paraseq benchmark

This is a benchmark of different FASTQ parsers in rust. It is not meant to be an exhaustive benchmark, but rather a simple comparison of the speed of different parsing strategies.

Mainly:

1. Copying
2. Zero-copy
3. Minimal-copy parsing

The libraries tested are:

Library	Sequential Processing Strategy	Parallel Processing Strategy
bio	Copying	N/A
fastq-rs	Zero-copy	One-copy
seq_io	Zero-copy	One-copy
needletail	Zero-copy	N/A
paraseq	Minimal-copy	Minimal-copy

The benchmark is done on files of the same number of records but with each file having a different record length.

Benchmarking

This benchmark was taken on my local computer using a AMD Ryzen 7 7700X and an NVMe SSD.

Running the benchmark

To run the benchmark you will need to have the following installed:

1. Rust
2. Hyperfine
3. Just
4. Nucgen

Then you can install the benchmarking tool by cloning the repository and running:

cargo install --path .

# Check that the tool is installed
psb --version

Then you can run the benchmark by running:

# Generate the nucleotide data
just gen_data

# Run the benchmark
just run_benchmarks