The SolverMAM class surface: @@SolverMAM/runAnalyzer.m and the gates around it. More...

#include <algorithm>
#include <cmath>
#include <string>
#include <vector>
#include "line/lang/qn/network_struct.h"
#include "line/lang/qn/solver_feature_sets.h"
#include "line/api/sn/sn_is_discrete_time.h"
#include "line/api/sn/sn_nonmarkov_toph.h"
#include "line/solvers/mam/mam_dispatch.h"
#include "line/solvers/mam/mam_types.h"
#include "line/solvers/mam/solver_mam_fj.h"
#include "line/solvers/mam/solver_mam_ldqbd_transient.h"
#include "line/solvers/mam/solver_mam_passage_time.h"
#include "line/solvers/mam/solver_mam_prob.h"
#include "line/solvers/mam/solver_mam_transient_qbd.h"
#include "line/solvers/mva/solver_mva_runner.h"
#include "line/util/error.h"

Include dependency graph for solver_mam_runner.h:

Namespaces
namespace	line
namespace	line::mam

Functions
std::vector< std::string >	line::mam::list_valid_methods ()
	Port of SolverMAM.listValidMethods.
bool	line::mam::rcat_moved_to_ag (const std::string &method, std::string &why)
	True for the four RCAT names that moved to SolverAG, with the redirect message.
void	line::mam::check_method (const std::string &method)
	An unlisted method is refused; one that MOVED is redirected by name.
template<class T>
std::string	line::mam::mam_model_method_refusal (const qn::NetworkStruct< T > &L, const std::string &method)
	check_model_method asked WITHOUT raising: the same verdict as a sentence.
template<class T>
MamSolution< T >	line::mam::solver_mam_solve (const qn::NetworkStruct< T > &L, const MamOptions &opt)
	The gates and the dispatch of @@SolverMAM/runAnalyzer.m, without the metric filter.
template<class T>
mva::AvgResult< T >	line::mam::solver_mam_run_analyzer (const qn::NetworkStruct< T > &L, const MamOptions &opt)
	Port of @@SolverMAM/runAnalyzer.m for the lang='matlab' path: solve, then apply the metric filter @@NetworkSolver/getAvg puts between the analyzer and the caller.
template<class T>
ProbTable< T >	line::mam::solver_mam_get_prob (const qn::NetworkStruct< T > &L, const MamOptions &opt, std::size_t node, const mva::AvgResult< T > &avg)
	@@SolverMAM/getProb.m: the joint (level, phase) table at a node.
template<class T>
std::vector< T >	line::mam::solver_mam_get_prob_marg (const qn::NetworkStruct< T > &L, const MamOptions &opt, std::size_t ist, std::size_t jobclass, const mva::AvgResult< T > &avg)
	@@SolverMAM/getProbMarg.m: P(n jobs of one class) at a station.
template<class T>
qsys::BmapM1Result< T >	line::mam::solver_mam_get_mam_result (const qn::NetworkStruct< T > &L)
	@@SolverMAM/getMAMResult.m: the M/G/1-type internals of a single queue.
template<class T>
std::vector< RespTCdf< T > >	line::mam::solver_mam_get_cdf_respt (const qn::NetworkStruct< T > &L, const MamOptions &opt)
	@@SolverMAM/getCdfRespT.m: the response-time CDF per class.
template<class T>
std::vector< RespTCdf< T > >	line::mam::solver_mam_get_sjrn_t (const qn::NetworkStruct< T > &L, const MamOptions &opt)
	@@SolverMAM/getSjrnT.m and sjrnT.m, both aliases of getCdfRespT.
template<class T>
bool	line::mam::mam_has_fj_percentiles (const qn::NetworkStruct< T > &L, const MamOptions &opt)
	Whether getPerctRespT reads the FJ_codes table rather than inverting a CDF.
template<class T>
std::vector< std::vector< T > >	line::mam::solver_mam_get_perct_respt (const qn::NetworkStruct< T > &L, const MamOptions &opt, const std::vector< double > &percentiles)
	@@SolverMAM/getPerctRespT.m: response-time percentiles per class.
template<class T>
TranResult< T >	line::mam::solver_mam_get_tran_avg (const qn::NetworkStruct< T > &L, const MamOptions &opt_in)
	Port of @@SolverMAM/getTranAvg.m: transient queue length, utilization and throughput.

Detailed Description

The SolverMAM class surface: @@SolverMAM/runAnalyzer.m and the gates around it.

Everything that happens BEFORE and AFTER one inner solve lives here: the method whitelist (SolverMAM.listValidMethods), the per-method structural gates (SolverMAM.supportsModelMethod), the process gate, and the metric filter. runAnalyzer's Finite Capacity Region rejection needs no counterpart: the C++ NetworkStruct has no region field, so such a model cannot be built.

THE METRIC FILTER IS SHARED WITH SolverMVA AND SolverNC VERBATIM (filter_metric, sn_get_residt_from_respt, sn_get_arvr_from_tput in solver_mva_runner.h), because @@NetworkSolver/getAvg is solver-independent: it is literally the same code path for all three in the reference.

NON-MARKOVIAN SERVICE IS CONVERTED, NOT REFUSED (since 2026-08-01). solver_mam_analyzer.m:40 runs sn_nonmarkov_toph on its own copy of the struct, and so does solver_mam_solve here: a Gamma, Weibull, Lognormal, Pareto or Uniform is replaced by the two-moment concentrated-ME fit, or by the 20-phase Bernstein density fit under phfit_cme = false, and a Det by a 20-phase Erlang when preserve_det is off. The conversion runs AFTER the slotted test, which would otherwise see a continuous fit where a Geometric was declared. What survives the conversion and still reaches check_processes is NHPP / MAPt / PHt, whose content is a schedule that no homogeneous surrogate can carry, and those are refused by name.

TWO GATES REMAIN STRICTER HERE THAN IN THE REFERENCE, and both are the honest outcome rather than a limitation:

A CLASS-SWITCHING CHAIN AT AN FCFS STATION. The reference's own chain-1 mark collapse makes the marking handed to the queue solver disagree with the class count, and MATLAB then fails on a dimension mismatch. The analyzer refuses by name with both counts; see solver_mam_basic.h.
LOAD DEPENDENCE UNDER 'default' OUTSIDE THE LD-QBD SHAPE. MATLAB has no counterpart because it has no LoadDependence declaration for MAM at all (mam_feature_set's declaration is a C++-only widening, solver_feature_sets.h). check_model_method refuses a load-dependent model here when its shape is not the one solver_mam_ldqbd requires, because the featset gate cannot see topology and 'default' would otherwise fall through to solver_mam_basic, which never reads st.lldscaling.

Definition in file solver_mam_runner.h.

Namespaces

Functions

Detailed Description