Replacing the atomic_inc_and_test Instruction

The function adds integers and checks whether the return value is 0.

• Code on x86:

  static inline int atomic_inc_and_test(atomic_t *v) 
  { 
  unsigned char c; 
  asm volatile(LOCK_PREFIX "incl %0; sete %1" : "+m" (v->counter), "=qm" (c) : : "memory"); 
  return c != 0; 
  }

• Alternatives for Kunpeng processors:

  Method 1: Use the built-in atomic operations provided by the GCC.

  __sync_add_and_fetch(&_value.counter,i)

  Method 2: Use the inline assembly (transfer 1 to i and check the result).

  static inline void atomic_fetch_add(int i, atomic_t *v) 
  { 
  unsigned long tmp; 
  int result, val; 
  // Write prefetch instructions.
  prefetchw(&v->counter); 
  __asm__ volatile( 
  "\n\t" 
  "@ atomic_fetch\n\t" 
  "1: ldrex %0, [%4]\n\t" @result, tmp 
  // Perform the v->counter+i (5%) operation and save the execution result to the register where val (%1) is located.
  " add %1, %0, %5\n\t" @result, 
  " strex %2, %1, [%4]\n\t" @tmp, result,tmp 
  " teq %2, #0\n\t" @tmp 
  " bne 1b" 
  : "=&r"(result), "=&r"(val), "=&r"(tmp), "+Qo"(v->counter) 
  : "r"(&v->counter), "Ir"(i) 
  : "cc"); 
  return result; 
  }