Skip to content

BENCHMARKS.md

Latest commit

 

History

History
66 lines (51 loc) · 3.27 KB

BENCHMARKS.md

File metadata and controls

66 lines (51 loc) · 3.27 KB

interval — Benchmarks

Source: benchmarks.cpp Full doc: docs/interval.md

Measures the cost of every Interval<T> operation: construction, queries, arithmetic operations, normalization, and equality. This answers the question: how much does each interval operation actually cost, and is the std::optional return from intersect() free?

Build and Run

g++ -std=c++20 -O2 benchmarks.cpp -o benchmarks && ./benchmarks

A Note on Measurement Resolution

On Apple Silicon (M-series), steady_clock::now() costs ~42 ns per call, creating a ~18–42 ns measurement floor. All Interval<T> operations complete in sub-tick time (mean 28–37 ns, min 0 ns). For these cases the mean is an upper bound, not the exact cost — the operations themselves are cheaper than one timer tick.

Suite Overview

Suite What is measured
1 · Construction Interval(a,b), make_empty(), make_universe()
2 · Queries contains(value), contains(Interval), overlaps(), is_empty(), length(), center()
3 · Operations clamp(), merge(), intersect(), expand(), translate()
4 · Normalization normalize(), denormalize(), round-trip
5 · Operators operator==, operator!=
6 · Mixed Realistic sequences combining construction, queries, and operations

Reference Results

Measured with -O2 on an Apple M-series CPU (2026-03-05). All operations are floor-dominated.

Operation Mean Min
Interval<int>(a, b) construction ~35 ns 0 ns
Interval<double>(a, b) construction ~33 ns 0 ns
Interval<int>::make_empty() ~29 ns 0 ns
Interval<int>::make_universe() ~29 ns 0 ns
contains(int value) ~30–31 ns 0 ns
contains(Interval) ~29 ns 0 ns
overlaps() ~29–37 ns 0 ns
is_empty() ~29 ns 0 ns
length() ~29 ns 0 ns
center() ~29 ns 0 ns
clamp() ~29 ns 0 ns
merge() ~29–32 ns 0 ns
intersect() — returns Some ~30 ns 0 ns
intersect() — returns nullopt ~30 ns 0 ns
expand() ~30 ns 0 ns
translate() ~30 ns 0 ns
normalize() ~29 ns 0 ns
denormalize() ~29 ns 0 ns
normalize + denormalize round-trip ~29 ns 0 ns
operator== ~29 ns 0 ns
operator!= ~29 ns 0 ns
construct + contains + clamp ~30 ns 0 ns
merge three intervals ~29 ns 0 ns
intersect chain (double) ~36 ns 0 ns

Interpretation

Every Interval<T> operation is a handful of arithmetic comparisons on two values stored inline. There is no heap allocation, no virtual dispatch, and no branching beyond the comparisons themselves. The compiler inlines all operations and elides the std::optional wrapper for intersect() in most cases.

The measurement floor (~42 ns timer tick on Apple Silicon) means we cannot distinguish operations from one another at this scale — they all complete in under one tick. The actual cost is in the low single-digit nanosecond range.

intersect() returning std::optional<Interval<T>> has no measurable overhead compared to plain accessors: the optional is typically returned in registers and the compiler elides any copy.