DSPF_sp_qrd_solver
[DSPF_sp_qrd, DSPF_sp_qrd_cmplx]

Collaboration diagram for DSPF_sp_qrd_solver:

---

Detailed Description

Functions
int	DSPF_sp_qrd_solver (const int Nrows, const int Ncols, float *restrict Q, float *restrict R, float *restrict b, float *restrict y, float *restrict x)
int	DSPF_sp_qrd_solver_cmplx (const int Nrows, const int Ncols, float *restrict Q, float *restrict R, float *restrict b, float *restrict y, float *restrict x)

---

Function Documentation

int DSPF_sp_qrd_solver	(	const int	Nrows,
		const int	Ncols,
		float *restrict	Q,
		float *restrict	R,
		float *restrict	b,
		float *restrict	y,
		float *restrict	x
	)

This function solves the system of linear equations A*x=b for x using the inputs produced by DSPF_sp_qrd where A*x=Q*R*x=b. The following procedure is performed.

1. Use matrix multiplication to generate y: y = transpose(Q)*b
2. Use backward substitution for solve for x: x = inv(R)*y

The values stored in the matrices are assumed to be single precision floating point values. This code is suitable for dense matrices. No optimizations are made for sparse matrices.

Parameters:

	Nrows	= number of rows in matrix A
	Ncols	= number of columns in matrix A
	Q	= pointer to lower triangular matrix Q[Nrows*Nrows]
	R	= pointer to upper triangular matrix R[Nrows*Ncols]
	b	= pointer to vector b[Nrows]
	y	= pointer to temporary vector y[Nrows]
	x	= pointer to final solver output vector x[Ncols]

Algorithm:

DSPF_sp_qrd_cn.c is the natural C equivalent of the optimized intrinsic C code without restrictions. Note that the intrinsic C code is optimized and restrictions may apply.

Assumptions:

The arrays P, L, U, and inv_A are stored in distinct arrays. In-place processing is not done.

Implementation Notes:

Interruptibility : The code is interruptible.
Endian support : supports both Little and Big endian modes.

int DSPF_sp_qrd_solver_cmplx	(	const int	Nrows,
		const int	Ncols,
		float *restrict	Q,
		float *restrict	R,
		float *restrict	b,
		float *restrict	y,
		float *restrict	x
	)

This function solves the system of linear equations A*x=b for x using the inputs produced by DSPF_sp_qrd_cmplx where A*x=Q*R*x=b. The following procedure is performed.

1. Use matrix multiplication to generate y: y = transpose(Q)*b
2. Use backward substitution for solve for x: x = inv(R)*y

The values stored in the matrices are assumed to be float precision floating point values. This code is suitable for dense matrices. No optimizations are made for sparse matrices.

Parameters:

	Nrows	= number of rows in matrix A
	Ncols	= number of columns in matrix A
	Q	= pointer to lower triangular matrix Q[Nrows*2*Nrows]
	R	= pointer to upper triangular matrix R[Nrows*2*Ncols]
	b	= pointer to vector b[Nrows]
	y	= pointer to temporary vector y[Nrows]
	x	= pointer to final solver output vector x[Ncols]

Algorithm:

DSPF_sp_qrd_complex_cn.c is the natural C equivalent of the optimized intrinsic C code without restrictions. Note that the intrinsic C code is optimized and restrictions may apply.

Assumptions:

1. The arrays P, L, U, and inv_A are stored in distinct arrays. In-place processing is not done.
2. The arrays consist of complex number entries with alternating real and imaginary parts: real0,imag0,real1,imag1,...

Implementation Notes:

Interruptibility : The code is interruptible.
Endian support : supports both Little and Big endian modes.

---

Copyright 2014, Texas Instruments Incorporated