Benchmark in Milliseconds
a day ago
- Tweak input size so the benchmark runs in about 300ms.
- Milliseconds (1-999) offer enough precision and are easy to scan without needing units or decimals.
- Faster than 10ms risks skew from fixed costs; 300ms makes one-off overheads irrelevant.
- 300ms is fast but noticeable for humans, aiding intuitive understanding of speed.
- Longer than 1 second slows iteration; running 10 times should be quick.
- The goal is to build intuition for decision-making, not precise measurement.