Numwise

Numwise is a mathematical library for exact safe integer calculations.

Benchmarks

Run npm run benchmark for Numwise's regression workload, or npm run benchmark:compare for a fair comparison with documented third-party packages. See docs/benchmarks.md for the compared APIs, fairness rules, and interpretation guidance.

Comparison snapshot

The following results were measured on September 22, 2026. Times are elapsed milliseconds for the listed iteration count; lower is better within the same table. Max is the slowest of seven samples. Package initialization is outside the timed region. Results vary with hardware, Node.js/V8, thermal conditions, and background activity, so run npm run benchmark:compare locally before making performance decisions.

| Environment | Value | | --- | --- | | Node | v24.15.0 | | Platform | win32 x64 10.0.26200 | | CPU | 12th Gen Intel(R) Core(TM) i5-12500H | | OS | Windows 11 Home |

| Package | Version | | --- | ---: | | numwise | 0.2.5-dev | | number-theory | 1.1.0 | | compute-gcd | 1.2.1 | | big-integer | 1.6.52 | | mathjs | 15.2.0 |

gcd/large — 20,000 iterations

| Implementation | Median | Max | Checksum | | --- | ---: | ---: | ---: | | numwise gcd | 1.86 ms | 2.87 ms | 200001 | | number-theory gcd | 1.65 ms | 1.95 ms | 200001 | | compute-gcd | 18.24 ms | 18.81 ms | 200001 | | mathjs gcd | 1384.83 ms | 1407.91 ms | 200001 | | big-integer gcd | 52.98 ms | 55.74 ms | 200001 |

lcm/safe-integer — 20,000 iterations

| Implementation | Median | Max | Checksum | | --- | ---: | ---: | ---: | | big-integer lcm* | 22.54 ms | 23.90 ms | 1.8014488154699787e+21 | | numwise lcm | 0.25 ms | 0.34 ms | 1.8014488154699787e+21 | | mathjs lcm* | 0.66 ms | 0.88 ms | 1.8014488154699787e+21 |

isPrime/medium — 2,000 iterations

| Implementation | Median | Max | Checksum | | --- | ---: | ---: | ---: | | number-theory isPrime | 3794.89 ms | 3903.23 ms | 20001 | | numwise isPrime | 0.62 ms | 0.81 ms | 20001 | | mathjs isPrime | 0.62 ms | 0.66 ms | 20001 | | big-integer isPrime | 29.91 ms | 38.24 ms | 20001 |

modPow/large — 2,000 iterations

| Implementation | Median | Max | Checksum | | --- | ---: | ---: | ---: | | number-theory powerMod* | 5.67 ms | 6.06 ms | 16734892762803 | | big-integer modPow | 2.84 ms | 3.20 ms | 16734892762803 | | numwise modPow | 1.40 ms | 1.71 ms | 16734892762803 |

primeFactors/semiprime — 100 iterations

| Implementation | Median | Max | Checksum | | --- | ---: | ---: | ---: | | number-theory primeFactors | 3022.69 ms | 3219.96 ms | 2010008 | | numwise primeFactors | 0.30 ms | 0.37 ms | 2010008 |

primesUpTo/medium — 20 iterations

| Implementation | Median | Max | Checksum | | --- | ---: | ---: | ---: | | numwise primesUpTo | 5.78 ms | 5.93 ms | 22026585 | | number-theory sieve | 33.68 ms | 38.57 ms | 22026585 |

gcd/small — 30,000 iterations

| Implementation | Median | Max | Checksum | | --- | ---: | ---: | ---: | | numwise gcd | 1.37 ms | 1.91 ms | 1800006 | | compute-gcd | 3.19 ms | 3.45 ms | 1800006 | | mathjs gcd | 537.10 ms | 593.48 ms | 1800006 | | number-theory gcd | 2.82 ms | 3.37 ms | 1800006 | | big-integer gcd | 19.54 ms | 19.89 ms | 1800006 |

An asterisk marks an adapter that may return an inexact Number instead of rejecting an unrepresentable result. big-integer rows include conversion from Number inputs and conversion back; mathjs rows include its normal numeric dispatch. See the benchmark guide for the complete fairness and API-compatibility notes.

What Numwise is NOT

Numwise is not a replacement for packages that provide advanced mathematical functions or symbolic computation. It focuses on high-performance numerical calculations.