'no assembly is recorded'

[Aleksey Shipilev]

On 1/15/19 9:52 PM, Сергей Цыпанов wrote:
> Thanks for explanation! And why there's no assembly print-out on C1 level as I see it on C2? Even
> if Array.newInstance is calling into VM right away there eventually should be assembly on the
> lowest level.
I don't understand the question. In c1.txt, the hottest block is this:

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

 <no assembly is recorded, native region>
....................................................................................................
 17,03%  <total for region 1>


It has no assembly, because it is VM native code.

The first "c1, level 1" entry is:

  1,67%         c1, level 1
com.luxoft.logeek.benchmark.array.generated.ArrayInstantiationBenchmark_newInstance_jmhTest::newInstance_avgt_jmhStub,
version 769 (85 bytes)

...which is too cold to be shown, as the line above suggests:

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

Drop hotThreshold if you want to see that 1.67% disassembled, but it would not matter much for
performance.

-Aleksey