harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From xiaoming gu <xiaoming...@gmail.com>
Subject Re: [drlvm][jitrino] write barrier was broken
Date Sat, 23 May 2009 10:00:39 GMT
Hi, Simon. The debug is about an peephole optimization. I assumed no control
flow branch in the middle of a basic block. But I find some CALL insts break
the assumption sometimes like following:

I4: t5:ref:cls:java/lang/ref/ReferenceQueue =ADD
t0:cls:java/lang/ref/Reference, t4(0):intptr
I5: CALL t7(0):I_32 (AU:t0:cls:java/lang/ref/Reference,
t5:ref:cls:java/lang/ref/ReferenceQueue,
t2:cls:java/lang/ref/ReferenceQueue) [bcmap:6]
I6: t9:ref:cls:java/lang/Object =ADD
t0:cls:java/lang/ref/Reference,t8(0):intptr
I7: CALL t10(0):I_32
(AU:t0:cls:java/lang/ref/Reference,t9:ref:cls:java/lang/Object,t1:cls:java/lang/Object)
[bcmap:11]
I8: RET t11(0):int16!

So far I don't know why CALL can sit there. Whatever, I'll file a bug on
JIRA and put a preliminary patch. Thanks.

Xiaoming

On Sat, May 16, 2009 at 11:25 PM, xiaoming gu <xiaoming.gu@gmail.com> wrote:

> Thanks for the helpful info, Simon. I'll look into it. -Xiaoming
>
>
> On Sat, May 16, 2009 at 10:56 PM, Simon Zhou <simon.harmony@gmail.com>wrote:
>
>> Xiaoming,
>>
>> I am sorry that maybe my previous guesswork is wrong.
>> it is not caused by global_prop, because the issue still exists after I
>> remove the global_prop in config file.
>> but after I revert all the modification of this patch, VM runs pretty
>> well,
>> so it maybe caused by other modification in this patch but global_prop, or
>> not just global_prop. I am not sure...
>> Thanks for your help!
>> Thanks
>> Simon
>> 2009/5/16 xiaoming gu <xiaoming.gu@gmail.com>
>>
>> > Hi, Simon. A workaround is to replace global_prop with early_prop in the
>> > used execution mode configuration file. I'll take care of this bug these
>> > days. Thanks.
>> >
>> > Xiaoming
>> >
>> > On Sat, May 16, 2009 at 5:04 PM, Simon Zhou <simon.harmony@gmail.com>
>> > wrote:
>> >
>> > > I think I found the origin of this issue by binary searching the
>> previous
>> > > patches, I foud that it is broken since version 747844 (747843 is OK),
>> so
>> > > I am guessing the patch of 5826 breaks this.
>> > >
>> > > To replay the error, just add -XX:gc.generate_barrier=true to command
>> > line
>> > > when running a Java application.
>> > > I added some of my guesswork in
>> > > https://issues.apache.org/jira/browse/HARMONY-5826
>> > >
>> > > Xiaoming, Would you like to give me advice for this issue? Or is there
>> > some
>> > > workaroud for it? Thank you very much!
>> > >
>> > > Thanks
>> > > Simon
>> > >
>> > > 2009/5/16 Simon Zhou <simon.harmony@gmail.com>
>> > >
>> > > > Hi All,
>> > > >
>> > > >   I am doing my GSoC2009 project of implementing weak reference in
>> > > > Concurrent GC. I found that as long as the gc.generate_barrier=true
>> is
>> > > set,
>> > > > VM will crash when it startups (int vm_init2() method). I was
>> following
>> > > this
>> > > > issue these days but did not work out.
>> > > >   This issue seem to come out after milestone8, so  I guess it could
>> be
>> > > > caused by some patches on Jitrino after M8.
>> > > >    Is there anyone else being aware of that? Or could you please
>> > provide
>> > > > some information about which patch could cause it?
>> > > >    Thank you very much!
>> > > >
>> > > > Thanks
>> > > > Simon
>> > > >
>> > > > --
>> > > > From : Simon.Zhou@PPI, Fudan University
>> > > >
>> > >
>> > >
>> > >
>> > > --
>> > > From : Simon.Zhou@PPI, Fudan University
>> > >
>> >
>>
>>
>>
>> --
>> From : Simon.Zhou@PPI, Fudan University
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message