outofmemory error on building
|
View:
New views
5 Messages
—
Rating Filter:
Alert me
|
 |
outofmemory error on building
by pankajsejwal
::
Rate this Message:
to find the control flow i was inserting VM.sysWriteln method and i didn have any problem initially but later it started showing Javaoutofmemory:java heap size sort of error.
is there a way to increase heap size permanently on openjdk coz after javac ,java is called instantly and there is no way to intercept. please share what ever info u have on it...! thanks.. |
|
|
Re: [rvm-research] outofmemory error on building
by Michael Bond
::
Rate this Message:
Hi,
You can increase the heap size for the build by modifying the maxmemory variable in the following snippet from build.xml: <echo message="Building bootimage. Output redirected to : ${build.base}/BootImageWriterOutput.txt"/> <echo message="MMTk properties = ${mmtk.properties}"/> <java classname="org.jikesrvm.tools.bootImageWriter.BootImageWriter" fork="yes" maxmemory="500M" failonerror="false" resultproperty="bootimage-writer.result" dir="${basedir}/tools/bootImageWriter/src"> <classpath> <pathelement location="${build.bootimage-writer.dir}"/> <path refid="rvm.class.path"/> </classpath> Could you please let us know your name and affiliation? Cheers, Mike On Mon, 28 Sep 2009, pankajsejwal wrote: > > to find the control flow i was inserting VM.sysWriteln method and i didn > have any problem initially but later it started showing Javaoutofmemory:java > heap size sort of error. > is there a way to increase heap size permanently on openjdk coz after javac > ,java is called instantly and there is no way to intercept. > please share what ever info u have on it...! > thanks.. > -- > View this message in context: http://www.nabble.com/outofmemory--error-on-building-tp25643001p25643001.html > Sent from the jikesrvm-researchers mailing list archive at Nabble.com. > > > ------------------------------------------------------------------------------ > Come build with us! The BlackBerry® Developer Conference in SF, CA > is the only developer event you need to attend this year. Jumpstart your > developing skills, take BlackBerry mobile applications to market and stay > ahead of the curve. Join us from November 9-12, 2009. Register now! > http://p.sf.net/sfu/devconf > _______________________________________________ > Jikesrvm-researchers mailing list > Jikesrvm-researchers@... > https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers > ------------------------------------------------------------------------------ Come build with us! The BlackBerry® Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9-12, 2009. Register now! http://p.sf.net/sfu/devconf _______________________________________________ Jikesrvm-researchers mailing list Jikesrvm-researchers@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers |
|
|
Re: [rvm-research] outofmemory error on building
by pankajsejwal
::
Rate this Message:
thanks for replying sir,
sir my name is:Pankaj Sejwal Affiliation(if I understand it right):Birla Institute of engineering and Science,Pilani,India. I am persuing ME(master of engineering) in Computer Science and we are using Jikes RVM to learn advanced compilation techniques. I am to implement " Thread specific heap" on this platform. some things i m not able to understand is that what is actual relation b/w space and allocators. What i want to do is that i want to find out if an object is local or global so that i can bring it around in same space but the escape analysis says that it sorts out registers and hashes them to put in buckets, what is the exact meaning of registers here where is should have been objects? Second thing I wanted to know is that allocation if I am to bother about then do I have to worry about any particular GC or I can pic anyone(coz it has been specified that every plan has its own way of working on heap). When i take NOGC, after initializing heap it proceeds to get monotone page resource which is showing "Currently unimplemented",if it is so then why control goes there? Third problem is that to find control i am employing VM.syswriteln and Log.write methods but even for very small programs I am getting some particular functions being called excessively and some not even once.I realized that there are two calls to stack allocation even if no class is provide to rvm to execute. thanks
