Crate log_benchmark

Simple benchmark suite for logging data. The goal is to get an estimate for the entire process of logging data, including serialization and processing by the recording stream.

Timings are printed out while running, it’s recommended to measure process run time to ensure we account for all startup overheads and have all background threads finish.

If not specified otherwise, memory recordings are used.

The data we generate for benchmarking should be:

• minimal overhead to generate
• not homogeneous (arrow, ourselves, or even the compiler might exploit this)
• not trivially optimized out
• not random between runs

Run all benchmarks:

cargo run -p log_benchmark --release

Run specific benchmark:

cargo run -p log_benchmark --release -- --benchmarks points3d_large_batch

For better whole-executable timing capture you can also first build the executable and then run:

cargo build -p log_benchmark --release
./target/release/log_benchmark

Modules

• image 🔒
• points3d_large_batch 🔒
• points3d_many_individual 🔒
• points3d_shared 🔒

Structs

• Args 🔒

Enums

• Benchmark 🔒

Functions

• lcg
  Very simple linear congruency “random” number generator to spread out values a bit.
• main 🔒