3 Here we list changes of Ipopt since the release of version 2.2.0.
4 More detailed information about incremental changes can be found in the
5 [commit history](https://github.com/coin-or/Ipopt/commits).
10 ### 3.14.10 (2022-10-11)
12 - Added option `grad_f_constant` to specify that objective function is linear.
13 If set, the gradient of the objective will be requested by Ipopt only once. [#597]
14 - Added `OrigIpoptNLP::orig_d_L()` and `OrigIpoptNLP::orig_d_U()` to get
15 original constraint sides (before relaxation due to bound_relax_factor > 0).
16 - `TNLP::get_curr_violations()` now returns the constraint violation and
17 complementarity w.r.t. the original (non-relaxed) constraint sides. [#603]
19 ### 3.14.9 (2022-07-21)
21 - Fixed mapping of meta data for variable bounds, e.g., variable names,
22 from TNLP to Ipopts internal NLP [#590].
24 ### 3.14.8 (2022-07-13)
26 - Added options ma27_print_level, ma57_print_level, and mumps_print_level
27 to enable output from these linear solvers.
29 ### 3.14.7 (2022-06-24)
31 - Fixed that ComputeSensitivityMatrix() of sIpopt assumed that there are
32 no more than two parameters [#578, by Andrea Vescovini].
33 - For completeness, added option `gradient_approximation` to enable approximation
34 of gradient of objective function by finite differences. Do not use. [#573]
35 - Added function `IPSETPROBLEMSCALING` to Fortran interface to set problem
36 scaling [#577, by Steven R. Hall]
38 ### 3.14.6 (2022-05-02)
40 - Fixed mapping of meta data for inequalities, e.g., constraint names,
41 from TNLP to Ipopts internal NLP [#570, by Daniel Oliveira].
42 - Fixed that MC68 ordering time was not accounted in symbolic factorization
43 time of HSL MA86 [#571].
44 - Include more header files in IpIpoptCalculatedQuantities.hpp for setups
45 where forward declarations are not sufficients [#572].
47 ### 3.14.5 (2022-02-09)
49 - Tried to fix recognition of JNI headers on macOS >= 11 [#516].
50 - Fixed that only primal variable values where passed to `finalize_solution()`
51 when a timelimit was reached [#552].
53 ### 3.14.4 (2021-09-20)
55 - Skip build of Java interface if either java or jar is not found [#510].
56 Only give warning if javac and jar are found, but no java or javadoc.
57 - Fixed that `--with-lapack-lflags` was ignored if `--with-lapack` was not
58 specified explicitly [#512,#515].
60 ### 3.14.3 (2021-09-03)
62 - Fixed timing for iterate initialization if initialization fails due to
64 - Fixed possible integer overflow when reserving space for indices of Jacobian
65 belonging to fixed variables (introduced with 3.14.0) and reduced memory
66 usage for indices of Jacobian belonging to fixed variables.
68 ### 3.14.2 (2021-07-21)
70 - Added `OptionsList::UnsetValue()` to remove an option setting.
71 - Added missing translation of some Ipopt status codes into AMPL solve result codes.
72 - If using the MPI-parallel version of MUMPS: Moved calls to
73 `MPI_Init()`/`MPI_Finalize()` in MUMPS interface into global constructor/destructor
74 of Ipopt library (if building with GCC/clang). Use configure flag
75 `--disable-mpiinit` to disable. [#500]
77 ### 3.14.1 (2021-06-25)
79 - Fixed invalidation of cached Hessians when reoptimizing with same structure.
80 - Added `OptionsList::SetBoolValue()` and `OptionsList::SetBoolValueIfUnset()`. [#492]
81 - Skip check for and link against libdl if linear solver loader is disabled.
82 - Fixed missing initialization of `BacktrackingLineSearch::in_watchdog_`.
83 - Fixed a problem with the current solution not being reset when
84 initialization of a NLP fails in reoptimization.
85 - Fixed that C++11 capability was not correctly identified with MS/Intel compilers.
87 ### 3.14.0 (2021-06-15)
91 - Due to a contribution by Mitchell Clement [#428], it is now possible
92 to build Ipopt in a variant that uses single-precision floating
93 point arithmetic instead of the default double-precision.
94 This can be enabled by specifying the configure flag `--with-precision=single`.
95 Doing so has a number of consequences on Ipopt interfaces and Ipopt dependencies.
96 See the Ipopt installation instructions for more details.
97 - The name of all functions in `IpBlas.hpp` and `IpLapack.hpp` has been
98 changed to drop the precision-specifier "D" in the name.
99 Wrapper using the old names are available, but their use is deprecated.
100 - It is now possible to build Ipopt in a variant that uses 64-bit
101 integers for its `Index` type. This can be enabled by specifying
102 the configure flag `--with-intsize=64`. Doing so has a number of
103 consequences on Ipopt interfaces and Ipopt dependencies. See
104 the Ipopt installation instructions for more details. [#259]
105 - Added new header `IpTypes.h` that defines Ipopt types for integer
106 and real numbers for C users.
107 - Deprecated almost never used type `Ipopt::Int`, use `int` instead.
108 - Deprecated `Number`, `Index`, and `Int` (on global namespace) in
109 `IpStdCInterface.h`, use `ipnumber`, `ipindex`, and `int` instead, respectively.
110 - Deprecated `Bool` in `IpStdCInterface.h` and replaced it by `bool`
111 from `stdbool.h`. Note, that `Bool` was defined to be `int`, but `bool`
112 is likely a shorter type, e.g., `signed char`. Deprecated `TRUE` and
113 `FALSE`, use `true` and `false` instead.