|
|
|
Re: [rvm-research] outofmemory error on building
by Robin Garner
::
Rate this Message:
pankajsejwal wrote:
> thanks for replying sir, > sir my name is:Pankaj Sejwal > Affiliation(if I understand it right):Birla Institute of engineering and > Science,Pilani,India. I am persuing ME(master of engineering) in Computer > Science and we are using Jikes RVM to learn advanced compilation techniques. > I am to implement " Thread specific heap" on this platform. > some things i m not able to understand is that what is actual relation b/w > space and allocators. Essentially a Space is a global, shared object that represents a virtual memory region managed under the same allocation and collection policy. An Allocator is a thread-local object that allows a thread to allocate from a Space. Typically the Allocator will allocate one or more chunks of memory from the global Space, and then return smaller chunks to the application. This way the majority of allocations are unsynchronized (being thread-local), periodically performing synchronized allocation from the global Space. You can see a correspondence with the components of a plan, for example in MS.java (the global mark-sweep plan object), we have public static final MarkSweepSpace msSpace = new MarkSweepSpace("ms", DEFAULT_POLL_FREQUENCY, VMRequest.create()); ie a global MarkSweepSpace is created. Then in the per-thread MSMutator class, protected MarkSweepLocal ms = new MarkSweepLocal(MS.msSpace); each thread acquires its own MarkSweepLocal, a subclass of Allocator. > What i want to do is that i want to find out if an object is local or global > so that i can bring it around in same space but the escape analysis says > that it sorts out registers and hashes them to put in buckets, what is the > exact meaning of registers here where is should have been objects? > > Second thing I wanted to know is that allocation if I am to bother about > then do I have to worry about any particular GC or I can pic anyone(coz it > has been specified that every plan has its own way of working on heap). > When i take NOGC, after initializing heap it proceeds to get monotone page > resource which is showing "Currently unimplemented",if it is so then why > control goes there? The NoGC plan is not designed for running real programs, since it can't perform garbage collection. In the first instance you're probably best off looking at one of the stable full-heap collectors, MarkSweep or SemiSpace. I'm not sure I understand your specific question - could you point to a particular line of code ? > Third problem is that to find control i am employing VM.syswriteln and > Log.write methods but even for very small programs I am getting some > particular functions being called excessively and some not even once.I > realized that there are two calls to stack allocation even if no class is > provide to rvm to execute. Remember that JikesRVM itself uses the heap, and since JikesRVM is a compile-only VM, you will see a lot of allocation even when running a simple "hello world" application. Another approach you might try if you want to understand the call hierarchy of MMTk is to use a profiler like jip (http://jiprof.sourceforge.net/) against the MMTk harness. Here's an example from some profiling I was doing recently (omitting some of the harness internals and vmmagic calls) 30199 9184.6 163.3 9.8 0.2 | | | | | | +--GenMSMutator:alloc (org.mmtk.plan.generational.marksweep) 30199 9021.3 149.5 9.6 0.2 | | | | | | | +--GenMutator:alloc (org.mmtk.plan.generational) 30199 8871.8 449.6 9.4 0.5 | | | | | | | | +--BumpPointer:alloc (org.mmtk.utility.alloc) 30199 2333.9 157.0 2.5 0.2 | | | | | | | | | +--Allocator:alignAllocationNoFill (org.mmtk.utility.alloc) 30199 2176.8 1207.2 2.3 1.3 | | | | | | | | | | +--Allocator:alignAllocation (org.mmtk.utility.alloc) 30140 1351.6 243.8 1.4 0.3 | | | | | | | | | +--Allocator:fillAlignmentGap (org.mmtk.utility.alloc) 59 4496.6 1.2 4.8 | | | | | | | | | +--BumpPointer:allocSlow (org.mmtk.utility.alloc) 59 4492.7 1.2 4.8 | | | | | | | | | | +--Allocator:allocSlowInline (org.mmtk.utility.alloc) 59 0.1 0.1 0.0 | | | | | | | | | | | +--Stats:gcCount (org.mmtk.utility.statistics) 59 0.1 0.1 0.0 | | | | | | | | | | | +--BumpPointer:getSpace (org.mmtk.utility.alloc) 59 4491.1 3.5 4.8 | | | | | | | | | | | +--BumpPointer:allocSlowOnce (org.mmtk.utility.alloc) 59 11.5 1.5 0.0 | | | | | | | | | | | | +--Conversions:bytesToPages (org.mmtk.utility) 59 5.0 1.1 0.0 | | | | | | | | | | | | | +--Conversions:bytesToPagesUp (org.mmtk.utility) 59 1.9 0.8 0.0 | | | | | | | | | | | | | +--Conversions:pagesToAddress (org.mmtk.utility) 59 4431.3 2.9 4.7 | | | | | | | | | | | | +--Space:acquire (org.mmtk.policy) 59 0.1 0.1 0.0 | | | | | | | | | | | | | +--Plan:gcInProgress (org.mmtk.plan) 59 0.1 0.1 0.0 | | | | | | | | | | | | | +--Plan:isInitialized (org.mmtk.plan) 59 0.1 0.1 0.0 | | | | | | | | | | | | | +--Collection:isEmergencyAllocation (org.mmtk.harness.vm) 59 4.2 0.9 0.0 | | | | | | | | | | | | | +--PageResource:reservePages (org.mmtk.utility.heap) 59 1.7 0.9 0.0 | | | | | | | | | | | | | | +--PageResource:lock (org.mmtk.utility.heap) 59 0.1 0.1 0.0 | | | | | | | | | | | | | | | +--Plan:gcInProgress (org.mmtk.plan) 59 0.7 0.4 0.0 | | | | | | | | | | | | | | | +--JavaLock:acquire (org.mmtk.harness.scheduler.javathreads)> thanks 59 0.1 0.1 0.0 | | | | | | | | | | | | | | +--MonotonePageResource:adjustForMetaData (org.mmtk.utility.heap) 59 1.5 0.7 0.0 | | | | | | | | | | | | | | +--PageResource:unlock (org.mmtk.utility.heap) 59 0.1 0.1 0.0 | | | | | | | | | | | | | | | +--Plan:gcInProgress (org.mmtk.plan) 59 0.7 0.5 0.0 | | | | | | | | | | | | | | | +--JavaLock:release (org.mmtk.harness.scheduler.javathreads) 59 4404.6 0.4 4.7 | | | | | | | | | | | | | +--PageResource:getNewPages (org.mmtk.utility.heap) 59 4404.1 4.0 4.7 | | | | | | | | | | | | | | +--MonotonePageResource:allocPages (org.mmtk.utility.heap) 59 1.3 0.7 0.0 | | | | | | | | | | | | | | | +--PageResource:lock (org.mmtk.utility.heap) 59 0.1 0.1 0.0 | | | | | | | | | | | | | | | | +--Plan:gcInProgress (org.mmtk.plan) 59 0.4 0.2 0.0 | | | | | | | | | | | | | | | | +--JavaLock:acquire (org.mmtk.harness.scheduler.javathreads) 59 2.5 0.9 0.0 | | | | | | | | | | | | | | | +--Space:chunkAlign (org.mmtk.policy) 59 3.4 0.7 0.0 | | | | | | | | | | | | | | | +--Conversions:pagesToBytes (org.mmtk.utility) 59 2.7 0.8 0.0 | | | | | | | | | | | | | | | +--PageResource:commitPages (org.mmtk.utility.heap) 59 0.1 0.1 0.0 | | | | | | | | | | | | | | | | +--MonotonePageResource:adjustForMetaData (org.mm The first column is the invocation count - you can see a sharp difference between the fast-path (thread-local, Allocator) and the slow-path (global, Space). You can get this profile by running < download and unpack jip into $HOME/tools > $ ant mmtk-harness $ export JIP=$HOME/tools/jip-1.1.1/profile/profile.jar $ java -Xbootclasspath/a:$JIP -javaagent:$JIP \ -jar target/mmtk/mmtk-harness.jar \ MMTk/harness/test-scripts/Quicksort.script \ plan=org.mmtk.plan.generational.marksweep.GenMS initHeap=5m There could well be other tools Regards, Robin > > Michael Bond wrote: >> Hi, >> >> You can increase the heap size for the build by modifying the maxmemory >> variable in the following snippet from build.xml: >> >> <echo message="Building bootimage. Output redirected to : >> ${build.base}/BootImageWriterOutput.txt"/> >> <echo message="MMTk properties = ${mmtk.properties}"/> >> <java