[Intel-gfx] [RFC 06/28] drm/i915: Convert i915_gem_init_swizzling to intel_gt

[Chris Wilson]

Quoting Tvrtko Ursulin (2019-06-14 10:06:41)
> 
> On 13/06/2019 14:49, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2019-06-13 14:35:17)
> >> From: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> >>
> >> Start using the newly introduced struct intel_gt to fuse together correct
> >> logical init flow with uncore for more removal of implicit dev_priv in
> >> mmio access.
> >>
> >> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> > 
> > Looks fine, I might move it again later next to the fence registers, or
> > at least pull this and the detection into its own intel_gt_swizzling.c
> > 
> > Hmm, now that I said that, does that seem like a reasonable thing to do
> > right away, see i915_gem_fence_regs.c for the swizzle probe?
> 
> Is swizzling global for the memory controller or applicable only for 
> fenced regions?

As far as my understanding goes, it used to be only for fenced regions
when the gpu was the gmch, but completely migrated to the memory
controller around snb (with the ring architecture, the GPU was just
another client). This coincides with swizzling becoming defunct as part
of tiling. To further muddy the pictures, all the time the memory
controller is interleaving across the channels. I am pretty certain
around gen3 this was explicitly controlled by the GPU for its pages, but
by gen5 this is transparent to the GPU. (See the issues with L-shaped
memory configurations where the magic was not hidden from the GPU.)

So, aiui, for our world view tiling and swizzling are complicit.
-Chris