[Mono-dev] Using libJIT in Mono Instead of LLVM

[Kirill Kononenko]

Hi Rodrigo,


Thanks for your reply.

> LLVM is been experimented as a AOT only compiler.
> The exeuction time for LLVM is huge and not suitable for a JIT VM such as
> mono.

>
> So are you suggesting to use libjit in a similar fashion to LLVM or to
> replace the regular
> mono JIT?

Mono JIT rather for this moment. But if an integration with gcc can be
done and it is interesting to other people too, than both.

>
> Without published numbers on JIT and compiled code speed (and how to
> reproduce
> the experiment) it's hard to follow you on the "really beat all Mono
> benchmarks" beyond
> faith.
>

> Does libjit support the whole spectrum of ECMA 335 for CLR types and
> operations? Does it support
> async exception handling? And precise stack marking? Multiple custom call
> conventions? There are
> just a few of the many walls we have reached on LLVM and have to work on.

Yes it does support this all.

>
> There is some code on mono's svn to integrate llvm, you could take it and
> plug libjit in. If it shows some
> promissing results, we could really talk into integrating it.
>
It is unlikely I will have time to write some code any soon, because
of a lot of work documentating my current work. But I think this could
be interesting to someone else too, for example under a google summer
of code, or a diploma for example.


Anyways, let continue sharing of ideas, which come to us both. I am
interested to know which features to libJIT may benefit Mono and its
users. Thanks for your reply again.


Thanks,
Kirill

> On Sun, Apr 5, 2009 at 7:15 AM, Kirill Kononenko
> <kirill.kononenko at gmail.com> wrote:
>>
>> Hello Everyone,
>>
>>
>> I have already sent two emails in this mailing list. Most likely their
>> will be not accepted here. So I want to ask your opinion, about a
>> possible cooperation of using libJIT as another engine for Mono
>> instead of LLVM. It has already been tested with the Portable.NET JIT,
>> and has proven that it is a good concept. I am sure that libJIT can be
>> extended very easily to really beat all Mono benchmarks, and feet
>> other user needs. Also if an integration with gcc will be considered,
>> even more people will benefit. I think with current economic situation
>> it is to the best of everyone to share code, and cooperate. We could
>> explore various options, for example how we could suit better needs of
>> Mono users, and .NET users on open source/free software platforms. I
>> am really sure that a cooperation could benefit everyone, and everyone
>> wins at the end.
>>
>>
>>
>>
>> Thanks,
>> Kirill
>> _______________________________________________
>> Mono-devel-list mailing list
>> Mono-devel-list at lists.ximian.com
>> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>
>