A2410 in NetBSD

[mark_k]

I thought I'd go back and test various graphics cards in NetBSD. NetBSD automatically uses a supported graphics card for its text-mode console.

A few years ago I created a NetBSD 3.1 HDF. With an A4000-like config + A2410, WinUAE 4.10.1 boots the HDF OK: NetBSD switches to A2410 (1024x768) and console text appears OK. I can log in then halt/shutdown.

Trying that in WinUAE 5.2.0, NetBSD switches to the A2410 but no text appears, just an all-grey screen. Same in 5.1.0. With WinUAE 5.0.0, A2410 console text does appear but there are some bogus black horizontal lines across the screen below the cursor.

I uploaded the HDF back then. NetBSD_3.1_installed_A4000_060.tar.xz, ~69MB archive.

https://www.4!shared.com/s/fxmOtXyHz

Configs:

https://www.media!fire.com/file/m44kew16xc991m4/NetBSD_3.1_configs.7z/file

(Remove the ! after pasting, the EAB forum deletes any unaltered mention of the site names...)

Edit to add: Changing graphics card to e.g. Piccolo, that seems a bit broken in WinUAE 5.2.0. Console text appears OK, but when the screen first needs to scroll, it blanks. Works fine in WinUAE 3.6.1.

[Toni Wilen]

Picasso II (and other GD-CL54xx based cards) scrolling fixed. VRAM reads (that are usually very rare) were not working correctly in some situations.

A2410: No idea, need to check the sources. NetBSD clears the VRAM, sets RAMDAC palette and other settings etc. But then it never writes anything to overlay (part of "program" RAM bank) RAM. If I remember correctly, NetBSD uses RAMDAC overlay feature as a text console screen.

[mark_k]

Thanks.

Another issue I noticed. With Cybervision 64/3D, booting switches to RTG, white screen, with a huge number of repeated lines in the log:
01008504 no mapping


Also check oMniBus ET4000W32 and ET4000AX. They both seem to work OK, but for the ET4000W32 the colours are unusual: dark blue background, mid-grey foreground/text. Not sure if that's due to different DAC (AT&T ATT20c491), or maybe whoever wrote the driver just liked that palette? The ET4000AX by comparison has dark blue-green background, grey text. Its boot text mentions Sierra SC11483 DAC.

[Toni Wilen]

EDIT: A2410 also fixed.

Originally Posted by mark_k

Another issue I noticed. With Cybervision 64/3D, booting switches to RTG, white screen, with a huge number of repeated lines in the log:
01008504 no mapping

Fixed (was fixed in 86box update)

But my NetBSD 6 HDF fails with "panic: uvm_km_suballoc: unable to allocate space in parent map". (both CV64 and CV64/3D. Others work)

btw, NetBSD has a bug in Amiga CV64/3D driver.

grf_cv3dreg.h:

#define CMD_NOP (7 << 27) /* %1111 << 27 */

Comment is correct. Actual value is wrong. (%0111 is undefined blitter command)

ET4000W32 the colours are unusual: dark blue background, mid-grey foreground/text.

Yes, it detects wrong RAMDAC (6bit/8bit difference) but I am not sure if fixing this is worth the trouble because all other Amiga software does not do any detection. Probably needs yet another RAMDAC emulation..

[mark_k]

More testing with 5.3.0b5 now...

Retina (original Zorro II) display shows mostly black screen with random coloured character-width lines.

Retina BLT Z-III display shows all-grey screen. WinUAE may not actually be switching to/enabling the RTG display, that could be the native output.

[mark_k]

Originally Posted by Toni Wilen

But my NetBSD 6 HDF fails with "panic: uvm_km_suballoc: unable to allocate space in parent map". (both CV64 and CV64/3D. Others work)

I saw that testing with a NetBSD 5.2 HDF just now.

On NetBSD 1.3.3, CyberVision 64 seems to work OK. CyberVision 64/3D initially seems to work OK, but once the screen fills with text (so scrolling is needed), screen fills with black vertical stripes on grey background. HDF (uploaded years ago): https://www.4!shared.com/s/fnHXx4kZ8

[Toni Wilen]

Originally Posted by mark_k

More testing with 5.3.0b5 now...

Retina (original Zorro II) display shows mostly black screen with random coloured character-width lines.

Retina BLT Z-III display shows all-grey screen. WinUAE may not actually be switching to/enabling the RTG display, that could be the native output.

This needs text mode support. NCR chip emulation used in Retina was coded by me and I didn't care about text modes (or any other legacy PC stuff).

Probably simple enough but I am not that interested. Maybe later...

Originally Posted by mark_k

I saw that testing with a NetBSD 5.2 HDF just now.

On NetBSD 1.3.3, CyberVision 64 seems to work OK. CyberVision 64/3D initially seems to work OK, but once the screen fills with text (so scrolling is needed), screen fills with black vertical stripes on grey background. HDF (uploaded years ago): https://www.4!shared.com/s/fnHXx4kZ8

Probably need S3Virge update from 86box. Not going to happen in this beta series.

[mark_k]

Back to the A2410, this time with Amix. (Figured I may as well post here rather than make a new thread.)

In Amix the A2410 mouse cursor is all white.

Testing with the hardfile from this post by vintageBytes. The link to amix_hardfile.7z no longer works, but the file is available via web.archive.org (go there and enter the URL from the post).

After booting the HDF, log in as root then do either "xinit -- -tiga -tm 3" or "olinit -- -tiga -tm 3"

[Toni Wilen]

RAMDAC overlay colors are set to 000000, FFFFFF, FFFFFF, 000000. Write is done by the TMS CPU so not that easy to debug. I thought the desktop was much more colorful when I last saw it. (Few years ago?)

It does look like Amix is in some strange grayscale mode.

[mark_k]

I'm not sure of the specifics, but I think Amix shipped with a monochrome-only X server for A2410. I don't know whether the X11R5 server on the Amix 2.1 tape (not installed/used by default AFAIK) is capable of colour.

There are Amix HDFs which have X11R5 and third-party Cirrus Logic card support, there the X desktop is in colour (with Piccolo/Picasso configured in WinUAE). E.g. see this other post by vintageBytes (use web.archive.org to download that).

Edit: It is possible to use A2410 in colour. Do e.g. "xinit -- -tiga". Not very colourful but not black & white. [The monochrome-only X server is probably just for native chipset modes.]

[Toni Wilen]

Second plane of overlay is completely zeroed which explains single color mouse cursor. Maybe it is supposed to be like this.. (CGX4/Picasso96 etc have usual 3 color overlay mouse)

[mark_k]

I can't imagine it would be like that on real hardware; an all-white mouse pointer is basically unusable, e.g. it vanishes when over the white background of a terminal window. People would have been moaning about that on Usenet back in the day...

A2410_HWSW_Overview.pdf mentions "The next function takes advantage of the Flash Write Enable signal (FWE) available in many commercial Video RAMs (VRAMS). By allowing the utilization of this feature, the screen can be cleared in a small fraction of the time that would be required by the conventional method of writing to the frame buffer memory."

Does WinUAE emulate that? Could the driver be using that to set or clear the overlay bitplanes?

By the way, if you didn't already notice there are some DMI Resolver-related source files in /usr/x11r5/server/ddx/amix/