All of lore.kernel.org
 help / color / mirror / Atom feed
From: Daniel Vetter <daniel.vetter@ffwll.ch>
To: Intel Graphics Development <intel-gfx@lists.freedesktop.org>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>,
	Ben Widawsky <benjamin.widawsky@intel.com>
Subject: [PATCH] drm/i915: Only bind each object rather than for every execbuffer
Date: Mon, 27 Jan 2014 23:36:16 +0100	[thread overview]
Message-ID: <1390862176-12543-1-git-send-email-daniel.vetter@ffwll.ch> (raw)
In-Reply-To: <1390846259-14915-10-git-send-email-daniel.vetter@ffwll.ch>

From: Chris Wilson <chris@chris-wilson.co.uk>

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
inside the execbuffer routine.

v2: Split out prep work and unrelated changes.

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

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72906
Tested-by: jianx.zhou@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> (v1)
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_gem.c            | 20 ++++++++++++--------
 drivers/gpu/drm/i915/i915_gem_execbuffer.c |  2 --
 2 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index c732f20a86ca..05e79fa02db2 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3333,6 +3333,9 @@ 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;
 
@@ -3535,7 +3538,9 @@ 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)
@@ -3886,22 +3891,21 @@ 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 9399a6fa4c2f..d7229ad2bd22 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -589,8 +589,6 @@ i915_gem_execbuffer_reserve_vma(struct i915_vma *vma,
 		obj->base.pending_write_domain = I915_GEM_DOMAIN_RENDER;
 	}
 
-	vma->bind_vma(vma, obj->cache_level, flags);
-
 	return 0;
 }
 
-- 
1.8.5.2

  reply	other threads:[~2014-01-27 22:36 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-01-27 18:10 Split-up of "drm/i915: Only bind each object rather than for every execbuffer" Daniel Vetter
2014-01-27 18:10 ` [PATCH 1/9] drm/i915: Consolidate binding parameters into flags Daniel Vetter
2014-01-27 21:47   ` Chris Wilson
2014-01-27 21:55     ` Daniel Vetter
2014-01-27 22:34   ` [PATCH] " Daniel Vetter
2014-01-28 13:09     ` Jani Nikula
2014-01-27 18:10 ` [PATCH 2/9] drm/i915: Don't set PIN_MAPPABLE for legacy ringbuffers Daniel Vetter
2014-01-27 18:10 ` [PATCH 3/9] drm/i915: Don't pin the status page as mappable Daniel Vetter
2014-01-27 18:10 ` [PATCH 4/9] drm/i915: Handle set_cache_level errors in the status page setup Daniel Vetter
2014-01-28 13:11   ` Jani Nikula
2014-01-27 18:10 ` [PATCH 5/9] drm/i915: Don't allocate context pages as mappable Daniel Vetter
2014-01-27 18:10 ` [PATCH 6/9] drm/i915: Allow blocking in the PDE alloc when running low on gtt space Daniel Vetter
2014-01-27 18:10 ` [PATCH 7/9] drm/i915: Simplify i915_gem_object_ggtt_unpin Daniel Vetter
2014-01-28 13:13   ` Jani Nikula
2014-01-28 14:21     ` Chris Wilson
2014-01-27 18:10 ` [PATCH 8/9] drm/i915: Directly return the vma from bind_to_vm Daniel Vetter
2014-01-28 13:21   ` Jani Nikula
2014-01-27 18:10 ` [PATCH 9/9] drm/i915: Only bind each object rather than for every execbuffer Daniel Vetter
2014-01-27 22:36   ` Daniel Vetter [this message]
2014-01-28 13:29 ` Split-up of "drm/i915: Only bind each object rather than for every execbuffer" Jani Nikula

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1390862176-12543-1-git-send-email-daniel.vetter@ffwll.ch \
    --to=daniel.vetter@ffwll.ch \
    --cc=benjamin.widawsky@intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.