ILOG CPLEX 8.1 Release Notes
Contents
1. ILOG CPLEX 7 5 and earlier ILOG CPLEX 8 0 Symbolic Constant Value Symbolic Constant Value CPX OPTIMAL 1 CPX STAT OPTIMAL 1 CPX IT LIM FEAS 5 CPX STAT ABORT IT LIM 10 CPX IT LIM INFEAS 6 CPX STAT ABORT IT LIM 10 CPX TIME LIM FEAS 7 CPX STAT ABORT TIME LIM 11 CPX TIME LIM INFEAS 8 CPX STAT ABORT TIME LIM 11 CPX OPTIMAL INFEAS 11 CPX STAT OPTIMAL INFEAS 5 CPX ABORT DUAL INFEAS 14 CPX STAT ABORT USER 3 CPX ABORT PRIM INFEAS 15 CPX STAT ABORT USER 3 CPX ABORT PRIM DUAL INFEAS 16 CPX STAT ABORT USER 3 CPX ABORT PRIM DUAL FEAS 17 CPX STAT ABORT USER 13 CPX PRIM INFEAS 32 CPX STAT INFEASBLE 3 CPX DUAL INFEAS 33 CPX STAT UNBOUNDED 2 CPX PRIM DUAL INFEAS 34 CPX STAT INFEASIBLE 2 CPX_PRIM_OBJ_LI 35 CPX_STAT_ABORT_PRIM_OBJ_LIM 21 CPX_DUAL_OBJ_LIM 36 CPX_STAT_ABORT_DUAL_OBJ_LIM 22 CPX_OPTIMAL_FACE_UNBOUNDED 37 CPX_STAT_OPTIMAL_FACE_UNBOUNDED 20 CPX_NUM_BEST_PRIM_DUAL_FEAS 38 CPX_STAT_NUM_BES 6 CPX_NUM_BEST_PRIM_INFEAS 39 CPX_STAT_NUM_BES 6 CPX_NUM_BEST_DUAL_INFEAS 40 CPX_STAT_NUM_BES 6 CPX NUM BEST PRIM DUAL INFEAS 41 CPX STAT NUM BES 6 CPX BARRIER NUM ERROR 42 CPX STAT NUM BES 6 CPX BARRIER INCONSISTENT 43 no longer used 10 ILOG CPLEX 8 1 RELEASE NOTES RELEASE NOTES FOR CPLEX 8 0 REPRODUCED2. ILOG CPLEX 7 5 and earlier ILOG CPLEX 8 0 Symbolic Constant Value Symbolic Constant Value CPX ABORT INFEAS 13 CPX STAT ABORT USER 13 CPX ABORT CROSSOVER 18 CPX STAT ABORT USER 13 Table 2 Conversion of CPXget stat return value when CPXgetmethod is CPX ALG DUAL ILOG CPLEX 7 5 and earlier ILOG CPLEX 8 0 Symbolic Constant Value Symbolic Constant Value CPX OPTIMAL 1 CPX STAT OPTIMAL T CPX INFEASIBLE 2 CPX STAT UNBOUNDED 2 CPX UNBOUNDED 3 CPX STAT INFEASIBLE 3 CPX_OBJ_LIM 4 CPX_STAT_ABORT_OBJ_LIM 12 CPX_IT_LIM_FEAS 5 CPX STAT ABORT IT LIM 10 CPX IT LIM INFEAS 6 CPX STAT ABORT IT LIM 10 CPX TIME LIM FEAS 7 CPX STAT ABORT TIME LIM 11 CPX TIME LIM INFEAS 8 CPX STAT ABORT TIME LIM 11 CPX NUM BEST FEAS 9 CPX STAT NUM BES 6 CPX NUM BEST INFEAS 0 CPX STAT NUM BES 6 CPX OPTIMAL INFEAS 1 CPX STAT OPTIMAL INFEAS 5 CPX_ABORT_FEAS 2 CPX STAT ABORT USER 13 CPX ABORT INFEAS 3 CPX STAT ABORT USER 13 CPX ABORT CROSSOVER 8 CPX STAT ABORT USER 13 ILOG CPLEX 8 1 RELEASE NOTES 9 RELEASE NOTES FOR CPLEX 8 0 REPRODUCED Table 3 Conversion of CPXget stat return value when CPXgetmethod is CPX ALG BARRIER
3. Table 4 Conversion of error return values from any of the optimizers ILOG CPLEX 7 5 and earlier ILOG CPLEX 8 0 Symbolic Constant Value Symbolic Constant Value CPXERR_BOUNDS_INFEAS 1100 CPX_STAT_INFEASIBLE 3 CPXERR_PRESLV_INForUNBD 1101 CPX_STAT_INForUNBD 4 CPXERR_PRESLV_INE 1117 CPX STAT INFEASIBLE 3 CPXERR PRESLV UNBD 1118 CPX STAT UNBOUNDED 2 CPXERR DEMAND BALANCE 1533 CPX STAT INFEASIBLE 3 Solution Status Codes for Concert Technology Users Table 5 contains the new IloCplex getCplexStatus return values when the model has been solved with the primal simplex or dual simplex algorithm that is IloCplex getAlgorithm returns IloCplex Primal or IloCplex Dual Table 5 Conversion of IloCplex getCplexStatus Primal or Dual Simplex ILOG CPLEX 7 5 and earlier ILOG CPLEX 8 0 Optimal Optimal Infeasible Infeasible Unbounded Unbounded InforUnbd InforUnbd ObjLim AbortObjLim ItLimFeas AbortItLim ItLimInfeas AbortItLim TimeLimFeas AbortTimeLim TimeLimInfeas AbortTimeLim NumBestFeas NumBest NumBestInfeas NumBest OptimalInfeas OptimalInfeas ILOG CPLEX 8 1 RELEASE NOTES 11 RELEASE NOTES FOR CPLEX 8 0 12 Table 5 Conversion of IloCplex getCplexStatus Primal or Dual Simplex REPRODUCED ILOG CPLEX 7
4. ILOG CPLEX 8 1 Release Notes December 2002 Copyright 2001 2002 by ILOG This document and the software described in this document are the property of ILOG and are protected as ILOG trade secrets They are furnished under a license or non disclosure agreement and may be used or copied only within the terms of such license or non disclosure agreement No part of this work may be reproduced or disseminated in any form or by any means without the prior written permission of ILOG S A Printed in France Table of Contents Conversion Notes i u ars ee ne 5 New Features 0 0c ccc tennin eee rennen nennen nennen en nenn 5 Concert Technology Version mmn 6 Advanced Documentation and Examples lt 6 Additional informati0N ooooocoocnccnna Rh hh hh hn 6 Release Notes for CPLEX 8 0 reproduced oooococococnoco e eee eens 7 Conversion Notes for CPLEX 8 0 0 00 cece ete rn 7 Solution Status Codes 0 0 eens 7 Problem Types 242 2 004 ov ed Era ERE Re ea en 14 Other Changes Affecting All Users 222 2 n onen een seen I 14 Other Changes Affecting Component Library Users 2 222222 aa 16 New Features in CPLEX 8 0 oococcccccoc rh rn 17 Algorithmic Features LL kas 17 Component Libraries Usage o oooocococcccoc RR teens 20 ILOG CPLEX 8 1 RELEASE NOTES iii TABLE OF CONTENTS iv ILOG CPLEX 8 1 RELEASE NOTES ILOG CPLEX 8 1 Release Notes Please review th
