cajeta.lang.Math — the one stop for scalar and vector numeric functions. It has two kinds of members: the declared, method-level-templated max / min / clamp below (each monomorphized per T at the call site, zero dispatch cost), and compiler intrinsics — sqrt, floor, ceil, trunc, round, abs, fma, plus the transcendentals sin, cos, tan, asin, acos, atan, atan2, exp, exp2, log, log2, log10, pow, rsqrt — which are not declared in the class; the compiler intercepts the call and emits the right instruction on the host and inside an @Kernel. Intrinsics also apply element-wise to a Vector<T,N>, and @FastMath on a kernel relaxes IEEE FP for its whole body.

int32 a = Math.max(3, 7);               // 7   (max<int32>)
int32 b = Math.min(3, 7);               // 3
float64 c = Math.clamp(1.5, 0.0, 1.0);  // 1.0 (clamp<float64>)

Methods

Signature
`static T max<T>(T a, T b)` ⚑	The larger of `a` and `b`; per-`T` monomorphization
`static T min<T>(T a, T b)` ⚑	The smaller of `a` and `b`
`static T clamp<T>(T x, T lo, T hi)` ⚑	`x` clamped to `[lo, hi]`

⚑ = @EntryPoint

v1 limitation: the bodies use > / <, which work for built-in primitives but won’t link for user classes until operator overloading flows through method-level template specialization.

Math

Methods

See also