JSON-driven runners for the Twin microbenchmarks. Prefer the repo-root workflow (./install.sh, ./run.sh); this file covers only bash/main.sh and bash/config.json.
From the repository root:
./bash/main.sh help
./bash/main.sh list -c bash/config.json
./bash/main.sh bash -c bash/config.json # throughput runs
./bash/main.sh counts -c bash/config.json # stats / counters runs
./bash/main.sh all -c bash/config.json # both modes (default)Options:
| Option | Meaning |
|---|---|
-c FILE |
Config file (default: bash/config.json) |
-i INDEX |
Config selection: filtered (default; only configs whose run_in_modes include the mode), all, or a 0-based index |
config.json has:
global_settings- defaults shared by every experiment (num_runs,cores,max_threads,threads,op_between,time_main,time_secondary,pinning,numa)configurations- one entry per paper workload (inserts/increase,rand/0, ...)
Per-configuration fields (override globals when set):
| Field | Role |
|---|---|
name |
Results directory under results/<name>/ |
programs |
Binaries to run (names without .out) |
threads / op_counts / op_percentages / del_percentages |
Workload sweep |
test_cases |
{num, name, plots...} entries passed as -i |
run_in_modes |
Which modes run this config (bash, counts) |
count_stats_flags |
Compile-time -D... flags for counts mode |
less_op |
Passed through as -l |
Validate JSON with jq . bash/config.json.
Optional configs in this directory:
config_main.json-- recommended figures for./run.sh(default)
- bash - builds without stats macros; writes throughput JSONL (
stats_*.jsonl) - counts - builds with
count_stats_flags; writes counter JSONL (counts_*.jsonl)
Each run also writes a plot_config.json next to those files for generatePlots.sh.
results/<configuration_name>/
stats_*.jsonl
counts_*.jsonl
plot_config.json
Reruns that would collide get a numeric suffix (results/<name>_1/, ...).