Simulator build for research of virtual reality optimization. The power consumption constants is from here.
- Client-side Optimizations
- Make use of threshold
- lower threshold means client side need to draw something by guessing
- to make threshold lower (~= 0.8), we need to draw more precisely, therefore:
- (static) find frame (closest in time) that covers the missing part
- (dynamic) use motion vector to draw the missing part
- Make use of threshold
-
First of all, you need a recent version of rust compiler (maybe >= 1.25). For using the simulator, you need a
user viewport fileand atracedump fileand these two are the first argument and second argument that the simulator need to run. -
The next two argument is the
thresholdandsegment sizewhich is simple, but notice that the threshold isa floating point >= 0 && <= 1. -
The next four argument is the width and height for level one and level two viewport respectively. And if the level one and level two viewport has the same dimension (ex: 2000 1000 2000 1000) then the simulator simulate with only level one frame and the full size frame.
-
The last two argument is power and O0, which is the content to dump and the flag of optimization level, but now the simulator only support with
power O0. -
To conclude, you can simply change the dimension in following line:
cargo run user_viewport_result/Elephant-training-2bpICIClAIg tracedump/elephant.txt object_as_cluster/elephant.json 0.96 20 1440 1440 1440 1440 power O0 >> elephant_power.txt &
cargo run user_viewport_result/Rhino-training-7IWp875pCxQ tracedump/rhinos.txt object_as_cluster/rhinos.json 0.96 20 1440 1440 1440 1440 power O0 >> rhino_power.txt &
cargo run user_viewport_result/Rollercoaster-8lsB-P8nGSM tracedump/roller.txt object_as_cluster/roller.json 0.96 20 1440 1440 1440 1440 power O0 >> roller_power.txt &
cargo run user_viewport_result/Timelapse-CIw8R8thnm8 tracedump/nyc.txt object_as_cluster/nyc.json 0.96 20 1440 1440 1440 1440 power O0 >> nyc_power.txt &
cargo run user_viewport_result/Paris-sJxiPiAaB4k tracedump/paris.txt object_as_cluster/paris.json 0.96 20 1440 1440 1440 1440 power O0 >> paris_power.txt &- Output format:
wifi soc threshold level-1-hit level-2-hit level-3-hit segment-hit- For instance:
336.30700560388595 3449.5758177213975 0.96 0.25491287519697325 0 0.745087124803027 0.20603531096488842
- For instance:
- TODO: should fill this part as detail as possible!
- Counting power consumption by using the hit rate calculated in the simulate
function in
src/simulator.rs
The data is visualized in the vros-dataset. And you might be interested in the summary of the data.