[Intel-gfx] [PATCH v4 3/3] drm/i915: optimise i915_gem_object_vmap_range() for small objects

Chris Wilson chris at chris-wilson.co.uk
Tue Feb 23 10:16:16 UTC 2016


On Mon, Feb 22, 2016 at 03:18:28PM +0000, Dave Gordon wrote:
> Now that we use this function for ringbuffers and other "small" objects,
> it's worth avoiding an extra kmalloc()/kfree() cycle if the page array
> is small enough to put on the stack. Here we've chosen an arbitrary
> cutoff of 32 (4k) pages, which is big enough for a ringbuffer (4 pages)
> or a context image (currently up to 22 pages).
> 
> Signed-off-by: Dave Gordon <david.s.gordon at intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> Cc: Alex Dai <yu.dai at intel.com>
> Cc: Chris Wilson <chris at chris-wilson.co.uk>
> ---
>  drivers/gpu/drm/i915/i915_gem.c | 18 +++++++++++++-----
>  1 file changed, 13 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 14942cf..effb69b 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -2422,6 +2422,7 @@ void *i915_gem_object_vmap_range(struct drm_i915_gem_object *obj,
>  				 unsigned int npages)
>  {
>  	struct sg_page_iter sg_iter;
> +	struct page *mypages[32];

Use stack or stack_pages, that's the pattern we've used elsewhere.
Though pages_on_stack would be more in line with the rest of the kernel.

>  	struct page **pages;
>  	void *addr;
>  	int i;
> @@ -2431,10 +2432,16 @@ void *i915_gem_object_vmap_range(struct drm_i915_gem_object *obj,
>  		return NULL;
>  	}
>  
> -	pages = drm_malloc_ab(npages, sizeof(*pages));
> -	if (pages == NULL) {
> -		DRM_DEBUG_DRIVER("Failed to get space for pages\n");
> -		return NULL;
> +	if (npages <= ARRAY_SIZE(mypages))
> +		pages = mypages;

If one branch requires braces, add them to all.

> +	else {
> +		pages = kmalloc(npages*sizeof(*pages), GFP_TEMPORARY | __GFP_NOWARN);

drm_malloc_gfp

Otherwise, seems an ok idea. The deepest vmap is the
ioctl/execbuffer/cmdparser to which using a further 256 bytes of stack is
acceptable.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre


More information about the Intel-gfx mailing list