5. 5 and earlier ILOG CPLEX 8 0 AbortFeas AbortUser AbortInfeas AbortUser AbortCrossover AbortUser Table 6 contains the new IloCplex getCplexStatus return values when the model has been solved with the barrier algorithm that is IloCplex getAlgorithm returns IloCplex Barrier Table 6 Conversion of IloCplex getCplexStatus Barrier ILOG CPLEX 7 5 and earlier ILOG CPLEX 8 0 Optimal Optimal InforUnbd InforUnbd ItLimFeas AbortItLim ItLimInfeas AbortItLim TimeLimFeas AbortTimeLim TimeLimInfeas AbortTimeLim Optimallnfeas OptimalInfeas AbortDualInfeas AbortUser AbortPrimInfeas AbortUser AbortPrimDualInfeas AbortUser AbortPrimDualFeas AbortUser PrimInfeas Infeasible Duallnfeas Unbounded PrimDuallnfeas Infeasible PrimObjLim AbortPrimObjLim DualObjLim AbortDualObjLim OptimalFaceUnbounded OptimalFaceUnbounded NumBestPrimDualFeas NumBest ILOG CPLEX 8 1 RELEASE NOTES RELEASE NOTES FOR CPLEX 8 0 REPRODUCED Table 6 Conversion of IloCplex getCplexStatus Barrier ILOG CPLEX 7 5 and earlier ILOG CPLEX 8 0 NumBestPrimnfeas NumBest NumBestDualInfeas NumBest NumBestPrimDualInfeas NumBest BarrierNumError NumBest BarrierInconsistent no longer used New Solution Query Routines The new Callable Library functions CPXsolninfo and CPXNETsolninfo are provided as a mean
6. first describes changes to solution status codes These changes will affect most users so we recommend that everyone reads this section The second section describes changes to problem types that will also affect many users The third section describes other changes affecting all users The fourth section describes additional changes that affect users who embed ILOG CPLEX optimizers in their applications Solution Status Codes Solution status codes for LP and QP have been completely revised in ILOG CPLEX 8 0 making the interface simpler and more unified This section describes the changes that must be made to existing applications to work with the new status codes The new codes are described in more detail in Appendix B of the ILOG CPLEX 8 0 Reference Manual MIP status codes have been changed too but the changes are much smaller in scope than those for LP and QP The MIP changes are discussed in MIP Solution Status Codes below The major change in status codes is that their meaning no longer depends on the optimizer used solution status codes now relate to the problem being solved This section includes Tables 1 to 6 below that show how to convert from the previous optimizer specific interpretations to the new problem interpretations Another change is that the optimizers will no longer return an error value when either presolve or consistency checks have found the problem to be infeasible or unbounded These conditions now result in a norma
7. the presolved model to a file via the routine CPXpreslvwrite SGI users need IRIX 6 5 9 or above to run ILOG CPLEX 8 0 Other Changes Affecting Component Library Users User callback functions written in C must be re defined to use const for certain parameters The function definition must use the const modifier and the pointer argument must be declared as type CPXCENVpt r the constant variant of CPXENVpt r Moreover the function itself must not change any const arguments or pass them to functions where they are not const that is const arguments cannot be changed within a user written callback Several features that were previously announced as obsolete have now been removed They are the CPLEX License Manager the Old Memory Model and several symbolic constant names appearing in C include file cplex h The Kernighan amp Ritchie style C function prototypes also have been removed Consult earlier CPLEX Release Notes documents if you still need to perform one of these conversions The r1oCplex member functions setRootAlgorithm setNodeAlgorithm getRootAlgorithm and getNodeAlgorithm have been replaced by parameters and are now deprecated features which will be removed in a future release For example setRootAlgorithm alg becomes setParam IloCplex RootAlg alg Seethe discussion related to these in New Features in CPLEX 8 0 on page 17 of these Release Notes To support a unified approach to LP and OP models functi
