Delegate PolyVanderKernel
One numpy.polynomial Vandermonde kernel call (see the file header): the pseudo-Vandermonde matrix of one basis
over cols points, 1-D or the fused 2-D / 3-D outer product.
public delegate void PolyVanderKernel(byte** srcs, long* srcCols, byte* dst, long dstRow, long cols, long* rows, long block, byte** scratch)
Parameters
srcsbyte**Per dimension, the first point to LOAD (C order of the points' shape). A NumSharp.Backends.Kernels.PolyVanderKey.NDims-long array.
srcColslong*Per dimension, the byte stride between consecutive points of that source (any sign; contiguous sources take the vector load).
dstbyte*Output row 0, point 0: the C-contiguous
(rows, cols)result buffer ((ideg + 1, npts)for 1-D,(prod(ideg_k + 1), npts)for 2-D / 3-D).dstRowlongByte stride between output rows (
cols * itemsize).colslongPoints (0 does nothing).
rowslong*Per dimension,
ideg_k + 1(≥ 1).blocklongPoints per block (≥ 1): the scratch holds one block of every slot and per-dimension row.
scratchbyte**Per dimension, the scratch area of that dimension: its NumSharp.Backends.Kernels.PolyVanderRoutines.Slots(NumSharp.Backends.Kernels.PolyBasis) slots, then (2-D / 3-D) its
rows[k]matrix rows, each slot / rowblock * itemsizebytes; for 3-D a fourth entry, the product row. The caller owns the memory.