classname="org.jikesrvm.tools.bootImageWriter.BootImageWriter" >> fork="yes" >> maxmemory="500M" >> failonerror="false" >> resultproperty="bootimage-writer.result" >> dir="${basedir}/tools/bootImageWriter/src"> >> <classpath> >> <pathelement location="${build.bootimage-writer.dir}"/> >> <path refid="rvm.class.path"/> >> </classpath> >> >> Could you please let us know your name and affiliation? >> >> Cheers, >> Mike >> >> >> On Mon, 28 Sep 2009, pankajsejwal wrote: >> >>> to find the control flow i was inserting VM.sysWriteln method and i didn >>> have any problem initially but later it started showing >>> Javaoutofmemory:java >>> heap size sort of error. >>> is there a way to increase heap size permanently on openjdk coz after >>> javac >>> ,java is called instantly and there is no way to intercept. >>> please share what ever info u have on it...! >>> thanks.. >>> -- >>> View this message in context: >>> http://www.nabble.com/outofmemory--error-on-building-tp25643001p25643001.html >>> Sent from the jikesrvm-researchers mailing list archive at Nabble.com. >>> >>> >>> ------------------------------------------------------------------------------ >>> Come build with us! The BlackBerry® Developer Conference in SF, CA >>> is the only developer event you need to attend this year. Jumpstart your >>> developing skills, take BlackBerry mobile applications to market and stay >>> ahead of the curve. Join us from November 9-12, 2009. Register >>> now! >>> http://p.sf.net/sfu/devconf >>> _______________________________________________ >>> Jikesrvm-researchers mailing list >>> Jikesrvm-researchers@... >>> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers >>> >> ------------------------------------------------------------------------------ >> Come build with us! The BlackBerry® Developer Conference in SF, CA >> is the only developer event you need to attend this year. Jumpstart your >> developing skills, take BlackBerry mobile applications to market and stay >> ahead of the curve. Join us from November 9-12, 2009. Register >> now! >> http://p.sf.net/sfu/devconf >> _______________________________________________ >> Jikesrvm-researchers mailing list >> Jikesrvm-researchers@... >> https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers >> >> > ------------------------------------------------------------------------------ Come build with us! The BlackBerry® Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9-12, 2009. Register now! http://p.sf.net/sfu/devconf _______________________________________________ Jikesrvm-researchers mailing list Jikesrvm-researchers@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-researchers |
|
|
Re: [rvm-research] outofmemory error on building
by pankajsejwal
() ::
Rate this Message:
thanks a lot for this precise information sir..
the question I was trying to put about escape analysis is simpleEscape.java if (escapes.threadLocal) { result.setThreadLocal(reg, true); } if (escapes.methodLocal) { result.setMethodLocal(reg, true); } from here the control is sent to FI_EscapeSummary.java void setThreadLocal(Register r, boolean b) { if (b) { hash.put(r, THREAD_LOCAL); } else { hash.put(r, MAY_ESCAPE_THREAD); then it says that it hashes them in buckets and sets the buckets accordingly.I am not able to understand that classification of buckets is on basis of what? Do they get set on the basis of thread active for the application? If it is so how can i find for which thread it is being done?? Is flow insensitive escape analysis the best variation of escape analysis? Another thing that I think is that even if i don't know the objects in their true form i can use scheduler to find thread id of currently executing thread and use a hashmap<threadid,register,THREAD_LOCAL> to make a combination?? I wanted one more info please if i try to divide space run time from memory manager what r the possible places i might have to maipulate like accounting..? please respond to my question...
|
| Free embeddable forum powered by Nabble | Forum Help |
