//! Demonstrator: show what the inference_thread_count tracing event
//! emits in production, across all eight (workload, on_battery,
//! gpu_offloaded) tuples.
//!
//! Run with:
//!   cargo run -p magnotia-core --example tuning_log_demo
//!
//! Output is to stderr (tracing's default). Each unique tuple emits
//! exactly one INFO line; subsequent calls with the same tuple are
//! silenced by the per-process log-once cache.

use lumotia_core::tuning::{inference_thread_count, Workload};

fn main() {
    tracing_subscriber::fmt()
        .with_env_filter("lumotia_core=info")
        .with_target(true)
        .with_writer(std::io::stderr)
        .init();

    let cores = num_cpus::get_physical();
    let logical = num_cpus::get();
    eprintln!("Host: {cores} physical / {logical} logical cores\n");

    for (label, override_value) in [
        ("AC override", Some("ac")),
        ("Battery override", Some("battery")),
        ("No override (real sysfs probe)", None),
    ] {
        match override_value {
            Some(v) => std::env::set_var("MAGNOTIA_POWER_STATE_OVERRIDE", v),
            None => std::env::remove_var("MAGNOTIA_POWER_STATE_OVERRIDE"),
        }
        // Cache invalidation so the live probe re-runs each section.
        // Override paths bypass the cache anyway; this is for the
        // no-override block that actually hits sysfs.
        eprintln!("--- {label} ---");
        for &(workload, gpu) in &[
            (Workload::Llm, false),
            (Workload::Llm, true),
            (Workload::Whisper, false),
            (Workload::Whisper, true),
        ] {
            let n = inference_thread_count(workload, gpu);
            let w = format!("{workload:?}");
            eprintln!("  {w:>8} gpu_offloaded={gpu:>5} -> {n} threads");
        }
        eprintln!();
    }
}