Measure pcmp against matplotlib

pcmp draws its own benchmark. bench/speed.py in the repository times pcmp drawing each chart kind, then matplotlib drawing the same data as a similar dark 1080 px chart, and prints the numbers.

Run it

On Linux or macOS, you need a clone of the repository, Rust 1.97 or newer, and Python with matplotlib installed:

git clone https://github.com/gmrdad82/pito-compare
cd pito-compare
cargo build --release
python3 bench/speed.py

For each chart kind it starts a new process per chart, 50 times, and takes the median, for pcmp and for matplotlib alike. Nothing but the benchmark needs Python.

What it measured

On 2026-10-14, on an AMD Ryzen 9 9950X (32 threads) under Linux, with Python 3.14 and matplotlib 3.11, while the machine was busy with other work (load 62 on 32 threads). bench/pcmp-vs-matplotlib.toml holds the numbers it printed, and pcmp drew them:

Paired bars, lower is better: start to PNG on disk, pcmp 22.6, 14.6 and 13.2 ms against matplotlib 871, 855 and 844 ms; 39×, 59× and 64× faster

The balanced ratio chart of the same numbers: pcmp 39×, 59× and 64× faster than matplotlib

	pcmp	matplotlib
Paired bars	22.6 ms	871 ms
Balanced ratios	14.6 ms	855 ms
Single bars	13.2 ms	844 ms
Peak memory	13.3 MB	82.5 MB

Reading the numbers

Most of matplotlib’s time is Python and its import, but not all: in one run, drawing and saving after the import took 235 to 282 ms per chart, over ten times pcmp’s whole run. The matplotlib chart is the plainer of the two (no tags, no pixel font), which favours it. A quiet machine is faster for both.

Chart your own benchmark

The bench’s numbers are an ordinary data file. To chart your own, write your timings the same way, with better = "lower" for times, and draw them:

$ pcmp pcmp-vs-matplotlib.toml
pcmp-vs-matplotlib-paired.png
pcmp-vs-matplotlib-ratio.png