I add the following expression in InstCodeSelector::addElemIndex():
Opnd* current_buffer = irManager.newImmOpnd(typeManager.getInt32Type(),
(unsigned)¤t_obj_acc_record_pointer);
Opnd * Ip_Opnd =
irManager.newMemOpndAutoKind(typeManager.getPrimitiveType(Type::UInt32),
current_buffer, irManager.ne
On 1/25/07, zouqiong <[EMAIL PROTECTED]> wrote:
>
> You are right. If the only change you do is a new extra call in
translator
> it will limit optimizations. The best place to add tracing calls could
be
> hir2lir conversion, after all HLO optimizations were finished, but in
this
> case your help
You are right. If the only change you do is a new extra call in translator
it will limit optimizations. The best place to add tracing calls could be
hir2lir conversion, after all HLO optimizations were finished, but in this
case your helper can't be inlined.
If you want your helper to be inlined
On 1/24/07, zouqiong <[EMAIL PROTECTED]> wrote:
En, I don't know whether OPT can eliminate such code:
aload_0
getfield 5 aload_0
aload_0 ==>getfield 5
getfield 5 dup
If it can eliminate them; I add tracing code as follows:
aload_0
On 1/24/07, zouqiong <[EMAIL PROTECTED]> wrote:
It means that Chip Multi-processor. I want to know whether there're some
optimization face to CMP.
Ok, now it's clear. It's sad to say but I'm not an expert in this area, so I
can't advise.
Ian,
Thank you for your link. The list of Java related
Hi Zuoqiong,
there's a lot of research and Java and CMPs. For example, the Jamaica
project at the University of Manchester:
http://www.cs.manchester.ac.uk/apt/projects/jamaica/
has been looking at runtime adaptive parallelizing compilers (see also
IBM's Ninja project). There's also significant
En, I don't know whether OPT can eliminate such code:
aload_0
getfield 5 aload_0
aload_0 ==>getfield 5
getfield 5 dup
If it can eliminate them; I add tracing code as follows:
aload_0 aload_0
getfield 5 call ObjectRecord
It means that Chip Multi-processor. I want to know whether there're some
optimization face to CMP.
2007/1/23, Mikhail Fursov <[EMAIL PROTECTED]>:
On 1/23/07, zouqiong <[EMAIL PROTECTED]> wrote:
>
> Hi Mikhail:
>I want to ask you a question: Is there any researches about Java in
> CMP?
So
On 1/23/07, zouqiong <[EMAIL PROTECTED]> wrote:
Hi Mikhail:
I want to ask you a question: Is there any researches about Java in
CMP?
Sorry, it's the end of the day, so I'm very slow..
What do you mean by "CMP" ? Container-managed persistence or 'compare' op?
--
Mikhail Fursov
On 1/23/07, zouqiong <[EMAIL PROTECTED]> wrote:
Hi, Mikhail:
I implement code to trace object access just follow WB's way. But I'm
afraid that the instrumented code in ops like getfield would effect the
liveness analysis optimize?
When we add new commands/types to optimizer we have to teac
Hi Mikhail:
I want to ask you a question: Is there any researches about Java in CMP?
2007/1/22, Mikhail Fursov <[EMAIL PROTECTED]>:
Hi!
Your task is very close to write barriers implementation in OPT.
I think that the best place today is IRBuilder.cpp that is called from
JavaByteCodeTranslat
Hi, Mikhail:
I implement code to trace object access just follow WB's way. But I'm
afraid that the instrumented code in ops like getfield would effect the
liveness analysis optimize?
2007/1/22, Mikhail Fursov < [EMAIL PROTECTED]>:
Hi!
Your task is very close to write barriers implementation
Hi!
Your task is very close to write barriers implementation in OPT.
I think that the best place today is IRBuilder.cpp that is called from
JavaByteCodeTranslator during HIR creation from Java bytecode.
See WB implementation as an example and add tracing code to object access
ops like genLdField..
Hi, all:
I want to instrument the Java application when it is compiled by the OPT
to gather object access sequences. However the code generation passes by
OPT are very complex. Which pass is the most suitable for me to instrument
the code ?
--
Best Regards,
Qiong,Zou
14 matches
Mail list logo
