Fetch-and-add

From Wikipedia, the free encyclopedia

In computer science, the fetch-and-add CPU instruction is a special instruction that atomically modifies the contents of a memory location. It is used to implement Mutual exclusion and concurrent algorithms in multiprocessor systems.

In uniprocessor systems, it is sufficient to disable interrupts before accessing a critical region. However, in multiprocessor systems, it is impossible and undesirable to disable interrupts on all processors at the same time; and even with interrupts disabled two or more processors could be attempting to access the same memory at the same time. The fetch-and-add instruction, allows any processor to atomically increment a value in memory location, preventing such multiple processor collisions.

Maurice Herlihy (1993) proved that fetch-and-add is inferior to compare-and-swap.

[edit] Implementation

The standard fetch and add -instruction behaves like the following function. Crucially the entire function is executed atomically: no process can interrupt the function mid-execution and hence see a state that only exists during the execution of the function. This code only serves to help explain the behaviour of fetch-and-add; atomicity requires explicit hardware support and hence can not be implemented as a simple high level function.

<< atomic >>
function FetchAndAdd(address location) {
    int value := *location
    *location := value + 1
    return value
}

With fetch-and-add primitive a mutual exclusion lock can be implemented as:

 record locktype {
    int ticketnumber
    int turn
 }
 procedure LockInit( locktype* lock ) {
    lock.ticketnumber := 0
    lock.turn  := 0
 }
 procedure Lock( locktype* lock ) {
    int myturn := FetchAndAdd( &lock.ticketnumber )
    while lock.turn ≠ myturn 
        skip // spin until lock is acquired
 }
 procedure UnLock( locktype* lock) {
    FetchAndAdd( &lock.turn )
 }

These routines provide a mutual-exclusion lock when following conditions are met:

Locktype data structure is initialized with function LockInit before use
Number of tasks waiting for the lock does not exceed INT_MAX at any time
Integer datatype used in lock values can 'wrap around' when continuously incremented

[edit] x86 implementation

In the x86 architecture, the instruction ADD with the first operand specifying a memory location is a fetch-and-add instruction that has been there since the 8086 (it just wasn't called that then), and with the LOCK prefix, is atomic across multiple processors. However, it could not return the original value of the memory location (though it returned some flags) until the 486 introduced the XADD instruction.