| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| README.textile | 2026-09-12 | 2.5 kB | |
| line-3.0.7-cpp.zip | 2026-09-11 | 10.5 MB | |
| line-3.0.7-matlab.zip | 2026-09-11 | 107.5 MB | |
| line-3.0.7-python.zip | 2026-09-11 | 35.1 MB | |
| line-3.0.7-java.zip | 2026-09-11 | 66.3 MB | |
| line-3.0.7.zip | 2026-09-11 | 191.5 MB | |
| Totals: 6 Items | 411.0 MB | 12 | |
- New native C++ codebase (`cpp/`) brought to parity with MATLAB, JAR and Python: full ports of MVA, NC, CTMC, SSA, FLD, MAM, ENV, LN, BA and AG, a `line-cli` front end, exact and multiprecision arithmetic, HiGHS-backed BA bounds, and a C++ manual.
- LDES gains a native C++ discrete-event engine (`common/ldes`) that reproduces the Java SSJ sample path, plus a layered LQN engine with replication, cache nodes, Petri nets, BAS/BBS blocking, capacity regions, busy periods, transient and slotted time, MAPt/PHt, breakdowns, retrial orbits and warm start.
- Cache modeling overhaul: delayed-hit retrieval systems in CTMC, SSA, NC, MVA, LDES and LN, cache networks via per-item classes, item sizes with cost caps, h-LRU/CLIMB/q-LRU/TTL policies, ray asymptotics for the cache normalizing constant, and FLD/ENV mean-field cache transients.
- New solvers and methods: SolverBA bounds (Majumdar-Woodside, QRF, mapqn linear/quadratic reduction, stochastic network calculus, Thomasian fork-join), SolverAG (RCAT), ranked and learned SolverAUTO selection, SolverQNS, interval-valued UQ, and the line-opt optimization, sensitivity and EKF-based LQN inference framework.
- Product-form expansion in NC/MVA: CoMoM, MCMC, importance sampling, order-independent and pass-and-swap networks, MEM, Morrison DPS, SQD/QSA/SCAT/Linearizer AMVA families, SJN, mixed load dependence, busy periods and queue-length moment tables.
- CTMC/SSA: NRM becomes the SSA default engine (PS family, buffers, FCR, KCHOICES, impatience, network signals, polling, SPN, caches); CTMC adds MDD state spaces, CFTP perfect sampling, Fox-Glynn, GMRES/BiCGSTAB, BSCC absorption, REPLY-signal synchronous calls, true BAS blocking, native fork-join, exact state-dependent routing, ME processes and a SageMath symbolic backend.
- Fluid and MAM: minnormal moment closure as the FLD default, TBI, RODAS/LSODA DAE integration, Ko-Pender kp with covariance and fluid Petri nets; MAM adds exact MAP/M/c, MAP/PH/c, MMAP[K]/G[K]/1, M/D/c, D/M/c, PH/M/c, finite buffers, bgchain as closed default, discrete-time Q-MAM and Geo/Geo/1, RAP QBDs, ETAQA and M3PP/MMAP closed-form fitters.
- Layered networks and Petri nets: flat and srvnph layering, replication with fan-in/fan-out, forwarding, async and phase-2 calls, quorum joins, JSQ/round-robin call groups, setup, function and cache tasks, admission constraints, heterogeneous pools, transient LN, LQN-in-ENV and LQN2QN; QPN queueing places, inhibitor arcs, marking-dependent firing rates and PNML interchange.