Numerical operations not involving communication. More...

Collaboration diagram for Local operations:

Functions
ghost_error	ghost_localdot (void res, ghost_densemat vec1, ghost_densemat *vec2)
	Compute the local dot product of two dense vectors/matrices. More...

ghost_error	ghost_axpy (ghost_densemat y, ghost_densemat x, void *a)
	Computes y := ax + y* with scalar a. More...

ghost_error	ghost_vaxpy (ghost_densemat y, ghost_densemat x, void *a)
	Computes column-wise y := a_ix + y* with separate scalar a_i. More...

ghost_error	ghost_axpby (ghost_densemat y, ghost_densemat x, void a, void b)
	Computes y := ax + by with scalar a and b. More...

ghost_error	ghost_vaxpby (ghost_densemat y, ghost_densemat x, void a, void b)
	Computes column-wise y := a_ix + b_iy with separate scalar a_i and b_i. More...

ghost_error	ghost_axpbypcz (ghost_densemat y, ghost_densemat x, void a, void b, ghost_densemat z, void c)
	Computes y := ax + by + cz* with scalar a, b, and c. More...

ghost_error	ghost_vaxpbypcz (ghost_densemat y, ghost_densemat x, void a, void b, ghost_densemat z, void c)
	Computes column-wise y := a_ix + b_iy + c_iz* with separate scalars a_i, b_i, and c_i. More...

ghost_error	ghost_scale (ghost_densemat x, void s)
	Computes x := sx* with scalar s. More...

ghost_error	ghost_vscale (ghost_densemat x, void s)
	Computes column-wise x := s_ix* with separate scalars s_i. More...

ghost_error	ghost_conj (ghost_densemat *x)
	Conjugates a densemat. More...

ghost_error	ghost_mult (ghost_densemat y, ghost_densemat x, void *a)
	Computes y(i,j) := ax(i,j)y(i,j) or y(i,j) := ax(i)* (if x has 1 column), with scalar a. More...

ghost_error	ghost_tsmm (ghost_densemat x, ghost_densemat v, ghost_densemat w, void alpha, void *beta)
	Multiply a distributed dense tall skinny matrix with a redundant dense matrix. More...

ghost_error	ghost_tsmm_inplace (ghost_densemat x, ghost_densemat w, void alpha, void beta)
	Multiply a distributed dense tall skinny matrix with a redundant dense matrix in-place. More...

Detailed Description

Numerical operations not involving communication.

Function Documentation

ghost_error ghost_axpby	(	ghost_densemat *	y,
		ghost_densemat *	x,
		void *	a,
		void *	b
	)

Computes y := a*x + b*y with scalar a and b.

Parameters

y	The in-/output densemat.
x	The input densemat
a	Points to the scale factor a.
b	Points to the scale factor b.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_axpbypcz	(	ghost_densemat *	y,
		ghost_densemat *	x,
		void *	a,
		void *	b,
		ghost_densemat *	z,
		void *	c
	)

Computes y := a*x + b*y + c*z with scalar a, b, and c.

Parameters

y	The in-/output densemat.
x	The input densemat x
z	The input densemat z
a	Points to the scale factor a.
b	Points to the scale factor b.
c	Points to the scale factor c.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_axpy	(	ghost_densemat *	y,
		ghost_densemat *	x,
		void *	a
	)

Computes y := a*x + y with scalar a.

Parameters

y	The in-/output densemat
x	The input densemat
a	Points to the scale factor.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_conj ( ghost_densemat * x )

Conjugates a densemat.

Parameters

x	The densemat.

Returns: GHOST_SUCCESS on success or an error indicator.

This function does nothing for real-valued densemats.

Here is the call graph for this function:

ghost_error ghost_localdot	(	void *	res,
		ghost_densemat *	vec1,
		ghost_densemat *	vec2
	)

Compute the local dot product of two dense vectors/matrices.

Parameters

res	Where to store the result.
vec1	The first vector/matrix.
vec2	The second vector/matrix.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_mult	(	ghost_densemat *	y,
		ghost_densemat *	x,
		void *	a
	)

Computes y(i,j) := a*x(i,j)*y(i,j) or y(i,j) := a*x(i) (if x has 1 column), with scalar a.

Parameters

y	The in-/output densemat.
x	The input densemat
a	Points to the scale factor a.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_scale	(	ghost_densemat *	x,
		void *	s
	)

Computes x := s*x with scalar s.

Parameters

x	The densemat.
s	The scale factor.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_tsmm	(	ghost_densemat *	x,
		ghost_densemat *	v,
		ghost_densemat *	w,
		void *	alpha,
		void *	beta
	)

Multiply a distributed dense tall skinny matrix with a redundant dense matrix.

Parameters

[in,out]	x
[in]	v
[in]	w
[in]	alpha
[in]	beta

Compute $x = \alpha \cdot v \cdot w + \beta \cdot x$ .

v is MxK, distributed.

w is KxN, redundant.

x is MxN, distributed.

M<<N

This kernel is auto-generated at compile time for given values of K and N.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_tsmm_inplace	(	ghost_densemat *	x,
		ghost_densemat *	w,
		void *	alpha,
		void *	beta
	)

Multiply a distributed dense tall skinny matrix with a redundant dense matrix in-place.

Parameters

[in,out]	x
[in]	w
[in]	alpha
[in]	beta

Compute $x(:,1:K) = \alpha \cdot x(:,1:M) \cdot w + \beta \cdot x(:,1:M)$ .

w is MxK, M>K, redundant, col-major.

x is NxM, distributed, row-major.

M,K<<N

This kernel is auto-generated at compile time for given values of M and K.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_vaxpby	(	ghost_densemat *	y,
		ghost_densemat *	x,
		void *	a,
		void *	b
	)

Computes column-wise y := a_i*x + b_i*y with separate scalar a_i and b_i.

Parameters

y	The in-/output densemat.
x	The input densemat
a	Points to the scale factors a. Length must be number of densemat columns.
b	Points to the scale factors b. Length must be number of densemat columns.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_vaxpbypcz	(	ghost_densemat *	y,
		ghost_densemat *	x,
		void *	a,
		void *	b,
		ghost_densemat *	z,
		void *	c
	)

Computes column-wise y := a_i*x + b_i*y + c_i*z with separate scalars a_i, b_i, and c_i.

Parameters

y	The in-/output densemat.
x	The input densemat x
z	The input densemat z
a	Points to the scale factors a. Length must be number of densemat columns.
b	Points to the scale factors b. Length must be number of densemat columns.
c	Points to the scale factors c. Length must be number of densemat columns.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_vaxpy	(	ghost_densemat *	y,
		ghost_densemat *	x,
		void *	a
	)

Computes column-wise y := a_i*x + y with separate scalar a_i.

Parameters

y	The in-/output densemat
x	The input densemat
a	Points to the scale factors a. Length must be number of densemat columns.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

ghost_error ghost_vscale	(	ghost_densemat *	x,
		void *	s
	)

Computes column-wise x := s_i*x with separate scalars s_i.

Parameters

x	The densemat.
s	The scale factors. Length must be number of densemat columns.

Returns: GHOST_SUCCESS on success or an error indicator.

Here is the call graph for this function:

Functions

Detailed Description

Function Documentation