JIT compiler information
Jookalook
vjucas2 at gmail.com
Fri Jun 25 08:13:05 PDT 2010
The main problem is that when a method is denied compilation, we have no way
of knowing which methods were denied and the reason for them being denied.
If there is a flag or tool out there that gives us this information, we
would then be able to look back at certain methods and see if we can make
them more suitable for compilation, or adjust some of the compilation flags
to encompass more of the methods that get denied.
If this sort of thing doesn't exist, then does anyone have any good ideas on
how to make this information available.
Ulf Zibis-2 wrote:
>
> Am 23.06.2010 21:33, schrieb Jookalook:
>> LogCompilation gives a great overview about which methods get compiled
>> and if
>> they were successful and such, but it doesn't tell you why a method
>> doesn't
>> pass the test to even be considered for compilation. Basically, I am
>> trying
>> to figure out why some methods in my codebase, that I know are called
>> very
>> often, are not being compiled. Is there some way of seeing information
>> for
>> every method that is called in the JVM, not just the ones that pass the
>> compilation requirements. That way I would be able to see the specifics
>> on
>> why a method is not chosen for compilation.
>>
>
> For compilation I can't tell you, but for inlining a special case can
> occur:
>
> At the time, where method a, which calls method b, becomes compiled, the
> threshold for inlining b might not have been reached.
> If later the threshold for inlining b theoretically will be reached,
> there is kinda bad chance to become inlined then. But in extrem cases
> there is some chance to become inlined later by result of the HotSpot
> profiler.
> This case may be missed by the HotSpot logger ???
>
> If I'm wrong, someone may correct me.
>
> -Ulf
>
>
>>
>> Remi Forax wrote:
>>
>>> Le 21/06/2010 23:12, Viktor Jucas a écrit :
>>> why:
>>> http://wikis.sun.com/display/HotSpotInternals/LogCompilation+overview
>>>
>>> how:
>>> http://wikis.sun.com/display/HotSpotInternals/PrintAssembly
>>>
>>> cheers,
>>> Rémi
>>>
>>>
>>>
>>
>
>
>
--
View this message in context: http://old.nabble.com/JIT-compiler-information-tp28953898p28993401.html
Sent from the OpenJDK Hotspot Compiler Development List mailing list archive at Nabble.com.
More information about the hotspot-compiler-dev
mailing list