All of lore.kernel.org
 help / color / mirror / Atom feed
From: Matthew Auld <matthew.auld@intel.com>
To: intel-gfx@lists.freedesktop.org
Cc: "Thomas Hellström" <thomas.hellstrom@linux.intel.com>,
	dri-devel@lists.freedesktop.org
Subject: [Intel-gfx] [PATCH 1/3] drm/i915/ttm: consider all placements for the page alignment
Date: Wed, 23 Jun 2021 12:26:35 +0100	[thread overview]
Message-ID: <20210623112637.266855-1-matthew.auld@intel.com> (raw)

Just checking the current region is not enough, if we later migrate the
object somewhere else. For example if the placements are {SMEM, LMEM},
then we might get this wrong. Another idea might be to make the
page_alignment part of the ttm_place, instead of the BO.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
---
 drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 21 ++++++++++++++++++++-
 1 file changed, 20 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
index c5deb8b7227c..5d894bba6430 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
@@ -753,6 +753,25 @@ void i915_ttm_bo_destroy(struct ttm_buffer_object *bo)
 		call_rcu(&obj->rcu, __i915_gem_free_object_rcu);
 }
 
+static u64 i915_gem_object_page_size(struct drm_i915_gem_object *obj)
+{
+	u64 page_size;
+	int i;
+
+	if (!obj->mm.n_placements)
+		return obj->mm.region->min_page_size;
+
+	page_size = 0;
+	for (i = 0; i < obj->mm.n_placements; i++) {
+		struct intel_memory_region *mr = obj->mm.placements[i];
+
+		page_size = max_t(u64, mr->min_page_size, page_size);
+	}
+
+	GEM_BUG_ON(!page_size);
+	return page_size;
+}
+
 /**
  * __i915_gem_ttm_object_init - Initialize a ttm-backed i915 gem object
  * @mem: The initial memory region for the object.
@@ -793,7 +812,7 @@ int __i915_gem_ttm_object_init(struct intel_memory_region *mem,
 	obj->base.vma_node.driver_private = i915_gem_to_ttm(obj);
 	ret = ttm_bo_init(&i915->bdev, i915_gem_to_ttm(obj), size,
 			  bo_type, &i915_sys_placement,
-			  mem->min_page_size >> PAGE_SHIFT,
+			  i915_gem_object_page_size(obj) >> PAGE_SHIFT,
 			  true, NULL, NULL, i915_ttm_bo_destroy);
 	if (!ret)
 		obj->ttm.created = true;
-- 
2.26.3

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

WARNING: multiple messages have this Message-ID (diff)
From: Matthew Auld <matthew.auld@intel.com>
To: intel-gfx@lists.freedesktop.org
Cc: "Thomas Hellström" <thomas.hellstrom@linux.intel.com>,
	dri-devel@lists.freedesktop.org
Subject: [PATCH 1/3] drm/i915/ttm: consider all placements for the page alignment
Date: Wed, 23 Jun 2021 12:26:35 +0100	[thread overview]
Message-ID: <20210623112637.266855-1-matthew.auld@intel.com> (raw)

Just checking the current region is not enough, if we later migrate the
object somewhere else. For example if the placements are {SMEM, LMEM},
then we might get this wrong. Another idea might be to make the
page_alignment part of the ttm_place, instead of the BO.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
---
 drivers/gpu/drm/i915/gem/i915_gem_ttm.c | 21 ++++++++++++++++++++-
 1 file changed, 20 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
index c5deb8b7227c..5d894bba6430 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_ttm.c
@@ -753,6 +753,25 @@ void i915_ttm_bo_destroy(struct ttm_buffer_object *bo)
 		call_rcu(&obj->rcu, __i915_gem_free_object_rcu);
 }
 
+static u64 i915_gem_object_page_size(struct drm_i915_gem_object *obj)
+{
+	u64 page_size;
+	int i;
+
+	if (!obj->mm.n_placements)
+		return obj->mm.region->min_page_size;
+
+	page_size = 0;
+	for (i = 0; i < obj->mm.n_placements; i++) {
+		struct intel_memory_region *mr = obj->mm.placements[i];
+
+		page_size = max_t(u64, mr->min_page_size, page_size);
+	}
+
+	GEM_BUG_ON(!page_size);
+	return page_size;
+}
+
 /**
  * __i915_gem_ttm_object_init - Initialize a ttm-backed i915 gem object
  * @mem: The initial memory region for the object.
@@ -793,7 +812,7 @@ int __i915_gem_ttm_object_init(struct intel_memory_region *mem,
 	obj->base.vma_node.driver_private = i915_gem_to_ttm(obj);
 	ret = ttm_bo_init(&i915->bdev, i915_gem_to_ttm(obj), size,
 			  bo_type, &i915_sys_placement,
-			  mem->min_page_size >> PAGE_SHIFT,
+			  i915_gem_object_page_size(obj) >> PAGE_SHIFT,
 			  true, NULL, NULL, i915_ttm_bo_destroy);
 	if (!ret)
 		obj->ttm.created = true;
-- 
2.26.3


             reply	other threads:[~2021-06-23 11:26 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-06-23 11:26 Matthew Auld [this message]
2021-06-23 11:26 ` [PATCH 1/3] drm/i915/ttm: consider all placements for the page alignment Matthew Auld
2021-06-23 11:26 ` [Intel-gfx] [PATCH 2/3] drm/i915: support forcing the page size with lmem Matthew Auld
2021-06-23 11:26   ` Matthew Auld
2021-06-23 12:54   ` [Intel-gfx] " Thomas Hellström
2021-06-23 12:54     ` Thomas Hellström
2021-06-23 11:26 ` [Intel-gfx] [PATCH 3/3] drm/i915/gtt: ignore min_page_size for paging structures Matthew Auld
2021-06-23 11:26   ` Matthew Auld
2021-06-23 11:51   ` [Intel-gfx] " Thomas Hellström
2021-06-23 11:51     ` Thomas Hellström
2021-06-23 12:25     ` [Intel-gfx] " Matthew Auld
2021-06-23 12:25       ` Matthew Auld
2021-06-23 12:44       ` [Intel-gfx] " Thomas Hellström
2021-06-23 12:44         ` Thomas Hellström
2021-06-23 13:32   ` [Intel-gfx] " Thomas Hellström
2021-06-23 13:32     ` Thomas Hellström
2021-06-23 13:38     ` [Intel-gfx] " Matthew Auld
2021-06-23 13:38       ` Matthew Auld
2021-06-23 13:39       ` [Intel-gfx] " Thomas Hellström
2021-06-23 13:39         ` Thomas Hellström
2021-06-23 12:50 ` [Intel-gfx] [PATCH 1/3] drm/i915/ttm: consider all placements for the page alignment Thomas Hellström
2021-06-23 12:50   ` Thomas Hellström
2021-06-23 14:28 ` [Intel-gfx] ✗ Fi.CI.SPARSE: warning for series starting with [1/3] " Patchwork
2021-06-23 14:43 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork

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=20210623112637.266855-1-matthew.auld@intel.com \
    --to=matthew.auld@intel.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=thomas.hellstrom@linux.intel.com \
    /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.