Increased the default and maximum size of the CMS marking stack and the size of the parallel workers' work queues in 64-bit mode. The latter was accomplished by an increase in the width of the Taskqueue's Age struct and its Tag field in 64-bit mode.
Reviewed-by: jmasa, tonyp
Avoid overflow list walk in CMS & ParNew when it is unnecessary. Fix a couple of correctness issues, including a C-heap leak, in ParNew at the intersection of promotion failure, work queue overflow and object array chunking. Add stress testing option and related assertion checking.
Reviewed-by: jmasa
DTrace probe "contended-exit" should be fired before unparking object, or context could be lost. Probe firing was moved to proper place.
Reviewed-by: coleenp, kamg
Make reguard_stack change access to RW, not execute and use os::protect_memory with the new parameter when change needed to X.
Reviewed-by: acorn, jcoomes
In os::Linux::rebuild_cpu_to_node_map() fix the size of the CPU bitmap. Fixed arithmetic in MutableNUMASpace::adaptive_chunk_size() that could cause overflows and underflows of the chunk_size variable.
Reviewed-by: apetrusenko
The default value of UseVMInterruptibleIO is changed to false for JDK 7, but the default isn't changed for JDK 6 and earlier.
Reviewed-by: never, acorn, dholmes, kamg, alanb
When we encounter marking stack overflow during precleaning of Reference lists, we were using the overflow list mechanism, which can cause problems on account of mutating the mark word of the header because of conflicts with mutator accesses and updates of that field. Instead we should use the usual mechanism for overflow handling in concurrent phases, namely dirtying of the card on which the overflowed object lies. Since precleaning effectively does a form of discovered list processing, albeit with discovery enabled, we needed to adjust some code to be correct in the face of interleaved processing and discovery.
Reviewed-by: apetrusenko, jcoomes
6660573: G1: BigApps Failure : guarantee(satb_mq_set.completed_buffers_num() == 0,"invariant")
When exiting a mutator thread is removed from the thread list before it has a chance to flush its SATB and barrier queues. If GC happens at this moment the objects that are refererred from these queues can be moved, which will case a crash. The fix is simply to flush the buffers before removing a thread from the list.
Reviewed-by: jcoomes, tonyp
