Struct CumProdAxisKernel<T>
public readonly struct CumProdAxisKernel<T> : INDAxisSameTypeKernel<T> where T : unmanaged, IMultiplyOperators<T, T, T>, IMultiplicativeIdentity<T, T>
Type Parameters
T
- Implements
- Inherited Members
Methods
Execute(T*, T*, long, long, long)
Writes the running products of one line along the accumulated axis (np.cumprod's per-line step).
public static void Execute(T* src, T* dst, long srcStride, long dstStride, long length)
Parameters
srcT*The line's first input element.
dstT*The line's first output element.
srcStridelongThe input stride along the axis, in elements.
dstStridelongThe output stride along the axis, in elements.
lengthlongThe axis length.
Remarks
complex128 multiplies with NumPy's accumulate form, the contracted loop_scalar
(PolyLoopScalarComplexMultiply(Complex, Complex)): NumPy's accumulate
hands each line to CDOUBLE_multiply with the output one element ahead of the running value it reads, an
overlap that rules out simd_cmul on every axis (parity QA-10).