假设我们有类PetersonLock,如下所示:
class PetersonLock {
AtomicBoolean flag[] = new AtomicBoolean[2];
volatile int victim;
public void Acquire(int id) {
flag[id].set(true);
victim = id;
while (flag[1-id].get() && victim == id);
}
public void Release(int id) {
flag[id].set(false);
}
}
其中id是1或0.我现在有以下类来分层使用3个PetersonLocks进行4个进程。
class Lock4Pete{
PetersonLock[] lock = new PetersonLock[3];
public void Acquire(int id) {
lock[0].Acquire(id/2);
lock[1+id/2].Aquire(id%2);
}
public void Release(int id) {
lock[1+id/2].Release(id%2);
lock[0].Release(id/2);
}
}
id为0,1,2或3时。
我不明白这背后的想法,我也不知道如何修复此代码。我不知道他们在这里要做什么。为什么我们需要为4个进程提供3个锁,为什么每个进程都使用lock [0]?
一些帮助将非常感激。这不是家庭作业,而是我不太懂的练习。
答案 0 :(得分:1)
4个线程分为2个分区,每个分区包含2个线程。 id/2
指定线程所属的bin,id%2
指定其bin中线程的索引。
让我们重写代码以将id/2
和id%2
作为单独的变量处理。
class Lock4Pete {
// This lock determines the active bin.
PetersonLock masterLock = new PetersonLock;
// Each of these locks guards the corresponding bin.
PetersonLock[] binLocks = new PetersonLock[2];
public void Acquire(int bin, int index) {
// After this line is executed in one of the threads,
// any thread from a *different* bin will have to wait
// until this thread calls masterLock.Release(bin);
// before it can execute masterLock.Acquire(another_bin);
masterLock.Acquire(bin);
// It is possible that more than one thread reaches this point
// simultaneously, but they are guaranteed to be from the same bin.
// Now we only need to make sure that threads from that bin can
// neither acquire the lock simultaneously nor come to a deadlock.
// After this line is executed,
// any thread from the *same* bin will have to wait until
// this thread calls binLocks[bin].Release(index);
// before it can execute binLocks[bin].Aquire(another_index);
binLocks[bin].Aquire(index);
// Thus, only one thread at a time can reach the end of this
// method and acquire the lock.
}
public void Release(int bin, int index) {
binLocks[bin].Release(index);
masterLock.Release(bin);
}
}