8. ARCROSSALG in the Callable Library and BARRIER CROSSOVER in the Interactive Optimizer if you wish to use barrier without Crossover The default value for the BarGrowth parameter CPX PARAM BARGROWTH or IloCplex BarGrowth has been increased from 10 to 10 to reduce the likelihood that a model will be declared to have an unbounded optimal face The barrier optimizer no longer has an explicit iteration count limit by default In other words BarItLim the barrier iteration limit parameter CPX PARAM BARITLIM or IloCplex BarItLim now defaults to a very large value The method for expressing the quadratic term of the objective function in ILOG CPLEX LP format for file input has changed The factor of 0 5 will no longer be assumed for the terms appearing within square brackets The factor is specified by appending the characters 2 after the closing bracket In order to ease conversion the implied factor will continue to be assumed for this release however a warning message will be printed that says Warning Line 234 deprecated format expecting 2 We recommend you begin using this new form for expressing the quadratic term of a QP now ILOG CPLEX 8 1 RELEASE NOTES 15 RELEASE NOTES FOR CPLEX 8 0 REPRODUCED 16 While error code CPXERR PRESLV INForUNBD Error 1101 is no longer returned by the optimizers as noted in the previous section on status code changes it can still arise when a user attempts to write
9. PLEX 8 1 RELEASE NOTES 17 RELEASE NorES FOR CPLEX 8 0 REPRODUCED 18 e Goals e Simplex optimizers for QP problems e Mixed Integer QP problems e Sifting e Concurrentopt for multiple processor computers e MIP parameters e Barrier parameters e Function to remove an SOS from the problem Goals Goals which have proven to be quite powerful and widely used in Constraint Programming are now available for ILOG CPLEX MIP Goals provide a new level of control for users who want to manage the way that ILOG CPLEX explores the MIP search tree See the ILOG CPLEX 6 0 User s Manual for full details on using goals in ILOG CPLEX Simplex Optimizers for Quadratic Programming Problems Dual Simplex and Primal Simplex optimizers have been added for solving Quadratic Programming QP problems in prior ILOG CPLEX releases QP problems were solved only by the Barrier optimizer which can not restart an optimization from a previous solution Additionally Presolve has been extended to perform new reductions on QP problems benefitting all the QP optimizers See the ILOG CPLEX 8 0 User s Manual for full details on these new capabilities for solving QP problems Mixed Integer Quadratic Programming Problems The branch amp cut algorithm has been extended to permit discrete variables in QP formulations so that Mixed Integer Quadratic Programming MIQP is now supported That is a problem of the new type CPXPROB MIOP may contain any combination of C
10. ds at once For Concert Technology users this capability is exercised by setting IloCplex RootAlg to IloCplex AutoAlg Or IloCplex Concurrent C or by setting IloCplex IntParam RootAlg to IloCplex Algorithm Auto or IloCplex Algorithm Concurrent Java and invoking method cplex solve The first thread uses Dual Simplex if a second thread is available it uses Barrier if your license supports Barrier and the next thread uses Primal Simplex If additional threads are available on the computer they are applied to the Barrier run by an invocation of Parallel Barrier Unfortunately this does not happen on SGI platforms due to a problem in the threading libraries On a machine with enough memory this will result in a solution being returned by the fastest of the available algorithms on each problem eliminating the need to choose a single optimizer for all purposes MIP Parameters The default MIP emphasis setting CPX PARAM MIPEMPHASIS or IloCplex MipEmphasis has been changed to a new value of Balanced CPX MIPEMPHASIS BALANCED or MIPEmphasisBalanced which seeks a balance between searching for feasible solutions and proving optimality There should be no conversion issue related to this change because the previous Optimality default has been renumbered to 2 and the new default remains numbered 0 The new BestBound choice CPX MIPEMPHASIS BESTBOUND or MIPEmphasisBestBound emphasizes moving the Best Bound as an aggressive techn
