diff options
| -rw-r--r-- | README.md | 8 | ||||
| -rw-r--r-- | doc/analysis-textual.md | 8 | ||||
| -rw-r--r-- | doc/analysis-visual.md | 10 |
3 files changed, 26 insertions, 0 deletions
@@ -31,6 +31,14 @@ Legacy documentation; may be outdated: * [Energy Benchmarks with Multipass](doc/energy-multipass.md) (DE) * [Performance Benchmarks for Multipass](doc/nfp-multipass.md) (DE) +## Data Analysis + +It can be helpful to visualize acquired data points to get a feel for how the observed performance attributes behave. +Most of the options and methods documented here work for all three scripts: analyze-archive, analyze-kconfig, and analyze-log. + +* [Textual Data Analysis](doc/analysis-textual.md) +* [Visual Data Analysis](doc/analysis-visual.md) + ## Model Generation dfatool supports six types of performance models: diff --git a/doc/analysis-textual.md b/doc/analysis-textual.md new file mode 100644 index 0000000..bea2099 --- /dev/null +++ b/doc/analysis-textual.md @@ -0,0 +1,8 @@ +# Textual Data Analysis + +## Parameter and NFP overview + +`--info` prints the names and ranges of observed **parameters** (configuration values) and **Observations** (performance attributes / NFPs). +It is also helpful to determine if and how parameter filters affect the subset of data used for analysis. + +## parameter filters diff --git a/doc/analysis-visual.md b/doc/analysis-visual.md new file mode 100644 index 0000000..eb1914b --- /dev/null +++ b/doc/analysis-visual.md @@ -0,0 +1,10 @@ +# Visual Data Analysis + +The parameter and NFP filters from [Textual Data Analysis](analysis-textual.md) apply here as well. + +## Raw Data Visualization + +There are two ways of visualizing all measured data independent of their parameters: + +* `--boxplot PREFIX` writes boxplots of all observations to PREFIX/(name)-(attribute).pdf. These may be helpful to see which observations are stable and which show a lot of variance, possibly due to the influence of parameters. +* `--plot-unparam=name:attribute:ylabel` plots all observations of name/attribute in the order in which they were observed. Useful to identify trends, especially when the parameter variation scheme is known as well. |