Benchmarks

The numbers, with their provenance.

Records imported from committed result files in the VNX-DNA repository at v9.0.0 (e76e24b). Each shows its source file, commit, host, channel model and statistics.

All decodes are simulated.None: no DNA has been synthesised, stored or sequenced. Timings come from one shared host and are not hardware-independent.

Dataset

Archive size scaling, clean channel

Simulated

Encode (s)Decode (s)

Input size	Exact recovery	Strands	Encode (s)	Decode (s)	Peak RSS (MiB)	nt per byte
20,000	yes	691	0.06	0.08	46	10.814
1,048,576	yes	32,838	0.21	0.44	87	9.802
10,485,760	yes	327,763	1.25	2.06	263	9.784
104,857,600	yes	3,277,129	10.94	18.39	287	9.782
1,073,741,824	yes	33,557,316	105.79	195.45	323	9.782

Evidence: SIMULATED clean channel
Release: v9.0.0
Commit: 3af1062e6e23
Run commit: adba34e6bf1b
Date: 2026-10-08
Hardware: 8 vCPU shared x86-64 development host
Operating system: Linux
Python: 3.12.3
Profile: v4-balanced
Channel model: clean (no errors)
Trials per point: 1
Seed: 93000
Statistics: One run per size; wall time on a shared host with load average recorded in the source file.

Source file in repository

Show the labelled interface-demonstration dataset