11. ds of class IloCplex in Concert Technology One use of this is to determine if a primal feasible solution exists when the solution status is Unbounded MIP Solution Status Codes The only solution status change for MIP is the addition of two new status codes CPXMIP UNBOUNDED and CPXMIP_INForUNBD ILOG CPLEX 7 5 would return an error code if it found a MIP model to be infeasible or unbounded ILOG CPLEX 8 1 RELEASE NOTES 13 RELEASE NorES FoR CPLEX 8 0 REPRODUCED 14 Note that in ILOG CPLEX 7 5 a model would be declared infeasible if the branch amp cut optimizer found no feasible solutions while in ILOG CPLEX 8 0 this model will be declared infeasible or unbounded The new status code more accurately describes the conclusion that can be reached from this outcome The user may turn off dual reductions using the presolve reduction parameter CPX PARAM REDUCE Or IloCplex Reduce ifa firm infeasible conclusion is desired Problem Types To accommodate the new MIQP problem class ILOG CPLEX problem types have been reorganized and streamlined The allowed values for problem type in the C interface are now CPXPROB LP CPXPROB MILP CPXPROB FIXEDMILP CPXPROB OP CPXPROB MIQP and 9 9 CPXPROB FIXEDMIOP Consult the ILOG CPLEX 6 0 Reference Manual entry CPXchgprobtype for details Concert Technology now interprets all problems as MIQPs and treats MILP QP and LP as special case
12. enMP binary The default parallel thread count has been changed to one It previously defaulted to its maximum possible value as determined by the number of processors in the machine and the number of threads licensed The behavior of the default Automatic setting of CPX PARAM LPMETHOD has been changed This setting will now choose Primal Simplex in cases where a primal feasible starting basis is available Also if the number of parallel threads global default thread count is set to a value greater than 1 and an advanced simplex basis is not available the new Concurrentopt feature see Concurrentopt on page 19 will be invoked Concert Technology will give this same behavior for the default Automatic settings of parameters IloCplex RootAlg C or I1oCplex IntParam RootAlg Java The MIP backtrack parameter CPX PARAM BTTOL Or IloCplex Bttol controls the amount of backtracking in the branch amp cut search its default value has been changed from 0 01 to 0 9999 because of various changes made in the overall branching strategy As a conversion issue this change would affect only users who for some reason manually set ILOG CPLEX parameters to their default values Most users will find the new default value works well Users may no longer choose Dual to iteration limit then barrier or Barrier without crossover as root or node algorithms in MIP Use the barrier crossover parameter BarCrossAlg in Concert Technology CPX PARAM B
13. ese brief release notes carefully before using ILOG CPLEX 8 1 They are organized in three sections e Conversion Notes a guide to the upgrade procedure from ILOG CPLEX 8 0 to ILOG CPLEX 8 1 e New Features an overview of the features introduced in ILOG CPLEX 8 1 e Additional information Conversion Notes No conversion of application source code is required when moving from CPLEX version 8 0 to version 8 1 CPLEX 8 0 license keys can be used with version 8 1 For details on conversion from CPLEX 7 5 to CPLEX 8 0 please refer to Conversion Notes for CPLEX 6 0 New Features This section lists the new features in ILOG CPLEX 8 1 ILOG CPLEX 8 1 RELEASE NOTES 5 ADDITIONAL INFORMATION Concert Technology Version CPLEX 8 1 uses ILOG Concert Technology 1 3 for compatibility with ILOG Solver 5 3 Advanced Documentation and Examples The ILOG CPLEX 8 1 Advanced Reference Manual is included with the standard documentation set Advanced examples are available in cp1ex81 examples src Additional information If you are upgrading from ILOG CPLEX 8 0 these notes should be all you need If you are upgrading from a version of CPLEX prior to 8 0 please refer first to the Release Notes for CPLEX 6 0 reproduced included for your convenience in this document and to any prior release notes ILOG CPLEX 7 5 ILOG CPLEX 7 1 and so forth to bring your ILOG CPLEX software up to date from your particular version The de