114 - Deprecated `IPOPT_EXPORT` macro and introduced `IPOPT_CALLCONV`.
115 - Deprecated `IPOPT_FORTRAN_INTEGER_TYPE` and `ipfint`. These were
116 always assumed to be `int`. Use `ipindex` or `ipopt::Index` instead.
118 #### Linear Solver Interfaces
120 - Due to a contribution by Byron Tasseff [#446], it is now possible to use
121 the linear solver [SPRAL](https://github.com/ralna/spral) (Sparse Parallel
122 Robust Algorithms Library) with Ipopt.
123 SPRAL is open-source and can, optionally, make use of NVIDIA GPUs.
124 If Ipopt has been build with SPRAL, then option `linear_solver` can be
125 set to `spral` to enable use of SPRAL.
126 See the installation instructions on how to build the Ipopt/SPRAL interface
127 and the options documentation for new options that are available for
128 the Ipopt/SPRAL interface.
129 - Added `IpoptLinearSolvers.h` with function `IpoptGetAvailableLinearSolvers()`
130 to retrieve information which linear solvers are available for Ipopt
131 (linked in or loaded at runtime). Options `linear_solver` and
132 `linear_system_scaling` can now only be set to values which corresponding
133 code is available (linked in or for load at runtime).
134 - Revised and streamlined implementation of feature that loads libraries
135 with HSL or Pardiso routines at runtime. Added options `hsllib` and `pardisolib`
136 to specify name of of HSL and Pardiso libraries, respectively.
137 - Separated interfaces for Pardiso from [pardiso-project.org](https://pardiso-project.org)
138 and Pardiso from Intel MKL and allow to have both Pardiso versions available
139 with the same Ipopt libraries.
140 - Pardiso from pardiso-project.org can be selected via `linear_solver=pardiso`.
141 configure option <b>`--with-pardiso` should now specify only a Pardiso
142 library to be loaded at runtime</b> (the value for `--with-pardiso` decides
143 the default value for option `pardisolib`).
144 To avoid conflicts with Pardiso from MKL, it is **no longer possible to
145 link against Pardiso from pardiso-project.org**.
146 - Pardiso from MKL can be selected via <b>`linear_solver=pardisomkl`</b>.
147 - Options that influence Pardiso from MKL are named <b>`pardisomkl_msglvl`, `pardisomkl_order`, etc.</b>
148 - configure option `--disable-pardisomkl` can be used to disable the check
149 for MKL Pardiso. [#454]
150 - Fixed that return code (`info%flag`) of `ma97_solve()` was not checked
151 in `Ma97SolverInterface::MultiSolve()`.
152 - Calls into MUMPS are now protected by a mutex if compiling for C++11 or higher.
153 This prevents Ipopt from calling MUMPS concurrently from several threads.
154 - An insufficient memory return status is now also given if the memory
155 required for a working space array of a linear solver exceeds the maximal
156 value for the working space length variable, e.g., if MA27 requires a
157 working space array of length higher than 2^31.
158 - Avoid floating point overflow when computing memory increase in interfaces
163 - Changed treatment of NLPs with all variables fixed and `fixed_variable_treatment`
164 set to `make_parameter`:
165 Ipopt will still terminate early, but initialize its data structures
166 first and print a solve summary (49a0d3a56).
167 Changed treatment of NLPs with inconsistent variable bounds or constraint
168 sides: These will now result in an invalid problem definition error (-11)
170 As a consequence, solve statistics should now always be available if the
171 return status from `(Re)Optimize()` is larger than -10.
172 - Bound multipliers are now computed for fixed variables if
173 `fixed_variable_treatment` is set to `make_parameter` (the default).
174 This can trigger a reevaluating of the gradient of the objective
175 function or the Jacobian of the constraint functions. If this
176 is not desired, then option `fixed_variable_treatment` can be set
177 to the new value `make_parameter_nodual`. [#308]
178 - Changed default for `honor_original_bounds` to no.
179 Variable bounds (see option `bound_relax_factor`) are now relaxed
180 by at most the value of `constr_viol_tol`.
181 The solve summary now reports the violation of original bounds
182 and `SolveStatistics::(Scaled)Infeasibilities()` has been extended
183 to report the violation of original variable bounds. [#312]
184 - If Ipopt hits a time or iteration limit during watchdog phase,
185 the iterate from before the watchdog phase is now restored and
186 passed to `finalize_solution`. Note that this does not apply if
187 Ipopt is stopped due to a user-interrupt (`intermediate_callback`).
189 - When `BacktrackingLinesearch` could not find a trial point that
190 provided sufficient progress, it may resort to call the feasibility
191 restoration phase. If, however, the current (scaled) constraint
192 violation was below tol/100, this would result in the infamous
193 "Restoration phase is called at almost feasible point..." abort
194 with status code `Restoration_Failure`. Now, this message has been
195 changed to "Linesearch failed, but no restoration phase or other
196 fall back is available." and the status code to `Error_In_Step_Computation`
197 to better reflect that the linesearch failed and not the restoration
198 phase. Further, the unscaled constraint violation tolerance now
199 needs to be below `constr_viol_tol/10` as well in order to trigger
201 - When a square problem is solved and the restoration phase only succeeded to
202 find a point that is feasible w.r.t. constr_viol_tol, but not w.r.t. tol,
203 then status Feasible_Point_Found is returned now.
