switch和if的性能比较

      一直认为switch的效率高于if，特别对于条件大于3的时候，以前也貌似看过资料说：switch的查找类似二叉树，if的则是线性。

      最近做了一个简单的测试，对有多个条件式的switch和if做比较，发现两者的效率几乎相同，if的效率甚至高于switch；<测试环境sun jdk6.1.13>

     40个条件式的测试，测试1000 0000次，if耗时219ms， switch耗时234ms，平均都在4-5k/s；

     30个条件式的测试，测试1000 0000次，if耗时188 switch耗时172平均都在5-6kw/s，

     20个条件式，测试1000 0000次，if耗时109mswitch耗时125ms，平均在8-9kw/s

 

      总结：

      1 经过jdk的改进，目前switch和if的性能上差别微乎其微

      2 条件式的多少对于一般应用的整体性能来说，影响也非常小。

      3 影响效率的最大因子是条件式的数量，而非if或switch

测试代码

private static void testSitchAndIf(){
  int testCount = 10000000;
  int flag = 0;
  
  Random random = new Random(testCount);
  int[] flags = new int[testCount];
  for(int i = 0; i < testCount; i++){
   flags[i] = random.nextInt() + 1;
  }
  // 先统一取一遍值，保证测试都用缓存值
  for(int i = 0; i < testCount; i++){
   flag = flags[i];
  }
  long time1 = System.currentTimeMillis();
  for(int i = 0; i < testCount; i++){ 
   //flag = random.nextInt() + 1;
   flag = flags[i];
   if(flag == 1){}
   else if(flag == 2){}
   else if(flag == 3){}
   else if(flag == 4){}
   else if(flag == 5){}
   else if(flag == 6){}
   else if(flag == 7){}
   else if(flag == 8){}
   else if(flag == 9){}
   else if(flag == 10){}
   else if(flag == 11){}
   else if(flag == 12){}
   else if(flag == 13){}
   else if(flag == 14){}
   else if(flag == 15){}
   else if(flag == 16){}
   else if(flag == 17){}
   else if(flag == 18){}
   else if(flag == 19){}
   else if(flag == 20){}
   else if(flag == 21){}
   else if(flag == 22){}
   else if(flag == 23){}
   else if(flag == 24){}
   else if(flag == 25){}
   else if(flag == 26){}
   else if(flag == 27){}
   else if(flag == 28){}
   else if(flag == 29){}
   else if(flag == 30){}
//   else if(flag == 31){}
//   else if(flag == 32){}
//   else if(flag == 33){}
//   else if(flag == 34){}
//   else if(flag == 35){}
//   else if(flag == 36){}
//   else if(flag == 37){}
//   else if(flag == 38){}
//   else if(flag == 39){}
//   else if(flag == 40){}   
  }
  
  long time2 = System.currentTimeMillis();  
  for(int i = 0; i < testCount; i++){
   flag = flags[i];
   switch(flag){
    case 1: break;
    case 2: break;
    case 3: break;
    case 4: break;
    case 5: break;
    case 6: break;
    case 7: break;
    case 8: break;
    case 9: break;
    case 10: break;
    case 11: break;
    case 12: break;
    case 13: break;
    case 14: break;
    case 15: break;
    case 16: break;
    case 17: break;
    case 18: break;
    case 19: break;
    case 20: break;
    case 21: break;
    case 22: break;
    case 23: break;
    case 24: break;
    case 25: break;
    case 26: break;
    case 27: break;
    case 28: break;
    case 29: break;
    case 30: break;
//    case 31: break;
//    case 32: break;
//    case 33: break;
//    case 34: break;
//    case 35: break;
//    case 36: break;
//    case 37: break;
//    case 38: break;
//    case 39: break;
//    case 40: break;
//    case 41: break;
   }   
  }
  
  long time3 = System.currentTimeMillis();
  System.out.println("loop count:" + testCount);
  System.out.println("if consume time:" + (time2 - time1) + ",avg:" + testCount / (time2 - time1) * 1000);
  System.out.println("switch consume time:" + (time3 - time2) + ",avg:" + testCount / (time3 - time2) * 1000);
 }

 

 

 

评论

2 楼 fishermen 2010-08-27  

循环几十次，可计量时间短，而且样本数量也少，不容易看出实际效果。

1 楼 jarry-li 2010-08-23  

楼主为什么不是循环几十次，if..else 有10 000 000次呢