[Bug 773546] leak in ensure_profiles (double free or corruption)

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=773546

--- Comment #24 from Florent Thiéry <florent.thiery at ubicast.eu> ---
Here are my results with and without this patch on SKL/GT2 (i3-6100U) of
vaapih264enc 1080p encoding performance (NV12); to me there is no major
performance regression.

+-----------+---------+----------+------------+-----------+---------+
|   tune    | pattern | encoders | fps before | fps after |  delta  |
+-----------+---------+----------+------------+-----------+---------+
| none      | black   |        1 |        107 |       107 | 0,00%   |
| low-power | black   |        1 |        196 |       176 | -10,20% |
| none      | black   |        2 |        133 |       138 | 3,76%   |
| low-power | black   |        2 |        198 |       195 | -1,52%  |
| none      | black   |        3 |        141 |       141 | 0,00%   |
| low-power | black   |        3 |        198 |       197 | -0,51%  |
| none      | smpte   |        1 |        101 |        98 | -2,97%  |
| low-power | smpte   |        1 |        186 |       187 | 0,54%   |
| none      | smpte   |        2 |        128 |       129 | 0,78%   |
| low-power | smpte   |        2 |        186 |       190 | 2,15%   |
| none      | smpte   |        3 |        131 |       133 | 1,53%   |
| low-power | smpte   |        3 |        189 |       190 | 0,53%   |
| none      | snow    |        1 |        103 |       104 | 0,97%   |
| low-power | snow    |        1 |        119 |       118 | -0,84%  |
| none      | snow    |        2 |        130 |       131 | 0,77%   |
| low-power | snow    |        2 |        126 |       126 | 0,00%   |
| none      | snow    |        3 |        137 |       136 | -0,73%  |
| low-power | snow    |        3 |        121 |       120 | -0,83%  |
+-----------+---------+----------+------------+-----------+---------+

Here are the results on SKL/GT3e (i5-6260U)

+-----------+---------+----------+------------+-----------+--------+
|   tune    | pattern | encoders | fps before | fps after | delta  |
+-----------+---------+----------+------------+-----------+--------+
| none      | black   |        1 |        115 |       113 | -1,74% |
| low-power | black   |        1 |        269 |       266 | -1,12% |
| none      | black   |        2 |        130 |       134 | 3,08%  |
| low-power | black   |        2 |        230 |       222 | -3,48% |
| none      | black   |        3 |        138 |       137 | -0,72% |
| low-power | black   |        3 |        220 |       236 | 7,27%  |
| none      | smpte   |        1 |        105 |       107 | 1,90%  |
| low-power | smpte   |        1 |        235 |       237 | 0,85%  |
| none      | smpte   |        2 |        121 |       127 | 4,96%  |
| low-power | smpte   |        2 |        235 |       239 | 1,70%  |
| none      | smpte   |        3 |        127 |       129 | 1,57%  |
| low-power | smpte   |        3 |        203 |       209 | 2,96%  |
| none      | snow    |        1 |        109 |       110 | 0,92%  |
| low-power | snow    |        1 |        110 |       108 | -1,82% |
| none      | snow    |        2 |        128 |       129 | 0,78%  |
| low-power | snow    |        2 |        112 |       112 | 0,00%  |
| none      | snow    |        3 |        131 |       132 | 0,76%  |
| low-power | snow    |        3 |        110 |       110 | 0,00%  |
+-----------+---------+----------+------------+-----------+--------+

Note that this is a single run per test because of the leak. I think this patch
does not harm performance (at least not on SKL)

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.