205 #### Ipopt interfaces
207 - Added `TNLP::get_curr_iterate()` and `TNLP::get_curr_violations()`
208 to request the current iterate (primal and dual variable values)
209 and primal and dual infeasibility w.r.t. the TNLP. The methods
210 are meant to be called during `intermediate_callback` to inspect
211 the current iterate. The C, Fortran, and Java interfaces have
212 been extended by corresponding functions, too. The hs071 examples
213 have been extended to show use of the new functions. Added test
214 `getcurr` to test new functions. [#382, #451]
215 - Extended the Java interface by the possibility to specify an `intermediate_callback`.
216 - Added flag to `TNLPAdapter::ResortX()` to specify how to handle fixed variables.
217 Added flag to `TNLPAdapter::ResortG()` to specify whether to correct by
218 right-hand-side of equality constraints.
219 Added `TNLPAdapter::ResortBoundMultipliers()` to generate correct duals
220 for fixed variables if `fixed_variable_treatment` is `make_constraint`.
221 Added `TNLPAdapter::GetFullDimensions()`, `TNLPAdapter::GetFixedVariables()`,
222 `TNLPAdapter::GetPermutationMatrices()`, and `TNLPAdapter::GetC_Rhs()`
223 to retrieve more information of the transformation in a `TNLPAdapter`.
224 - Added `TNLPAdapter::ResortBounds()` and deprecated `TNLPAdapter::ResortBnds()`.
225 - `AmplTNLP` constructor and `AmplTNLP::get_options()` now expect to
226 receive a pointer to a `RegisteredOptions` object as well.
227 Previous versions of these methods are still available but deprecated.
228 Using a NULL pointer for the `RegisteredOptions` argument is
230 - `std::overflow_error` exceptions are now caught by Ipopt even if rethrowing
231 of non-Ipopt exceptions is enabled.
235 - Added option `max_wall_time` to specify a wallclock time limit.
236 Added `SolverReturn` code `WALLTIME_EXCEEDED` and `ApplicationReturnStatus`
237 code `Maximum_WallTime_Exceeded`.
238 - Changed default for `max_cpu_time` to 1e20 to indicate no CPU timelimit.
239 - Detailed timing statistics are no longer collected by default.
240 To reenable, set the new option `timing_statistics` to yes or set the
241 option `print_timing_statistics` to yes [#299].
242 - Removed `IpoptData::ResetCpuStartTime()`. Deprecated `IpoptData::cpu_time_start()`,
243 use `IpoptData::TimingStats()::OverallAlgorithm()::StartCpuTime()` instead.
244 - Deprecated `SolveStatistics::TotalCPUTime()`, use `SolveStatistics::TotalCpuTime()`
249 - All Ipopt options are now also available via the AMPL interface,
250 that is, can be set in AMPL via an option `ipopt_options "..."` statement.
251 - Added flag "advanced" for `RegisteredOption` to store whether an
252 option is rather meant for expert users.
253 - Added class `RegisteredCategory` to store information on category
254 of registered option. Next to the name, this is a priority, which
255 determines the order in which categories are printed. As a
256 consequence, methods `RegisteredOption::RegisteringCategory()` and
257 `RegisteredOptions::RegisteringCategory()` now return a
258 `RegisteredCategory` object instead of a string. Further,
259 `RegisteredOption::SetRegisteringCategory()` has been removed.
260 - Deprecated existing `RegisteredOptions::OutputXyz()` methods and
261 added new variant of `RegisteredOptions::OutputOptionDocumentation()`.
262 Added parameter `print_advanced_options` to indicate whether to
263 print documentation for advanced options, too.
264 - Added previously undocumented advanced options to options reference.
268 - Updated buildsystem, including improved checks for dependencies,
269 use of current autotools, and skipping the build of intermediate
270 non-distributed libraries.
271 - Fixed missing initialization of `IpoptApplication::smart_jnlst` in
272 second `IpoptApplication` constructor, if `ipopt verbosity > 0`.
273 - Add GCC format attribute to `Journalist::Printf` functions to enable
274 `printf`-formatter check.
275 - Fixed `DenseVector::SumLogsImpl()` such that it returns 0.0 for
276 a vector of dimension 0. Returned `nan` for homogeneous 0-vector
277 of dimension 0 before, which may have caused the restoration phase to
278 fail for problems with only equality or only inequality constraints.
279 Also other `DenseVector` methods now skip calculations when dimension
280 is 0 to avoid (probably harmless) divisions by zero.
281 - Fixed a problem where moving slack away from 0 did not succeed
282 when mu was very small. [#212]
283 - Fixed a problem where moving slacks away from 0 resulted in `nan`
284 if multipliers were zero. Added `Vector::ElementWiseSelect()`.
285 - Various tiny bugfixes and improvements in performance and code style
286 by following suggestions of `cppcheck`.
287 - Added documentation on some available C preprocessor flags for expert users.
288 - Fixed static build of sIpopt without GCC. Fixed that installed sIpopt
289 headers were not usable (`SIPOPTLIB_EXPORT` not defined).
290 - Fixed wrong gradient of objective function and Lagrangian Hessian in
291 restoration problem [#478, by Nai-Yuan Chiang].
292 - If Ipopt is compiled for checklevel 2 or higher and the GLIBC extension `feenableexcept()`
293 is available, then floating-pointing exceptions divbyzero, overflow, and
294 invalid are raised while `IpoptAlgorithm::Optimize()` is running.
