Ian,
Per your suggestions, I reran my original timing experiments using:
(a) mach_absolute_time() for time measurements, and
(b) random (uniformly distributed) input data:
- 100 * drand48() for sqrt -vice- vvsqrt tests, &
- 20π * drand48() for sin -vice- vvsin tests.
The results are similar to the original set but showing somewhat better vForce/scalar performance in both cases;
- ~22.5 nsec/input value for vvsqrt() vice ~30 nsec for scalar sqrt() loop, &
- ~44 nsec/input value for vvsin() vice ~53 nsec for scalar sin() loop.
These were run, as before, using 10,000 pt input sets and (to my satisfaction) seem in general agreement with your & Jeff Biggus' graphs.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
PerfOptimization-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
