examples: add parallel-assets demos for Rust and Node

Python's parallel_assets.py demoed GIL-release multi-core throughput;
Rust and Node both lacked a sibling that shows their own native
parallelism. Close the gap with two real, runnable examples.

* examples/rust/src/bin/parallel_assets.rs — synthesises an (assets,
  bars) panel with a deterministic per-asset LCG, runs a serial baseline,
  then `Sma::batch_parallel` / `Rsi::batch_parallel` via rayon, asserts
  the two outputs are element-wise identical and prints the speedup.
  Toggle indicator with `--indicator sma|rsi`.
* examples/node/parallel_assets.js — same shape, but the parallel run is
  a `worker_threads` pool that re-loads the native binding in each
  worker. Each worker computes the last non-null indicator value for its
  slice; the main thread aggregates and verifies serial == parallel
  per asset.

Both examples report timings and the serial-vs-parallel sanity check
passes. Defaults (200 × 5000) keep the example fast on dev hardware;
larger `--assets`/`--bars` is where the speedup numbers move (Node's
worker spawn cost dominates the smallest sizes, which is honest and
educational).

examples/README.md gains the two new rows.
This commit is contained in:
kingchenc
2026-05-23 00:18:46 +02:00
parent 962ced0712
commit 43b0b26736
3 changed files with 312 additions and 0 deletions
+141
View File
@@ -0,0 +1,141 @@
//! Parallel multi-asset indicator computation via rayon.
//!
//! Builds a synthetic `(assets, bars)` panel, runs a serial baseline, then
//! runs the same computation with [`BatchExt::batch_parallel`] (which is
//! gated behind wickra's default `parallel` feature) and reports the
//! speedup. The Rust counterpart of `examples/python/parallel_assets.py`.
//!
//! Run with:
//! ```text
//! cargo run --release -p wickra-examples --bin parallel_assets -- \
//! --assets 200 --bars 5000
//! ```
use std::env;
use std::time::Instant;
use wickra::{BatchExt, Rsi, Sma};
#[derive(Debug, Clone)]
enum Which {
Sma,
Rsi,
}
impl Which {
fn parse(s: &str) -> Result<Self, Box<dyn std::error::Error>> {
match s {
"sma" => Ok(Self::Sma),
"rsi" => Ok(Self::Rsi),
other => Err(format!("--indicator: expected 'sma' or 'rsi', got {other}").into()),
}
}
}
struct Args {
assets: usize,
bars: usize,
indicator: Which,
}
fn parse_args() -> Result<Args, Box<dyn std::error::Error>> {
let mut assets = 200_usize;
let mut bars = 5_000_usize;
let mut indicator = Which::Sma;
let mut it = env::args().skip(1);
while let Some(arg) = it.next() {
match arg.as_str() {
"--assets" => {
assets = it.next().ok_or("--assets needs a value")?.parse()?;
}
"--bars" => {
bars = it.next().ok_or("--bars needs a value")?.parse()?;
}
"--indicator" => {
indicator = Which::parse(&it.next().ok_or("--indicator needs a value")?)?;
}
other => return Err(format!("unexpected argument: {other}").into()),
}
}
if assets == 0 || bars == 0 {
return Err("--assets and --bars must be positive".into());
}
Ok(Args {
assets,
bars,
indicator,
})
}
/// Deterministic synthetic `(assets, bars)` panel. Each asset uses an
/// independent LCG seed so the series are uncorrelated but reproducible.
fn synthesize_panel(n_assets: usize, n_bars: usize) -> Vec<Vec<f64>> {
let mut series = Vec::with_capacity(n_assets);
for a in 0..n_assets {
let mut s = Vec::with_capacity(n_bars);
let mut price = 100.0_f64;
let mut state: u32 = 1_234_567_u32
.wrapping_add(a as u32)
.wrapping_mul(2_654_435_761);
for _ in 0..n_bars {
state = state.wrapping_mul(1_103_515_245).wrapping_add(12_345) & 0x7FFF_FFFF;
let r = f64::from(state) / f64::from(0x7FFF_FFFF_u32);
price += (r - 0.5) * 0.4;
s.push(price);
}
series.push(s);
}
series
}
fn run_serial(panel: &[Vec<f64>], indicator: &Which) -> Vec<Vec<Option<f64>>> {
panel
.iter()
.map(|prices| match indicator {
Which::Sma => Sma::new(14).expect("SMA(14)").batch(prices),
Which::Rsi => Rsi::new(14).expect("RSI(14)").batch(prices),
})
.collect()
}
fn run_parallel(panel: &[Vec<f64>], indicator: &Which) -> Vec<Vec<Option<f64>>> {
match indicator {
Which::Sma => Sma::batch_parallel(panel, || Sma::new(14).expect("SMA(14)")),
Which::Rsi => Rsi::batch_parallel(panel, || Rsi::new(14).expect("RSI(14)")),
}
}
fn main() -> Result<(), Box<dyn std::error::Error>> {
let args = parse_args()?;
println!("Generating {}×{} synthetic panel…", args.assets, args.bars);
let panel = synthesize_panel(args.assets, args.bars);
let t0 = Instant::now();
let serial = run_serial(&panel, &args.indicator);
let t_serial = t0.elapsed();
println!(
"Serial: {:>8.3} s ({} assets, indicator={:?})",
t_serial.as_secs_f64(),
args.assets,
args.indicator
);
let t0 = Instant::now();
let parallel = run_parallel(&panel, &args.indicator);
let t_parallel = t0.elapsed();
let speedup = t_serial.as_secs_f64() / t_parallel.as_secs_f64().max(1e-9);
println!(
"Parallel: {:>8.3} s (rayon, speedup ~{:.2}x)",
t_parallel.as_secs_f64(),
speedup
);
assert_eq!(serial.len(), parallel.len(), "asset count mismatch");
for (i, (s, p)) in serial.iter().zip(parallel.iter()).enumerate() {
if s != p {
return Err(format!("asset {i}: serial and parallel results disagree").into());
}
}
println!("Parallel results match serial results — OK.");
Ok(())
}