295 - Fixed that norm on unscaled complementarity or scaled complementarity tolerance
296 were negative when maximizing (by using a negative scaling factor for the
298 - Changed formula for relative error in derivative checker. The absolute
299 error is now scaled up if the approximate derivative value is between
300 derivative_test_tol and 1. [#487].
301 - The second-order derivative checker now uses values for obj_factor and
302 lambda that are different from 1.
307 ### 3.13.4 (2021-02-24)
309 - Fixed a linking issue for `ipopt_sens` [#418]
310 - Fixed Makefile for Java example regarding location of jar file
311 - Fixed build of R interface if using `-fvisibility=hidden`.
313 ### 3.13.3 (2020-10-16)
315 - Members of `AmplTNLP` class are now protected instead of private.
316 - Updated Eclipse Public License from 1.0 to 2.0.
317 - Fixed dangling pointer problems with Journalist used for debugging
318 (`--with-ipopt-verbosity > 0`) when more than one `IpoptApplication`
319 is used. [#393, thanks to Brad Bell]
320 - Fixed build problem when using HSL library that does not include
321 MA27, MA57, or MC19. [#395]
322 - Added example `recursive_nlp` that uses Ipopt to solves an
323 optimization problem for the evaluation of the objective function.
324 [contributed by Brad Bell]
325 - Fixed build of linear-solver loader on Windows [#408]
327 ### 3.13.2 (2020-04-30)
329 - The C-preprocessor defines `COIN_IPOPT_CHECKLEVEL`,
330 `COIN_IPOPT_VERBOSITY`, and `FORTRAN_INTEGER_TYPE`, which are defined
331 by `IpoptConfig.h`, have been renamed to `IPOPT_CHECKLEVEL`,
332 `IPOPT_VERBOSITY`, and `IPOPT_FORTRAN_INTEGER_TYPE`, respectively.
333 They are still available under their previous name, but these
334 will be removed in Ipopt 3.14.
335 - Changed dependencies as used by coinbrew to use new versions (2.1)
336 of ThirdParty/HSL and ThirdParty/MUMPS and dropped ThirdParty/Metis.
337 The new versions of the HSL and MUMPS build scripts now look
338 for a Metis library in the system and should work with both
340 - Changed location where Java interface jar gets installed from
341 `$libdir` to `$datadir/java/`.
342 - minor fixes to buildsystem
344 ### 3.13.1 (2020-03-11)
346 - Added asserts that check whether sparsity pattern of Jacobian
347 and Hessian as returned by TNLP are within range w.r.t. number
348 of variables and constraints. [#350]
349 - `TNLPAdapter::ResortBnds` now initializes complete output arrays
350 with 0.0 before filling in values for non-fixed variables. Use
351 new argument `clearorig` to turn this off. [#352]
352 - bring back configure variables `ADD_{C,CXX,F}FLAGS`
353 - added configure option `--enable-relocatable` to make prefix in
354 pkg-config files relative to pcfiledir (assuming that `--libdir`
356 - bring back `configall_system.h` for build without config header
357 - minor fixes to buildsystem
359 ### 3.13.0 (2019-10-19)
361 This major release comes with a larger renovation of the build
362 system and a changed directory structure (eliminated top directory),
363 which is the result of a long and still on-going effort to use
364 recent autotools versions for various COIN-OR projects, reduce
365 future maintenance efforts, and adapting behaviors of standard
366 autotools-based projects.
367 As a consequence, a monolithic build of Ipopt, which builds Ipopt
368 with all its dependencies in one run of configure and make is no
369 longer possible. Dependencies should now be build and installed
370 before building Ipopt.
371 Additionally, support for some outdated versions of dependencies
372 and unmaintained components of Ipopt has been dropped and some
373 improvements that may require changes on the users side have been
376 A more detailed, probably incomplete, list of changes follows:
377 - Removed git submodules. Dependencies (HSL, Mumps, ASL, etc) now
378 need to be build and installed in advance, either manually or
380 - Dropped support for HSL < 2013.
381 - Dropped support for MA28 in the linear solver loader.
382 - Dropped support for Pardiso < 4.0 from pardiso-project.org.
383 - Added support for Mumps 5.2.x, though initial experiments on
384 CUTEst indicated that, on average, performance is worse than
385 when using Mumps 4.10.0.
386 - Dropped CUTEr interface, the successor CUTEst includes an
388 - Dropped Matlab interface as it is unmaintained and it was
389 reported that it stopped functioning.
390 Use https://github.com/ebertolazzi/mexIPOPT instead.
391 - Dropped MSVS project files as unmaintained and not functioning
392 with current Ipopt anymore.
393 - Integrated Java interface into the main Ipopt library, that is,
394 it is handled equivalently to the C and Fortran interfaces:
395 - The source moved into `src/Interfaces`.
396 - The JNI functions are now included in the main Ipopt library,
397 thus an extra jipopt library is no longer build or necessary.
398 - The Java class and `org.coinor.ipopt.jar` package are build and
399 installed as part of the main Ipopt build.
400 - The examples moved into `examples/*_java`.
401 - A Java interface test is executed by `make test`.
402 - To build javadoc, run `make javadoc` in the main build directory.
403 - The configure flag `--disable-java` can be used to disable the
404 check for Java and build of the Java interface.
405 - `DLLPATH` and `DLLNAME` have been removed from the Ipopt class and
406 constructors that works without arguments and with only one
407 argument (specifying the Ipopt library namestem) have been added.
