Is graal VM going to be part of a future JDK release?

Fri Mar 13 23:01:25 UTC 2015

Awesome! Ok, now need to find some time to play with it.

Thanks Thomas

sent from my phone
On Mar 13, 2015 6:42 PM, "Thomas Wuerthinger" <thomas.wuerthinger at oracle.com>
wrote:

> Vitaly,
>
> It does. Running your example with -XX:+UnlockDiagnosticVMOptions
> -XX:+PrintAssembly and Graal’s OTN binary release, I get the following
> assembly snippets:
>
> ...
>   0x0000000106516d74: vmovsd 0x10(%rdx,%rbx,8),%xmm0  ;*daload
>                                                 ; - Test::calc at 12 (line
> 21)
> ...
>   0x0000000106516d86: vmovsd 0x10(%rcx,%rbx,8),%xmm1  ;*daload
>                                                 ; - Test::calc at 15 (line
> 21)
> ...
>   0x0000000106516d8c: vmulsd %xmm1,%xmm0,%xmm0
> ...
>   0x0000000106516d9c: vmovsd %xmm0,0x10(%rsi,%rbx,8)  ;*dastore
> …
>
> We are still working on generalising the set of recognized patterns and
> creating the optimal machine code representation for the target
> architecture once we understand the vector operation structure.
>
> - thomas
>
>
> On 13 Mar 2015, at 17:32, Vitaly Davidovich <vitalyd at gmail.com> wrote:
>
> Thomas,
>
> AFAIK, graal has some intrinsics that use vector instructions (I think
> array copy/equals?), but does it also auto-vectorize, say, arithmetic?
>
> for (int i = 0; i < some_trip_count; ++i)
>     a[i] = b[i] * c[i];
>
> Does it unroll and vectorize that?
>
> Thanks
>
>
> On Fri, Mar 13, 2015 at 12:07 PM, Thomas Wuerthinger <
> thomas.wuerthinger at oracle.com> wrote:
>
>> Vitaly,
>>
>> We are equally committed to Java performance as we are to Truffle guest
>> language performance. While we are currently making the largest difference
>> for JavaScript, Ruby, and R, we expect to also make a significant
>> difference for Java in the future. The mentioned OTN release of Graal
>> already includes auto-vectorisation features. We have several larger areas
>> where we try to further push the current limits of Java performance.
>>
>> - thomas
>>
>>
>> On 13 Mar 2015, at 16:38, Vitaly Davidovich <vitalyd at gmail.com> wrote:
>>
>> Thanks Thomas.
>>
>> I've been lurking on this mailing list for a while, so sort of have a big
>> picture view of what's happening.  However, as most conversations here tend
>> to center around guest languages, I wasn't quite sure what the "value-add",
>> so to speak, was for plain old java code.
>>
>> Regarding (c) in your email, yes, the somewhat naive inlining policy is
>> annoying (e.g. assert statements possibly screwing with inlining
>> decisions).  However, in C2, there's also a different inlining threshold
>> for very hot code (i.e. FreqInlineSize flag), which by default is 325 or
>> so.  I believe there's also ongoing work on incremental inlining.  In light
>> of that, does graal's late inlining policy still help somewhere? Just
>> curious.
>>
>> Do you guys have any plans to support some things that C2 currently
>> lacks, such as autovectorization of arithmetic/logic/etc operations on,
>> say, arrays? That would be pretty sweet.  Or is the (near to mid-term, at
>> least) focus going to remain on supporting guest languages?
>>
>> Thanks
>>
>> On Fri, Mar 13, 2015 at 11:29 AM, Thomas Wuerthinger <
>> thomas.wuerthinger at oracle.com> wrote:
>>
>>> Vitaly,
>>>
>>> Graal has a very different design from C2 and therefore it has distinct
>>> performance characteristics. We invite developers to try our latest OTN
>>> release [1] on their applications and give us feedback.
>>>
>>> The largest three distinguishing factors are:
>>> a) Guard optimizations for efficient handling of speculative code - this
>>> has its biggest impact on the JavaScript, Ruby [2], and R [3]
>>> implementations built on Graal’s multi-language framework Truffle.
