Messages in this thread | | | Date | Wed, 19 Aug 2009 12:58:18 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH 4/4][RFC] perf_counter: Allow sharing of output channels |
| |
* Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
> Provide the ability to configure a counter to send its output to > another (already existing) counter's output stream. > > [ compile tested only ]
very nice!
two comments:
> Index: linux-2.6/include/linux/perf_counter.h > =================================================================== > --- linux-2.6.orig/include/linux/perf_counter.h > +++ linux-2.6/include/linux/perf_counter.h > @@ -216,6 +216,7 @@ struct perf_counter_attr { > #define PERF_COUNTER_IOC_REFRESH _IO ('$', 2) > #define PERF_COUNTER_IOC_RESET _IO ('$', 3) > #define PERF_COUNTER_IOC_PERIOD _IOW('$', 4, u64) > +#define PERF_COUNTER_IOC_SET_OUTPUT _IO ('$', 5)
Time to add a new sys_perf_counter_chattr() syscall and deprecate the ioctls?
> @@ -415,6 +416,9 @@ enum perf_callchain_context { > PERF_CONTEXT_MAX = (__u64)-4095, > }; > > +#define PERF_FLAG_FD_NO_GROUP (1U << 0) > +#define PERF_FLAG_FD_OUTPUT (1U << 1)
Why not extend the size of perf_counter_attr and add an output_fd parameter? Zero would mean no fd (and this is also backwards compatible behavior).
FD_NO_GROUP is a bit unclean API as it aliases group_fd to two purposes: the real group_fd and this new output_fd. I think we should move output_fd to the attribute structure.
That way one could also create a group counter which shares the output channel with another counter, in the same perf_counter_open() call.
Ingo
| |