HotSpot compilation latency

[Liu Xin]

Hi, Stefan,
There're 2 counters: invocation and back-edge branch.  check out
methodData.hpp.
  InvocationCounter _invocation_counter;
  InvocationCounter _backedge_counter;

Don't you think it's easier to count discrete events than measure elapsed
time?
Indeed,  time is more accurate profile data in runtime. My guess is that
the instrumentation code will be very expensive.

@Tobias,
if I use an external profiler to measure runtime of methods, is that
possible JIT or AOT compilers can make use those profile data?

thanks,
--lx




On Mon, Aug 12, 2019 at 7:39 AM Stefan Reich <
stefan.reich.maker.of.eye at googlemail.com> wrote:

> Huh. So the compilation results (which method is compiled and when in the
> program flow) should typically not very between machines of different
> speed.
>
> (For a single-threaded program.)
>
> That's interesting, I always thought there was a time component involved
> ("profile for 100 ms" or something). Optimization makes more sense without
> it I guess.
>
> Thanks
>
> On Mon, 12 Aug 2019 at 14:52, Tobias Hartmann <tobias.hartmann at oracle.com>
> wrote:
>
> >
> > On 12.08.19 14:41, Stefan Reich wrote:
> > > So none of this is dependent on clock time at all?
> >
> > Yes, exactly.
> >
> > Best regards,
> > Tobias
> >
>
>
> --
> Stefan Reich
> BotCompany.de // Java-based operating systems
>