| Timer Function Performance
As Mike Abrash points out, when timing various parts of you code, it is essential to understand how long the actual time fetch call itself takes. You'll have no hope of timing something that takes one millionth of a second if your timer call takes one ten-thousandth of a second to return the time. The time required to return from the various timeGetTime(), GetTickCount(), QuerryPerformanceCounter(), and special assembly _emit 0x0F timing methods can vary by a factor of 150. This difference can obscure your results and reduce the accuracy of your measurements. It often essential to use the fastest timer available, especialy when monitoring for communication time-outs. In one instance, replacing GetTickCount() with a faster timer resulted in transactions being sped up by a factor of more than 100. This simple test application runs a variety of timing functions and reports how quickly the timing functions are executed. It generates a "times.txt" report in the directory where the application is run. It is also easy to modify the application code and add your own timing functions for comparison. To generate the report, just double-click the TimerTiming.exe and hit retun when the DOS window prompts you to. ** Note: This code does not strictly follow Intel's guidelines for CPU Identification and the RDTSC (Read Time-Stamp Counter) instruction. For more information, see Developer.Intel.com AP-485 Intel Processor Identification and the CPUID Instruction
| |||||||||||||||||||||||||||||||||
This simple application runs a number of timing functions to determine how long each call takes. This reveals that some methods of getting the elapsed time can take 150 times longer than others. The difference in timer function call speed can make a large difference in your timing accuracy!
