Hook the vma itself into the i915_gem_request_retire() so that we can
accurately track when a solitary vma is inactive (as opposed to having
to wait for the entire object to be idle). This improves the interaction
when using multiple contexts (with full-ppgtt) and eliminates some
frequent list walking.
A side-effect is that we get an active vma reference for free. The
consequence of this is shown in the next patch...
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
---
drivers/gpu/drm/i915/i915_debugfs.c | 2 +-
drivers/gpu/drm/i915/i915_gem.c | 36 ++++++++++++++++--------------
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 2 ++
drivers/gpu/drm/i915/i915_gem_gtt.c | 20 +++++++++++++++++
drivers/gpu/drm/i915/i915_gem_gtt.h | 5 +++++
5 files changed, 47 insertions(+), 18 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 179e3c5c5022..4df4ebbd56d6 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -356,7 +356,7 @@ static int per_file_stats(int id, void *ptr, void *data)
continue;
}
- if (obj->active) /* XXX per-vma statistic */
+ if (vma->active)
stats->active += vma->node.size;
else
stats->inactive += vma->node.size;
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 8a824c5d5348..1d21c5b79215 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2040,7 +2040,6 @@ i915_gem_object_retire__read(struct drm_i915_gem_request_active *active,
int ring = request->engine->id;
struct drm_i915_gem_object *obj =
container_of(active, struct drm_i915_gem_object, last_read[ring]);
- struct i915_vma *vma;
RQ_BUG_ON((obj->flags & (1 << (ring + I915_BO_ACTIVE_SHIFT))) == 0);
@@ -2052,12 +2051,9 @@ i915_gem_object_retire__read(struct drm_i915_gem_request_active *active,
* so that we don't steal from recently used but inactive objects
* (unless we are forced to ofc!)
*/
- list_move_tail(&obj->global_list, &request->i915->mm.bound_list);
-
- list_for_each_entry(vma, &obj->vma_list, obj_link) {
- if (!list_empty(&vma->vm_link))
- list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
- }
+ if (!list_empty(&obj->vma_list))
+ list_move_tail(&obj->global_list,
+ &request->i915->mm.bound_list);
drm_gem_object_unreference(&obj->base);
}
@@ -2567,7 +2563,19 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
{
struct drm_i915_gem_object *obj = vma->obj;
struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
- int ret;
+ int ret, i;
+
+ /* First wait upon any activity as retiring the request may
+ * have side-effects such as unpinning or even unbinding this vma.
+ */
+ if (vma->active && wait) {
+ for (i = 0; i < ARRAY_SIZE(vma->last_read); i++) {
+ ret = i915_wait_request(vma->last_read[i].request);
+ if (ret)
+ return ret;
+ }
+ RQ_BUG_ON(vma->active);
+ }
if (list_empty(&vma->obj_link))
return 0;
@@ -2582,12 +2590,6 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
BUG_ON(obj->pages == NULL);
- if (wait) {
- ret = i915_gem_object_wait_rendering(obj, false);
- if (ret)
- return ret;
- }
-
if (vma->is_ggtt && vma->ggtt_view.type == I915_GGTT_VIEW_NORMAL) {
i915_gem_object_finish_gtt(obj);
@@ -3023,9 +3025,8 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
/* And bump the LRU for this access */
vma = i915_gem_obj_to_ggtt(obj);
- if (vma && drm_mm_node_allocated(&vma->node) && !obj->active)
- list_move_tail(&vma->vm_link,
- &to_i915(obj->base.dev)->gtt.base.inactive_list);
+ if (vma && drm_mm_node_allocated(&vma->node) && !vma->active)
+ list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
return 0;
}
@@ -3874,6 +3875,7 @@ struct i915_vma *i915_gem_obj_to_ggtt_view(struct drm_i915_gem_object *obj,
void i915_gem_vma_destroy(struct i915_vma *vma)
{
WARN_ON(vma->node.allocated);
+ RQ_BUG_ON(vma->active);
/* Keep the vma as a placeholder in the execbuffer reservation lists */
if (!list_empty(&vma->exec_list))
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 6de8681bb64c..1d4378a4501e 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1099,6 +1099,8 @@ void i915_vma_move_to_active(struct i915_vma *vma,
}
}
+ vma->active |= 1 << engine;
+ i915_gem_request_mark_active(req, &vma->last_read[engine]);
list_move_tail(&vma->vm_link, &vma->vm->active_list);
}
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 130ccefb2491..5505603f52af 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3225,12 +3225,30 @@ void i915_gem_restore_gtt_mappings(struct drm_device *dev)
i915_ggtt_flush(dev_priv);
}
+static void
+i915_vma_retire(struct drm_i915_gem_request_active *active,
+ struct drm_i915_gem_request *rq)
+{
+ const unsigned engine = rq->engine->id;
+ struct i915_vma *vma =
+ container_of(active, struct i915_vma, last_read[engine]);
+
+ RQ_BUG_ON((vma->obj->active & (1 << engine)) == 0);
+
+ vma->active &= ~(1 << engine);
+ if (vma->active)
+ return;
+
+ list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
+}
+
static struct i915_vma *
__i915_gem_vma_create(struct drm_i915_gem_object *obj,
struct i915_address_space *vm,
const struct i915_ggtt_view *ggtt_view)
{
struct i915_vma *vma;
+ int i;
if (WARN_ON(i915_is_ggtt(vm) != !!ggtt_view))
return ERR_PTR(-EINVAL);
@@ -3242,6 +3260,8 @@ __i915_gem_vma_create(struct drm_i915_gem_object *obj,
INIT_LIST_HEAD(&vma->vm_link);
INIT_LIST_HEAD(&vma->obj_link);
INIT_LIST_HEAD(&vma->exec_list);
+ for (i = 0; i < ARRAY_SIZE(vma->last_read); i++)
+ init_request_active(&vma->last_read[i], i915_vma_retire);
vma->vm = vm;
vma->obj = obj;
vma->is_ggtt = i915_is_ggtt(vm);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 4e9553ace33f..c2f2c62ac88d 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -34,6 +34,8 @@
#ifndef __I915_GEM_GTT_H__
#define __I915_GEM_GTT_H__
+#include "i915_gem_request.h"
+
struct drm_i915_file_private;
typedef uint32_t gen6_pte_t;
@@ -180,10 +182,13 @@ struct i915_vma {
struct drm_i915_gem_object *obj;
struct i915_address_space *vm;
+ struct drm_i915_gem_request_active last_read[I915_NUM_RINGS];
+
/** Flags and address space this VMA is bound to */
#define GLOBAL_BIND (1<<0)
#define LOCAL_BIND (1<<1)
unsigned int bound : 4;
+ unsigned int active : I915_NUM_RINGS;
bool is_ggtt : 1;
/**
--
2.6.4
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
