Originally Posted by gulliver
I know how:
It says he showed video at 15fps
It says he used HAM on an A2000, so HAM6 it is.
It says the show was in NY, USA so probably used some NTSC image size
Then we do the math with little guessing:
Datarate 320x200x6bitx15fps= 704 KB/s rounded up
So, it is pretty doable, with, of course, the right storage media and its appropiate interface. You could even do it better!
It's 48000 bits for 320x200 per 6 bitplane image (or 64000 for 256 colour AGA). so that's 46.875kb per HAM6/EHB frame uncompressed we are talking?
So 15fps on most old original drives is possible/close but 930kb/s is very achievable for a much more credible and pleasant 20FPS even with a standard £1 costing 6gb 2.5" IDE drive of decent quality in an A600 or A1200. And these are uncompressed sizes of individual frames+palette swap per frame so the 7mhz A500/600/1000/2000 should be fine with EHB and HAM6.
The real question after watching that video again is this...as the video being parsed is not an actual AVI after listening to what the guy is saying but a series of HAM6 images...is there actually enough grunt in the OCS chipset to pull a 48k bitmap data for the image from a fast medium...stick it in screen memory...and swap the palette colours?
I ask this because if this guy's routine is state of the art and still requires an accelerator to achieve 15fps then clearly there is not enough time to do this on a 7mhz OCS Amiga then right? Which is disappointing to say the least if you need an 030 just to do that. Surely there are OCS games which blit more than 48k of graphics every 1/25th of a second in total?
All that we need is some machine code routine to load each uncompressed frame, put into the display area as 6 bits as is and change the palette colours...so I guess what I am asking is what is the actual DMA bandwidth on an OCS Amiga to handle this...is 1mb/s not possible on an Amiga bus using Agnus to blit the data in..data that should be read in via DMA too?
Anyway I am glad I fired up all your imagination so far...see sometimes it is good to dream...impressive reality may come from this.
Also another thought....would it help if you load in seperate bitplanes and write them to the respective bitplane on an alternate double buffered screen?So while you are loading one segment of 8k per frame for bitplane 2 of frame 1 you can be writing via the blitter bitplane 1 data for frame 1 from the previous loaded data. Can you do this in DMA or will the blitter hog the DMA and prevent this working.
Because that is 6 writes per frame so for 15 frames so that's 90 separate operations in the OCS chipset to write 1 bitplane of data to screen memory. Personally HAM is a nightmare to compute from an AVI I bet, if you want it to look optimal it will need Digi-view quality of processing and optimisation for each frame. Stick to a dithered EHB screen per frame I think for testing the throughput of the OCS chipset.