Vamiga: Test ddf2 fails

[dirkwhoffmann]

In the reference image, the first couple of red lines are not continuous at the right border.

[dirkwhoffmann]

Strange. In line 48, background color is red all along the line (this is why vAmiga draws the line continuously). I have no clue why the display should become black early in the scanline as it is the case in the SAE reference image. Would be interesting to see how it looks on a real Amiga.

Reference:

[mithrendal]

I would like to test this on my real Amiga. But I first have to look for a new serial adapter to transfer the ADF to the Amiga 1000. I need a kind of female DB25 -> usb adaption. I will report screenshots here when I am done with it. ;-) I did this transfer in the past with an old PC with a db9 serial port but that PC is no more, it has already been thrown away.
EDIT: just bought the adapter, it should be arriving on thursday...

[dirkwhoffmann]

That would be awesome!!

At the moment we completely rely on the assumption that SAE / UAE is 100% accurate. Although I have a lot of confidence in the (functional) quality of UAE, some bug might still be hiding there, too.

[dirkwhoffmann]

Looks good now:

[mithrendal]

ddf2 on real A1000

[dirkwhoffmann]

As far as I can tell, the emulator images match the original. Thanks again, Mithrendal!

Original image is now part of the vAmigaTS repo.

[dirkwhoffmann]

For reference: This test looks differently on ECS machines.

Findings: Based on recent experiments I assume that the Amiga is equipped with a vertical and a horizontal DDF flipflop. When the horizontal raster beam passes DDFSTRT, the horizontal flipflop is set. When it passes DDFSTOP, the flipflop is cleared.

Test case ddf2 sets DDFSTOP to coordinate 0xE4 which is never reached by the horizontal counter (counter wraps over after 0xE2).

• OCS Agnus (Mithrendal's A1000):

From the image I conclude that the horizontal DDF flipflop gets cleared at the hardware stop, no matter what DDFSTOP says.

• ECS Agnus (My A500 MMSE):

From the image I conclude that the horizontal DDF flipflop is not cleared which causes the data fetch window to open at the earliest possible coordinate in the next line.

I don’t know if a real Amiga really has such a thing as a horizontal DDF flipflop, but this model allows me to reproduce both images 😎:

vAmiga with OCS Agnus:

vAmiga with ECS Agnus: