用 RDTSC 和 CPUID 測量效能的範例

說明

• 根據 How to Benchmark Code Execution Times on Intel® IA-32 and IA-64 Instruction Set Architectures 這篇文章

下載

• http://adf.ly/1Bp51b

結果

• Loading hello module...
  loop_size:0 >>>> variance(cycles): 36990; max_deviation: 60472 ;min time: 476
  loop_size:1 >>>> variance(cycles): 30499; max_deviation: 52876 ;min time: 480
  loop_size:2 >>>> variance(cycles): 2974050; max_deviation: 540140 ;min time: 476
  .........
  .........
  loop_size:997 >>>> variance(cycles): 931008; max_deviation: 304056 ;min time: 472
  loop_size:998 >>>> variance(cycles): 989013; max_deviation: 304684 ;min time: 460
  loop_size:999 >>>> variance(cycles): 2738977; max_deviation: 302940 ;min time: 448
  
  total number of spurious min values = 268
  total variance = 8371656
  absolute max deviation = 63107276
  variance of variances = 17465975709446
  variance of minimum values = 195
  

心得

• 結果比原文中的還差，大概是因為我把避免 CPU 被搶奪和中斷的部分拔掉且在 user mode 執行的關係
• 改進版:
  用 RDTSC，CPUID 和 RDTSCP 測量效能的範例

資源

• How to Benchmark Code Execution Times on Intel® IA-32 and IA-64 Instruction Set Architectures