GFX9000 extension bug related to R#22

[albs-br]

Hi guys, me and some fellows on MSX.org found a bug on the GFX9000 implementation of V9990 chip, more specifically on R#22:

The two higher bits (bits 7 and 6) are supposed to disable layers AND sprites, not just the tile patterns layer.

The documentation states:
SDA: Set to "1" to disable layer "A" and sprites.
SDB: Set to "1" to disable layer "B" and sprites.

The discussion can be viewd here (page 4 onwards).

If necessary I can provide a test ROM showing the bug.

[MBilderbeek]

The simplest possible test ROM would be ideal, including a video what's happening on real MSX.
The reference to the discussion is good to have, but the core of the points in the ticket here is most helpful. :)

[MBilderbeek]

@albs-br Any news on a demonstration ROM?

[albs-br]

Hi, sorry, I forgot about this issue.

Now I made 4 ROMs for testing:

v9990-openmsx-issue-1466-A.rom ; ----------- Enable both layers and sprites
v9990-openmsx-issue-1466-B.rom ; ----------- Disable layer A and sprites
v9990-openmsx-issue-1466-C.rom ; ----------- Disable layer B and sprites
v9990-openmsx-issue-1466-D.rom ; ----------- Disable both layers and sprites

Green is layer A, Blue layer B and the red square is a sprite.

They change the two higher bits of R#22 (called SDA and SDB).
All 4 ROMS shows the same output on openMsx, but on real hardware will lead to different results, I suppose (I don't have physical V9990 cart).

This behaviour is undocumented, there is no info about it on the official Yamaha manual, only on MSX Banzai site.

Source code and ROMs are here (v.0.41.2): https://github.com/albs-br/msx-tests

Hope that helps and congrats for the great work on openMsx.

[MBilderbeek]

Thanks. Of course we will first need someone to show this on a real MSX with V9990 before we can change any openMSX code.

[albs-br]

Probably the best way is asking for help on msx.org forum.

[MBilderbeek]

Done: https://www.msx.org/forum/msx-talk/openmsx/test-some-v9990-roms

[MBilderbeek]

Attached ROMs confirmed working on real machine in that thread. In openMSX, they all give the same output indeed. Source code is available as mentioned in the above post.
testroms.zip

@m9710797 can you take a quick look?