408 - Method `Ipopt::finalize` has been marked as deprecated and will
409 be removed in some future Ipopt version. Users must call
410 `dispose()` explicitly.
411 - Integrated sIpopt into the main Ipopt build, that is, it is now
412 build together with Ipopt, but installed as separate library
413 and executable. Use `--disable-sipopt` to disable building sIpopt.
414 - `IPOPT_THREAD_LOCAL` now uses C++11's `thread_local` keyword if C++11
416 - When using a GCC-compatible compiler, Ipopt and sIpopt interface
417 functions are now declared with `visibility(default)`-attribute,
418 thus building Ipopt with `-fvisibility=hidden` still produces a
420 - When using a MSVC-compatible compiler, Ipopt and sIpopt interface
421 functions are now declared with `dllimport`-attribute, so that an
422 Ipopt C++ DLL can be used.
423 - Under Windows/Msys2, DLLs are now build by default.
424 - Cygwin and MSys1 are not supported.
425 - pkg-config is now mandatory to use dependencies like ASL or HSL.
426 On Windows, make sure to use a pkg-config version that produces
428 - Script "`compile`" is now used to wrap around calls of cl/icl/ifort
429 and translate GCC-style compiler flags to MSVC style.
430 - "Addlibs" files have been removed, pkg-config should be used instead.
431 - Header files are now installed in the better named
432 `$prefix/include/coin-or` instead of `$prefix/include/coin`.
433 - The default for `--prefix` is no longer the build directory, but
434 the autotools-default, probably `/usr/local`.
435 - The check for a Fortran compiler can be disabled via `--disable-f77`
436 and Ipopt can easier be build without a Fortran compiler.
437 - Lapack is no longer optional, but required. The separate check
438 for Blas and the `--with-blas` flags have been removed.
439 - `--enable-debug` does not imply `--disable-shared` anymore.
440 - Removed `--enable-debug-ipopt`, use `--enable-debug` instead.
441 - Removed configure variables `{ADD,OPT,DBG}_{C,CXX,F77}FLAGS`.
442 Use `{C,CXX,F77}FLAGS` instead.
443 - Silent build output is now enabled by default, use configure
444 flag `--disable-silent-rules` or call make with `V=1` to disable.
445 - Also for static builds, PIC objects are now generated by default,
446 use `--without-pic` to disable.
447 - The `--with-*-incdir` and `--with-*-lib` configure flags have been
448 replaced by corresponding `--with-*-cflags` and `--with-*-lflags`
449 flags. Note that the include directories need to be specified
450 via `-I<dir>` in `--with-*-cflags`.
451 - Fixed handling of `ma77_default_control` in `LSL_setMA77()`.
452 - Fixed calculation of quality function when setting option
453 `quality_function_centrality` to `reciprocal`.
454 - Fixed compiler warnings, in particular when using `-Wunused-parameter`.
455 - Changed default for `ma97_print_level` to -1. This avoids messages
456 about numerical singular systems written to stdout by default.
460 ### 3.12.13 (2019-04-08)
462 - fixed Pardiso settings when using Pardiso from Pardiso project
463 website (by Olaf Schenk): the new settings should provide much
464 better performance; the default for option `pardiso_order` changed
465 from `five` to `metis`.
466 - changed distinction of MKL and Basel Pardiso in configure: to
467 use MKL Pardiso, only specify MKL for Blas; to use Basel Pardiso,
470 ### 3.12.12 (2018-11-17)
472 - allow for `--without-matlab-home` to disable check for Matlab [r2748]
473 - add `dppsv` to `v8-ifort` [r2746]
474 - disable error in `LibraryHandler.c` if `snprintf` detection failed [r2751]
476 ### 3.12.11 (2018-09-16)
478 - fill MUMPS struct with zeros when allocating in MUMPS interface [r2724]
479 - minor fix in build-system of ThirdParty/ASL
481 ### 3.12.10 (2018-06-02)
483 - fixed setting for parallel solve when using MKL Pardiso
484 (by t1393988511) [r2711]: parallel solve was disabled (which
485 is not the default); note, that the setting for parallel
486 factorization was not affected
487 - fixed invalid read in AMPL interface for problems without
488 objective function [r2715, #305]
489 - updated ThirdParty/ASL to retrieve updated ASL (20180528) [#305]
490 - name JIpopt library `libjipopt.dylib` on Mac OS X [r2718, #275]
492 ### 3.12.9 (2018-01-15)
494 - fixed memory leak in MA86 interface (by mhahn) [r2700,#283]
495 - fixed handling of time limit when reoptimizing: CPU time spend
496 was accumulated when reoptimizing, while it should have been
497 reset for each solve (by paul-scott) [r2702,r2703]
498 - fixed sign in Jacobian finite-difference approximation when point
499 was close to variable upper bounds (by Enrico Bertolazzi) [r2704]
501 ### 3.12.8 (2017-06-12)
503 - add define for `FORTRAN_INTEGER_TYPE` to `config_ipopt_default.h`
504 - `IpoptApplication::RethrowNonIpoptException()` now returns whether
505 non-ipopt exceptions were rethrown before the method was called.
