The Latest Changes With Phoronix Test Suite 7.0-Ringsaker
Previously covered as one of the usability enhancements for Phoronix Test Suite 7.0 was a new system information view from the CLI to present the system hardware/software in a much nicer fashion. A few tweaks have been done to that since the last posting.
Tweaks to that are still ongoing and feedback welcome as always. Also still working on a few Phodevi detection improvements for BSD/macOS too this cycle.
One of the new sub-commands this cycle is phoronix-test-suite winners-and-losers and you pass that a result file. So far this command just spits out a few statistics about the test runs with the most wins and losses in a given result file. Feedback is welcome on any other result file statistics you might find useful.
Making use of colors in supported terminals was initially added in the Phoronix Test Suite 6.8 release, but it's been extended in Phoronix Test Suite 7.0. Now there are some handy additions like when displaying results of multiple test runs, the best result will automatically be highlighted green while the worst result of that run will be colored red, etc.
One of the big set of changes this week in Git was a low-level rework of the result parsing functionality. The end goal of this is now that one test run can natively generate multiple test results at the same time, i.e. multiple graphs of data generated off that first run at the same time. There's been ways to do that in a non-ideal way previously like making use of a result cache share to capture and share that log file among multiple test runs and then exposing each output as its own test run option, but now it can be done natively via the conventional result-definition.xml file by simply specifying multiple matches. So in the first example, showing both the IOPS and MB/s output to the terminal at the same time and storing them to separate graphs. Aside from the result cache share, there were also other "extra/linked data" options for generating additional graphs, but instead it's now all using this native single-run-to-multiple-result-objects capability. Long story short, much nicer (while allowing test profiles to still maintain forwards and backwards compatibility) and allows for more easily running of complex workloads that may generate multiple but independent and interesting data points.
There is also now a phoronix-test-suite debug-result-parser sub-command for being able to easily verify the result parsing behavior on previously-generated result file logs.
More features are still being worked on for the Phoronix Test Suite 7.0 release -- including Phoromatic and OpenBenchmarking.org. The 7.0-Ringsaker release will come later this quarter.
11 Comments
