Scientific Computing

Implemented — numerical-method examples in ordinary Karkain arithmetic live under examples/12-scientific-computing/ and run byte-identically on both engines.

Again with the honest caveat: there is no scientific-computing framework, no tensor language surface and no arbitrary-precision type in Karkain. What exists is IEEE double arithmetic through the generated C runtime plus the implemented array/loop surface. These examples show iterative methods that avoid the unreliable builtin float helpers where possible — which is both correct and more instructive.

examples/12-scientific-computing/

File

Demonstrates

01_sqrt_newton.kark

Newton–Raphson square root iteration (√2 → 1.41421 …)

02_numerical_integration.kark

Trapezoid rule converging to 1/3 on ∫₀¹ x² dx

03_statistics.kark

Mean, variance and standard deviation over a fixed sample

04_matrix_multiply.kark

3×3 integer matrix product

Newton square root (excerpt)

func sqrt_newton(x) {
    let guess = x
    let i = 0
    while (i < 10) {
        guess = (guess + x / guess) / 2.0
        i = i + 1
    }
    return guess
}

Run:

$ karkain run examples/12-scientific-computing/01_sqrt_newton.kark

Output format: floats print at 6 significant figures, identically on both engines (1.41421, 0.333333, …).

See also

Machine Learning — gradient-descent shares these numeric idioms. Type Reference — the float type used by the runtime.