v2.0.0 profiling notes

On the v2.0.0 branch:

Setup

Make sure you have mise installed

curl https://mise.run | sh

And give permissions to mise.toml:

cd ixa
mise trust mise.toml

You can always run mise run to get autocomplete for all the commands:

Running the reference SIR

The baseline version:

cargo run --bin run_bench -p ixa-bench --release -- --group large_sir --bench baseline

The the entities version:

cargo run --bin run_bench -p ixa-bench --release -- --group large_sir --bench entities

Running the SIR reference benchmark

mise run bench:hyperfine

Running the sample_entity benchmark

This is a scaling analysis on sample_entity

mise run bench:criterion sample_entity

At the end, you'll get a summary:

=== Scaling summary: sample_entity_whole_population ===
  baseline: n=1000, t=6.82 ns/sample
  ratios vs baseline:
    n=   1000:       6.82 ns/sample  (x1.000)
    n=  10000:       6.71 ns/sample  (x0.984)
    n= 100000:       6.78 ns/sample  (x0.995)

=== Scaling summary: sample_entity_single_property_indexed ===
  baseline: n=1000, t=51.15 ns/sample
  ratios vs baseline:
    n=   1000:      51.15 ns/sample  (x1.000)
    n=  10000:      51.23 ns/sample  (x1.002)
    n= 100000:      51.45 ns/sample  (x1.006)

=== Scaling summary: sample_entity_multi_property_indexed ===
  baseline: n=1000, t=102.70 ns/sample
  ratios vs baseline:
    n=   1000:     102.70 ns/sample  (x1.000)
    n=  10000:     102.07 ns/sample  (x0.994)
    n= 100000:     102.31 ns/sample  (x0.996)