Issues with dtraceasm, and a workaround

Henri Tremblay henri.tremblay at gmail.com
Tue Feb 23 14:42:28 UTC 2021 

Seems to work perfectly. As I said, I can't get hsdis to work but I see the
normal address output so it looks good to me.
I tried with 1.8.0_275, 11.0.9 and 16-ea-30. I also tried GraalVM CE
20.1.0. It works as well. The funny part is that it resolves the addresses
a bit more. Here are the outputs:

Hotspot:

Perf output processed (skipped 1.364 seconds):

 Column 1: sampled_pc (837 events)


Hottest code regions (>10.00% "sampled_pc" events):


....[Hottest Region
1]..............................................................................

 0x107740320, [unknown] (18 bytes)


 <no assembly is recorded, native region>

....................................................................................................

 99.40%  <total for region 1>


....[Hottest
Regions]...............................................................................

 99.40%   0x107740320  [unknown] (18 bytes)

  0.24%   0x10588fcd8  [unknown] (0 bytes)

  0.12%   0x10551e39e  [unknown] (0 bytes)

  0.12%   0x105888c4e  [unknown] (0 bytes)

  0.12%   0x1058907fb  [unknown] (0 bytes)

....................................................................................................

100.00%  <totals>


....[Hottest Methods (after
inlining)]..............................................................

 99.40%   0x107740320  [unknown]

  0.24%   0x10588fcd8  [unknown]

  0.12%   0x105888c4e  [unknown]

  0.12%   0x1058907fb  [unknown]

  0.12%   0x10551e39e  [unknown]

....................................................................................................

100.00%  <totals>


....[Distribution by
Source]........................................................................

 99.40%   0x107740320

  0.24%   0x10588fcd8

  0.12%   0x105888c4e

  0.12%   0x10551e39e

  0.12%   0x1058907fb

....................................................................................................

100.00%  <totals>


WARNING: The perf event count is suspiciously low (837). The performance
data might be

inaccurate or misleading. Try to do the profiling again, or tune up the
sampling frequency.

With some profilers on Mac OS X, System Integrity Protection (SIP) may
prevent profiling.

In such case, temporarily disabling SIP with 'csrutil disable' might help.



# Run complete. Total time: 00:00:02


REMEMBER: The numbers below are just data. To gain reusable insights, you
need to follow up on

why the numbers are the way they are. Use profilers (see -prof, -lprof),
design factorial

experiments, perform baseline and negative tests that provide experimental
control, make sure

the benchmarking environment is safe on JVM/OS/HW level, ask for reviews
from the domain experts.

Do not assume the numbers tell you what you want them to tell.


Benchmark                                     Mode  Cnt           Score
Error  Units

JMHSample_01_HelloWorld.wellHelloThere       thrpt       3960115374.548
      ops/s

JMHSample_01_HelloWorld.wellHelloThere:·asm  thrpt                  NaN
        ---

GraalVM:

Hottest code regions (>10.00% "sampled_pc" events):


....[Hottest Region
1]..............................................................................

 0x11e7581c0, [unknown] (20 bytes)


 <no assembly is recorded, native region>

....................................................................................................

 98.19%  <total for region 1>


....[Hottest
Regions]...............................................................................

 98.19%                0x11e7581c0  [unknown] (20 bytes)

  0.36%               libjvm.dylib  outputStream::update_position(char
const*, unsigned long) (3 bytes)

  0.12%               libjvm.dylib  SpinPause (0 bytes)

  0.12%               libjvm.dylib  LIRGenerator::do_getClass(Intrinsic*)
(0 bytes)

  0.12%               libjvm.dylib
G1BarrierSetRuntime::write_ref_array_post_entry(HeapWord*,
unsigned long) (0 bytes)

  0.12%               libjvm.dylib  jni_GetArrayLength (0 bytes)

  0.12%               libjvm.dylib  Arena::Amalloc(unsigned long,
AllocFailStrategy::AllocFailEnum) (0 bytes)

  0.12%               libjvm.dylib  nmethod::print_nmethod(bool) (0 bytes)

  0.12%               libjvm.dylib
AccessInternal::PostRuntimeDispatch<G1BarrierSet::AccessBarrier<802934ull,
G1BarrierSet>, (AccessInternal::BarrierType)1,
802934ull>::oop_access_barrier(oopDesc*, long, oopDesc*) (0 bytes)

  0.12%               libjvm.dylib  VMThread::execute(VM_Operation*) (0
bytes)

  0.12%                0x116c71ca2  [unknown] (0 bytes)

  0.12%          libsystem_c.dylib  DYLD-STUB$$pthread_mutex_unlock (0
bytes)

  0.12%   libsystem_platform.dylib  _platform_strlen (0 bytes)

....................................................................................................

 99.88%  <totals>


....[Hottest Methods (after
inlining)]..............................................................

 98.19%                0x11e7581c0  [unknown]

  0.36%               libjvm.dylib  outputStream::update_position(char
const*, unsigned long)

  0.12%               libjvm.dylib
AccessInternal::PostRuntimeDispatch<G1BarrierSet::AccessBarrier<802934ull,
G1BarrierSet>, (AccessInternal::BarrierType)1,
802934ull>::oop_access_barrier(oopDesc*, long, oopDesc*)

  0.12%               libjvm.dylib  VMThread::execute(VM_Operation*)

  0.12%               libjvm.dylib  Arena::Amalloc(unsigned long,
AllocFailStrategy::AllocFailEnum)

  0.12%                0x116c71ca2  [unknown]

  0.12%               libjvm.dylib  jni_GetArrayLength

  0.12%               libjvm.dylib  nmethod::print_nmethod(bool)

  0.12%               libjvm.dylib  SpinPause

  0.12%               libjvm.dylib  LIRGenerator::do_getClass(Intrinsic*)

  0.12%          libsystem_c.dylib  DYLD-STUB$$pthread_mutex_unlock

  0.12%   libsystem_platform.dylib  _platform_strlen

  0.12%               libjvm.dylib
G1BarrierSetRuntime::write_ref_array_post_entry(HeapWord*,
unsigned long)

....................................................................................................

 99.88%  <totals>


....[Distribution by
Source]........................................................................

 98.19%                0x11e7581c0

  1.33%               libjvm.dylib

  0.12%                0x116c71ca2

  0.12%   libsystem_platform.dylib

  0.12%          libsystem_c.dylib

....................................................................................................

 99.88%  <totals>

On Tue, 23 Feb 2021 at 06:02, Aleksey Shipilev <shade at redhat.com> wrote:

> On 2/23/21 5:27 AM, Henri Tremblay wrote:
> > Yes. This looks fine:
> >
> > Collection<String> messages
> =Utils.tryWith("sudo","kill","-TERM",Long.toString(dtraceProcess.pid()));
> > if (!messages.isEmpty()) {
> >      throw new IllegalStateException(messages.toString());
> > }
> >
> > However, Paul is right, the pid() method doesn't exist so you will need
> something like this:
>
> Yes. In fact, there is already a similar code in the Utils that does
> similar thing. The additional
> wrinkle is that the code is supposed to compile and run on JDK 7 and JDK
> 8, which requires a bit
> more fiddling.
>
> I have updated the PR with the latest version, please try it?
>    https://github.com/openjdk/jmh/pull/22
>
> --
> Thanks,
> -Aleksey
>
>

More information about the jmh-dev mailing list