sn_get_buffer_size.h File Reference

Physical buffer size of a station, in jobs, the one in service included. More...

#include <algorithm>
#include <cstddef>
#include <limits>
#include "line/lang/qn/network_struct.h"

Include dependency graph for sn_get_buffer_size.h:

Go to the source code of this file.

Namespaces
namespace	line
namespace	line::sn

Functions
template<class T>
double	line::sn::sn_get_buffer_size (const qn::NetworkStruct< T > &sn, std::size_t ist)
	Physical buffer size of a station, in jobs, the one in service included.

Detailed Description

Physical buffer size of a station, in jobs, the one in service included.

Port of matlab/src/api/sn/sn_get_buffer_size.m. Kendall's K: the tighter of the station capacity sn.cap and the SUM of the per-class capacities sn.classcap, +inf when the station is unbounded. Both fields already fold setCapacity, setClassCapacity, a finite orbit and the closed-chain population, so this is the single place that decides whether a buffer BINDS.

Two details that are easy to get wrong and both matter. The per-class capacities are SUMMED, not minimised: a zero marks a class the station does not serve and is dropped first. And the "unreachable capacity" suppression compares against the population that can actually REACH this station, i.e. the classes it serves, not the model's total, without which a MIXED model reads as finite-buffered at every station an open class never visits.

This lived inside solvers/nc/solver_nc_mem.h as nc::detail::buffer_size. It is the gate of every finite-buffer branch, not an NC internal, so it now sits on the api surface. The BODY is NetworkStruct::buffer_size, which this forwards to: the struct's own has_blocking (the product-form conjunct the cpp solvers read through has_product_form) needs the same rule, and two copies of a predicate this load-bearing is exactly how the member and the free function drift apart.

Definition in file sn_get_buffer_size.h.