Amiga 1000 vs SUN 2/120 running globe/Framedemo

[pandy71]

Originally Posted by Gorf

The MSX2+ Platform used the Yamaha V9958 Video Chip.
It had a "YJK" mode:
YJK is composed of these three components:

Y = B/2 + R/4 + G/8
J = R - Y
K = G -Y

So Y is quite similar to Luminance and I think this format would have been quite nice for HAM, when only/mainly changing the Y component.

Both a component signal and a RGB signal could have been derived easily from an YJK source (essentially just some adders for RGB)

To get enough coulor-space, at least the Y component probably needs to be 5 bits wide internally.

Technically better (empirically proven) and probably simpler will be YCoCg https://en.wikipedia.org/wiki/YCoCg?useskin=vector color space - easier to implement (in software and in hardware).
It could be implemented in fast opamps externally and switched on the fly for example with help of ZD pin (genlock status).

[Locutus]

Originally Posted by Gorf

No it is using bitmaps:

Framedemo
framedemo displays a series of frames, each of which contains a 256 by 256 image one-bit-deep pixels
(that is, the image is a square monochrome bitmap, with 256 bits on a side). framedemo looks for the frames in the files frame.1 through frame.n in the current working directory, and displays them in numerical order.

http://bitsavers.informatik.uni-stut...3.4_198706.pdf

But the speed can be adjusted:

If you move the cursor onto the image surface, you can type certain commands to affect the rate at which the frames are displayed.
The initial rate is one frame per second:
r removes l/20th of a second from the interval.
F removes one second from the interval.
Ff makes the interval as small as possible.
s adds 1/20th of a second.
S adds one second.

@tschak909

Maybe you give it an other try with "Ff" to get the max framrate?

Aha! Thanks for actually digging this up, I tried to look for it but somehow failed.


Reading frames from individual files from the filesystem is probably also not great for performance, but that's on the implementor

[Gorf]

I assume it will buffer the sequence after the first run … or at least the file-system will keep it buffed.
But what we see in the video is simply the default 1 second delay after each frame
In this mode one could probably open 20 windows simultaneously without any further slowdown

[Gorf]

Originally Posted by tschak909

Amiga vs Sun2/120, two systems of comparable power (sun2/120s were used for bring-up of the Amiga), running "frame demos". This second try pushes the speed of the sun framedemo as high as it will go. #retrocomputing https://youtube.com/shorts/UmPryBACkl0

Well … now we see that bitmaps are overwritten multiple times each vertical screen refresh. So it is simply not synced to the monitor.
And we see Sun’s multitasking at work, giving each window a much longer uninterrupted time, than AOS would do.

What we still don’t see is how good or bad SUN’s software blitter actually is.

The best optical result with this dema would probably come from the smallest delay step (0.2 sec) resulting in 5 frames per second.
And more than 3 windows in parallel

[Bruce Abbott]

Originally Posted by Gorf

we see Sun’s multitasking at work, giving each window a much longer uninterrupted time, than AOS would do.

Pretty poor 'multitasking'.

What we still don’t see is how good or bad SUN’s software blitter actually is.

What we do see is that the Amiga animates the globes smoothly, while the Sun doesn't.

Pretty impressive for a machine that cost 20 times less.

Originally Posted by Locutus

...Also the Sun is running in 1152x900 at 72Hz, also a 'minor' difference there :-)

In 2 colors. Video refresh rate is irrelevant, except to note that the Sun can't produce composite video for displaying on a TV or recording to videotape.

The Amiga is running 4 colors. In 2 colors it should be able to run the same speed in 640x512. So the Sun has ~3 times more work to do, but its CPU is 40% faster so it should run at about half the speed.

[Gorf]

Originally Posted by Bruce Abbott

Pretty poor 'multitasking'.

What we do see is that the Amiga animates the globes smoothly, while the Sun doesn't.

More likely just the way this particular program was written, as it does apparently not yield its quantum of cpu-time after each frame-update, but simply keeps updating until it is preempted.

The Amiga is running 4 colors.

The globe on Amiga is just 2 colors, so it is just one bitplane, that needs updating. Same as on the SUN. Only the Amiga's globe is a much smaller area and therefore much less pixels.

This is really comparing apples and oranges. Not useful.