On 11.07.2018 18:22, Maxime Villard wrote:
> Right now we have three (or more?) different implementations for
> Performance
> Monitoring Counters:
>  * PMC: this one is MI. It is used only on one ARM model (xscale I think).
>    There used to be an x86 code for it, but it was broken, and I removed
> it.
>    The implementation comes with libpmc, a library we provide. The code
>    hasn't moved these last 15 years. I don't like this implementation,
> it is
>    really invasive (see the numerous pmc.h files that are all empty).
>  * X86PMC: this one is MD, and only available for x86. I wrote it myself.
>    The code is small (x86/pmc.c), and functional. The PMCs are system-wide,
>    and retrieved on a per-cpu basis. But this implementation does not
>    support tracking, that is, we get numbers (about the cache misses for
>    example), but we don't know where they happened.
>  * TPROF: this one is MI, but only x86 support is present. TPROF provides
>    the backend needed to support tracking: via a device, that userland can
>    read from, in order to absorb the event samples produced by the kernel.
>    The backend is pretty good, but the frontend (where the user chooses
>    which PMC etc) is inexistent - the CPU/event detection is not there
>    either. The backend is MI (/dev/tprof/tprof.c), and can be used on other
>    architectures. The module already exists to dynamically modload.
> I think it would be good to:
>  * Remove PMC entirely. Then remove libpmc too.
>  * Merge X86PMC into the x86 part of TPROF. That is to say, into
>    x86/tprof_*. Then remove X86PMC.
>  * Later, maybe, someone will want to add other architectures in TPROF,
> like
>    all the recent ARMs.
> Maxime

I'm not familiar with the internals myself, but from API point of view,
something usable for porting rr (https://github.com/mozilla/rr) or even
Linux perf-top is highly desirable. I treat personally perf-top as a
gold standard.

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to