Re: [Patch v2] drm/ttm: Schedule delayed_delete worker closer

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 2023-11-08 09:49, Christian König wrote:
Am 08.11.23 um 13:58 schrieb Rajneesh Bhardwaj:
Try to allocate system memory on the NUMA node the device is closest to
and try to run delayed_delete workers on a CPU of this node as well.

To optimize the memory clearing operation when a TTM BO gets freed by
the delayed_delete worker, scheduling it closer to a NUMA node where the
memory was initially allocated helps avoid the cases where the worker
gets randomly scheduled on the CPU cores that are across interconnect
boundaries such as xGMI, PCIe etc.

This change helps USWC GTT allocations on NUMA systems (dGPU) and AMD
APU platforms such as GFXIP9.4.3.

Acked-by: Felix Kuehling <Felix.Kuehling@xxxxxxx>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@xxxxxxx>

Reviewed-by: Christian König <christian.koenig@xxxxxxx>

Going to push this to drm-misc-next.

Hold on. Rajneesh just pointed out a WARN regression from testing. I think the problem is that the bdev->wq is not unbound.

Regards,
  Felix



Thanks,
Christian.

---

Changes in v2:
  - Absorbed the feedback provided by Christian in the commit message and
    the comment.

  drivers/gpu/drm/ttm/ttm_bo.c     | 8 +++++++-
  drivers/gpu/drm/ttm/ttm_device.c | 3 ++-
  2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 5757b9415e37..6f28a77a565b 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -370,7 +370,13 @@ static void ttm_bo_release(struct kref *kref)
              spin_unlock(&bo->bdev->lru_lock);
                INIT_WORK(&bo->delayed_delete, ttm_bo_delayed_delete);
-            queue_work(bdev->wq, &bo->delayed_delete);
+
+            /* Schedule the worker on the closest NUMA node. This
+             * improves performance since system memory might be
+             * cleared on free and that is best done on a CPU core
+             * close to it.
+             */
+            queue_work_node(bdev->pool.nid, bdev->wq, &bo->delayed_delete);
              return;
          }
  diff --git a/drivers/gpu/drm/ttm/ttm_device.c b/drivers/gpu/drm/ttm/ttm_device.c
index 43e27ab77f95..72b81a2ee6c7 100644
--- a/drivers/gpu/drm/ttm/ttm_device.c
+++ b/drivers/gpu/drm/ttm/ttm_device.c
@@ -213,7 +213,8 @@ int ttm_device_init(struct ttm_device *bdev, struct ttm_device_funcs *funcs,
      bdev->funcs = funcs;
        ttm_sys_man_init(bdev);
-    ttm_pool_init(&bdev->pool, dev, NUMA_NO_NODE, use_dma_alloc, use_dma32);
+
+    ttm_pool_init(&bdev->pool, dev, dev_to_node(dev), use_dma_alloc, use_dma32);
        bdev->vma_manager = vma_manager;
      spin_lock_init(&bdev->lru_lock);




[Index of Archives]     [Linux DRI Users]     [Linux Intel Graphics]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]
  Powered by Linux