>>> b) Partial escape analysis to reduce object allocations even if the
>>> object escapes on some paths.
>>> c) Improved inlining strategy that explores the call tree and performs
>>> late inlining as default instead of inlining during bytecode parsing.
>>>
>>> We expect to publish a new improved version of Graal in another OTN
>>> release in the next couple of months.
>>>
>>> Regards, thomas
>>>
>>> [1]
>>> http://www.oracle.com/technetwork/oracle-labs/program-languages/overview/index-2301583.html
>>>
>>> [2] https://github.com/jruby/jruby/wiki/Truffle
>>> [3] https://bitbucket.org/allr/fastr
>>>
>>> On 13 Mar 2015, at 16:01, Vitaly Davidovich <vitalyd at gmail.com> wrote:
>>>
>>> Chris (or anyone else really),
>>>
>>> Are there any optimizations that graal does better (or at all) than
>>> Hotspot's C2? I know it has flow sensitive EA - are there other things?
>>>
>>> Thanks
>>>
>>> sent from my phone
>>> On Mar 13, 2015 9:53 AM, "Chris Seaton" <chris.seaton at oracle.com> wrote:
>>>
>>> Hi Bráulio,
>>>
>>> There’s no specific date yet of when Graal might be integrated into
>>> OpenJDK.
>>>
>>> You can of course install a binary build of GraalVM, which isn’t much
>>> different from installing an OpenJDK build
>>> http://lafo.ssw.uni-linz.ac.at/builds <
>>> http://lafo.ssw.uni-linz.ac.at/builds>. For the specific use case of
>>> JRuby, the ruby-build plugin to rbenv can already install JRuby+Truffle,
>>> with GraalVM bundled so all you have to do is run ‘run’ and you’ll get
>>> Graal.
>>>
>>> $ rbenv install jruby-9.0.0.0+graal-dev
>>> $ rbenv shell jruby-9.0.0.0+graal-dev
>>> $ ruby -X+T -e 'puts Truffle.graal?'
>>> true
>>>
>>> I think you can do something similar in RVM but not sure of the specifics
>>> at the moment.
>>>
>>> If you are interested in JRuby+Truffle, join us in #jruby on Freenode
>>> IRC.
>>>
>>> Regards,
>>>
>>> Chris
>>>
>>> On 13 Mar 2015, at 00:45, Bráulio Bhavamitra <brauliobo at gmail.com>
>>>
>>> wrote:
>>>
>>>
>>> Hello all,
>>>
>>> Sorry for asking this out of my anxiety...
>>>
>>> Having seen the incredible results when using the Graal VM with JRuby
>>> Truffle, I wonder if and when Graal code will be merged into OpenJDK
>>> mainline. Is there a roadmap or it is not going to happen soon (next
>>>
>>> years)?
>>>
>>>
>>> cheers,
>>> bráulio
>>>
>>> --
>>> "Lute pela sua ideologia. Seja um com sua ideologia. Viva pela sua
>>> ideologia. Morra por sua ideologia" P.R. Sarkar
>>>
>>> EITA - Educação, Informação e Tecnologias para Autogestão
>>> http://cirandas.net/brauliobo
>>> http://eita.org.br
>>>
>>> "Paramapurusha é meu pai e Parama Prakriti é minha mãe. O universo é meu
>>> lar e todos nós somos cidadãos deste cosmo. Este universo é a imaginação
>>>
>>> da
>>>
>>> Mente Macrocósmica, e todas as entidades estão sendo criadas,
>>>
>>> preservadas e
>>>
>>> destruídas nas fases de extroversão e introversão do fluxo imaginativo
>>> cósmico. No âmbito pessoal, quando uma pessoa imagina algo em sua mente,
>>> naquele momento, essa pessoa é a única proprietária daquilo que ela
>>> imagina, e ninguém mais. Quando um ser humano criado mentalmente caminha
>>> por um milharal também imaginado, a pessoa imaginada não é a propriedade
>>> desse milharal, pois ele pertence ao indivíduo que o está imaginando.
>>>
>>> Este
>>>
>>> universo foi criado na imaginação de Brahma, a Entidade Suprema, por isso
>>> a propriedade deste universo é de Brahma, e não dos microcosmos que
>>>
>>> também
>>>
>>> foram criados pela imaginação de Brahma. Nenhuma propriedade deste mundo,
>>> mutável ou imutável, pertence a um indivíduo em particular; tudo é o
>>> patrimônio comum de todos."
>>> Restante do texto em
>>> http://cirandas.net/brauliobo/blog/a-problematica-de-hoje-em-dia
>>>
>>>
>>>
>>>
>>>
>>
>>
>
>