Re: [RFC][PATCH] perf: Rewrite core context handling

[Alexey Budankov <alexey.budankov@linux.intel.com>]

Hi Alex,

On 17.10.2018 18:01, Alexander Shishkin wrote:
> Alexey Budankov <alexey.budankov@linux.intel.com> writes:
> 
>> Since it reduces to single cpu context (and single task context) at all times, 
>> ideally, it would probably be coded as simple as this: 
>>
>> 	perf_rotate_context()
>> 	{
>>             cpu = this_cpu_ptr(&cpu_context)
>>             for_every_pmu(pmu, cpu)
>>                     for_every_event_ctx(event_ctx, pmu)
>> 	                    rotate(event_ctx, pmu)
>> 	}
>>
>> so rotate(event_ctx, pmu) would operate on common events objects semantics 
>> and memory layout, and PMU specific code handle SW/HW programming differences.
> 
> Ok, what's event_ctx and how does that simplify things?

Currently, rotate_ctx() is called twice from perf_rotate_context() 
for cpu and task contexts:

struct perf_cpu_context {
	struct perf_event_context	ctx;
	struct perf_event_context	*task_ctx;

If it would be implemented in a loop that could, probably, reduce 
complexity of perf_rotate_context(), partly pushing the complexity 
*down* to SW/HW pmu specific code and perf_rotate_context() would 
become scalable for any number of contexts.

Thanks,
Alexey

> 
> Regards,
> --
> Alex
>