On Tue, 14 Jun 2016, Nicolas Pitre wrote:
> On Tue, 14 Jun 2016, Daniel Lezcano wrote:
> > If the storage is a bit larger (let's say 16 values) and there is no memset
> > and the sum is not computed, at least we need a count for the number of
> > values
> > in the array before this one is fulfilled, otherwise the statistics will be
> > wrong as we will take into account the entire array with old values, no ?
>
> The point is not to change from 8 to 16 entries, but to store raw 64-bit
> timestamps instead of computed 32-bit deltas. Whether or not those
> timestamps are too far apart and discarded can be done at idle entry
> time.
Correct, and you don't have to know how many timestamps are in the array
simply because if it is cleared at init time, then any not yet set value will
create a large gap, which you filter out.
The point is to make the fast path overhead as small as possible. And if
that's just a store and index increment, then it can be inline and not a
function call.
Thanks,
tglx
