pfqn_mushift.h File Reference

Shift the load-dependent service-rate lattice of selected stations. More...

#include <cstddef>
#include <vector>
#include "line/num/number.h"
#include "line/util/error.h"
#include "line/util/matrix.h"

Include dependency graph for pfqn_mushift.h:

Go to the source code of this file.

Namespaces
namespace	line
namespace	line::pfqn

Functions
template<class T>
Matrix< T >	line::pfqn::pfqn_mushift (const Matrix< T > &mu, const std::vector< std::size_t > &iset)
	Shift the load-dependent service-rate lattice of selected stations.
template<class T>
Matrix< T >	line::pfqn::pfqn_mushift (const Matrix< T > &mu, std::size_t i)
	Single-station overload, the form the reference is actually called with.

Detailed Description

Shift the load-dependent service-rate lattice of selected stations.

Templated port of matlab/src/api/pfqn/pfqn_mushift.m, cross-checked against jar/src/main/java/jline/api/pfqn/ld/Pfqn_mushift.java.

The load-dependent convolution and MVA recursions need mu(i, n+1) when a job is added at station i. This drops the first column of that station's rate row and the last column of every other row, giving an (M x N-1) lattice in which station i is "one job ahead":

mushifted(i, j) = mu(i, j+1) for the shifted stations mushifted(m, j) = mu(m, j) otherwise, j = 1 .. N-1

Arithmetic: EXACT-CAPABLE. The routine only copies entries, so it is exact in every arithmetic and carries no transcendental gate.

Note on the MATLAB loop. The reference recomputes the whole matrix inside a loop over iset, so with more than one index only the LAST one is actually shifted; every earlier one is overwritten. That is a defect, not a convention – the routine's own name and its single caller (one station at a time) say each listed station should be shifted – so this port shifts every station in iset. With the single-element iset that the reference is called with, the two agree exactly.

Definition in file pfqn_mushift.h.