Mark-Copy's first phase marks all live objects, and constructs per-block unidirectional Rsets and count the vol of live data of each block.引自 Ch10 Other partitioned schemes
The illustration is very similar to G1. Is the mark the "pre write barrier(SATB)" ?
Two advantage arise from using marker rather than mutator construct rsets: 1. precise rsets; 2.no duplicate引自 Ch10 Other partitioned schemes
The above is not very clear to me...
Because the marker counts the vol of live data, hence it could determine how many blocks can be evacuated in each pass.引自 Ch10 Other partitioned schemes
It'd better figure out how it determines the CSet and the algorithm of how the collector tries to meet the soft pause time goal.
Mark-copy has disadvantages: each full collection scans every object twice, once to mark and once to copy. Marking reuires space for a mark-stack and for Rsets. Each copying pass may require thread sstacks and global data to be rescanned.引自 Ch10 Other partitioned schemes
表示其中内容是对原文的摘抄