There is a race between the concurrent refinement threads and the humongous object allocation that can cause the concurrent refinement threads to corrupt the part of the BOT that it is being initialized by the humongous object allocation operation. The solution is to do the humongous object allocation in careful steps to ensure that the concurrent refinement threads always have a consistent view over the BOT, region contents, and top. The fix includes some very minor tidying up in sparsePRT.
Reviewed-by: jcoomes, johnc, ysr
The remembered set iterator for sparse tables incorrectly assumes that index 0 corresponds to the bottom of the heap, not address 0 as it is the case.
Reviewed-by: ysr, jmasa
Implemented block-based work stealing. Moved copying during the rset scanning phase to the main copying phase. Made the size of rset table depend on the region size.
Reviewed-by: apetrusenko, tonyp
The fix addresses two memory leaks in G1 code: (1) _evac_failure_scan_stack - a resource object allocated on the C heap was not freed; (2) RSHashTable were linked into deleted list which was only cleared at full GC.
Reviewed-by: tonyp, iveresov
For heaps larger than 32Gb, the number of heap regions overflows the data type used to hold the region index in the SparsePRT structure. Changed the region indexes, card indexes, and RSet hash table buckets to ints and added some size overflow guarantees.
Reviewed-by: ysr, tonyp
The _expanded flag of the sparse RSets is not reset and this can leave a RSet in an inconsistent state if it is expanded more than once. Also, we should be iterating over the _cur, instead of the _next, sparse table
Reviewed-by: apetrusenko, iveresov
