I did some tests with a loop that constantly wrote to VRAM using the VDP, and it seemed to work fine, have I misunderstood something?
Did you do this on a real MSX? IF so then you have a very strange MSX 
If you write to vram at same time the VDP is building the screen you'll get distortions.
ei
halt
di
This works fine if only the VDP generates interrupts. But if you plug in additional hardware which also generates INTS you wont be able to tell if it comes from the VDP on not.
Then I do have some issues with setting gfx modes. The BIOS way works fine to set Screen 1 (32x24 text mode). However when I set Screen 0, I get
40x24 as expected in emulator, but on my real MSX (NMS8245) instead I get some kind of MSX2 mode with 80 characters wide screen.
The 80x24 is set by setting:
ld a,80
ld (&hf3ae),a ;width 80
when calling the bios for textmode,it will look at that address to determine the width.
I also tried setting modes with the VDP directly, using the code hitchhiker posted, but then I only get garbage on screen no matter what mode I try. Strange...
If setting the mode yourself through the vdp-regs then you'll have to wipe clean the vram yourself. If you do it through the bios,the bios wipes it for you.
when working on a msx2 @3.5Mhz (V9938 VDP) you can output data to vdp at full z80 speed, even outside VBLANK, no problem at all.
when working on a msx1, (TMS VDP), the maximum speed you can achieve depends if you are in vblank or not.
outside vblank, every byte write or read should wait almost 26 z80 T-States (including the instruction that write data to VDP port 98h)
in vblank, no limit or delay.
Ahh! OK, that clears things up and makes more sense. The code I was testing with was my plasma code that actually does a bit of calculation for each output:
ld a,ix+XP add a,b out (c),a
Haven't counted how many T-states that is, but if I recall correctly the ix instructions takes quite a few.
And then I was testing it on my MSX2, but it would be nice to be able to write code that is MSX1 compatible.
Am I right in assuming that no emulator emulates this correctly, so the only way to know for sure that you don't stress the VDP too much is to test on a real computer?
For waiting for interrupts the tecnique below is OK, however i do not know why you disable ints
ei
halt
My thought was that I didn't want an interrupt going off when I was doing my VDP stuff, since I read that it could mess things up, but as long as I finish my code before next VBL, I guess it shouldn't be a problem having interrupts on all the time.
to get a good setup of the screen mode, you need also to intialize the VRAM....
The rest of the program was the same as before, so the VRAM has stuff written to it.
Ie. before I had like this:
1) Set Screen 1 using BIOS
2) Do effect
and I instead changed to
1) Set Screen1 using VDP
2) Do effect
And only got garbage.
This works fine if only the VDP generates interrupts. But if you plug in additional hardware which also generates INTS you wont be able to tell if it comes from the VDP on not.
OK, what is the correct procedure for a VBL wait if you want to be compatible with expanded units?
Also, thanks to everyone for helping out. I apologize for bombarding you with all these newbie questions! 
