Improve Libjpeg-turbo tjbench tool
Improve benchmark tool such that there is a standardized benchmark run option, and remove deps on speed of IO system in use, such that tmpfs, disk or SD media does not affect results.
Blueprint information
Related branches
Related bugs
Sprints
Whiteboard
Headline:
tjbench the benchmark tool packaged with libjpeg-turbo now contains improvements such that the benchmark is more reliable and accurately measuring processor usage during libjpeg-turbo operations. Differences in IO speed which would previously cause swings in benchmark results have been removed. Further a standard benchmark mode has been added so users no longer need worry about which options they should pick to benchmark.
Acceptance:
tjbench no longer sees variability in results due to IO speeds, such as tmpfs vs SD media. A new default mode to run recognized measurements to detect regressions in library allowing for standardization when reporting performance results.
Work Items
Work items:
Gain consensus on the list on what standardized options should be used for performance measurements: TODO
Examine and recode to remove IO variability: TODO
Code standardized options: TODO
Submit code upstream: TODO
