On 30 September 2014 12:02, Ola Liljedahl <[email protected]> wrote:
> I updated Yogesh's document with some paragraphs at the end: > > https://docs.google.com/a/linaro.org/document/d/1dN0jy61VvoLjStd1KTD1N3hViT1NJVZdztnM7i2HmXU > > I think we need the input from some folks more accustomed to upstreaming > kernel changes. What do they think of our ideas that it should be possible > to exclusively reserve PMU counters using the existing kernel mechanisms > (e.g. perf_event_open) and read the PMU counter directly from user space > (possibly using a kernel-exported vDSO or some other standard mechanism). > Extend and enhance existing kernel facilities instead of duplicating the > functionality in user space. Co-exist with other usages of the PMU > resources (they will just have fewer PMU counters to use). > Agree, but we need to solidify what it is we propose first I think. If we get the vDSO overhead metrics out shortly (1-2 weeks ?), maybe we scratch together a patch that reserves the counters for use with vDSO and take that patch for comment within the Linaro Kernel community as a kickoff. > -- Ola > > > > _______________________________________________ > lng-odp mailing list > [email protected] > http://lists.linaro.org/mailman/listinfo/lng-odp > > -- *Mike Holmes* Linaro Sr Technical Manager LNG - ODP
_______________________________________________ lng-odp mailing list [email protected] http://lists.linaro.org/mailman/listinfo/lng-odp
