请教likely和unlikely

turbo99cn

#define likely(x)        __builtin_expect(!!(x), 1)
#define unlikely(x)        __builtin_expect(!!(x), 0)

#if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
#define __builtin_expect(exp,c) (exp)
#endif


看了好久没看明白，怎么回事。
likely,unlikely的作用是什么？
那个__builtin_expect的宏定义有什么作用？
为什么  ！！(x)   这么用？后面的  1 是干什么的？

albcamus

__builtin_expect在gcc的info手册中：

-- Built-in Function: long __builtin_expect (long EXP, long C)
     You may use `__builtin_expect' to provide the compiler with branch
     prediction information.  In general, you should prefer to use
     actual profile feedback for this (`-fprofile-arcs'), as
     programmers are notoriously bad at predicting how their programs
     actually perform.  However, there are applications in which this
     data is hard to collect.

     The return value is the value of EXP, which should be an integral
     expression.  The value of C must be a compile-time constant.  The
     semantics of the built-in are that it is expected that EXP == C.
     For example:

          if (__builtin_expect (x, 0))
            foo ();

     would indicate that we do not expect to call `foo', since we
     expect `x' to be zero.  Since you are limited to integral
     expressions for EXP, you should use constructions such as

          if (__builtin_expect (ptr != NULL, 1))
            error ();

     when testing pointer or floating-point values.

jeffshia

只记得《lkd》一书中说，gcc有内嵌的指令，这些指令可以优化很可能发生的和很不可能发生的条件分支，至于为什么这样，我也不清楚，估计是和编译器相关的东西。

jeffshia

在gcc的内嵌优化指令中，会把最可能发生的事情放到流水线的最前面，而把发生比较少可能性的事情放在比较靠后的位置。大概就是这样子的。