Re: [hatari-devel] New cycle exact bus sharing for blitter

[ Thread Index | Date Index | More Archives ]

On Sun, 13 Aug 2017, Nicolas Pomarède wrote:

Le 13/08/2017 à 14:32, Anders Eriksson a écrit :
The invitation works on Mega STe, but as the doc writes it's one write shorter per line. Question is what you mean with shifted? As the plasma is 8 px shorted, I centered it with an extra nop, is that what you mean by shifted?

What the hell, let's try Troeds latest Hatari and see what it looks like in MSTe mode...

Now tested, it looks like expected with 5 px of bg from the previous line. In hindsight, for the emulator sake, it would perhaps have been better to not have the center-nop and let the colour at the right border become "long". At the time of doing the demo, I'm not sure Hatari had a (working) Mega STe mode, if it did, I was unaware of it. 505 did the Mega STe testings and it's very likely that his monitor didn't see that far out in the border.


this is what I see when I set Hatari to mega STE mode : instead of being "vertical" (starting from the left border), we see the result is diagonal, which means that there're more cycles per line than the expected 512 cycles.

Is that what you see under the OSX build too ? It's not that the color transitions are visible, but it shouldn't be diagonal like this I guess ?

Maybe there something wrong in the way Hatari "respond" to the mege STE detection in this demo. Do you test ff8e21 to detect mega STE, or sthg else ?


That looks like it's using the regular STe codepath rather than the MSTe one. This is how it looks when I run in Troeds Hatari from Aug13 with Hatari in MSTe mode:

And in regular STe mode:

It is what I would expect on real hardware as well, a little bit narrower plasma on MSTe.

Thinking for a second what could make my detection of MSTe to fail (I'm using standard cookie jar) would be if the MSTe is using a TOS that doesn't know of the MSTe. So I tried it with 1.62, and then I get your failed plasma.

So I think Hatari allowing for 1.06/1.62 on Mega STe is the problem. The machine was delivered with 2.05 or 2.06 and I guess 1.06/1.62 will wrongly set _MCH to STe.

Anders Eriksson

Mail converted by MHonArc 2.6.19+