hiplarSet: ~~ Methods for Function 'hiplarSet' in Package 'HiPLARM' ~~

In HiPLARM: High Performance Linear Algebra in R

Description

hiplarSet allows the user to access internal variables within HiPLAR, these are generally set automatically but should the user wish they can over ride these settings by following the instructions below. The hiplarSet function is useful if users want to choose either the MAGMA or PLASMA libraries or if they wish to set the crossover point within the function. The crossover point (for users with GPU and CPU support) is the matrix size at which the function switches between the PLASMA, CPU libraries and the MAGMA, GPU libraries. This is generally automatic but the user may change if they so wish.

Details

The var argument accesses the particular settings that the user wishes to change In setting the xover_<function_name> the val argument should be set to a suitable value. These values are generally set automatically during setup so the user should be wary when changing it. When accessing the library setting the user should choose values from 1 to 3. 1 sets the PLASMA library for use, 2 sets the MAGMA library and 3 sets it to automatic.

Methods

signature(var = "character", val = "numeric")

Examples

	
		# Sets the PLASMA libary to be used exclusively #
		hiplarSet("hiplar_library",  1)
		# Sets the MAGMA library to be used exclusively #
		hiplarSet("hiplar_library", 2)
		# Enables autotune which selects PLASMA or MAGMA depending on the problem size.#
		hiplarSet("hiplar_library", 3)

		## Methods for setting crossover values ##
		optsize <- 512
		hiplarSet("xover_dgeMatrix_LU", optsize)
		hiplarSet("xover_dgeMatrix_crossprod", optsize)
		hiplarSet("xover_dgeMatrix_dgeMatrix_crossprod", optsize)
		hiplarSet("xover_dgeMatrix_matrix_crossprod", optsize)
		hiplarSet("xover_dgeMatrix_determinant", optsize)
		hiplarSet("xover_dgeMatrix_matrix_mm", optsize)
		hiplarSet("xover_dgeMatrix_matrix_mm",optsize)
		hiplarSet("xover_dgeMatrix_norm",optsize)
		hiplarSet("xover_dgeMatrix_solve", optsize)
		hiplarSet("xover_dgeMatrix_matrix_solve", optsize)
		hiplarSet("xover_dgeMatrix_rcond", optsize)
		hiplarSet("xover_dgeMatrix_LU", optsize)
		hiplarSet("xover_dgeMatrix_crossprod", optsize)
		hiplarSet("xover_dgeMatrix_dgeMatrix_crossprod", optsize)
		hiplarSet("xover_dgeMatrix_matrix_crossprod", optsize)
		hiplarSet("xover_dgeMatrix_determinant", optsize)
		hiplarSet("xover_dgeMatrix_matrix_mm", optsize)
		hiplarSet("xover_dgeMatrix_matrix_mm",optsize)
		hiplarSet("xover_dgeMatrix_norm",optsize)
		hiplarSet("xover_dgeMatrix_solve", optsize)
		hiplarSet("xover_dgeMatrix_matrix_solve", optsize)
		hiplarSet("xover_dgeMatrix_rcond", optsize)
		hiplarSet("xover_dpoMatrix_chol",optsize)
		hiplarSet("xover_dpoMatrix_rcond", optsize)
		hiplarSet("xover_dpoMatrix_solve", optsize)
		hiplarSet("xover_dpoMatrix_dgeMatrix_solve", optsize)
		hiplarSet("xover_dpoMatrix_matrix_solve", optsize)
		hiplarSet("xover_dtrMatrix_chol2inv", optsize)
		hiplarSet("xover_dtrMatrix_dtrMatrix_mm", optsize)
		hiplarSet("xover_dtrMatrix_matrix_mm", optsize)
		hiplarSet("xover_dtrMatrix_solve", optsize)
		hiplarSet("xover_dtrMatrix_matrix_solve", optsize)
		hiplarSet("xover_dsyMatrix_matrix_mm", 0)
		hiplarSet("xover_dsyMatrix_norm", 0)

HiPLARM documentation built on May 29, 2017, 10:42 p.m.