Re: Memory measurement with Async Profiler

[Andrei Pangin]

¯\_(ツ)_/¯

Just don't know how it should be used.

вторник, 6 февраля 2018 г., 14:54:48 UTC+3 пользователь Maxim Terletsky 
написал:
>
> One more question - why the original AsyncProfiler does not use tlab_size 
> in this method we talking about?
>
> https://github.com/jvm-profiling-tools/async-profiler/blob/master/src/allocTracer.cpp#L65
>

-- 
You received this message because you are subscribed to the Google Groups 
"mechanical-sympathy" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to mechanical-sympathy+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Memory measurement with Async Profiler

[Maxim Terletsky]

One more question - why the original AsyncProfiler does not use tlab_size 
in this method we talking about?
https://github.com/jvm-profiling-tools/async-profiler/blob/master/src/allocTracer.cpp#L65

-- 
You received this message because you are subscribed to the Google Groups 
"mechanical-sympathy" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to mechanical-sympathy+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: JITed very simple method.

[Aleksey Shipilev]

On 02/05/2018 09:58 PM, John Hening wrote:
> 1. Why do you find it is a bug in tierred compilation machinery? It is not on 
> my eye.  The
> compilation level is: C2, level 4

Wait a minute, where exactly does it say "C2, level 4" for you?

I am saying the disassembly you have provided in the original message is 
probably level 2/3, and if
it is hot, that might be a bug in tiered compilation machinery: we are not 
supposed to spend a lot
of time in methods with profiling enabled.

But you haven't verified that disassembly is actually on the hot path. Level 4 
code would be on
hotpath, and it would be without profiling.

> 2. I have still a doubt: why profiling counter is not synchronized. What if 2 
> or more threads
> executing a function

That is a race, so profile is not very accurate. We (mostly) do not care about 
that: there is a
tradeoff between profiling overhead and profile accuracy. (Yes, I know racy 
updates are quirky and
potentially lose the unbounded number of updates).

-Aleksey

-- 
You received this message because you are subscribed to the Google Groups 
"mechanical-sympathy" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to mechanical-sympathy+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


signature.asc
Description: OpenPGP digital signature