Mappign gfx memory

[mike-]

Hi,

I heard somewhere that it should be possible to access the gfx memory and map it as memory... But how? And is it really possibly. I've googled a bit but i havent found a thing.

[Piru]

Reading gfx card memory is really slow. Even writing is in most setups much slower than normal memory, and the memory is cache inhibited (CPU won't cache the memory accesses).

However technically it would go something like this: Allocate a large chunk of the graphics memory permanently (probably only possible with Picasso96). Add the memory to system with exec AddMemList() call.

This only works for setups that have the video memory mapped to amiga address space in linear fashion (this rules out A1200 Mediator + Radeon 9200 as system memory, for example).

[mike-]

Ok, so i guess mapping the unused memory of voodoo5 on a mediator is out of the question then.

[Piru]

It is out (for A1200 mediator at least).

[mike-]

Yeah, i read that somewhere, i wonder if the 4k mediator will get an update as well. atleast its a promising sign that elbox is doing stuff again, i guess that means there still might be dragons and sharks coming our way. A shark would be nice anyway.

[Piru]

It is out as in "is out of the question".

The real manufacturer of "SharkPPC" is no longer producing the cards (Sonnet Crescendo 7200), so you can forget ever seeing "SharkPPC".

Dragon release is now over 3 years overdue. Don't hold your breath.

[mike-]

Never say never, i never did manage to get a hold of that Sonnet, same really, would be fun to see if i could actually use it for something somehow.

[zipper]

mike- wrote:
Hi,

I heard somewhere that it should be possible to access the gfx memory and map it as memory... But how? And is it really possibly.

For  Picasso II card there was a program to do it ( http://main.aminet.net/pub/aminet/util/boot/apm3311.lha ) .

[Hans_]

Piru wrote:
Reading gfx card memory is really slow. Even writing is in most setups much slower than normal memory, and the memory is cache inhibited (CPU won't cache the memory accesses).

However technically it would go something like this: Allocate a large chunk of the graphics memory permanently (probably only possible with Picasso96). Add the memory to system with exec AddMemList() call.

I think that there's a problem with doing that. IIRC, Picasso96 (and Cybergraphics) reserves the right to move bitmaps around unless they're locked. But if they're locked, then the layers are locked too (or maybe it's just screen switching that's disabled, I can't remember). The only way to allocate the memory, would be if Picasso96 isn't managing that graphics memory (i.e., the card isn't in use as a graphics card).

[Piru]

IIRC with Picasso96 you could allocate vmem in a way that it stays locked to specific address. I could be wrong, I last looked into the API over 7 years ago.