Jdk1.6 JUC源码解析(2)-atomic-AtomicXXXArray
作者:大飞
功能简介:
- 数组原子量。
源码分析:
- 和原子量一样,数组原子量内部有一个Unsafe的静态引用。
private static final Unsafe unsafe = Unsafe.getUnsafe();
- 首先先看下AtomicIntegerArray。
AtomicIntegerArray的构造方法如下:
/** * Creates a new AtomicIntegerArray of given length. * * @param length the length of the array */ public AtomicIntegerArray(int length) { array = new int[length]; // must perform at least one volatile write to conform to JMM if (length > 0) unsafe.putIntVolatile(array, rawIndex(0), 0); } /** * Creates a new AtomicIntegerArray with the same length as, and * all elements copied from, the given array. * * @param array the array to copy elements from * @throws NullPointerException if array is null */ public AtomicIntegerArray(int[] array) { if (array == null) throw new NullPointerException(); int length = array.length; this.array = new int[length]; if (length > 0) { int last = length-1; for (int i = 0; i < last; ++i) this.array[i] = array[i]; // Do the last write as volatile unsafe.putIntVolatile(this.array, rawIndex(last), array[last]); } }注:当前源码来之jdk1.6,在里面会看到第一个构造方法最后就会添加一个volatile write,但在jdk1.8(从jdk1.7某个小版本开始)里面就看不到这个volatile write了,与群友讨论得知这应该是一个遗留代码,因为final完全可以保证这个语义(其他线程可以看到完全构造的内部array);第二个构造方法首先对内部final修饰的array进行赋值,然后进行数组元素copy,为了保证其他线程可以看到完全构造(内部元素copy完成)的array,所以要在copy最后加一个volatile write。
相对于AtomicInteger来说,AtomicIntegerArray里面的方法都带有下标:
/** * Atomically increments by one the element at index {@code i}. * * @param i the index * @return the updated value */ public final int incrementAndGet(int i) { while (true) { int current = get(i); int next = current + 1; if (compareAndSet(i, current, next)) return next; } }接下来看一下AtomicIntegerArray根据下标取值的get方法:
/** * Gets the current value at position {@code i}. * * @param i the index * @return the current value */ public final int get(int i) { return unsafe.getIntVolatile(array, rawIndex(i)); }方法里调用了unsafe的getIntVolatile方法。在hotspot/src/share/vm/classfile/vmSymbols.hpp中找到:
do_intrinsic(_getIntVolatile, sun_misc_Unsafe, getIntVolatile_name, getInt_signature, F_RN) \接着在hotspot/src/share/vm/opto/library_call.cpp中找到实现:
case vmIntrinsics::_getIntVolatile: return inline_unsafe_access(!is_native_ptr, !is_store, T_INT, true); ... bool LibraryCallKit::inline_unsafe_access(bool is_native_ptr, bool is_store, BasicType type, bool is_volatile) { if (callee()->is_static()) return false; // caller must have the capability! ...//省略不重要部分 if (is_volatile) { if (!is_store) insert_mem_bar(Op_MemBarAcquire); else insert_mem_bar(Op_MemBarVolatile); } if (need_mem_bar) insert_mem_bar(Op_MemBarCPUOrder); return true; }在方法的最后可以看到,如果是volatile且不是写操作的话,会加入一个Op_MemBarAcquire的内存屏障,再看下hotspot/src/cpu/x86/vm/x86_64.ad:
instruct membar_acquire() %{ match(MemBarAcquire); ins_cost(0); size(0); format %{ "MEMBAR-acquire ! (empty encoding)" %} ins_encode(); ins_pipe(empty); %}可见在x86_64下也相当于是对一个普通域的读取。
最后看一下AtomicIntegerArray根据下标设置值的set方法:
/** * Sets the element at position {@code i} to the given value. * * @param i the index * @param newValue the new value */ public final void set(int i, int newValue) { unsafe.putIntVolatile(array, rawIndex(i), newValue); }方法里调用了unsafe的putIntVolatile方法。在hotspot/src/share/vm/classfile/vmSymbols.hpp中找到:
do_intrinsic(_putIntVolatile, sun_misc_Unsafe, putIntVolatile_name, putInt_signature, F_RN) \接着在hotspot/src/share/vm/opto/library_call.cpp中找到实现:
case vmIntrinsics::_putIntVolatile: return inline_unsafe_access(!is_native_ptr, is_store, T_INT, true);和上面getIntVolatile的实现一样:
if (is_volatile) { if (!is_store) insert_mem_bar(Op_MemBarAcquire); else insert_mem_bar(Op_MemBarVolatile); }可见,如果是写操作做的话,加入Op_MemBarVolatile内存屏障,继续看hotspot/src/cpu/x86/vm/x86_64.ad:
instruct membar_volatile(rFlagsReg cr) %{ match(MemBarVolatile); effect(KILL cr); ins_cost(400); format %{ $$template if (os::is_MP()) { $$emit$$"lock addl [rsp + #0], 0\t! membar_volatile" } else { $$emit$$"MEMBAR-volatile ! (empty encoding)" } %} ins_encode %{ __ membar(Assembler::StoreLoad); %} ins_pipe(pipe_slow); %}如果是多核CPU,就会加入lock addl... 这个指令。其实就相当于是对一个volatile修饰的域的写操作喽。