14. have a function isQuadratic which indicates whether the current problem contains a quadratic term in the objective The new C function clearModel is available and is used to remove the contents of a model without destroying it Java users now can use the static method IloCplex putenv String envstr ina manner analogous to C and C usage For compatibility with the Java version of Concert Technology for C users can now use the following methods getNbarrierIterations getNSOSs getNNZs getNbinVars getNsemiContVars getNsemiIntVars getRangeSA getAlgorithm getSubAlgorithm getNintVars Visual Basic users no longer need to declare the functions in the CPLEX DLL themselves They can use the new cplex bas include file distributed in the include ilcplex subdirectory There are also several new examples that illustrate how the CPLEX DLL can be called from Visual Basic programs These can be found in the examples vb ILOG CPLEX 8 1 RELEASE NOTES 21 RELEASE NOTES FOR CPLEX 8 0 22 ILOG CPLEX 8 1 REPRODUCED RELEASE NOTES ILOG WORLDWIDE INFORMATION CENTER TEL 1 775 831 7744 1 800 367 4564 FRANCE ILOG S A Gentilly Tel 33 0 1 49 08 35 00 E mail info ilog fr GERMANY ILOG Deutschland GmbH Bad Homburg v d H Tel 49 6172 40 60 0 E mail info ilog de JAPAN ILOG CO LTD Tokyo Tel 81 3 5211 5770 E mail info ilog co jp SINGAPORE ILOG S Pte Ltd Singapo
15. ique for proving optimality on extremely difficult models The other choice for this parameter CPX MIPEMPHASIS FEASIBILITY is unchanged Most models will reach the optimal solution fastest using the Optimality emphasis but the default Balanced setting produces good feasible solutions earlier and more frequently while only slightly degrading time to proven optimality The Concert Technology enumerations for parameters IloCplex MIPEmphasis C and IloCplex IntParam MIPEmphasis Java have been changed accordingly Refer to the ILOG CPLEX include files for details The branch amp cut search can be fine tuned with the new dive type parameter CPX PARAM DIVETYPE Or IloCplex DiveType This parameter controls how much look ahead or probing is done while processing nodes The default value is the automatic setting 0 which will do some probing The setting of 1 activates the strategy used in previous ILOG CPLEX versions and the setting of 2 does probing all the time ILOG CPLEX 8 1 RELEASE NOTES 19 RELEASE NOTES FOR CPLEX 8 0 REPRODUCED 20 The new parameter CPX PARAM SYMMETRY controls symmetry breaking cuts IloCplex Symmetry C and IloCplex IntParam Symmetry Java for Concert users a Preprocessing feature These cuts are off by default but turning them on can be helpful on some models Barrier Parameters The feasibility algorithms for Barrier CPX PARAM BARALG setting
16. l termination Note that a zero return from an optimizer or a nonzero return from CPXgetstat no longer means that a solution exists and can be queried as it did in previous versions of ILOG CPLEX For non optimal solutions one must check for the existence of a solution with the new solution information routines described ILOG CPLEX 8 1 RELEASE NOTES 7 RELEASE NorES FoR CPLEX 8 0 REPRODUCED below The handling of these conditions by Concert Technology has not changed an exception is thrown when a solution query is made and there is no solution We have also changed the meaning of the unbounded return code slightly ILOG CPLEX 8 0 will declare a problem to be unbounded when the problem has an unbounded ray In previous versions an unbounded status also implied the existence of a feasible solution The new solution information routines can be used to determine whether a feasible solution was found as described below See the note in Appendix B of the ILOG CPLEX 6 0 Reference Manual for more information on the meaning of the unbounded status Information about the status of an aborted optimization that is feasibility of the current solution was previously encoded in the solution status This information must now be queried separately through new solution information routines described below Solution Status Codes for Callable Library Users The following tables provide a mapping from status codes in ILOG CPLEX 7 5 to the new sta
