Re: [rvm-research] Create new Exception type

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 454-5900

Hi Mohamed,

On Thu, Aug 9, 2012 at 9:51 PM, Mohamed Ahmed <mam...@gm...> wrote:

> Hi Minjia Zhang,
>
> Thanks for your reply.
>
> Yes, I'm using the optimizing compiler and ExpandRuntimeService. Using
> your hint, I've added a new dummy method
>
> @org.vmmagic.pragma.NoInline
> public static void dummy() throws MyExceptionClass{}
>
> Shouldn't your MyExceptionClass a Runtime exception if you are going to
throw it in your read and write barrier?

> and called it from the method that was throwing the uncaught exception.
>  public int getValue(){
> dummy();
> return value;
>  }
>
>
> public boolean add(int key) {
> do{try{
>  ....
>                          node.getValue();
>  ....
> return true;
>  }catch(MyExceptionClass e){}
> }while(true);
>  }
> Without calling dummy() the try...catch in add method is ignore.
>
> Ok, does the following make things simpler? Then you still can get the
benefit of inlining trivial methods like getValue() called in add().
public boolean add(int key) {
do{try{
 ....
                         boolean result = atomicAdd()
 ....
return boolean;
}catch(MyExceptionClass e){}
}while(true);
 }

> Can you please point me how to control the exception handlers. So that I
> have a correct implementation instead of this hack.
>

You might want to take a look at rectifyStateWithExceptionHandler in BC2IR,
but then you may need to treat each read and write access as a PEI, and
rectify exception handlers to all accesses. I'm not sure if it is an
efficient way to solve your problem.

