Blackfire Continuous Profiler
Low-overhead continuous profiling for app performance optimization.
Palanteer
Lean, high-resolution instrumentation for C++ and Python applications
Palanteer provides lightweight, nanosecond-resolution logging and automatic instrumentation for C++ and Python, enabling real-time event collection, visual analysis, and remote scripting for testing and performance monitoring.
Overview
Overview
Palanteer is a lean instrumentation suite for C++ and Python applications. By embedding a single‑header library in C++ and importing a Python module, developers gain nanosecond‑resolution event logging with an overhead of roughly 25 ns per event. The system automatically captures function entry/exit, memory allocations, exceptions, garbage‑collection cycles and coroutine activity, while also allowing manual logging of custom data.
Capabilities
Recorded events can be streamed to a Python script for remote control, configuration via CLI handlers, and real‑time visualisation through hierarchical logs, timelines, histograms and flame graphs. The C++ side offers compile‑time string hashing, selective instrumentation groups and optional automatic instrumentation on Linux GCC. Python instrumentation works out‑of‑the‑box and supports multithreading, asyncio and gevent, handling up to eight concurrent streams. This combination makes Palanteer suitable for performance profiling, automated testing, and monitoring of complex, concurrent software.
Highlights
Nanosecond-resolution event logging with ~25 ns overhead
Automatic instrumentation for Python functions, memory, exceptions and coroutines
Single-header, cross-platform C++ library with compile-time string hashing and stripping
Interactive visual tools: timelines, flame graphs, histograms, and hierarchical logs
Pros
- Very low runtime overhead
- Supports both C++ and Python out of the box
- Rich visual analysis suite
- Remote control and scripting via Python
Considerations
- Automatic C++ instrumentation limited to Linux GCC
- Manual integration required for non‑GCC compilers
- Python instrumentation may add overhead in high‑frequency code
- Visualization tools depend on external viewer or server
Managed products teams compare with
When teams consider Palanteer, these hosted platforms usually appear on the same shortlist.
Looking for a hosted option? These are the services engineering teams benchmark against before choosing open source.
Fit guide
Great for
- Performance engineers needing precise timing data
- QA teams building automated test harnesses
- Developers of multithreaded or async applications
- Projects that value a single-header C++ solution
Not ideal when
- Environments without GCC on Linux for automatic C++ instrumentation
- Applications where any added overhead is unacceptable
- Teams requiring a fully integrated GUI without scripting
- Projects that need built-in support for languages beyond C++/Python
How teams use it
Real-time performance profiling
Capture nanosecond timestamps of function calls and memory usage to identify bottlenecks.
Automated regression testing
Scripted CLI commands adjust program parameters and verify event streams across runs.
Concurrent system debugging
Log lock wait times and context switches across threads to detect deadlocks.
Async coroutine monitoring
Track entry/exit and garbage‑collection events of asyncio tasks for resource analysis.
Tech snapshot
C++90%
Python8%
C1%
CMake1%
Frequently asked questions
Does Palanteer support Windows?
The C++ header is cross‑platform, but automatic instrumentation is only available on Linux with GCC.
How is the overhead measured?
Typical cost is about 25 ns per event, yielding up to 5 M events/s on a standard x64 machine.
Do I need to modify my code to use Palanteer?
For Python, most instrumentation is automatic; C++ requires including the header and optional macros, with compile‑time selection of instrumentation groups.
Can I visualize data without writing Python scripts?
Palanteer provides visual tools such as timelines, flame graphs and histograms that can be launched from recorded log files.
Is the library thread‑safe?
Yes, it supports multithreaded programs and can record up to eight concurrent streams.
Project at a glance
Stable- Stars
- 2,202
- Watchers
- 2,202
- Forks
- 96
Repo age4 years old
Last commit8 months ago
Primary languageC++
Last synced yesterday