kernels.cl File Reference

Macros
#define	NDIAG   ((2*ORDER+1)*(2*ORDER+1))
#define	NROW   (NSPLINE * NSPLINE)
#define	NCOL   (NSPLINE * NSPLINE)

Functions
double2	cpxm (double2 a, double2 b)
	Complex multiplication. More...
kernel void	a_vec_b_vec (private double2 a, global double2 *x, private double2 b, global double2 *y)
	AXBY operation. More...
kernel void	vec_mul_vec (global double2 *a, global double2 *b, global double2 *c)
	Vector-vector multiplication. More...
kernel void	vec_norm (global double2 *v, global double *n)
	Per-element square of a vector. More...
kernel void	scalar_product (global double2 *u, global double2 *v, global double2 *z)
	Full scalar product. More...
kernel void	norm (global double2 *v, global double *z)
	Full vector norm. More...
kernel void	DIA_dot_vec (global double2 *A, global double2 *x, global double2 *y)
	DIA-matrix-vector multiplication. More...

Macro Definition Documentation

#define NCOL   (NSPLINE * NSPLINE)

#define NDIAG   ((2*ORDER+1)*(2*ORDER+1))

#define NROW   (NSPLINE * NSPLINE)

Function Documentation

kernel void a_vec_b_vec	(	private double2	a,
		global double2 *	x,
		private double2	b,
		global double2 *	y
	)

Computes the linear combination \( ax + by \) of vectors \( x \) and \( y \) and stores the result in \( x \).

double2 cpxm ( double2  a, double2  b  )

inline

Multiplies two complex numbers and returns the product.

kernel void DIA_dot_vec	(	global double2 *	A,
		global double2 *	x,
		global double2 *	y
	)

The input matrix is supplied in the form of zero-padded concatenated diagonals. For example, the matrix

\[ A = \pmatrix { a_{11} & a_{12} & & & \cr a_{21} & a_{22} & a_{23} & & \cr & a_{32} & a_{33} & a_{34} & \cr & & a_{43} & a_{44} & a_{45} \cr & & & a_{54} & a_{55} \cr } \]

would be column-padded to

\[ A' = \pmatrix { 0 & a_{11} & a_{12} \cr a_{21} & a_{22} & a_{23} \cr a_{32} & a_{33} & a_{34} \cr a_{43} & a_{44} & a_{45} \cr a_{54} & a_{55} & 0 \cr } \]

and sent in as a 1D array

\[ A'' = [0, a_{21} , a_{32}, a_{43}, a_{54}; a_{11}, a_{22}, a_{33}, a_{44}, a_{55}; a_{12}, a_{23}, a_{34}, a_{45}, 0] \ . \]

kernel void norm	(	global double2 *	v,
		global double *	z
	)

kernel void scalar_product	(	global double2 *	u,
		global double2 *	v,
		global double2 *	z
	)

kernel void vec_mul_vec	(	global double2 *	a,
		global double2 *	b,
		global double2 *	c
	)

Computes per-element vector-vector multiplication \( a * b \) and stores the resulting vector in \( c \),

kernel void vec_norm	(	global double2 *	v,
		global double *	n
	)

Similarly to vec_mul_vec, computes a per-element square of a given vector \( v \) and stores the result in the vector given as the second argument.