Awesome

xtr

What is it?

XTR is a C++ logging library aimed at applications with low-latency or real-time requirements. The cost of log statements is minimised by delegating as much work as possible to a background thread.

It is designed so that the cost of a log statement is consistently fast---i.e. every call is fast, not just the average case. No allocations or system calls are made when a log statement is made.

Features

• Fast (please see benchmark results).
• No allocations when logging, even when logging strings.
• Formatting, I/O etc are all delegated to a background thread. Work done at the log statement call-site is minimized---for example a no argument log statement only involves writing a single pointer to a ring buffer.
• Safe: No references taken to arguments unless explicitly requested.
• Comprehensive suite of unit tests which run cleanly under AddressSanitizer, UndefinedBehaviourSanitizer, ThreadSanitizer and LeakSanitizer.
• Log sinks with independent log levels (so that levels for different subsystems may be modified independently).
• Ability to modify log levels via an external command.
• Non-printable characters are sanitized for safety (to prevent terminal escape sequence injection attacks).
• Type-safe---formatting is done via fmtlib.
• io_uring support.
• Support for custom I/O back-ends (e.g. to log to the network, write compressed files, etc).
• Support for logrotate integration.
• Support for systemd journal integration.
• CMake and Conan integration supported.
• Fully documented.

Supported platforms

• Linux (x86-64)
• FreeBSD (x86-64)

Documentation

https://choll.github.io/xtr

Benchmarks

Below is the output of PRODUCER_CPU=2 CONSUMER_CPU=1 make benchmark_cpu on a stock Ryzen 5950X with SMT disabled, isolated cores and g++ version 11.2.0

sudo cpupower --cpu 2,1 frequency-set --governor performance
Setting cpu: 1
Setting cpu: 2
build/g++-lto-release/benchmark/benchmark
2022-04-26T22:54:43+01:00
Running build/g++-lto-release/benchmark/benchmark
Run on (16 X 5084 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x16)
  L1 Instruction 32 KiB (x16)
  L2 Unified 512 KiB (x16)
  L3 Unified 32768 KiB (x2)
Load Average: 3.32, 2.74, 1.99
***WARNING*** CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.
---------------------------------------------------------------------------------
Benchmark                                       Time             CPU   Iterations
---------------------------------------------------------------------------------
logger_benchmark                             2.19 ns         2.19 ns    318972090
logger_benchmark_tsc                         7.81 ns         7.81 ns     89307646
logger_benchmark_clock_realtime_coarse       8.05 ns         8.05 ns     87732373
logger_benchmark_int                         3.31 ns         3.31 ns    210259973
logger_benchmark_long                        3.33 ns         3.33 ns    209935480
logger_benchmark_double                      3.26 ns         3.26 ns    212240698
logger_benchmark_c_str                       4.86 ns         4.86 ns    146577914
logger_benchmark_str_view                    3.12 ns         3.12 ns    224958501
logger_benchmark_str                         4.36 ns         4.36 ns    158750720
logger_benchmark_non_blocking                2.01 ns         2.01 ns    349312408

Benchmarks Comparing Against Other loggers

Integer Benchmark, One Thread

Integer Benchmark, Four Threads

String Benchmark, One Thread

String Benchmark, Four Threads

See logger_benchmarks for more details.

Installation notes

See INSTALL.md