Use clock_gettime instead of clock for benchmarks #126
Conversation
src/apps/applib/include/benchmark.h
Outdated
| BODY; \ | ||
| } \ | ||
| END_TIMER(duration); \ | ||
| printf("\t-- %s: %Le ns per iteration (%d iterations)\n", name, \ |
There was a problem hiding this comment.
Can we change the formatting for nanoseconds here? In the example you posted on #125 I find it really hard to easily see order-of-magnitude differences. I'd prefer to still express this in ms, or maybe translate this to ops/second?
There was a problem hiding this comment.
I get that. I was considering working out a way to change the unit dynamically.
src/apps/applib/include/benchmark.h
Outdated
| elapsed.tv_sec--; \ | ||
| elapsed.tv_nsec = 1E9 + elapsed.tv_nsec; \ | ||
| } \ | ||
| const long double var = elapsed.tv_sec * 1E9 + elapsed.tv_nsec |
There was a problem hiding this comment.
This still looks like nanoseconds. Am I missing something?
There was a problem hiding this comment.
Wow good catch. I must have forgotten about the final conversion back to microseconds.
src/apps/applib/include/benchmark.h
Outdated
| elapsed.tv_sec--; \ | ||
| elapsed.tv_nsec = 1E9 + elapsed.tv_nsec; \ | ||
| } \ | ||
| const long double var = (elapsed.tv_sec * 1E9 + elapsed.tv_nsec) / 1E3 |
There was a problem hiding this comment.
Maybe add named constants for these scaling factors?
There was a problem hiding this comment.
I had thought that might be more straightforward, just got a bit verbose. Let me post the diff and you can tell me your opinion.
There was a problem hiding this comment.
Looks good to me, makes it clearer what the units are
Use clock_gettime instead of clock for benchmarks
No description provided.