OpenMSX - V9990PixelRenderer::updateScrollAYLow()

Paragon (1639)

10-05-2022, 13:25

Hi everyone,

I'm looking for information on "on the fly" scroll registers modifcation on v9990.
I have found this in the OpenMSX source:

Quote:

Changing high byte of vertical scroll registers only takes effect at the start of the next page. These members hold the current value of the scroll value. note: writing the low byte has effect immediatly (or at the next line), but the effect is not the same as on V99x8, see V9990PixelRenderer::updateScrollAYLow() for details

But looking at V9990PixelRenderer::updateScrollAYLow() code, I do not understand its meaning, nor do I see any additional information. What is the ' lastY ' value?

void V9990PixelRenderer::updateScrollAYLow(EmuTime::param time)
{
	if (displayEnabled) {
		sync(time);
		// happens in all display modes (verified)
		// TODO high byte still seems to be wrong .. need to investigate
		verticalOffsetA = lastY;
	}
}

Can someone shed some light on this ?

To give some context : I'm trying to make a scroll-split on the v9990. After setting an HBLANK interruption at Y=188, I change the value of the layer_B Y scroll register (only 0-255, so writing only on R#21), and restore it at VBLANK. It does not work, and I don't understand why. Furthermore, I tested writing 0 at HBLANK, and writing again 0 at VBLANK (basically not changing the value at all), and all I get is that the part below Y=188 disappears. Which does not make sense at all ... I really don't understand what's happening.

Por Metalion

Paragon (1639)

10-05-2022, 14:56

After multiple tests, I found out how it worked:
Y_scroll_register_value = hblank_line + 1 - (screen_position - real_position)
The '+1' is needed to because the register value change actually happens on the next line.

So for example, if you have an HBLANK interrupt at line 186, and you want the screen area below that line to be scrolled down 8 pixel, your Y scroll register value needs to be 186 + 1 - (186+8 - 186) = 179

I'll test tonight on real hardware.

Por Metalion

Paragon (1639)

10-05-2022, 16:51

Metalion wrote:

After multiple tests, I found out how it worked: ...

Not so easy, after all ... I have found out it does not work like that
I continue my tests.

Por ToriHino

Paladin (956)

10-05-2022, 17:05

Interesting indeed, also curious about the behavior on real hardware.

Por wouter_

Hero (540)

10-05-2022, 17:43

Hi,
First of all openMSX v9990-emulation is much less mature than the V99x8 emulation. Always verify on real V9990 hardware (and then please report any discrepancies).

I did some digging in the openMSX code, and apparently the implementation is based on the information provided by 'GhostwriterP' in this MRC forum topic from 2008.

Por Metalion

Paragon (1639)

10-05-2022, 19:28

As I should have guessed from the OpenMSX code, it's actually simpler than that, but also counterintuitive. When you change the scroll Y register value after an HBLANK interruption, the v9990 considers that the starting line of the screen (y=0) is located at that HBLANK line.

Por Metalion

Paragon (1639)

11-05-2022, 20:10

Metalion wrote:

I'll test tonight on real hardware.

I just tested it on a real V9990, and I confirm that the scroll split works like it should.

Por GhostwriterP

Paladin (718)

12-05-2022, 17:43

Scrolling little bits up or down with vertical scroll will work fine AFAIK. Only in case of full vertical wrap around or crossing 256 line border you may get issues, this because the most significant scroll bit(s) (bit 8 and higher) is not looked at by the VDP during display period... at least that is how I remembered it.