Benchmark: Corpse Reviver
This module provides a command-line benchmarking script built for the purpose of measuring the performance benefit of SCV-CR. Two types of JSON output are produced: analysis results and run-time results. Analysis results collect information during SCV-CR optimization such as blame data. Run-time results measure benchmark performance after optimization.
1 Command Line
Short |
| Long |
| Description |
-b |
| --baseline |
| Run baseline measurements, in other words, don’t optimize with SCV-CR. |
-c |
| --cutoff |
| Set the maximum number of components to measure exhaustively before sampling. |
-g |
| --gc-log |
| Log garbage collection statistics during analysis. This can cause the analysis to slow down significantly; that’s why it’s disabled by default. |
-i |
| --iterations |
| Set the number of iterations to run for each configuration. |
-n |
| --no-skip |
| Don’t skip the analysis of modules prefixed with _. This the convention for indicating modules that should be treated as opaque. |
-o |
| --output |
| Set the directory where analysis results will be placed. By default, it is the current working directory. |
-r |
| --resume |
| Resume an existing setup. In other words, run the experiment described by existing in files in $XDG_DATA_HOME/gtp-measure. After running a baseline experiment you should use this option for the optimized experiment so it uses the same samples as the baseline. |
-w |
| --worker-count |
| Set the number of parallel workers to use for measuring. By default, this is 1. Each worker is given one benchmark to process at a time. |
-R |
| --num-samples |
| Set the number of samples to take when measuring a benchmark approximately. |
-S |
| --sample-factor |
| Set the sample size as a factor of the number of components in the benchmark. |
See GTP configuration for more details on some of these options.