Re: [RFC PATCH] perf_counter: dynamically allocate tasks'perf_counter_context struct

[Ingo Molnar]

* Paul Mackerras <paulus@xxxxxxxxx> wrote:

> Ingo Molnar writes:
> 
> > * Paul Mackerras <paulus@xxxxxxxxx> wrote:
> > 
> > > This replaces the struct perf_counter_context in the task_struct 
> > > with a pointer to a dynamically allocated perf_counter_context 
> > > struct.  The main reason for doing is this is to allow us to 
> > > transfer a perf_counter_context from one task to another when we 
> > > do lazy PMU switching in a later patch.
> > 
> > Hm, i'm not sure how far this gets us towards lazy PMU switching.
> > 
> > In fact i'd say that the term "lazy PMU switching" is probably 
> > misleading, we should use: "equivalent PMU context switching" or 
> > instead.
> 
> Yes, that's what I mean.
> 
> As you say, we need to be able to detect when two tasks have 
> equivalent contexts - that is, when their counters are all 
> inherited from a common ancestor.  My idea is that in that 
> situation we simply swap the contexts: move the context of the 
> outgoing task onto the incoming task, and give the incoming task's 
> context to the outgoing task.  With my patch, that involves simply 
> swapping the pointers over and adjusting the task pointers in the 
> two contexts.
> 
> That means that all the counters get transferred over to the 
> incoming task, so there is nothing in the PMU or the arch code 
> that needs to changed or adjusted.  The outgoing task still has a 
> perfectly valid context, so it doesn't matter if it migrates to 
> another CPU.  The nice thing is that there is nothing special or 
> unusual about the state after we have swapped the contexts - 
> nothing that needs to be remembered or undone later.

Yeah, agreed - your scheme is simpler than the scheme i thought of!

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/