tsort: write a benchmark

Author: sylvestre

Cargo.lock

@@ -26,3 +26,11 @@ fluent = { workspace = true }
 [[bin]]
 name = "tsort"
 path = "src/main.rs"
+
+[dev-dependencies]
+divan = { workspace = true }
+tempfile = { workspace = true }
+
+[[bench]]
+name = "tsort_bench"
+harness = false

src/uu/tsort/Cargo.toml

@@ -0,0 +1,226 @@
+// This file is part of the uutils coreutils package.
+//
+// For the full copyright and license information, please view the LICENSE
+// file that was distributed with this source code.
+
+use divan::{Bencher, black_box};
+use std::fs::File;
+use std::io::{BufWriter, Write};
+use tempfile::TempDir;
+
+/// Generate topological sort test data with different characteristics
+fn generate_linear_chain(num_nodes: usize) -> Vec<u8> {
+    let mut data = Vec::new();
+
+    for i in 0..num_nodes.saturating_sub(1) {
+        data.extend_from_slice(format!("node{} node{}\n", i, i + 1).as_bytes());
+    }
+
+    data
+}
+
+/// Generate a DAG with more complex dependencies
+fn generate_tree_dag(depth: usize, branching_factor: usize) -> Vec<u8> {
+    let mut data = Vec::new();
+    let mut node_id = 0;
+
+    // Generate a tree-like DAG
+    for level in 0..depth {
+        let nodes_at_level = branching_factor.pow(level as u32);
+
+        for parent in 0..nodes_at_level {
+            let parent_id = node_id + parent;
+            for child in 0..branching_factor {
+                if level + 1 < depth {
+                    let child_id = node_id + nodes_at_level + parent * branching_factor + child;
+                    data.extend_from_slice(format!("node{parent_id} node{child_id}\n").as_bytes());
+                }
+            }
+        }
+        node_id += nodes_at_level;
+    }
+
+    data
+}
+
+/// Generate a more complex graph with cross-dependencies
+fn generate_complex_dag(num_nodes: usize) -> Vec<u8> {
+    let mut data = Vec::new();
+
+    // Create a diamond-like pattern with multiple levels
+    let levels = ((num_nodes as f64).sqrt() as usize).max(4);
+    let nodes_per_level = num_nodes / levels;
+
+    for level in 0..levels - 1 {
+        let start_current = level * nodes_per_level;
+        let start_next = (level + 1) * nodes_per_level;
+        let end_current = ((level + 1) * nodes_per_level).min(num_nodes);
+        let end_next = ((level + 2) * nodes_per_level).min(num_nodes);
+
+        for i in start_current..end_current {
+            // Each node connects to 1-3 nodes in the next level
+            let connections = ((i % 3) + 1).min(end_next - start_next);
+            for j in 0..connections {
+                let target = start_next + ((i + j) % (end_next - start_next));
+                data.extend_from_slice(format!("node{i} node{target}\n").as_bytes());
+            }
+        }
+    }
+
+    data
+}
+
+/// Generate a random-like DAG that stresses the algorithm
+fn generate_wide_dag(num_nodes: usize) -> Vec<u8> {
+    let mut data = Vec::new();
+
+    // Create many parallel chains that occasionally merge
+    let num_chains = (num_nodes / 50).max(5);
+    let chain_length = num_nodes / num_chains;
+
+    for chain in 0..num_chains {
+        let chain_start = chain * chain_length;
+        let chain_end = ((chain + 1) * chain_length).min(num_nodes);
+
+        // Build the chain
+        for i in chain_start..chain_end.saturating_sub(1) {
+            data.extend_from_slice(
+                format!(
+                    "chain{}_{} chain{}_{}\n",
+                    chain,
+                    i - chain_start,
+                    chain,
+                    i + 1 - chain_start
+                )
+                .as_bytes(),
+            );
+        }
+
+        // Occasionally connect chains
+        if chain > 0 && chain % 3 == 0 {
+            let prev_chain = chain - 1;
+            let prev_end = (prev_chain * chain_length + chain_length / 2).min(num_nodes - 1);
+            let curr_mid = chain_start + chain_length / 4;
+            data.extend_from_slice(
+                format!(
+                    "chain{}_{} chain{}_{}\n",
+                    prev_chain,
