wip: vram
The most natural way of managing PCI allocations is through the drm_gem_vram_helper
. If we're going to use this method, we'll have do various changes in the common code as well -- here's an initial pass at it. If we're not going to use this method, I guess we'll have to switch back to explicit TTM (I'm assuming drm_gem_vram_helper
will deprecate TTM eventually)?
Edited by Gurchetan Singh