- 再看下AtomicLongArray。
AtomicLongArray的内部结构和AtomicIntegerArray类似,这里不做分析,只看一下get和set方法中内部调用的unsafe的getLongVolatile和putLongVolatile方法。
首先看下get方法中的getLongVolatile方法:
/** * Gets the current value at position {@code i}. * * @param i the index * @return the current value */ public final long get(int i) { return unsafe.getLongVolatile(array, rawIndex(i)); }看下内联实现,vmSymbols.hpp,里面有如下代码:
do_intrinsic(_getLongVolatile, sun_misc_Unsafe, getLongVolatile_name, getLong_signature, F_RN) \然后找到library_call.cpp中找到相应实现:
case vmIntrinsics::_getLongVolatile: return inline_unsafe_access(!is_native_ptr, !is_store, T_LONG, true); ... bool LibraryCallKit::inline_unsafe_access(bool is_native_ptr, bool is_store, BasicType type, bool is_volatile) { if (callee()->is_static()) return false; // caller must have the capability! ... if (!is_store) { Node* p = make_load(control(), adr, value_type, type, adr_type, is_volatile); ... if (is_volatile) { if (!is_store) insert_mem_bar(Op_MemBarAcquire); else insert_mem_bar(Op_MemBarVolatile); } if (need_mem_bar) insert_mem_bar(Op_MemBarCPUOrder); return true; }首先,后面那个MemBarAcquire之前看过,在x86-64下没什么用,看下前面的make_load,找到hotspot/src/share/vm/opto/graphKit.cpp:
Node* GraphKit::make_load(Node* ctl, Node* adr, const Type* t, BasicType bt, int adr_idx, bool require_atomic_access) { assert(adr_idx != Compile::AliasIdxTop, "use other make_load factory" ); const TypePtr* adr_type = NULL; // debug-mode-only argument debug_only(adr_type = C->get_adr_type(adr_idx)); Node* mem = memory(adr_idx); Node* ld; if (require_atomic_access && bt == T_LONG) { ld = LoadLNode::make_atomic(C, ctl, mem, adr, adr_type, t); } else { ld = LoadNode::make(_gvn, ctl, mem, adr, adr_type, t, bt); } return _gvn.transform(ld); }可见,至少对这个long值得加载是原子的(这里的原子操作应该指的是将long的高4字节和低4字节的操作合并成一个原子操作,比如某些平台不支持非volatile的long/double域的原子操作)。
然后看下set方法中的setLongVolatile方法:
/** * Sets the element at position {@code i} to the given value. * * @param i the index * @param newValue the new value */ public final void set(int i, long newValue) { unsafe.putLongVolatile(array, rawIndex(i), newValue); }直接到library_call.cpp中看实现:
case vmIntrinsics::_putLongVolatile: return inline_unsafe_access(!is_native_ptr, is_store, T_LONG, true); ... bool LibraryCallKit::inline_unsafe_access(bool is_native_ptr, bool is_store, BasicType type, bool is_volatile) { if (callee()->is_static()) return false; // caller must have the capability! ... if (is_volatile) { if (!is_store) insert_mem_bar(Op_MemBarAcquire); else insert_mem_bar(Op_MemBarVolatile); } if (need_mem_bar) insert_mem_bar(Op_MemBarCPUOrder); return true; }之前说过,这里加入了一个lock addl ... 的内存屏障。
- 最后看下AtomicReferenceArray。
AtomicReferenceArray的内部结构和AtomicIntegerArray类似,有一点点细节上的区别:
public class AtomicReferenceArray<E> implements java.io.Serializable { private static final long serialVersionUID = -6209656149925076980L; private static final Unsafe unsafe; private static final int base; private static final int shift; private static final long arrayFieldOffset; private final Object[] array; // must have exact type Object[] static { int scale; try { unsafe = Unsafe.getUnsafe(); arrayFieldOffset = unsafe.objectFieldOffset (AtomicReferenceArray.class.getDeclaredField("array")); base = unsafe.arrayBaseOffset(Object[].class); scale = unsafe.arrayIndexScale(Object[].class); } catch (Exception e) { throw new Error(e); } if ((scale & (scale - 1)) != 0) throw new Error("data type scale not a power of two"); shift = 31 - Integer.numberOfLeadingZeros(scale); } private long checkedByteOffset(int i) { if (i < 0 || i >= array.length) throw new IndexOutOfBoundsException("index " + i); return byteOffset(i); } private static long byteOffset(int i) { return ((long) i << shift) + base; }首先,AtomicReferenceArray里面多了一个arrayFieldOffset,这个域用来支持反序列化的。其次,与AtomicIntegerArray不同,AtomicReferenceArray并没有scale域,取而代之的是shift域。阅读代码可知,其实目的都是计算rawIndex = base + index * scale,只不过AtomicReferenceArray里面把一部分运算转换为等价的位操作(当然前提是scala为2的幂)。(了解更多位操作技巧,可参考Hacker's Delight)。
其他涉及到unsafe的操作前面都分析过,这里不做分析了。
好了,源码就分析到这里!