[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-hotspot-gc-dev
Subject:    custom compacting GC implementation
From:       psockali () gmail ! com (Pressenna Sockalingasamy)
Date:       2010-09-28 22:50:52
Message-ID: 4CA2714C.3090403 () gmail ! com
[Download RAW message or body]

  Hi,

I'm trying to implement a (so far) simple heap, which stops the world and compacts \
spaces. It is all working, but as soon as I compact spaces and get new oop addresses \
to be used, the JVM SIGSEGVs. For some reasons, the backtrace (via DDD) shows me that \
the VM is executing code in the heap address space. Is there something special I have \
to do / consider when adjusting pointers during a safepoint? Currently I am just \
using the same approach as the SharedHeap::process_strong_roots(...). My code base is \
the b20 source drop.

Current code is as follows:

	rmsAdjustClosure rac(_adjustmentMap);
	Universe::oops_do(&rac);
	ReferenceProcessor::oops_do(&rac);
	JNIHandles::oops_do(&rac);
	Threads::oops_do(&rac, NULL);
	ObjectSynchronizer::oops_do(&rac);
	FlatProfiler::oops_do(&rac);
	Management::oops_do(&rac);
	JvmtiExport::oops_do(&rac);
	SystemDictionary::oops_do(&rac);
	vmSymbols::oops_do(&rac);
	JNIHandles::oops_do(&rac);
	CodeCache::oops_do(&rac);
	SymbolTable::oops_do(&rac);
	StringTable::oops_do(&rac);
	// adjust pointers in the permanent space / generation
	_heap->permanent_oop_iterate(&rac);


Regards,
Pressenna Sockalingasamy


PS: sorry for the double post... forgot the sample code


[prev in list] [next in list] [prev in thread] [next in thread]