "blanking viewport" or part of display will be clipped.

For example in many modes hblank ends before bitplane fetches start (hblank end < DDFSTRT), part of left edge is filled with border color.

EDIT: winuae.zip updated, does IDE messages still appear? (I assume this was introduced in b9?)

mark_k · 23 February 2014, 15:57

Quote:

Originally Posted by Toni Wilen

EDIT: winuae.zip updated, does IDE messages still appear? (I assume this was introduced in b9?)

No, not in b9. I first noticed it in the winuae.exe you uploaded earlier today, datestamp 2014-02-23 11:49 (maybe 13:49 your local time).

The IDE messages don't appear with the 2014-02-23 13:42 (15:42 your time?) winuae.exe.

amilo3438 · 23 February 2014, 18:12

Quote:

Originally Posted by mark_k

No, not in b9. I first noticed it in the winuae.exe you uploaded earlier today, datestamp 2014-02-23 11:49 (maybe 13:49 your local time).

Hmm, seems you probably missed this:

"Betas are supposed to be the ones to test, not random test versions." (quote taken from here)

p.s. Just hope the same applies for all, right?!

By the way, thanks for the effort you have put into improving the emulator.

Toni Wilen · 23 February 2014, 18:42

You misunderstood. It means winuae.zip is only allowed to used if I say so. (something got fixed and it needs confirmation, like above). Using it because it is newer is not acceptable reason!

amilo3438 · 23 February 2014, 19:12

@Toni

Quote:

Originally Posted by mark_k

No, not in b9. I first noticed it in the winuae.exe you uploaded earlier today

The new problem introduced with test version not present in the last official beta 9 ... so I understood.

... but if you say so, then I understand it.

22 February 2014, 19:47	#1
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,336	Debian AGA console display issues Quick time test booting Debian HDF; WinUAE 2.7.0 223 secs vs 2.7.1b8 179 secs. Not bad, about 20% faster. There are a couple of problems with the Debian AGA display (both console and X): - Since WinUAE 2.4.0 the rightmost ~24 pixels are blanked, maybe fewer in later versions (~20???). With WinUAE 2.3.3 only about 3-4 pixels are blanked. See screenshot, rightmost pixels of letter d and all of e and f are blanked. I mentioned this before, but thought I'd re-test with the latest 2.7.1 beta (no change). - During the boot sequence (using the Debian HDF I mentioned here, type boot-hda1 at the prompt with A4000 config) the screen blanks and reappears several times. I assume that's normal, it happens with ECS too. However with AGA when the screen reappears it's squashed/distorted, see pic. Happens with 2.3.3, 2.4.0, 2.6.0 and later (2.5.x just gives an always-blank screen). With 2.3.3 and 2.4.0 the distorted display looks a bit different. Log output near that time: Code: BEAMCON0=5BB3 VTOTAL=01C0 HTOTAL=0071 HSSTOP=001A HBSTRT=0001 HBSTOP=0023 VSSTOP=0004 VBSTRT=01C0 VBSTOP=0027 HSSTRT=000C VSSTRT=0001 HCENTER=0045 HSYNCSTART=00E3 HSYNCEND=0041 BEAMCON0=5BB3 VTOTAL=01C0 HTOTAL=0071 HSSTOP=001A HBSTRT=0001 HBSTOP=0023 VSSTOP=0004 VBSTRT=01C0 VBSTOP=0030 HSSTRT=000C VSSTRT=0001 HCENTER=0045 HSYNCSTART=00E3 HSYNCEND=0041 BEAMCON0=5BB3 VTOTAL=01C0 HTOTAL=0071 HSSTOP=001A HBSTRT=0001 HBSTOP=0023 VSSTOP=0004 VBSTRT=01C0 VBSTOP=0030 HSSTRT=000C VSSTRT=000C HCENTER=0045 HSYNCSTART=00E3 HSYNCEND=0041 BEAMCON0=5BB3 VTOTAL=01C0 HTOTAL=0071 HSSTOP=001A HBSTRT=0001 HBSTOP=0023 VSSTOP=000E VBSTRT=01C0 VBSTOP=0030 HSSTRT=000C VSSTRT=000C HCENTER=0045 HSYNCSTART=00E3 HSYNCEND=0041 warning: 5BB6 written to BEAMCON0 PC=000EF14A If I open the settings window and change the line mode or interlaced line mode settings, the display corrects itself. Attached Thumbnails

