On Tue, Dec 15, 2009 at 7:58 PM, Donnie Fang <donnie.f...@gmail.com> wrote:
> Hmm, in function ttm_bo_mem_space:
> for (i = 0; i <= placement->num_busy_placement; ++i) {
> ret = ttm_mem_type_from_flags(placement->placement[i],
> &mem_type);
> Is the "placement->placement[i]" typo error or something else? I think it
> should be "busy_placement"
The patch to fix this was sent to the list yesterday.
Dave.
>
> 2009/12/15 Jerome Glisse <gli...@freedesktop.org>
>>
>> On Tue, Dec 15, 2009 at 02:04:01PM +0800, Donnie Fang wrote:
>> > Hi Jerome,
>> > The field busy_placement in struct ttm_placement hasn't been used
>> > yet. what is this field exact usage? Do you mean TTM core can reference
>> > this
>> > field when do bo evict? But i can't find out any clue about this.
>> >
>>
>> This field purpose was explain in a mail on dri-devel. It's used
>> when we need to evict buffer to allow driver to choose alternates
>> placements order in this case. For instance:
>> placement= 1 vram, 2 gtt
>> busy_placement=1 gtt, 2 vram
>>
>> If there is not enough place in vram or gtt, ttm will start evicting
>> bo from gtt and put the bo their, if after evicting from gtt it still
>> hasn't enough place it will evict from vram (2 in busy_placement) and
>> try there.
>>
>> Cheers,
>> Jerome
>
>
> ------------------------------------------------------------------------------
> Return on Information:
> Google Enterprise Search pays you back
> Get the facts.
> http://p.sf.net/sfu/google-dev2dev
>
> --
> _______________________________________________
> Dri-devel mailing list
> Dri-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/dri-devel
>
>
------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel
