Re: [Intel-gfx] [PATCH 2/2] drm/i915: Track unbound pages

2012-08-20 Thread Daniel Vetter
On Mon, Aug 20, 2012 at 10:36:09AM +0100, Chris Wilson wrote: > On Mon, 20 Aug 2012 11:00:39 +0200, Daniel Vetter > wrote: > > int > > -i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj, > > - gfp_t gfpmask) > > +i915_gem_object_get_pages_gtt(struct drm_i915_g

Re: [Intel-gfx] [PATCH 2/2] drm/i915: Track unbound pages

2012-08-20 Thread Chris Wilson
On Mon, 20 Aug 2012 11:00:39 +0200, Daniel Vetter wrote: > int > -i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj, > - gfp_t gfpmask) > +i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj) > { > + struct drm_i915_private *dev_priv = obj->ba

[Intel-gfx] [PATCH 2/2] drm/i915: Track unbound pages

2012-08-20 Thread Daniel Vetter
When dealing with a working set larger than the GATT, or even the mappable aperture when touching through the GTT, we end up with evicting objects only to rebind them at a new offset again later. Moving an object into and out of the GTT requires clflushing the pages, thus causing a double-clflush p