17. lation so that its quadratic matrix is positive semi definite PSD A new function CPXqpindefcertificate IloCplex qpIndefCertificate for Concert Technology users is available In cases where the quadratic matrix in the objective function of a QP model violates the assumption of positive semi definiteness it returns a vector that demonstrates the violation Two new query routines are provided to support the new sifting algorithm These are CPXgetsiftitcnt IloCplex getNsiftingIterations C and IloCplex getNsiftingIterations Java for Concert Technology users and CPXgetsiftphaselcnt IloCplex getNsiftingPhaseOneIterations ILOG CPLEX 8 1 RELEASE NOTES RELEASE NOTES FOR CPLEX 8 0 REPRODUCED C and I1oCplex getNsiftingPhaseOneIterations Java for Concert users CPXgetsiftitcnt returns the number of sifting iterations performed during the most recent optimization and CPXgetsiftphaselcnt returns the number of sifting phase 1 iterations performed by the most recent optimization before a feasible solution was found C parameters IloCplex RootAlg and IloCplex NodeAlg now are available as is the case for Java users The parameters replace the functionality previously provided by the IloCplex member functions setRootAlgorithm setNodeAlgorithm getRootAlgorithm and getNodeAlgorithm These four functions are now deprecated features and will be removed in a future release Concert Technology users now
18. ns Old New BasisStatus getStatus const IloNumVar con BasisStatus getBasisStatus const IloNumVar con const const BasisStatus getStatus const IloRange con BasisStatus getBasisStatus const IloRange con const const void getStatuses BasisStatusArray stat void getBasisStatuses BasisStatusArray stat const IloNumVarArray con const IloNumVarArray con const const void getStatuses BasisStatusArray stat void getBasisStatuses BasisStatusArray stat const IloRangeArray con const IloRangeArray con const const void getStatuses BasisStatusArray cstat void getBasisStatuses BasisStatusArray cstat const IloNumVarArray var const IloNumVarArray var BasisStatusArray rstat BasisStatusArray rstat const IloRangeArray con const IloRangeArray con const const New Features in CPLEX 8 0 Here are the new features found in the Application Programming Interface API that you may find useful in solving particular kinds of models ILOG CPLEX 8 0 contains many performance enhancements that benefit all problem types without any need to change or add to your present solution procedures But in addition various features have been added that provide new functionality Here is an outline of what you may wish to try Refer to the ILOG CPLEX 8 0 User s Manual and ILOG CPLEX 8 0 Reference Manual for complete details on these features Algorithmic Features The algorithmic features added in this version are ILOG C
19. on I1oCplex LPCallbackI has been renamed to 11oCplex ContinuousCallbackI Also the callbacks IloCplex PrimalSimplexCallbackI and IloCplex DualSimplexCallbackI have been combined into a single callback IloCplex SimplexCallbackI IloCplex CallbackI makeClone has been renamed to IloCplex CallbackI duplicateCallback Programs that use the macros for defining callbacks will be unaffected by this change IloCplex Algorithm NetworkPrimal and IloAlgorithm NetworkDual have been merged into IloAlgorithm Network Routine CPXcangetbase has been removed Identical functionality is now available through the new CPXsolninfo routine To improve clarity and consistency the names of a few Concert Technology for C methods and objects have been changed ILOG CPLEX 8 1 RELEASE NOTES RELEASE NOTES FOR CPLEX 8 0 REPRODUCED Method isQP is now isQuadratic Methods getLb getLbs getUb and getUbs have been changed to getLB getLBs getUB and getUBs The return value of functions getCplexStatus and getCplexSubStatus has been changed from Status to CplexStatus The functions for querying basis information have been renamed to getBasisStatus and getBasisStatuses according to Table 7 New Names for Basis Query Functions on page 17 In all cases the old names are still supported but will be discontinued in the next release Table 7 New Names for Basis Query Functio
