|
CSE P527, Wi '18: The Course FAQ |
|
 CSE Home |
About Us |
Search |
Contact Info |
Common questions:
Q. How do I time my program?You've asked us to measure the time taken by our algorithms. How do I to this? Should I buy that $3.99 Rolex I keep getting emails about? --- T. Treasure
A. Dear Timeless,
No, save the Rolex to impress that Certain Someone. Computers generally have built-in clocks. Accesssing them
depends on the language. For example, in C/C++, you do:
#include <time.h>
int main(void) {
clock_t start_tick, end_tick;
double elapsed;
start_tick = clock();
my_really_fast_algorithm(42);
end_tick = clock();
elapsed = (end_tick - start_tick) / (double)CLOCKS_PER_SEC;
printf("My Really Fast Algorithm took %f seconds!\n", elapsed);
return 0;
}
CLOCKS_PER_SEC, defined in time.h, tells how to scale the system-dependent
clock() results to seconds. (It is often 1000, but that does not tell you the resolution of the
timer, which is often 10 or .001 milliseconds. Oldtimers remember 16.7 millisecond clocks, i.e., 60 hertz powerline
frequency...)
Similarly, in Java public static long currentTimeMillis() returns the current time in millisecond units
(but not necessarily millisecond accuracy).
Specific systems may have additional features. These will be less portable, but may be more convenient. E.g., a former student pointed out these "high performance timer" functions for Windows:
QueryPerformanceFrequency, returns a Boolean indicating whether the system supports HPCs, and takes a pointer to a 64 bit integer in which to store the clock
frequency, in ticks per second.
QueryPerformanceCounter also returns a Boolean indicating support, and takes a pointer to a 64
bit integer in which to store the current clock.
If you have a pointer to other useful faclities in these or other languages, please send them to me or post to Catalyst GoPost.
 |
Computer Science & Engineering University of Washington Box 352350 Seattle, WA 98195-2350 (206) 543-1695 voice, (206) 543-2969 FAX |