[15/41] drm/i915: Use radixtree to jump start intel_partial_pages()

Submitted by Chris Wilson on Oct. 14, 2016, 12:18 p.m.

Details

Message ID 20161014121833.439-16-chris@chris-wilson.co.uk
State Accepted
Commit d2a84a76a3b970fa32e6eda3d85e7782f831379e
Headers show
Series "Series without cover letter" ( rev: 1 ) in Intel GFX

Browsing this patch as part of:
"Series without cover letter" rev 1 in Intel GFX
<< prev patch [15/41] next patch >>

Commit Message

Chris Wilson Oct. 14, 2016, 12:18 p.m.
We can use the radixtree index of the obj->pages to find the start
position of the desired partial range.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 40 ++++++++++++++++++++++++-------------
 1 file changed, 26 insertions(+), 14 deletions(-)

Patch hide | download patch | download mbox

diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 2bbbda191e93..b3f341fe77bf 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3584,35 +3584,47 @@  intel_partial_pages(const struct i915_ggtt_view *view,
 		    struct drm_i915_gem_object *obj)
 {
 	struct sg_table *st;
-	struct scatterlist *sg;
-	struct sg_page_iter obj_sg_iter;
+	struct scatterlist *sg, *iter;
+	unsigned int count = view->params.partial.size;
+	unsigned int offset;
 	int ret = -ENOMEM;
 
 	st = kmalloc(sizeof(*st), GFP_KERNEL);
 	if (!st)
 		goto err_st_alloc;
 
-	ret = sg_alloc_table(st, view->params.partial.size, GFP_KERNEL);
+	ret = sg_alloc_table(st, count, GFP_KERNEL);
 	if (ret)
 		goto err_sg_alloc;
 
+	iter = i915_gem_object_get_sg(obj,
+				      view->params.partial.offset,
+				      &offset);
+	GEM_BUG_ON(!iter);
+
 	sg = st->sgl;
 	st->nents = 0;
-	for_each_sg_page(obj->pages->sgl, &obj_sg_iter, obj->pages->nents,
-		view->params.partial.offset)
-	{
-		if (st->nents >= view->params.partial.size)
-			break;
+	do {
+		unsigned int len;
 
-		sg_set_page(sg, NULL, PAGE_SIZE, 0);
-		sg_dma_address(sg) = sg_page_iter_dma_address(&obj_sg_iter);
-		sg_dma_len(sg) = PAGE_SIZE;
+		len = min(iter->length - (offset << PAGE_SHIFT),
+			  count << PAGE_SHIFT);
+		sg_set_page(sg, NULL, len, 0);
+		sg_dma_address(sg) =
+			sg_dma_address(iter) + (offset << PAGE_SHIFT);
+		sg_dma_len(sg) = len;
 
-		sg = sg_next(sg);
 		st->nents++;
-	}
+		count -= len >> PAGE_SHIFT;
+		if (count == 0) {
+			sg_mark_end(sg);
+			return st;
+		}
 
-	return st;
+		sg = __sg_next(sg);
+		iter = __sg_next(iter);
+		offset = 0;
+	} while (1);
 
 err_sg_alloc:
 	kfree(st);

Comments

On 14/10/2016 13:18, Chris Wilson wrote:
> We can use the radixtree index of the obj->pages to find the start
> position of the desired partial range.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
>   drivers/gpu/drm/i915/i915_gem_gtt.c | 40 ++++++++++++++++++++++++-------------
>   1 file changed, 26 insertions(+), 14 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index 2bbbda191e93..b3f341fe77bf 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -3584,35 +3584,47 @@ intel_partial_pages(const struct i915_ggtt_view *view,
>   		    struct drm_i915_gem_object *obj)
>   {
>   	struct sg_table *st;
> -	struct scatterlist *sg;
> -	struct sg_page_iter obj_sg_iter;
> +	struct scatterlist *sg, *iter;
> +	unsigned int count = view->params.partial.size;
> +	unsigned int offset;
>   	int ret = -ENOMEM;
>   
>   	st = kmalloc(sizeof(*st), GFP_KERNEL);
>   	if (!st)
>   		goto err_st_alloc;
>   
> -	ret = sg_alloc_table(st, view->params.partial.size, GFP_KERNEL);
> +	ret = sg_alloc_table(st, count, GFP_KERNEL);
>   	if (ret)
>   		goto err_sg_alloc;
>   
> +	iter = i915_gem_object_get_sg(obj,
> +				      view->params.partial.offset,
> +				      &offset);
> +	GEM_BUG_ON(!iter);
> +
>   	sg = st->sgl;
>   	st->nents = 0;
> -	for_each_sg_page(obj->pages->sgl, &obj_sg_iter, obj->pages->nents,
> -		view->params.partial.offset)
> -	{
> -		if (st->nents >= view->params.partial.size)
> -			break;
> +	do {
> +		unsigned int len;
>   
> -		sg_set_page(sg, NULL, PAGE_SIZE, 0);
> -		sg_dma_address(sg) = sg_page_iter_dma_address(&obj_sg_iter);
> -		sg_dma_len(sg) = PAGE_SIZE;
> +		len = min(iter->length - (offset << PAGE_SHIFT),
> +			  count << PAGE_SHIFT);
> +		sg_set_page(sg, NULL, len, 0);
> +		sg_dma_address(sg) =
> +			sg_dma_address(iter) + (offset << PAGE_SHIFT);
> +		sg_dma_len(sg) = len;
>   
> -		sg = sg_next(sg);
>   		st->nents++;
> -	}
> +		count -= len >> PAGE_SHIFT;
> +		if (count == 0) {
> +			sg_mark_end(sg);
> +			return st;
> +		}
>   
> -	return st;
> +		sg = __sg_next(sg);
> +		iter = __sg_next(iter);
> +		offset = 0;
> +	} while (1);
>   
>   err_sg_alloc:
>   	kfree(st);

Looks OK and I trust you when you say it is faster.

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Regards,

Tvrtko