Benchmarking

get_MNV includes a built-in benchmark binary for measuring variant annotation performance.

Synthetic benchmark

cargo run --release --bin bench_variants -- --warmup 5 --iters 30

Scale up the synthetic dataset:

cargo run --release --bin bench_variants -- \
  --warmup 3 --iters 20 --threads 4 --synthetic-scale 4

Dataset benchmark

Run against a real dataset with CSV export:

cargo run --release --bin bench_variants -- \
  --dataset example \
  --warmup 5 --iters 20 \
  --threads 4 \
  --csv benchmark.csv \
  --max-avg-ms 200

This appends one row per run to benchmark.csv.

Benchmark options

Argument	Default	Description
--warmup <N>	5	Warmup iterations (not measured)
--iters <N>	30	Measured iterations
--threads <N>	1	Worker threads
--dataset <DIR>	none	Path to dataset directory (FASTA + VCF + annotation)
--contig <NAME>	all	Restrict to a single contig
--csv <FILE>	none	Append CSV results
--max-avg-ms <MS>	none	Fail if average exceeds threshold (regression detection)
--synthetic-scale <N>	1	Scale factor for synthetic gene/SNP count

Reproducible scripts

See scripts/ for batch benchmark runners: - scripts/run_reproducible_benchmark.sh - scripts/reproduce_example_run.sh