Symbiotic

High-resolution performance profiling for Rust with hardware PMU counters, eBPF, and tracing integration.

Features

• 1000Hz CPU Sampling: Microsecond-level granularity for detailed performance analysis
• Hardware PMU Integration: Direct access to CPU performance counters
• Cache Hierarchy Analysis: L1D/L1I/L2/L3 cache hit/miss statistics
• Memory Bandwidth Monitoring: Track memory throughput and bottlenecks
• SIMD Utilization: Monitor AVX2/AVX512 instruction usage
• Tracing Integration: Correlate hardware events with application spans
• Low Overhead: < 5% performance impact even at aggressive sampling rates
• Criterion Integration: Seamless integration with Rust benchmarks

Quick Start

use symbiotic::{Profiler, ProfilerConfig};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let config = ProfilerConfig::default();
    let mut profiler = Profiler::new(config)?;

    let report = profiler.profile_named("workload", || {
        // Your code here
    })?;

    println!("{}", report);
    Ok(())
}

Criterion Integration

use criterion::{criterion_group, criterion_main, Criterion};
use symbiotic::criterion::DeepProfiler;

fn bench_function(c: &mut Criterion) {
    c.bench_function("example", |b| {
        b.iter(|| {
            // Your benchmark code
        });
    })
    .with_profiler(DeepProfiler::new());
}

criterion_group!(benches, bench_function);
criterion_main!(benches);

Requirements

• Linux kernel >= 5.8 (for eBPF support)
• perf_event_paranoid <= 1 for PMU access
• Optional: CAP_PERFMON capability for non-root profiling

License

This project is dual-licensed under MIT OR Apache-2.0.