pouët.net

Youtube link to the Revision 2023 Amiga Compo run https://m.youtube.com/watch?v=n95gUdRwGRA&t=2690s

Thanks for all the kind comments!

I'm really happy that so many people got emotional vibes from this .. melancholy, simplicity, quietness.. even silence in parts. Once again Xerxes' music blew me out of the water.

It hasn't been a group decision as such to include any symbolism in the demo, but for me there is a special symbolic significance in both the title and the narrative of the demo, because my father passed away about a week before Revision.

The candle blowing out in the middle of the demo makes me think of him every time I watch it - a very melancholy thing for sure, but I also feel like it's a little seed that's been planted, which can bear fruit and bring back memories of my dad whenever I happen to come across this demo in the future. Thanks for humoring me about that, and for timing it so well Espen <3

@xeron: yes, infinite bobs with ~25 ish color and depth buffers.

@Dan : Perhaps OCS with 060 and 64MB RAM on CDTV? First screen will be that FBI arcade warning "Winners don't use AGA". Perhaps both a laugh track and applause track playing from the CD. I've got all these brilliant OCS ideas!

@Jobbo:
The jittering may be from multiple sources. Perhaps the vertex transform precision is not enough? Those are in fixed point format while I try to have a perspective division done in parallel using the "FPU pipeline" (Never confirmed that this parallel action actually happens though)

Another could be lack of subtexel precision for the texture interpolation.

A third one could be that there is no perspective correction in the texture mapping. The jittering got reduced by subdividing the skybox mesh into more triangles so that was at least one confirmed source.

Lots of things to improve as always :)

Sensenstahl: the videos started out as just the “flex” sequence and then ballooned a bit out from there. Would’ve been much cooler to hand animate those sequences, or do them in 3d if course, but time :)

I definetly agree that pure c2p is a dead end if you want anything at decent framerate. Even on 060 amigas with 100s of MHz the typical c2p pass would limit you to 25fps just because of the 81920 bytes worth of chip copy that has to be done each frame for a 256 color display.

We try to mitigate by using lower color modes, copper rainbows and such. The cloud zoomer is 64 (or was it 32?) color and the effect is interleaved with Kalms' c2p so there is no fast ram chunky buffer pass in between. Then a copper gradient is added to make it more colorful.

It's not just enough to blame the c2p though. The effects may be very heavy. The c2p could be blamed for 50 -> 25 fps reduction, but anything below that is to be blamed on the effect code itself.

A few hardware tricks are present in the demo though.

The VHS distort at the beginning and end is 640x180 in HAM8. It is using the bitplane modulo to pick lines from a base image that contain 8 layers of pre-distorted images. Similar to how someone might implement a twister.

The music is using dynamic volume adjustment coupled with an 8-bit sample stream to get a sound fidelity that exceeds 8-bit, but without suffering the volume loss of the usual 14-bit modes (Cheetah's idea! Also mentioned in the HW ref manual on closer inspection!)

Anyway, people like to see "big 3D effects" with lots of details and pretty high color imagery. Tunnel tables and 2D bumps and whatever. Clearly there is a market for these effects even at low framerates :D

Btw: The demo can be run with --fps parameter to have it spit out fps for each part.