507 ### 3.12.7 (2017-02-25)
509 - removed compiler flag `-pedantic-errors` to avoid problems with some
510 configure tests when using recent GCC versions
511 - fixed rare bug in handling variable/constraint names in `AmplTNLP`
512 (by G. Hackebeil) [r2673]
513 - the `get.Mumps` script in ThirdParty/Mumps now renames `libseq/mpi.h`
514 to `libseq/mumps_mpi.h` to avoid conflicts when building in a MPI
515 environment (by T. Ralphs); note that if updating an existing
516 checkout/download of Ipopt, you may have to rerun get.Mumps
518 ### 3.12.6 (2016-07-20)
520 - better support for custom algorithm development [r2659] (by Gabriel Hackebeil):
521 "Reorganization of the `AlgorithmBuilder` class to allow easier
522 customization of the Ipopt algorithm. In particular, we wanted to
523 make use of the code that creates the `SymLinearSolver` object to
524 implement our own `SymLinearSolver` without copy-pasting everything
526 `AlgorithmBuilder::BuildBasicAlgorithm` now consists of 8 method calls
527 that build the core components passed into the arguments of the
528 `IpoptAlgorithm` class. These calls are ordered based on any dependencies
529 they might have. In addition, all code for creating the `PDSystemSolver`,
530 `AugSystemSolver`, and `SymLinearSolver` has been moved into separate factory
532 Also, included is a change to install a few more header files with Ipopt.
533 Some of these are required to subclass AlgorithmBuilder, and the others
534 are simply some matrix types that we require."
535 - extend build system to work without Fortran compiler [r2660,r2661]:
536 If no Fortran compiler is available (`F77=unavailable`), then
537 the build system checks for functions in Blas, Lapack, and
538 Pardiso via C linkage. This seems to work when using the Intel MKL,
539 thus allowing to build Ipopt with C/C++ compilers and MKL only.
540 The linear solver loader and the CuteR interface are disabled when
541 no Fortran compiler is available. A user may have to adjust the
542 definition of `F77_FUNC` in `Ipopt/src/Common/IpoptConfig.h`.
544 ### 3.12.5 (2016-04-30)
546 - changed `fptr` from `long` to `void*`: the Fortran side needs to
547 make sure that it uses a big enough integer type to store a
548 C pointer, thus `void*` can be used on the C side [r2599]
549 - added additional second-order-correction method, which can be
550 selected by setting the new option `soc_method` to 1 (by Wei Wan)
552 - added parameter `allow_clobber` with default value false to
553 `IpoptApplication::Initialize()` and `OptionsList::ReadFromStream()`
555 ### 3.12.4 (2015-08-09)
557 - option to use regularized Hessian when doing a curvature test
558 without inertia information (`neg_curv_test_tol` > 0), new
559 option `neg_curv_test_reg` to switch back to original behavior
560 (by N.-Y. Chiang and V. Zavala Tejeda) [r2579]
561 - sIpopt: Added access to sensitivity directional derivative
562 vector (`ds/dp*(p-p0)` Eq. 14 sIpopt implementation paper). Also,
563 added an option to compute the sensitivity matrix and provide
564 access to it. Finally, added an example that shows how to
565 access the new information. (by R. Lopez-Negrete)
566 - use workaround for failing check for random number generator
567 with any gcc 4.8.x, x >= 2
569 ### 3.12.3 and 3.11.11 (2015-04-15)
571 - fixed bug in MA97 interface that lead to conversion issues
572 (by J. Hogg) [r2566, #260]
574 ### 3.12.2 (2015-04-04)
576 - revised integration of doxygen-generated documentation into build system
579 ### 3.12.1 (2015-02-13)
581 - fixes to build system for dependency linking and library versioning
582 - Ipopt will now report an NLP with inconsistent variable bounds
583 or inconsistent constraints sides as infeasible instead of
584 throwing an invalid TNLP exception (by T. Kelman) [r2548]
586 ### 3.12.0 (2015-01-23)
588 - Library dependencies are now recorded in shared library builds,
589 which is intended to simplify linking against the Ipopt library.
590 However, the pkg-config and `ipopt_addlibs` files do not reflect
591 this change yet (it is rather experimental, imho). To restore
592 the previous behavior, use `--disable-dependency-linking` as
594 - If linking against Intel MKL for Blas/lapack, use of Pardiso
595 from MKL is now automatically enabled. Note, that this will
596 change the default solver on Ipopt builds without any of the
597 linear solvers MA27, MA57, MA97, and MA86 (these take preference
598 over Pardiso). [#216]
599 - dropped support for old HSL sources (<2013; ThirdParty/HSLold)
600 - updated ASL sources, now downloaded from AMPL-MP (github)
601 - some internal changes to data structures (improved use of compound
602 component spaces) and addition of `IpLapackDppsv` (by Gabe Hackebeil)
606 ### 3.11.10 (2015-01-18)
608 - fix a memory allocation in Java interface in cases where `jint`
609 has a different size than `int` [r2513]
610 - the buildsystem now tries the `Accelerate` framework instead of
611 `vecLib` for finding Blas/Lapack on MacOS X
613 ### 3.11.9 (2014-08-16)
615 - fix compilation issue of Java interface on systems where `Index`
616 and `jint` have different size [r2498, #241]
617 - work around failing check for random number generator with gcc
619 - readded `IpTaggedObject.cpp` to list of sources to compile in
620 MSVS `v8-ifort` project file [r2492]
621 - work around missing support for thread-local storage with gcc < 4.5
622 on MacOS X [r2491, #243]
623 - fix call to MKL Pardiso init function [r2489]
624 - speed up Triplet to CSR converter [r2487, #234]
625 - fixed a bug in equilibration scaling where average values were
626 computed incorrectly (by V. Zverovich) [r2483]
628 ### 3.11.8 (2014-04-08)
630 - fixed a bug, introduced with Ipopt 3.11.0, where the tag in the
631 Ipopt's caching mechanism was not unique over time, which lead
632 to failures of Ipopt that were difficult to debug or recognize
633 (e.g., Ipopt may have stopped with an restoration failure for
634 instances that solved fine with Ipopt 3.10) [r2472, r2473]
635 I'm very thankful to Gabriel Hackebeil and Kurt Majewski for
636 their debugging effort on this issue.
