[Gilbert]

Originally Posted by d4rk3lf

If you are interested in Akiko, here is some info from Gunnar from Apolo team.
http://www.apollo-core.com/knowledge...28256&z=ZrTXsF

I can't say he is right or wrong, because I am noob at hardware, but it' was an interesting read.. at least for me.

(p.s, let's avoid Vampire discussions, I gave link to Gilbert, only because Akiko).

Thanks for this. This guy rates Akiko higher than most do. An interesting read! It definitely sounds like you just call the chip and it does it's thing and puts the results in memory - so the CPU is free to do it's own things.

Originally Posted by Bruce Abbott

The way it works is the CPU writes 32 longwords worth of chunky pixels into it, then reads 32 longwords of bitplane data out. As far as I know there is no waiting for it do the conversion. The CPU has a bit of work to do copying from the chunky bitmap to Akiko and copying from Akiko to the bitplanes, but this is much less than the work required do the conversion with CPU instructions. IOW, it makes c2p viable on a stock CD32, when otherwise it would be much too slow.

However because Akiko is on the 7MHz chipRAM bus a faster CPU must slow down to access it. Some accelerator cards had unsophisticated ChipRAM bus interfaces that forced them wait for the whole time (or even longer if not synced to it) while others latched the write data so the CPU could continue processing while the ChipRAM bus write was in progress.

An interesting idea I have been thinking about is creating a c2p chip like Akiko but which runs on the CPU bus, perhaps even writing to ChipRAM automatically via DMA so the CPU only has to stuff it with chunky data. Or the ultimate, a full frame chunky bitmap in FastRAM which looks like an actual chunky screen to the CPU, coupled to ChipRAM via DMA for zero c2p overhead!

Ok so i understand this now. The CPU has to copy the results to chipram. But it sounds a lot faster with AKiko than without.

From that thread link above:

A CPU C2P routine does
1) LOAD of chunky data into the CPU,
2) 16 or so mathematical operations, /AND/OR/SHIFT...
3) STORE of planar data into chipmem for display.

A AKIKO C2P routine does
1) LOAD of chunky data into the CPU,
2) 1 WRITE and 1 READ to AKIKO
3) STORE of planar data into chipmem for display.

In a little more details, DoomAttack.040:

c2p_Akiko2: 48.6 PFS (SA_7376x12)
c2p_040: 48.6 FPS
c2p_Blitter: 11.5 FPS

Originally Posted by mcgeezer

What are you talking about, i’m not hating on any Amiga’s and i gave my answer in my response. Amiga was much more than a gaming platform, the CD32 is an A1200 with Cdrom drive and sub par 3d processing chip. It offered very little over and above what was already on the Amiga platform.

I was going to apologise but then you did hate on the CD32 by saying it has a sub par 3d processing chip! I feel a lot of A1200 owners do get insecure when the CD32 is nearby. They have to try to bring the CD32 down to A1200 level. Understandable though.