Yves,

High-level optimizer is not enabled on IPF yet so I think if you try
just to turn it on in emconf file this won't go smoothly. But if you'd
like to work on enabling high-level optimizations on IPF I believe JIT
guys are willing to help. ;-)

Thanks,
Pavel

On 5/31/07, Yves Ineichen <[EMAIL PROTECTED]> wrote:
Egor Pasko wrote:
> Interesting..
>
> On the 0x2E7 day of Apache Harmony Yves Ineichen wrote:
>> hi all,
>>
>> After I added the "inline" action to my optimizer path the release
>> compiler
>
> What do you mean by 'release compiler'? is that the one built in
> release mode or the one taken from our M1 'release'?

When building the drlvm one can choose between two 'working modes':
debug or release -- so what I called "release compiler" is actually the
drlvm built in "release mode".
I also should have mentioned that I build everything from the current
SVN trunk:

Apache Harmony Launcher : (c) Copyright 1991, 2006 The Apache Software
Foundation or its licensors, as applicable.
java version "1.5.0"
pre-alpha : not complete or compatible
svn = r543115, (May 31 2007), Linux/ipf/gcc 3.4.6, debug build
http://incubator.apache.org/harmony

> In both variants assertions are off, so, it is very interesting what
> version you use. Especially it is interesting because your path looks
> like the IPF one. On IPF the optimizer is currently off and not
> well-tested.
>> throws a not implemented FIXME and the debug version fails with
>> the following assertion:
>>
>> 
/data/home/iyves/src/harmony/working_vm/vm/jitrino/src/optimizer/inliner.cpp:1129:
>> static void
>> Jitrino::Inliner::runInlinerPipeline(Jitrino::CompilationContext&, const
>> char*): Assertion `p!=__null' failed.
>> SIGABRT in VM code.
>> Aborted
>>
>> Im using the following path definitions:
>>
>> 
-XX:jit.CS_OPT.path=opt_init,translator,optimizer,hir2lir,lock_method,ipf_codegen,unlock_method
>> 
-XX:jit.CS_OPT.path.optimizer=ssa,devirt,inline,purge,simplify,uce,dce,lazyexc,inline_helpers,purge,simplify,uce,dce,memopt,simplify,uce,dce,lower,dessa,statprof,markglobals
>>
>> Is the inliner currently not working as intended or is my path
>> configuration wrong?
>>
>> Thanks,
>>
>> --
>> Yves Ineichen
>> ::p = "This statement cannot be proven"::
>>
>


--
Yves Ineichen
::p = "This statement cannot be proven"::

Reply via email to