[PATCH] R600/SI: SI Control Flow Annotation bug fixed

[Michel Dänzer]

On 04.06.2014 06:45, Matt Arsenault wrote:
> On 06/03/2014 02:16 AM, Christian König wrote:
>> Am 03.06.2014 10:59, schrieb Matt Arsenault:
>>> On Jun 3, 2014, at 1:46 AM, Елена Денисова
>>> <denisova.lena at gmail.com> wrote:
>>> 
>>>> <0001-SI-Control-Flow-Annotation-bug-fix.patch>
>>> Needs a test
>> 
>> The control flow leading to this issue is rather complicated, and 
>> because of this I'm not sure if we want to have a test case for it 
>> cause it would break quite often on completely unrelated changes.
>> 
>> On the other hand I know that Lena doesn't have an environment to
>> run piglit tests with this, so somebody (Tim, Matt?) should do this
>> to make sure we don't break anything there.
>> 
>> [...]
> 
> [...] I can probably try running piglit on this tomorrow (although
> for me ~20 tests are failing already on trunk. Are those currently
> expected?)

Yes. The main purpose of piglit is to make sure a change doesn't cause
any regressions, by running piglit without and with the change and
comparing the results (by passing both results to the piglit summary
command).

I just gave this patch a spin on my Kaveri, doesn't cause any
regressions (nor fix any tests) for the gpu.py and quick_cl.py test sets.

Tested-by: Michel Dänzer <michel.daenzer at amd.com>


-- 
Earthling Michel Dänzer            |                  http://www.amd.com
Libre software enthusiast          |                Mesa and X developer