> You can see the address ranges if you go to the properties of your
> graphics card in device manager (under the "Resources" tab). I suspect
> the small IO address ranges are for sending commands to the GPU (which
> you can just use CPU store instructions to access), and the larger ones
> are a map of the RAM on the card (which you use DMA for).

That would be my guess also.

> Typically though I don't think you will find much documentation on *how*
> to communicate with a modern desktop GPU. You might find this
> interesting though, whilst not the same architecture as a PC it does
> give a bit of an insight on a modern GPU architecture (it's the one on
> the raspberry PI):
>
> http://www.broadcom.com/docs/support/videocore/VideoCoreIV-AG100-R.pdf

I'm not sure it answers my question, but it's certainly a very 
interesting read...

Sections 8 and 9 in particular seem to be saying that there's about 
two-dozen memory-mapped registers that tell the GPU where the control 
lists are, and then it sucks those into the GPU and goes away to do it's 
thing.

Post a reply to this message