- It is intended that benchmark_plot.sh be used to drive benchmark.py and plot.py
- It is recommended you experiment with commands like

  ./benchmark_plot.sh -t my_tag --level1 true --level2 false --level3 false --benchmark true --plot false

  ./benchmark_plot.sh -t my_tag --level1 true --level2 false --level3 false --benchmark false --plot true

  ./benchmark_plot.sh -t my_tag --level1 true --level2 false --level3 false --benchmark false --plot true --theo_max true

  look in directory blas1/my_tag for output .csv files and plots

- On a machine where you only want to run benchmarks, call benchmark_plot.sh with arguments
  --benchmark true --plot false
- There are .yaml files for each function that specify input argument sizes
- rocblas-bench is called, and output will be in .csv files
- If you are in your fork (not ROCmSoftwarePlatform) the .csv files can be checked in to github
  on a remote machine, and checked out for plotting on a local machine
- There is plotting of gflops vs size or gflops/theoretical_max_gflops vs size
- plots are in .png files
- For gfx90a here are checked in .csv result files and .png plot files
- to get the gfx90a .cvs and .png files the following thre calls were made to benchmark_plot.sh

  ./benchmark_plot.sh -t gfx90a  --level1 true  --level2 true --level3 true  --benchmark true --plot false

  ./benchmark_plot.sh -t gfx90a  --level1 true  --level2 true --level3 true  --benchmark false --plot true

  ./benchmark_plot.sh -t gfx90a  --level1 true  --level2 true --level3 true  --benchmark false --plot true --theo_max true
