Reviewed-by: rkennke, coleenp, kbarrett, dcubed
Reviewed-by: shade, eosterlund
Improve G1 worst-case latencies by making the full GC parallel. Reviewed-by: tschatzl, sangheki, ehelin