Simple Benchmark C-Header

A C-header-only library for simple benchmarking of a function call or code block. The following macros are available:

• BENCH(FCALL) for concise output (default).
• BENCH_VERBOSE(FCALL) for detailed output with wall/user/system/peak RSS delta.
• BENCH_N(FCALL, runs, warmup) for repeated wall-time stats.
• BENCH_CSV(FCALL, "file.csv", "comment") for machine-readable CSV rows.

FCALL can be a function call or a code block. Time metrics are wall/user/system and memory is peak RSS delta from rusage. Memory numbers are a practical approximation, not exact per-call allocation usage.

It should work on Linux and macOS.

Usage

#include "simple_bench.h"

void do_work(void) {
  /* your code */
}

int main() {
  BENCH(do_work());
  BENCH_VERBOSE(do_work());
  BENCH_N(do_work(), 10, 2);
  BENCH_CSV(do_work(), "bench.csv", "baseline");
  return 0;
}

results in:

Installation

Use the header directly, or depend on the Bazel target:

bazel test //...

Examples

Representative examples are available in examples/:

• cpu_bound_example: arithmetic-heavy loop workload.
• memory_bound_example: allocation + linear memory access workload.
• mixed_workload_example: data generation + sort + sampled reduction.

Build all examples:

bazel build //examples:all

Run one example:

bazel run //examples:cpu_bound_example

Notes about measurements

• Time is measured using clock_gettime(CLOCK_MONOTONIC).
• User/system CPU times use getrusage (ru_utime / ru_stime) deltas.
• Memory uses ru_maxrss from getrusage; this is peak RSS, so reported peak_rss_delta is an approximation.