Hi again,

I hope someone can clear up a few questions for me.
I'm having some trouble properly emulating the timing of things related to interrupts and video emulation which I couldn't find clear documentation for.

How I set up timing now (for PAL) is like this:

* 313 lines
* the first line of the display output is at line 40
* at the end of all 313 lines, the VDP (might) generate the VBlank IRQ
* the VBLANK timing bits in the VDP status registers are reset/set at the first cq last display lines (so not line 0 and 313, but line 40 and 232 or 252)
* HBLANK timing bits get reset/set as a rough guess (reset after 1/16 of one line, set after 15/16 of one line)
* the line number where a HBlank interrupt can take place counts from the first display line (40)onwards (so not the 1st line of the whole frame including borders)

With this set up, I can see that the screen split in Ashguine2 works fine, but Gryzor doesn't work properly and Vampire Killer shows sprite flicker when the player is somewhere high in the screen. The first display line of 40 was basically trial and error until the screen split in ashguine2 looked okay...

So my questions:
* Are VBLANK interrupts indeed generated after the whole frame, or are they generated after the last display line (so when the lower border begins)?
* Are above settings approx. ok?
* Is there any clear documentation about this that I missed?

Thanks!
Erik

Quite accurate documentation: the openMSX source code

Or the blueMSX source code

I was afraid you were going to say that :-)
Not exactly bed time reading material but hey, I guess I have to bite the bullet...

As an aside, I never thought MSX2 emulation would be this hard!
My own emulator JEmu2 already supports lots of different kinds of hardware, including a few 16 bit arcades (CPS1, snowbros), and Sega Master System but MSX2 emulation is much, much harder than any of them. Kudos to the people who were able to do it right!

Right is a big word, both the accurate emulators are still trying to get the details right that cause incompatibilities in some programs. I think the problem with MSX2 emulation is that the trickery that is timing sensitive is actually used.

about the number of lines in PAL/NTSC... There is no emulator capable of emulating that correctly... and with that I mean the vertical display stretch which needs to be rendered...

Quote:

* HBLANK timing bits get reset/set as a rough guess (reset after 1/16 of one line, set after 15/16 of one line)

Not really sure about that, they appear to be set 'late', but this may also simply be because of Z80-sluggishness. The best way to find this out is to put your emu and hardware side-by-side and change the 'delay value' until both display the same thing.

Quote:

So my questions: * Are VBLANK interrupts indeed generated after the whole frame, or are they generated after the last display line (so when the lower border begins)?

VBLANK starts after the last display line. (unless the program forces overscan mode)

Quote:

* Are above settings approx. ok?

They seem to be, although the numbers threw me off a little. (I'm used to seeing only the programmer's side of this)

Quote:

* Is there any clear documentation about this that I missed?

The V9938 manual has all the timing values in it. I remember using this once to create a high-speed interrupt using line-interrupts.

Thanks for that. I just found an important little bug regarding bit 7 of status register 0 (VBLANK IRQ flag) which seems to fix a lot

BTW, is there a V9938 databook somewhere to download which is complete? I only found an incomplete transcribed version in various places.
The best document I found on MSX hw is the excellent Portar document (which I used extensively for writing my emulator), but even that is missing many details which can only be found in source files of the better emulators. And the 'MILC' site was very useful too. Any more resources you know about?
Don't say the blue/openMSX emulator source I've seen the sources, but they seem to focus on functionality rather than documentation like for example the MAME source (which *is* very useful as documentation).

Freel free to join our openMSX IRC channel (#openMSX on irc.freenode.net) where you can get more info about where is what in the source and other knowledge of the developers. Also check the openMSX documentation.

Quote:

BTW, is there a V9938 databook somewhere to download which is complete? I only found an incomplete transcribed version in various places.

IIRC there's a large PDF on funet. It's a scanned version, and the page numbers don't match, but it is quite complete. It also has the timing values in miliseconds iirc.