nanopython benchmarks¶

For contributors evaluating optimization arcs, and for embedded/systems developers calibrating perf expectations. Snapshot from 2026-06-07. The current longitudinal data lives in metrics/history.ndjson (run make metrics-report for the trend). This page captures one point-in-time comparison for context.

Date: 2026-06-07 14:13:08 CEST
Platform: Darwin arm64 (macOS-14.8.4-arm64-arm-64bit-Mach-O)
Commit: 6f9ca14
Methodology: K=2 warmup, N=5 timed iterations per (interpreter, benchmark). Reported value is the median wall-clock; minimum shown in tooltip-style parens. Full methodology in ADR 004.

Interpreters¶

Name	Binary	Size	Version
nanopython	`nanopython`	954 KB	nanopython 0.4.0
MicroPython	`micropython`	652 KB	MicroPython v1.28.0-preview.80.gb14d129a16 on 2026-05-17; darwin (Clang 15.0.0) version
CPython	`python3`	51 KB	Python 3.14.5

nanopython is built with make build-release (-Doptimize=ReleaseFast -Dconfig=nano). MicroPython is the upstream unix-standard build. CPython is the system python3.

Results¶

benchmark	nanopython median (s)	MicroPython median (s)	CPython median (s)	nano / MicroPython	nano / CPython	output match
dictops	0.185 (min 0.184)	0.117 (min 0.115)	0.125 (min 0.123)	1.57×	1.48×	yes
fannkuch	0.554 (min 0.546)	0.558 (min 0.545)	0.338 (min 0.330)	0.99×	1.64×	yes
fib	0.545 (min 0.530)	0.430 (min 0.427)	0.272 (min 0.265)	1.27×	2.00×	yes
listops	0.150 (min 0.146)	0.109 (min 0.108)	0.113 (min 0.112)	1.37×	1.33×	yes
sieve	0.326 (min 0.315)	0.160 (min 0.155)	0.151 (min 0.148)	2.04×	2.16×	yes
strops	0.225 (min 0.183)	0.285 (min 0.272)	0.055 (min 0.052)	0.79×	4.10×	yes
geomean	n/a	n/a	n/a	1.28×	1.96×	n/a

Ratio reading: nano / X means nanopython's median divided by X's median. A value of 1.50× means nanopython took 1.5× as long as X (i.e. X is faster). Lower is better for nanopython.

Output equivalence¶

All benchmarks produce byte-identical stdout across all available interpreters.

Reproduce¶

# 1. Build a ReleaseFast nanopython (Debug is artificially slow).
make build-release

# 2. (Optional) Build MicroPython unix-standard if absent.
(cd sandbox/micropython/ports/unix && make submodules && make)

# 3. Run the suite.
python3 benchmarks/runner.py