'no assembly is recorded'

[Сергей Цыпанов]

Hello Aleksey,

here's the benchmark as I run it

---------------------------------------------------------------------------------------------------

@State(Scope.Thread)
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.NANOSECONDS)
@Fork(jvmArgsAppend = {"-XX:+UseParallelGC", "-Xms2g", "-Xmx2g"})
public class ArrayInstantiationBenchmark {

  @Param({"10"})
  private int length;

  @Benchmark
  public Object newInstance() {
    return Array.newInstance(Object.class, length);
  }
}
---------------------------------------------------------------------------------------------------

and this is runner class 

---------------------------------------------------------------------------------------------------
public class BenchmarkRunner {
  public static void main(String[] args) throws Exception {
    Options opt = new OptionsBuilder()
            .include(ArrayInstantiationBenchmark.class.getSimpleName())
            .warmupIterations(10)
            .warmupTime(TimeValue.seconds(1))
            .measurementIterations(5)
            .measurementTime(TimeValue.seconds(1))
            .forks(1) //0 makes debugging possible
            .shouldFailOnError(true)
            .jvmArgsAppend(
				        "-XX:TieredStopAtLevel=1"
            )
            .addProfiler(LinuxPerfAsmProfiler.class)
            .build();

    new Runner(opt).run();
  }
}
---------------------------------------------------------------------------------------------------


This file contains output for C1: https://yadi.sk/i/DGQuTsvRqdiTkg

This file contains output for C2 (no jvmArgsAppend): https://yadi.sk/i/mT1f6CNuSUjhCw



As one can see C2 file contains assembly, while for C1 we have this:

Hottest code regions (>10,00% "cycles" events):

....[Hottest Region 1]..............................................................................
libjvm.so, InstanceKlass::allocate_objArray (221 bytes) 

 <no assembly is recorded, native region>


Both are executed on Linux home 4.10.0-38-generic #42~16.04.1-Ubuntu SMP Tue Oct 10 16:32:20 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux


15.01.2019, 14:04, "Aleksey Shipilev" <shade at redhat.com>:
> On 1/2/19 8:58 PM, Сергей Цыпанов wrote:
>>  Alas, performance degradation I'm investigating is manifested only on -XX:TieredStopAtLevel=1 and
>>  for c1 there's no assembly captured.
>
> This is weird, C1 assembly should be printed in "perfasm", we do -XX:TieredStopAtLevel=1 all the
> time in JMH experiments. Please show the MCVE?
>
> Thanks,
> -Aleksey