20. ontinuous Binary General Integer Semi Continuous or Special Ordered Set variables in conjunction with an objective function that may contain a quadratic term See the ILOG CPLEX 6 0 User s Manual for full details on these new capabilities for solving MIQP problems Sifting A new method Sifting is now provided for solving continuous LP models It is activated by setting the parameter CPX PARAM LPMETHOD to the value CPX ALG SIFTING 5 and calling CPX1popt For Concert it is activated by setting parameter IloCplex RootAlg to IloCplex Sifting C Or IloCplex IntParam RootAlg to IloCplex Algorithm Sifting Java and invoking method cplex solve Sifting solves a sequence of LP subproblems where the results from one subproblem are used to select columns from the original model for inclusion in the next subproblem This iterative ILOG CPLEX 8 1 RELEASE NOTES RELEASE NOTES FOR CPLEX 8 0 REPRODUCED sifting process eventually converges to an optimal solution for the original model Sifting is especially applicable to models with many more columns than rows Concurrentopt A new method is provided for solving LP models on multiple CPU computers with appropriate ILOG CPLEX licensing If the CPX PARAM LPMETHOD parameter is set to CPX ALG AUTOMATIC 0 and no advanced simplex basis is loaded or if the parameter is set to CPX ALG CONCURRENT 6 subsequent CPX1popt calls will launch LP optimizations on multiple threa
21. re Tel 65 773 06 26 E mail info ilog com sg SPAIN ILOG S A Madrid Tel 34 91 710 2480 E mail info ilog es UK ILOG Ltd Bracknell Tel 44 0 1344 66 16 00 E mail info ilog co uk USA ETATS UNIS ILOG Inc Mountain View CA Tel 1 650 567 8000 E mail info ilog com Representatives and distributors in other countries WWW ILOG COM ILOG CPLEX and the ILOG logotype are registered trademarks and all ILOG product names are trademarks of ILOG All other brand product and company names are trademarks or registered trademarks of their respective holders 20n1e06
22. s The problem types CPXPROB_RELAXED now CPXPROB RELAXEDMILP and CPXPROB_ZEROEDOP and their counterparts solveRelaxed and solveZeroedQP in C and Java or relaxed mip and zeroed gp in the Interactive Optimizer will be removed in the next release We recommend you now convert any procedures with these problem types to use typically problem type CPXPROB LP in Callable Library programs Concert Technology users should use IloConversion objects instead of using solveRelaxed and should replace a quadratic objective with a linear objective model remove quadraticObj model add linearObj instead of using solveZeroedoP Removal of other problem types is not contemplated at this point Contact ILOG Technical Support if you need assistance in identifying the appropriate changes Other Changes Affecting All Users Parallel algorithms are now included in the standard ILOG CPLEX binaries and libraries Parallelism is controlled through user parameters in conjunction with limits set by your ILOG CPLEX license key Most users of the ILOG CPLEX Technologies Interactive Optimizer and the various Component Libraries will now have just one version to be concerned with instead of distinct parallel and serial versions The exception is users ILOG CPLEX 8 1 RELEASE NOTES RELEASE NOTES FOR CPLEX 8 0 REPRODUCED interested in the parallel dual simplex optimizer that will be included in a separate Op
23. s 1 and 2 now support solution of QP models Function to Remove an SOS CPXdelsetsos is a new function which removes one or more special ordered sets SOS from the problem Component Libraries Usage The use of const in C function prototypes allows the specification of arguments that will not be changed by execution of a function This modifier has been added throughout cplex h Several new typedefs have been added to ease the use of const with ILOG CPLEX structure types including CPXCENVptr CPXCLPptr and CPXCNETpt r This addition will not affect existing programs except ones that use callback functions as mentioned in the Conversion Notes section of this document Two new functions are provided to work with ILOG CPLEX parameters in C e CPXgetparamname Which provides the string corresponding to a parameter number and e CPXgetparamnum Which provides the number corresponding to a string matching a parameter name A new parameter CPX PARAM OPMETHOD and new routine CPXgpopt Callable Library only are available to control the choice of optimizer for the solution of QP models This same functionality is already provided in Concert through the solve function which now no longer issues a warning message if a Simplex optimizer is selected for QP A new parameter CPX_PARAM_OPMAKEPSDIND IloCplex QPmakePSDInd for Concert Technology users determines whether ILOG CPLEX will attempt to adjust a binary MIQP formu
