Is VDP1 manual change possible with double-density interlace?

[mrkotfw]

I'm running some tests and I have the following (VDP1+VDP2 normal resolution, non-interlace):

0. Set up list of commands in H-WRAM
1. Transfer commands from H-WRAM to VDP1 VRAM
2. Draw. PTMR=1
3. If drawing is done (Sprite End IRQ fired) before VBLANK OUT IRQ, perform FB change. FBCR=3
4. Go back to step 0

Essentially the above would allow me to perform as many "draw calls" as I like, and only change framebuffers when I'm done.

As opposed to 1-cycle mode where I must always finish drawing before V-Blank OUT IRQ is fired.


However, this doesn't seem to work for double-density interlace, as expected. Note above, I DO NOT call to erase the framebuffers (FBCR=2). Instead, the command list has a large polygon to erase the FB.

How would I approach this? Do I check VDP2(TVSTAT) for even/odd field scan and set FBCR=8 (DIE) and FBCR=12 (DIE|DIL), respectively during VBLANK-OUT?

I ran a test where I draw as normal (as explained above), and I go through two frames, each time setting FBCR=8 for even field, and FBCR=12 for odd field. No swapping of framebuffer.


Though, this isn't working. I get massive flickering.

How do I approach this? Draw twice? Does calling FBCR=8 and FBCR=12 swap the framebuffers for me?



Any bit of info would help!

 

[mrkotfw]

I'm completely in the dark here. Is it only possible to render via PTMR=2 when double-density interlace is enabled?

My goal is attempting to allow sub-30 FPS when double-density interlace is enabled by not swapping frame buffers. But now that I think about it, such a thing is not possible since both buffers are required to display a single frame.

At VBLANK-IN, I'm seeing if VDP1 is still drawing. If so, I set PTMR=0, which causes flickering.

Can anyone convince me that this is, or is not the case?

 

[mrkotfw]

Detect "transfer over" and perform pseudo draw continuation.

I'm making this more complicated than it needs to be. What I proposed above would be good if I had a third buffer. But I don't. It really sucks that I can't choose where to render.

It's really not easy to keep both VDP1 and VDP2 synced. It's a major PITA with timing, features, and all the edge cases...

Any other suggestions/ideas/features, I'd like to hear them.


Here is vdp_sync(). It doesn't yet include support for double-density interlace video modes.

The idea is to send a command list via vdp1_sync_draw(). You can send the VDP1 as many lists, and for every call, it'll draw immediately. All command lists are sent via SCU-DMA as well.

Then a call to vdp_sync() will "commit" VDP2 state and request to change VDP1 frame buffers.

Is there a way to know what the cost is to stop/start VDP1 drawing?