+                    prev_end - prev_chain * chain_length,
+                    chain,
+                    curr_mid - chain_start
+                )
+                .as_bytes(),
+            );
+        }
+    }
+
+    data
+}
+
+/// Create a temporary file with test data
+fn create_test_file(data: &[u8], temp_dir: &TempDir) -> std::path::PathBuf {
+    let file_path = temp_dir.path().join("test_data.txt");
+    let file = File::create(&file_path).unwrap();
+    let mut writer = BufWriter::new(file);
+    writer.write_all(data).unwrap();
+    writer.flush().unwrap();
+    file_path
+}
+
+/// Run uutils tsort with given arguments
+fn run_uutils_tsort(args: &[&str]) -> i32 {
+    use std::process::{Command, Stdio};
+
+    // Use the binary instead of calling uumain directly to avoid stdout issues
+    let output = Command::new("../../../target/release/coreutils")
+        .args(["tsort"].iter().chain(args.iter()))
+        .stdout(Stdio::null())
+        .stderr(Stdio::null())
+        .status()
+        .expect("Failed to execute tsort command");
+
+    i32::from(!output.success())
+}
+
+/// Benchmark linear chain graphs of different sizes
+/// This tests the performance improvements mentioned in PR #8694
+#[divan::bench(args = [1_000, 10_000, 100_000, 1_000_000])]
+fn tsort_linear_chain(bencher: Bencher, num_nodes: usize) {
+    let temp_dir = tempfile::tempdir().unwrap();
+    let data = generate_linear_chain(num_nodes);
+    let file_path = create_test_file(&data, &temp_dir);
+    let file_path_str = file_path.to_str().unwrap();
+
+    bencher.bench(|| {
+        black_box(run_uutils_tsort(&[file_path_str]));
+    });
+}
+
+/// Benchmark tree-like DAG structures
+#[divan::bench(args = [(4, 3), (5, 3), (6, 2), (7, 2)])]
+fn tsort_tree_dag(bencher: Bencher, (depth, branching): (usize, usize)) {
+    let temp_dir = tempfile::tempdir().unwrap();
+    let data = generate_tree_dag(depth, branching);
+    let file_path = create_test_file(&data, &temp_dir);
+    let file_path_str = file_path.to_str().unwrap();
+
+    bencher.bench(|| {
+        black_box(run_uutils_tsort(&[file_path_str]));
+    });
+}
+
+/// Benchmark complex DAG with cross-dependencies
+#[divan::bench(args = [1_000, 5_000, 10_000, 50_000])]
+fn tsort_complex_dag(bencher: Bencher, num_nodes: usize) {
+    let temp_dir = tempfile::tempdir().unwrap();
+    let data = generate_complex_dag(num_nodes);
+    let file_path = create_test_file(&data, &temp_dir);
+    let file_path_str = file_path.to_str().unwrap();
+
+    bencher.bench(|| {
+        black_box(run_uutils_tsort(&[file_path_str]));
+    });
+}
+
+/// Benchmark wide DAG with many parallel chains
+/// This should stress the hashmap optimizations from PR #8694
+#[divan::bench(args = [10_000, 50_000, 100_000])]
+fn tsort_wide_dag(bencher: Bencher, num_nodes: usize) {
+    let temp_dir = tempfile::tempdir().unwrap();
+    let data = generate_wide_dag(num_nodes);
+    let file_path = create_test_file(&data, &temp_dir);
+    let file_path_str = file_path.to_str().unwrap();
+
+    bencher.bench(|| {
+        black_box(run_uutils_tsort(&[file_path_str]));
+    });
+}
+
+/// Benchmark input parsing vs computation by using files with different edge densities
+#[divan::bench(args = [10_000, 50_000])]
+fn tsort_input_parsing_heavy(bencher: Bencher, num_edges: usize) {
+    let temp_dir = tempfile::tempdir().unwrap();
+    // Create a scenario with many edges but relatively few unique nodes
+    // This stresses the input parsing and graph construction optimizations
+    let num_unique_nodes = (num_edges as f64).sqrt() as usize;
+    let mut data = Vec::new();
+
+    for i in 0..num_edges {
+        let from = i % num_unique_nodes;
+        let to = (i / num_unique_nodes) % num_unique_nodes;
+        if from != to {
+            data.extend_from_slice(format!("n{from} n{to}\n").as_bytes());
+        }
+    }
+
+    let file_path = create_test_file(&data, &temp_dir);
+    let file_path_str = file_path.to_str().unwrap();
+
+    bencher.bench(|| {
+        black_box(run_uutils_tsort(&[file_path_str]));
+    });
+}
+
+fn main() {
+    divan::main();
+}