galitime 0.3.0

Introduction

galitime benchmarks commands and records timing and resource statistics in a simple tab-delimited format. It wraps the system time command, normalizes the output into stable column names, and can repeat commands across multiple runs. This is particularly helpful for benchmarking research tools and workflows that need to be tested and evaluated on different platforms.

The tool is inspired by the benchmarking script in Phylign, originally developed by Leandro Lima.

Quick example

Install from PyPI:

pip install -U galitime

Benchmark a command and print the result to standard output:

galitime -l stdout "sleep 0.1"

Write the benchmark output to a file:

galitime --log time.log "ls"

Use GNU Time explicitly:

galitime --gtime -l stdout "sleep 0.1"

Installation

brew install gnu-time

conda install -y -c bioconda -c conda-forge galitime

pip install -U galitime

Standalone usage

The top-level galitime file is the canonical standalone executable.

chmod +x ./galitime
./galitime -l stdout "sleep 0.1"

This is useful when copying a single executable into another repository, container image, or remote environment.

CLI

Program: galitime (benchmarking of computational experiments using GNU time)
Version: 0.3.0
Contact: Karel Brinda <karel.brinda@inria.fr>

usage: galitime [-r INT] [-g] [-l FILE] [-n STR] [-s STR] command

positional arguments:
  command              the command to be benchmarked

optional arguments:
  -h                   show this help message and exit
  -v                   show program's version number and exit
  -r INT, --reps INT   number of repetitions [1]
  -g, --gtime          call gtime instead of time (useful on MacOS)
  -l FILE, --log FILE  output (filename/stderr/stdout) [stderr]
  -n STR, --name STR   name of the experiment (for output)
  -s STR, --shell STR  shell for execution [/bin/bash]

Output columns

galitime writes tab-delimited output with these columns:

1. experiment – experiment name supplied with -n/--name; otherwise empty

2. run – repetition number when -r/--reps is greater than 1; otherwise empty

3. real_s – wall-clock time reported by time, in seconds

4. real_s_py – wall-clock time measured by Python around the whole execution

5. user_s – user CPU time in seconds

6. sys_s – system CPU time in seconds

7. percent_cpu – CPU usage percentage reported by time

8. max_ram_kb – maximum resident memory in kilobytes

9. fs_inputs – file system input operations

10. fs_outputs – file system output operations

11. exit_code – exit status of the benchmarked command

12. command – normalized command string that was executed

Development

python -m pip install .
python -m build
make test

Issues

Please use GitHub issues.

Changelog

See Releases.

License

MIT

Contact

• Karel Brinda <karel.brinda@inria.fr>