< prev index next >
src/share/vm/gc/parallel/psTasks.hpp
Print this page
rev 12906 : [mq]: gc_interface
*** 147,157 ****
// exceed the top of the generation. Note that having fewer GC threads
// than stripes works because all the tasks are executed so all stripes
// will be covered. In this example if 4 tasks have been created to cover
// all the stripes and there are only 3 threads, one of the threads will
// get the tasks with the 4th stripe. However, there is a dependence in
! // CardTableExtension::scavenge_contents_parallel() on the number
// of tasks created. In scavenge_contents_parallel the distance
// to the next stripe is calculated based on the number of tasks.
// If the stripe width is ssize, a task's next stripe is at
// ssize * number_of_tasks (= slice_stride). In this case after
// finishing stripe 0 in slice 0, the thread finds the stripe 0 in slice1
--- 147,157 ----
// exceed the top of the generation. Note that having fewer GC threads
// than stripes works because all the tasks are executed so all stripes
// will be covered. In this example if 4 tasks have been created to cover
// all the stripes and there are only 3 threads, one of the threads will
// get the tasks with the 4th stripe. However, there is a dependence in
! // PSCardTable::scavenge_contents_parallel() on the number
// of tasks created. In scavenge_contents_parallel the distance
// to the next stripe is calculated based on the number of tasks.
// If the stripe width is ssize, a task's next stripe is at
// ssize * number_of_tasks (= slice_stride). In this case after
// finishing stripe 0 in slice 0, the thread finds the stripe 0 in slice1
< prev index next >