24. s to access solution information produced by LP and QP optimization routines Some of this information was previously encoded in the solution status values In Concert Technology for C and Java users this functionality is provided with methods IloCplex isPrimalFeasible and IloCplex isDualFeasible The functions are described in the ZLOG CPLEX 8 0 Reference Manual and their usage is explained in the LOG CPLEX 8 0 User s Manual If an Infeasible or Unbounded code is returned there may not be a solution available Check for the existence of a solution with the so1ntype parameter CPXsolninfo for example status CPXsolninfo env lp amp solntype amp method amp pfeasind amp dfeasind If the solntype is CPX NO SOLN then there is no solution and any of the solution query routines for example CPXgetitcnt CPXsolution will result in an error When a solution is available the solntype value indicates whether the solution is basic value CPX BASIC SOLN or not CPX NONBASIC SOLN As mentioned above Concert Technology users will get an exception if a solution is queried but no solution is available This behavior is the same as in previous versions The solution information functions also provide a simple means to find out whether the solution is primal or dual feasible through the p easind and dfeasind arguments of routine CPXsolninfo in the Callable Library or the isPrimalFeasible and isDualFeasible metho
25. scriptions in these release notes are summaries full descriptions of the features mentioned will be found in e ILOG CPLEX 6 1 Getting Started e ILOG CPLEX 8 1 User s Manual e ILOG CPLEX 8 1 Reference Manual e ILOG CPLEX 6 1 Java Reference Manual and e ILOG CPLEX 6 1 Advanced Reference Manual All these manuals are provided in the distributed files in various online forms Getting Started and the User s Manual are available in printed form at the version 8 0 level 6 ILOG CPLEX 8 1 RELEASE NOTES RELEASE NOTES FOR CPLEX 8 0 REPRODUCED Release Notes for CPLEX 8 0 REPRODUCED Please review these release notes carefully before using ILOG CPLEX 8 0 They are organized into two sections e Conversion Notes a guide to the upgrade procedure from ILOG CPLEX 7 5 to ILOG CPLEX 8 0 and e New Features an overview of the features introduced in ILOG CPLEX 8 0 If you are upgrading from ILOG CPLEX 7 5 then these notes should be all you need If you are upgrading from a version of CPLEX prior to 7 5 please refer first to the ILOG CPLEX 7 5 Release Notes as well as any prior release notes ILOG CPLEX 7 1 ILOG CPLEX 7 0 and so forth to bring your ILOG CPLEX software up to date from your particular version Conversion Notes for CPLEX 8 0 These Conversion Notes are intended to help you convert from ILOG CPLEX 7 5 to ILOG CPLEX 8 0 as quickly and easily as possible We have divided the notes into four sections The
26. tus codes in ILOG CPLEX 8 0 Note that the mapping depends on the specific algorithm used in ILOG CPLEX 7 5 so be sure to consult the appropriate table Table 1 Conversion of CPXgetstat return value when CPXgetmethod is CPX ALG PRIMAL ILOG CPLEX 7 5 and earlier ILOG CPLEX 8 0 Symbolic Constant Value Symbolic Constant Value CPX OPTIMAL 1 CPX STAT OPTIMAL 1 CPX INFEASIBLE 2 CPX STAT INFEASIBLE 3 CPX UNBOUNDED 3 CPX STAT UNBOUNDED 2 CPX_OBJ_LIM 4 CPX STAT ABORT OBJ LIM 12 CPX IT LIM FEAS 5 CPX STAT ABORT IT LIM 10 CPX IT LIM INFEAS 6 CPX STAT ABORT IT LIM 10 CPX TIME LIM FEAS 7 CPX STAT ABORT TIME LIM 11 CPX TIME LIM INFEAS 8 CPX STAT ABORT TIME LIM 11 CPX NUM BEST FEAS 9 CPX STAT NUM BES 6 CPX NUM BEST INFEAS 10 CPX STAT NUM BES 6 CPX OPTIMAL INFEAS 11 CPX STAT OPTIMAL INFEAS 5 CPX ABORT FEAS 12 CPX STAT ABORT USER 13 ILOG CPLEX 8 1 RELEASE NOTES RELEASE NOTES FOR CPLEX 8 0 Table 1 Conversion of CPXgetstat return value when CPXgetmethod is CPX ALG PRIMAL REPRODUCED
Download Pdf Manuals
Related Search