23 February 2014, 11:34	#3
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,336	That mostly fixed the right side blanking, however there are still a few (2 or 3?) pixels missing. The display reappearing distorted after blanking problem is still there. My guess is, WinUAE detects a programmed mode change part-way through Debian writing the registers, so it picks up invalid/bogus settings. And for some reason WinUAE doesn't update the display when the last programmed-mode register is written. And manually changing line mode causes WinUAE to re-evaluate the programmed mode registers which fixes the display. Attached Thumbnails

23 February 2014, 12:03	#4
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,506	I don't get any flickering or corruption. Missing pixels are probably bug in console display mode setup and it may not work with all monitors (without also having clipped pixels). Not going to bother with it. EDIT: programmed mode recalculation may now detect changes that it missed previously. Still can't duplicate any glitches (Different config?) Last edited by Toni Wilen; 23 February 2014 at 12:51.

23 February 2014, 14:54	#8
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,336	The HDF I'm booting from is in the archive Debian_4GB_HD_CDorHDinstall_CL_installed5.tar.xz: Edit: The old Multiupload links for that archive have expired, I'll re-upload it. New Multiupload link Last edited by mark_k; 23 February 2014 at 15:20.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
How to do AGA copperbars in Blitz using display library	Coagulus	Coders. Blitz Basic	9	25 January 2016 15:53
Display issues with 2.3.3	squirminator2k	support.WinUAE	2	28 September 2011 08:10
WinUAE display issues	Roman0	support.WinUAE	4	18 April 2011 12:57
WinUAE 1.0 display issues	_Steve_	support.WinUAE	2	11 July 2005 20:32
AGA display	Dela	support.WinUAE	4	06 October 2002 21:59

23 February 2014, 10:30	#2
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,506	I found one more bug in programmed mode horizontal pulse to display positioning calculations. Should be correct now, without breaking other special monitor modes..

23 February 2014, 14:48	#7
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,336	How is the displayable area determined for programmed modes? I checked various ECS modes (the monitor drivers use slightly different settings on AGA machines, maybe something related to fetch mode?). It seems that HTOTAL + 1 - (HBSTOP - HBSTART) gives the number of displayable colour clocks (so multiply that by 8 to get number of 35ns pixels). At least for all modes I checked except EURO36. Calculating that for the Debian mode (HTOTAL=0071, HBSTRT=0001, HBSTOP=0023): 71 + 1 - (23 - 1) = $50. $50 × 8 = 640 (At the moment WinUAE shows 638 pixels for that mode I think.) By "max horizontal overscan" I mean the largest width you can set using Overscan prefs. EURO72 (max horizontal overscan 680): HTOTAL=0071, HBSTRT=0001, HBSTOP=001E HTOTAL + 1 - (HBSTOP - HBSTART) = 71 + 1 - (1E - 1) = 72 - 1D = $55. $55 × 8 = 680 EURO36 (max horizontal overscan 720): HTOTAL=00E2, HBSTRT=0008, HBSTOP=002F HTOTAL + 1 - (HBSTOP - HBSTART) = E2 + 1 - (2F - 8) = $BC. $BC × 4 = 752 For EURO36 the maximum selectable overscan width is 32 fewer than that. Multiscan (max horizontal overscan 680): HTOTAL=0071, HBSTRT=0001, HBSTOP=001E HTOTAL + 1 - (HBSTOP - HBSTART) = 71 + 1 - (1E - 1) = $55. $55 × 8 = 680 SUPER72 (max horizontal overscan 936): HTOTAL=0091, HBSTRT=0001, HBSTOP=001E HTOTAL + 1 - (HBSTOP - HBSTART) = 91 + 1 - (1E - 1) = $75. $75 × 8 = 936 DBLNTSC and DBLPAL (max horizontal overscan 744): HTOTAL=0079, HBSTRT=0001, HBSTOP=001E HTOTAL + 1 - (HBSTOP - HBSTART) = 79 + 1 - (1E - 1) = $5D. $5D × 8 = 744 Example of different register values used on AGA machines by the DBLNTSC monitor driver: DBLNTSC: HTOTAL=0081, HBSTRT=0001, HBSTOP=0021 HTOTAL + 1 - (HBSTOP - HBSTART) = 81 + 1 - (21 - 1) = $62. $63 × 8 = 784 Max selectable AGA horizontal overscan is 720; 64 shres pixels (= 8 colour clocks) less. Is the difference related to 4x fetch mode and/or sprites?

23 February 2014, 18:42	#12
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,506	You misunderstood. It means winuae.zip is only allowed to used if I say so. (something got fixed and it needs confirmation, like above). Using it because it is newer is not acceptable reason!

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