637 - building Mumps with pthreads is now disabled by default [#229]
638 - fixed setting of `LD` on Windows (now set to link only iff using
639 MS/Intel compilers) [#230]
640 - fixed download link for Gnumex [r2471]
641 - for some messages about too-few-degrees-of-freedom and restoration
642 failure, the message level changed from error to strong-warning
644 - revised calls to `MPI_Init` and `MPI_Finalize` in MUMPS interface [r2467]
645 (`MPI_Init` is now called only if function `MPI_Initialized` is available
646 and MPI has not been initialized already; `MPI_Finalize` is only called
647 if Ipopt also called `MPI_Init`; ...)
649 ### 3.11.7 (2013-12-18)
651 - adapted PARDISO parameters when using MKL PARDISO to be close
652 to using Basel PARDISO
653 - added options `pardiso_max_iterative_refinement_steps` and
654 `pardiso_order`; the former defaults to 1 in case of MKL PARDISO,
655 which may help on instances that otherwise fail due to numerical issues
656 - removed duplicate code in `IpQualityFunctionMuOracle.cpp` [#225, r2445]
657 - fixed bug in triplet to csr converter [#226, r2446]
658 - minor changes in buildsystem
660 ### 3.11.6 (2013-11-16)
662 - updates to Matlab Interface build system (by T. Kelman)
663 - fix to updates of R Interface [r2416, #223]
664 - fixed `SHAREDLIBEXT` in `v8-ifort`'s `config.h` [r2426, #224]
665 - minor fixes to the buildsystem
667 ### 3.11.5 (2013-10-26)
669 - added method `IpoptApplication::RethrowNonIpoptException()` to enable
670 rethrowing of non-ipopt and non-`bad_alloc` exceptions catched in
671 the `*Optimize()` and `Initialization()` methods; default is still to
672 return with `NonIpopt_Exception_Thrown` status
673 - minor fixes to the buildsystem [#215, #222]
675 ### 3.11.4 (2013-09-12)
677 - hopefully fixed non-working linear solver loader in DLLs build with
678 MSVS/`v8-ifort` project files [r2365]
679 - allow MC19 to be loaded via linear solver loader (by J. Currie) [r2366]
680 - fixed new point flag when running dependency detector [r2368]
681 - experimental: adapt Pardiso interface to work with MKL Pardiso
682 (by J. Currie, T. Kelman) [r2369, #216]:
683 - in a few tests it has been found that Pardiso from Intel MKL nowadays
684 seems to work fine with Ipopt
685 - to use Intel MKL with Ipopt 3.11, one has to specify the MKL libs via
686 `--with-pardiso` and add `-DHAVE_PARDISO_MKL -DHAVE_PARDISO_PARALLEL`
687 to the compiler flags
688 - note that this is still an experimental feature (and thus not enabled
690 - updated Ipopt/R interface to version 0.8.4 [r2373]
691 - additional variables have been included in the object returned from `ipoptr`:
692 - `z_L`: final values for the lower bound multipliers
693 - `z_U`: final values for the upper bound multipliers
694 - `constraints`: final values for the constraints
695 - `lambda`: final values for the Lagrange multipliers
696 - removed `ipoptr_environment` as argument in `ipoptr` (see also r2372)
697 - fixed bug in penalty term for centrality in quality function (not used by
698 default) [#219, r2374]
699 - minor bugfixes in AMPL interface, debug print statements, and compound matrix
700 (by G. Hackebeil) [#218, r2371, r2377, r2378, r2379]
701 - download scripts for ASL, Blas, and Lapack now first try to download tarball
702 copies from the COIN-OR server
704 ### 3.11.3 (2013-08-08)
706 - `get.*` scripts for ThirdParty/{ASL,Blas,Lapack} now work around broken
707 ftp access to www.netlib.org.
