Re: [Mesa-dev] [PATCH 7/8] radeonsi: implement AMD_performance_monitor for CIK+

2015-11-25 Thread Nicolai Hähnle
On 25.11.2015 14:56, Marek Olšák wrote: R600_PC_SHADER_WINDOWING seems unused. What is it good for? Ah yes, that's totally non-obvious in the code, sorry about that. It makes query->shader non-zero, which will cause SQ_PERFCOUNTER_CTRL to be written. I've added a short comment about this.

Re: [Mesa-dev] [PATCH 7/8] radeonsi: implement AMD_performance_monitor for CIK+

2015-11-25 Thread Marek Olšák
R600_PC_SHADER_WINDOWING seems unused. What is it good for? BTW, I think the EVENT_WRITE_EOP workaround is indeed required. The series is: Reviewed-by: Marek Olšák Marek On Tue, Nov 24, 2015 at 5:00 PM, Nicolai Hähnle wrote: > Expose most of the performance counter groups that are exposed by

[Mesa-dev] [PATCH 7/8] radeonsi: implement AMD_performance_monitor for CIK+

2015-11-24 Thread Nicolai Hähnle
Expose most of the performance counter groups that are exposed by Catalyst. Ideally, the driver will work with GPUPerfStudio at some point, but we are not quite there yet. In any case, this is the reason for grouping multiple instances of hardware blocks in the way it is implemented. The counters