Expand description
Performance profiling for Polars benchmarks
This crate collects process-specific performance metrics:
- Process CPU time and usage
- Process memory (RSS, virtual, heap, stack)
- Page faults and allocator statistics
- Thread pool metrics
- Hardware performance counters (optional)
Re-exports§
pub use layer::ProfilingLayer;pub use metrics::MetricsSnapshot;pub use metrics::ProfileReport;pub use metrics::CollectorData;pub use bench_integration::BenchProfiler;
Modules§
- analysis
- Advanced analysis algorithms for profiling data
- bench_
integration - Integration with criterion benchmarks via RAII profiler
- collectors
- Collectors for various performance metrics
- correlation
- Correlation engine for combining sampling data with PMU counters
- diff
- Differential profiling: compare two profiling runs to identify regressions.
- display
- Display utilities for performance metrics visualization
- ebpf
- eBPF kernel-side intelligence layer.
- export
- Export profiling data to machine-readable formats.
- layer
- Tracing layer that collects performance metrics
- metrics
- Metrics data structures
- pmu
- PMU (Performance Monitoring Unit) collector coordination
- report
- Report generation for profiling results
- senses
- Sensory system — the organism’s complete self-awareness.
- source_
registry - Global registry of captured function source text.
Macros§
- region
- No-op version when profiling is disabled.
Structs§
- Profiler
Config - Configuration for the profiler
- Profiler
Handle - Handle to the profiler for getting reports
Functions§
- init
- Awaken the organism’s nervous system.
- init_
profiling - Initialize the profiling subscriber and return a handle to get the report
- init_
with_ config - Initialize profiling with custom configuration
Attribute Macros§
- profile
- Attribute macro for lightweight per-function hardware counter profiling.
- profile_
bench - Attribute macro that automatically profiles a Criterion benchmark function.