Mojo function

isclose

isclose[dtype: DType, width: Int, *, symmetrical: Bool = True](a: SIMD[dtype, width], b: SIMD[dtype, width], *, atol: SIMD[float64, 1] = __init__[__mlir_type.!pop.float_literal](1.0E-8), rtol: SIMD[float64, 1] = __init__[__mlir_type.!pop.float_literal](1.0000000000000001E-5), equal_nan: Bool = False) -> SIMD[bool, width]

Returns a boolean SIMD vector indicating which element pairs of a and b are equal within a given tolerance.

For floating-point dtypes, the following criteria apply:

Symmetric (Python math.isclose style), when symmetrical is true:

|a - b| ≤ max(atol, rtol * max(|a|, |b|))

|a - b| ≤ max(atol, rtol * max(|a|, |b|))

Asymmetric (NumPy style), when symmetrical is false:

|a - b| ≤ atol + rtol * |b|

|a - b| ≤ atol + rtol * |b|

NaN values are considered equal only if equal_nan is true.

Parameters:

dtype (DType): Element type of the input and output vectors.
width (Int): Number of lanes in each SIMD vector.
symmetrical (Bool): If true, use the symmetric comparison formula (default: true).

Args:

a (SIMD[dtype, width]): First input vector.
b (SIMD[dtype, width]): Second input vector.
atol (SIMD[float64, 1]): Absolute tolerance.
rtol (SIMD[float64, 1]): Relative tolerance.
equal_nan (Bool): If true, treat NaNs as equal (default: false).

Returns:

A boolean vector where a and b are equal within the given tolerance.