These rough tests give an idea about which components will hold up a real-time processing chain and which components are good targets for optimization. Only frame processing components are tested because they are orders of magnitude slower than position processing components.
Custom Desktop
Intel Core i7-5820K CPU @ 3.30GHz
GeForce GTX 970 GPU with CUDA 7.5
1000x 1MP frames served using oat-frameserve test which has
the following runtime characteristics with no listening
components:
- real 0m0.073s
- user 0m0.049s
- sys 0m0.026s
-
bsub- real 0m0.881s
- user 0m0.079s
- sys 0m0.037s
-
mask- real 0m1.300s
- user 0m0.074s
- sys 0m0.041s
-
mog- real 0m1.744s
- user 0m0.077s
- sys 0m0.039s
-
undistort- real 0m29.946s
- user 0m0.066s
- sys 0m0.068s
- Note: slower than laptop...
- Note: replace open-cv implementation with shader. There are lots of tutorials on this.
-
diff- real 0m2.835s
- user 0m0.064s
- sys 0m0.056s
-
hsv- real 0m4.683s
- user 0m0.071s
- sys 0m0.067s
-
trsh- real 0m0.545s
- user 0m0.028s
- sys 0m0.012s
Lenovo ThinkPad X1 Carbon 3rdi
Intel Core i7-5600U CPU @ 2.60GHzi
1000x 1MP frames served using oat-frameserve test which has the following
runtime characteristics with no listening components:
- real 0m0.051s
- user 0m0.036s
- sys 0m0.008s
-
bsub- real 0m1.278s
- user 0m0.080s
- sys 0m0.008s
-
mask- real 0m2.049s
- user 0m0.060s
- sys 0m0.020s
-
mog- real 0m13.213s
- user 0m0.056s
- sys 0m0.068s
- Note: no CUDA support
-
undistort- real 0m24.106s
- user 0m0.096s
- sys 0m0.060s
-
No position source, just date and sample added to frame.
- real 0m0.934s
- user 0m0.064s
- sys 0m0.024s
-
Single position source with all display options turned on.
- real 0m6.274s
- user 0m0.132s
- sys 0m0.040s
-
diff- real 0m3.701s
- user 0m0.048s
- sys 0m0.024s
-
hsv- real 0m7.422s
- user 0m0.064s
- sys 0m0.028s