AccelerateSupport module

Classes
class	AccelerateCholeskyAtA
	A QR factorization and solver based on Accelerate without storing Q (equivalent to A^TA = R^T R) More...
class	AccelerateLDLT
	The default Cholesky (LDLT) factorization and solver based on Accelerate. More...
class	AccelerateLDLTSBK
	A direct Cholesky (LDLT) factorization and solver based on Accelerate with Supernode Bunch-Kaufman and static pivoting. More...
class	AccelerateLDLTTPP
	A direct Cholesky (LDLT) factorization and solver based on Accelerate with full threshold partial pivoting. More...
class	AccelerateLDLTUnpivoted
	A direct Cholesky-like LDL^T factorization and solver based on Accelerate with only 1x1 pivots and no pivoting. More...
class	AccelerateLLT
	A direct Cholesky (LLT) factorization and solver based on Accelerate. More...
class	AccelerateQR
	A QR factorization and solver based on Accelerate. More...

Detailed Description

This module provides an interface to the Apple Accelerate library. It provides the seven following main factorization classes:

• class AccelerateLLT: a Cholesky (LL^T) factorization.
• class AccelerateLDLT: the default LDL^T factorization.
• class AccelerateLDLTUnpivoted: a Cholesky-like LDL^T factorization with only 1x1 pivots and no pivoting
• class AccelerateLDLTSBK: an LDL^T factorization with Supernode Bunch-Kaufman and static pivoting
• class AccelerateLDLTTPP: an LDL^T factorization with full threshold partial pivoting
• class AccelerateQR: a QR factorization
• class AccelerateCholeskyAtA: a QR factorization without storing Q (equivalent to A^TA = R^T R)

#include <Eigen/AccelerateSupport>

In order to use this module, the Accelerate headers must be accessible from the include paths, and your binary must be linked to the Accelerate framework. The Accelerate library is only available on Apple hardware.

Note that many of the algorithms can be influenced by the UpLo template argument. All matrices are assumed to be symmetric. For example, the following creates an LDLT factorization where your matrix is symmetric (implicit) and uses the lower triangle:

AccelerateLDLT<SparseMatrix<float>, Lower> ldlt;