|

»

[rvm-core] Thread TID in Linux when using -X:processors=all

View: New views

4 Messages — Rating Filter: Alert me

	[rvm-core] Thread TID in Linux when using -X:processors=all by frederick.ryckbosch-2 :: Rate this Message: Reply to Author \| View Threaded \| Show Only this Message Hi, I'm Frederick Ryckbosch, a computer science student at the university of Gent. For my master thesis I'm trying to integrate the IBS (instruction based sampling) of AMD in Jikes RVM. I have a couple of questions about this. I'm using a Linux 32 bit OS and jikes-3.0.0. The first thing I have to do is initialize the counters in each thread. * Can I do this in GreenScheduler and GreenProcessor ? * In the same way the affinity of the threads is set ? There is one signal handler that will handle the overflow of the counters for all my threads. To read the correct counter, I use the thread's TID. Normally in Linux, when starting a number of threads, their TID is pid, pid+1, pid+2 and so on. This does not seem to be the case in the Jikes RVM. When starting with -X:processors=all (on QuadCore), 5 threads are started in stead of 4 thread. Only 4 threads are used by Java. The TIDS's are : pid, pid+12, pid+13, pid+14. Is this always the case ? What happend to the other 11 TID's ? I hope somebody can answer these questions. Thanks in advance ! Greetings, Frederick Ryckbosch ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Jikesrvm-core mailing list Jikesrvm-core@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-core
	Re: [rvm-core] Thread TID in Linux when using -X:processors=all by David P Grove :: Rate this Message: Reply to Author \| View Threaded \| Show Only this Message `Frederick Ryckbosch <frederick.ryckbosch@...> wrote on 10/05/2008 09:06:12 AM: > I'm using a Linux 32 bit OS and jikes-3.0.0. The first thing I have to > do is initialize the counters in each thread. > * Can I do this in GreenScheduler and GreenProcessor ? > * In the same way the affinity of the threads is set ?` `My guess would be yes. It might be useful to look at this old patch (http://sourceforge.net/tracker/index.php?func=detail&aid=1613215&group_id=128805&atid=723235) that was HPM support for AIX/PowerPC. It's not likely to be directly useful, and it's a couple years out-of-date, but it might point to places in the code where you can might want to make changes for what you are doing.` `--dave` ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Jikesrvm-core mailing list Jikesrvm-core@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-core
	Re: [rvm-core] Thread TID in Linux when using -X:processors=all by itkovian :: Rate this Message: Reply to Author \| View Threaded \| Show Only this Message Frederik, On 05 Oct 2008, at 15:06, Frederick Ryckbosch wrote: > Hi, > > I'm Frederick Ryckbosch, a computer science student at the university > of Gent. For my master thesis I'm trying to integrate the IBS > (instruction based sampling) of AMD in Jikes RVM. I have a couple of > questions about this. > > I'm using a Linux 32 bit OS and jikes-3.0.0. The first thing I have to > do is initialize the counters in each thread. > * Can I do this in GreenScheduler and GreenProcessor ? > * In the same way the affinity of the threads is set ? > > There is one signal handler that will handle the overflow of the > counters for all my threads. To read the correct counter, I use the > thread's TID. Normally in Linux, when starting a number of threads, > their TID is pid, pid+1, pid+2 and so on. This does not seem to be the > case in the Jikes RVM. When starting with -X:processors=all (on > QuadCore), 5 threads are started in stead of 4 thread. Only 4 threads > are used by Java. The TIDS's are : pid, pid+12, pid+13, pid+14. Is > this always the case ? What happend to the other 11 TID's ? I might be wrong, but is there not an extra thread started that wakes up each time the RVM has to check for a threadswitch? This used to be the case in the VM back from April 10th, 2006 (the VM we used in our paper 'Using HPM-Sampling to Drive Dynamic Compilation'). This thread executes the libc nanosleep() fuction. Upon awakening it sets a bit that can be checked in a yieldpoint, thereby notifying the VM that it should execute a threadswitch. At this moment, I have no idea why the tid's should be off 11 places. What configuration are you running? -- Andy ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Jikesrvm-core mailing list Jikesrvm-core@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-core
	Re: [rvm-core] Thread TID in Linux when using -X:processors=all by Andy Georges :: Rate this Message: Reply to Author \| View Threaded \| Show Only this Message Frederik, On 05 Oct 2008, at 15:06, Frederick Ryckbosch wrote: > Hi, > > I'm Frederick Ryckbosch, a computer science student at the university > of Gent. For my master thesis I'm trying to integrate the IBS > (instruction based sampling) of AMD in Jikes RVM. I have a couple of > questions about this. > > I'm using a Linux 32 bit OS and jikes-3.0.0. The first thing I have to > do is initialize the counters in each thread. > * Can I do this in GreenScheduler and GreenProcessor ? > * In the same way the affinity of the threads is set ? > > There is one signal handler that will handle the overflow of the > counters for all my threads. To read the correct counter, I use the > thread's TID. Normally in Linux, when starting a number of threads, > their TID is pid, pid+1, pid+2 and so on. This does not seem to be the > case in the Jikes RVM. When starting with -X:processors=all (on > QuadCore), 5 threads are started in stead of 4 thread. Only 4 threads > are used by Java. The TIDS's are : pid, pid+12, pid+13, pid+14. Is > this always the case ? What happend to the other 11 TID's ? I might be wrong, but is there not an extra thread started that wakes up each time the RVM has to check for a threadswitch? This used to be the case in the VM back from April 10th, 2006 (the VM we used in our paper 'Using HPM-Sampling to Drive Dynamic Compilation'). This thread executes the libc nanosleep() fuction. Upon awakening it sets a bit that can be checked in a yieldpoint, thereby notifying the VM that it should execute a threadswitch. At this moment, I have no idea why the tid's should be off 11 places. What configuration are you running? -- Andy ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Jikesrvm-core mailing list Jikesrvm-core@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-core