On 13/10/2016 10:23, Chris Wilson wrote:
On Thu, Oct 13, 2016 at 10:04:03AM +0100, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin <tvrtko.ursu...@intel.com>

*** BROKEN *** RFC only *** BROKEN ***

In cases where it is hard to know to minimum numbers of ents
table will need to hold at sg_alloc_time, and we end up with
a table with unused sg entries at its end, this function will
trim (free) the unused sg entry blocks and adjust the
table->orig_nents down.
An alternative I had in mind, was to allocate the new chunks
iteratively, i.e. as we run out of room in i915_sg_add_page.

Thought about it but did not like having to handle another source of failures. Hm, don't know. Will think about it.



Intel-gfx mailing list

Reply via email to