>
> Thanks again,
> Mohamed
>
>
Best,
Minjia

>
> On Thu, Aug 9, 2012 at 12:09 PM, Minjia Zhang <zha...@cs...
> > wrote:
>
>> Hi Mohamed,
>> Do you have this uncaught exception issue in BaseBase or in FullAdaptive?
>> My answer is based on my assumption that you have already added
>> instrumentation of your read and write barriers in the optimizing compiler.
>> Then the problem is the optimizing compiler identifies a PEI at BC2IR phase
>> and connects exception handlers to the basic block containing that PEI. In
>> other words, if your read/write barrier is added in a later phase, for
>> example, during ExpandRuntimeService, it probably does not have the
>> exception handler you expected associated with that basic block which leads
>> to an uncaught exception error.
>>
>> A quick way to solve this problem is to enclose accesses you instrumented
>> that might throw exceptions in your application code into a method, use
>> try-catch block to catch your MyExceptionClass thrown from that method. You
>> might also want to annotate that method as @NoInline using Jikes
>> annotation. The BC2IR phase will treat any noinlined call as potential
>> throw of anything and your exception handler block will be connected to it.
>>
>> Or perhaps you can try to insert your read and write barriers during
>> BC2IR, mark them as PEIs and rectify their states with exception handlers.
>> I'm not sure whether Jikes support adding PEIs at some point after BC2IR.
>>
>> And of course, you might need to first address the issue Steve mentioned
>> to make sure it is not caused by infinite regress.
>>
>> Good luck,
>> Minjia Zhang
>>
>> Department of Computer Science and Engineering,
>> The Ohio State University
>> 674, Dresse Lab,
>> 2015 Neil Avenue,
>> Columbus, Ohio 43210-1277.
>>
>> On Wed, Aug 8, 2012 at 6:33 PM, Mohamed Ahmed <mam...@gm...>wrote:
>>
>>> Thanks a lot Steve
>>>
>>>
>>> On Wed, Aug 8, 2012 at 6:23 PM, Steve Blackburn <
>>> Ste...@an...> wrote:
>>>
>>>> You could probably just have a flag that you set once you identify the
>>>> problem, and guard the problem detection code so it does not trigger once
>>>> you're already in the process of throwing an exception.   In other words,
>>>> I'm saying that you need to be aware of the possibility of infinite regress
>>>> and you need to have some way of combatting that --- there are many ways.
>>>>
>>>> --Steve
>>>>
>>>> On 09/08/2012, at 7:49 AM, Mohamed Ahmed wrote:
>>>>
>>>> Hi Steve,
>>>>
>>>> I need to know that something wrong has happened (in my STM, an
>>>> inconsistent read occurred). Any other way to do so?
>>>>
>>>> Thanks,
>>>> Mohamed
>>>>
>>>> On Wed, Aug 8, 2012 at 5:22 PM, Steve Blackburn <
>>>> Ste...@an...> wrote:
>>>>
>>>>> Hi Mohamed,
>>>>>
>>>>> You can't throw exceptions from within read and write barriers.    You
>>>>> can't do anything that would potential lead to infinite regress.   An
>>>>> exception would easily do that (barrier causes exception, causes barrier,
>>>>> causes exception ... etc).
>>>>>
>>>>> --Steve
>>>>>
>>>>> On 09/08/2012, at 5:33 AM, Mohamed Ahmed wrote:
>>>>>
>>>>> > Dear All,
>>>>> >
>>>>> > I'm trying to create a new runtime exception to be thrown from
>>>>> inside the VM. I used a new class (mypackage.MyExceptionClass) and throw it
>>>>> inside the VM. From the java code, I'm catching the same exception class
>>>>> (and I copied the class to the java code classpath)
>>>>> >
>>>>> > Now when the VM run that java code, sometimes the exception is
>>>>> caught and other times the exception ignores the catch block and becomes an
>>>>> uncaught error. (and in other cases, it is caught and also propagate as
>>>>> uncaught exception). It depends on the java code program that I'm running.
>>>>> >
>>>>> > I'm throwing the new exception in a read/write barrier.
>>>>> >
>>>>> > Thanks a lot,
>>>>> > Mohamed
>>>>> >
>>>>> ------------------------------------------------------------------------------
>>>>> > Live Security Virtual Conference
>>>>> > Exclusive live event will cover all the ways today's security and
>>>>> > threat landscape has changed and how IT managers can respond.
>>>>> Discussions
>>>>> > will include endpoint security, mobile security and the latest in
>>>>> malware
>>>>> > threats.
>>>>> http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/_______________________________________________
>>>>> > Jikesrvm-researchers mailing list
>>>>> > Jik...@li...
>>>>> > https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
>>>>>
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> Live Security Virtual Conference
>>>>> Exclusive live event will cover all the ways today's security and
>>>>> threat landscape has changed and how IT managers can respond.
>>>>> Discussions
>>>>> will include endpoint security, mobile security and the latest in
>>>>> malware
>>>>> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>>>>> _______________________________________________
>>>>> Jikesrvm-researchers mailing list
>>>>> Jik...@li...
>>>>> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
>>>>>
>>>>
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Live Security Virtual Conference
>>>> Exclusive live event will cover all the ways today's security and
>>>> threat landscape has changed and how IT managers can respond.
>>>> Discussions
>>>> will include endpoint security, mobile security and the latest in
>>>> malware
>>>> threats.
>>>> http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/_______________________________________________
>>>> Jikesrvm-researchers mailing list
>>>> Jik...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
>>>>
>>>>
>>>>
>>>>
>>>> ------------------------------------------------------------------------------
>>>> Live Security Virtual Conference
>>>> Exclusive live event will cover all the ways today's security and
>>>> threat landscape has changed and how IT managers can respond.
>>>> Discussions
>>>> will include endpoint security, mobile security and the latest in
>>>> malware
>>>> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>>>> _______________________________________________
>>>> Jikesrvm-researchers mailing list
>>>> Jik...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
>>>>
>>>>
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Live Security Virtual Conference
>>> Exclusive live event will cover all the ways today's security and
>>> threat landscape has changed and how IT managers can respond. Discussions
>>> will include endpoint security, mobile security and the latest in malware
>>> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>>> _______________________________________________
>>> Jikesrvm-researchers mailing list
>>> Jik...@li...
>>> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
>>>
>>>
>>
>>
>> ------------------------------------------------------------------------------
>> Live Security Virtual Conference
>> Exclusive live event will cover all the ways today's security and
>> threat landscape has changed and how IT managers can respond. Discussions
>> will include endpoint security, mobile security and the latest in malware
>> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>> _______________________________________________
>> Jikesrvm-researchers mailing list
>> Jik...@li...
>> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
>>
>>
>
>
> ------------------------------------------------------------------------------
> Live Security Virtual Conference
> Exclusive live event will cover all the ways today's security and
> threat landscape has changed and how IT managers can respond. Discussions
> will include endpoint security, mobile security and the latest in malware
> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
> _______________________________________________
> Jikesrvm-researchers mailing list
> Jik...@li...
> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers
>
>