JMH and ocperf

[Aleksey Shipilev]

On 11/22/2016 05:15 PM, Aleksey Shipilev wrote:
>  6) Submit for further review.

In private communication, Sergey indicated that ocperf works even with the
initial patch, I pushed it under:
 https://bugs.openjdk.java.net/browse/CODETOOLS-7901842

However, the symbolic mapping looks not entirely complete. perfasm tries to
match the symbolic names given in "events" string to what perf report answers.
But are we having that mapping in perf report? Because it shows zero events with
e.g.:

-prof perfasm:event=cycles,br_inst_exec.all_branches

...

Secondary result "·asm":
PrintAssembly processed: 126267 total address lines.
Perf output processed (skipped 5.187 seconds):
 Column 1: cycles (8794 events)
 Column 2: br_inst_exec.all_branches (0 events)

Sergey, can you please follow up on this?

Thanks,
-Aleksey