[Intel-gfx] [PATCH 11/11] drm/i915: Only bind each object rather than for every execbuffer

[Daniel Vetter]

One side-effect of the introduction of ppgtt was that we needed to
rebind the object into the appropriate vm (and global gtt in some
peculiar cases). For simplicity this was done twice for every object on
every call to execbuffer. However, that adds a tremendous amount of CPU
overhead (rewriting all the PTE for all objects into WC memory) per
draw. The fix is to push all the decision about which vm to bind into
and when down into the low-level bind routines through hints rather than
performing the bind unconditionally in the execbuffer routine.

Note that this is a regression introduced in the full ppgtt feature
branch, before this we've only done re-bound objects when the relevant
has_(aliasing_ppgtt|global_gtt)_mapping flag was clear. But since
that's per-object and not per-vma that optimization broke.

v2: Split out prep work and unrelated changes.

v3: Bring back functional change around PIN_GLOBAL that I've
accidentally split out.

v4: Remove the temporary hack for the old binding logic to avoid
bisection issues.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72906
Tested-by: jianx.zhou at intel.com
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk> (v1)
Cc: Ben Widawsky <benjamin.widawsky at intel.com>
Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
Acked-by: Ben Widawsky <ben at bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter at ffwll.ch>
---
 drivers/gpu/drm/i915/i915_gem.c            | 20 ++++++++++++--------
 drivers/gpu/drm/i915/i915_gem_execbuffer.c |  5 -----
 2 files changed, 12 insertions(+), 13 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index ce7064d1ab30..3618bb0cda0a 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3285,6 +3285,9 @@ search_free:
 	WARN_ON(flags & PIN_MAPPABLE && !obj->map_and_fenceable);
 
 	trace_i915_vma_bind(vma, flags);
+	vma->bind_vma(vma, obj->cache_level,
+		      flags & (PIN_MAPPABLE | PIN_GLOBAL) ? GLOBAL_BIND : 0);
+
 	i915_gem_verify_gtt(dev);
 	return vma;
 
@@ -3487,7 +3490,9 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
 		}
 
 		list_for_each_entry(vma, &obj->vma_list, vma_link)
-			vma->bind_vma(vma, cache_level, 0);
+			if (drm_mm_node_allocated(&vma->node))
+				vma->bind_vma(vma, cache_level,
+					      obj->has_global_gtt_mapping ? GLOBAL_BIND : 0);
 	}
 
 	list_for_each_entry(vma, &obj->vma_list, vma_link)
@@ -3838,22 +3843,21 @@ i915_gem_object_pin(struct drm_i915_gem_object *obj,
 			ret = i915_vma_unbind(vma);
 			if (ret)
 				return ret;
+
+			vma = NULL;
 		}
 	}
 
-	if (!i915_gem_obj_bound(obj, vm)) {
-
+	if (vma == NULL || !drm_mm_node_allocated(&vma->node)) {
 		vma = i915_gem_object_bind_to_vm(obj, vm, alignment, flags);
 		if (IS_ERR(vma))
 			return PTR_ERR(vma);
 	}
 
-	vma = i915_gem_obj_to_vma(obj, vm);
-
-	vma->bind_vma(vma, obj->cache_level,
-		      flags & PIN_GLOBAL ? GLOBAL_BIND : 0);
+	if (flags & PIN_GLOBAL && !obj->has_global_gtt_mapping)
+		vma->bind_vma(vma, obj->cache_level, GLOBAL_BIND);
 
-	i915_gem_obj_to_vma(obj, vm)->pin_count++;
+	vma->pin_count++;
 	if (flags & PIN_MAPPABLE)
 		obj->pin_mappable |= true;
 
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index b35849bbc367..d7229ad2bd22 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -589,11 +589,6 @@ i915_gem_execbuffer_reserve_vma(struct i915_vma *vma,
 		obj->base.pending_write_domain = I915_GEM_DOMAIN_RENDER;
 	}
 
-	/* Temporary hack while we rework the binding logic. */
-	flags = (entry->flags & EXEC_OBJECT_NEEDS_GTT) &&
-		!vma->obj->has_global_gtt_mapping ? GLOBAL_BIND : 0;
-	vma->bind_vma(vma, obj->cache_level, flags);
-
 	return 0;
 }
 
-- 
1.8.5.2