709 ### 3.11.2 (2013-07-01)
711 - changed default for option `option_file_name` to `ipopt.opt`; specifying an
712 empty string for this option now disables reading of an option file [r2339]
713 - missing initial values are now set to 0.0, projected onto variable bounds,
714 in AMPL interface [r2340, #205]
715 - fixed missing variable initialization in MA97 interface [r2341, #206]
717 ### 3.11.1 (2013-06-14)
719 - the setup for the `v8-ifort` MSVS project changed to use dynamic runtime
720 DLLs instead of static linking, which caused crashes in debug mode
721 (by M. Roelofs) [r2301]
722 - fixed memory leaks in Java Interface (by javier) [#200, r2312]
723 - updates and fixes to MA77 and MA87 interfaces, adding support of
724 HSL 2013 codes (by J. Hogg);
725 HSL 2012 still supported when compiled with Ipopt, but the linear solver
726 loader to dynamically load a HSL library at runtime now assumes HSL 2013
727 - added option `ma97_solve_blas3` (by J. Hogg) [r2329]
728 - changed default for option `ma27_meminc_factor` from 10.0 to 2.0 [r2330]
729 - fixed bug in `ipopt_auxdata` of MATLAB Interface related to `iterfunc` [r2325]
731 ### 3.11.0 (2013-05-07)
735 - update and extension of Ipopt documentation
736 - updated build of doxygen-generated documentation to comply with other
738 - localized global variables in `TaggedObject` and `RegisteredOption`,
739 so that Ipopt should now be threadsafe as long as Ipopt objects
740 (esp. `SmartPtr`'s) are not shared between threads and a threadsafe
741 linear solver is used (e.g., MA27) [#167]
742 - no more need for whitespace character at end of options file
743 - added options `print_frequency_iter` and `print_frequency_time` to regulate
744 which iteration summary lines should be printed [#161]
745 - function evaluation timings are now available in `OrigIpoptNLP` [#86]
746 - some fixes to uncommon issues with the Ipopt `SmartPtr` [#162]
748 #### Linear Solver Interfaces
750 - new build system for Harwell codes (ThirdParty/HSL), which requires
751 the coin-hsl archives from http://www.hsl.rl.ac.uk/ipopt/;
752 previously downloaded HSL sources can still be used by placing them
753 into ThirdParty/HSLold, but this option will be removed in a future
755 - new interfaces for Harwell codes HSL_MA77, HSL_MA86, and HSL_MA97;
756 see http://www.hsl.rl.ac.uk/ipopt/ about help on when to use which solver;
757 especially MA57 and HSL_MA97 should be considered as replacement for MA27;
758 however, MA27 is still the default for now
759 - changed default of `ma57_automatic_scaling` to `no` (faster in general,
760 but for higher reliability, you may want to set this option to yes)
762 #### Ipopt Interfaces
764 - major improvements for building the MATLAB interface (see documentation)
765 - MATLAB interface now returns number of function evaluations, too
766 - the MA57 interface can now be used with the MA57 library that comes with
767 MATLAB (configure option `--enable-matlab-ma57`; cannot use Metis)
768 - `auxdata` is now handled by a wrapper function `ipopt_auxdata.m` instead
769 of internally within the MEX code (replace Matlab call to `ipopt` with
770 `ipopt_auxdata` if using auxiliary data in any callbacks) [r2282]
771 - exposed more intermediate Ipopt information to `iterfunc` callback [r2283]
773 - fixes to JIpopt buildsystem (now may work on windows and uses libtool)
774 - JIpopt now reads options file `ipopt.opt` by default, if present
775 - removed predefined `KEY_` strings in JIpopt
776 - renamed API functions that retrieve solution values in JIpopt
778 - simplified installation of R interface
782 ### 3.10.4 (2013-05-05)
784 - fixed sign of dual values in AMPL solution again (with help of Gabe)
785 [r2169, r2170, r2184, #183]
786 - fixed bugs with reoptimizing a TNLP with all variables fixed [r2172, r2173, #179]
787 - fixed more issues with sparse data structures and non-double numbers
788 in Matlab interface (by T. Kelman) [r2191]
789 - added missing `final int` for Ipopt return code `Maximum_CpuTime_Exceeded`
790 in Java interface [r2216]
791 - fixed bug when trying to warmstart Ipopt in Java interface [r2253]
792 - fixed wrong use of `SmartPtr`'s in Java interface [r2255, r2263]
793 - fixed bug in returning final solution in Java interface [r2258]
794 - included patch in ThirdParty/Mumps to work around bugs in Mumps
795 matrix ordering routines AMF and QAMD (now give preference to AMD and METIS)
797 ### 3.10.3 (2012-11-19)
799 - minor fixes in MA86 interface (by Jonathan Hogg) [r2069, r2086]
800 - fix in `IpTripletToCSRConverter` for CSR forms with
801 extra entries (by Jonathan Hogg) [r2087]
802 - workaround problems with Mac OS X Lion's blas library
803 (by Frederic Hetch) [r2102, #181]
804 - the C interface now catches also Ipopt exceptions thrown
805 within the `OptimizeTNLP` call and returns `Ipopt::Unrecoverable_Exception`
806 as status [r2094, #144]
807 - fixed segmentation fault in adaptive barrier parameter update rule
808 when using the mehrotra option on unconstrained problems [r2114, #114]
809 - fixed segmentation fault in case no iterate is available in case of
810 catastrophic failure in restoration phase [r2115]
811 - fixed default for `mumps_dep_tol` to work with current Mumps versions [r2116]
812 - fixed sign of dual values in AMPL solution [r2123, #183]
813 - fixed issue with sparse gradients in Matlab interface
814 (by T. Kelman) [r2133, #187]
815 - sIPOPT (by H. Pirnay):
816 - starting values in C++ version of parametric example now
817 match AMPL version [r2098]
818 - numerical values in parametric example now match publication [r2099]
819 - added options `n_sens_steps` and `sens_boundcheck` as AMPL options [r2099]
820 - any non-zero suffix value is now accepted for `sens_init_constr` [r2100]
821 - fix typo in AMPL interface (by Weifeng Chen) [r2110]
822 - fix bug when compiling without AMPL interface [r2113]
824 - updated instruction on using nowadays HSL sources (by T. Kelman)
825 - fixed issue with libdir being `<prefix>/lib64`
828 ### 3.10.2 (2012-02-12)
830 - updates to HSL interface (by Jonathan Hogg):
831 - MC68 can now be loaded dynamically, too
832 - MA86 exploits built-in scaling
833 - MA86 can choose best ordering from AMD and Metis
834 - fix for return code of MA86 for singular matrices
835 - corrected computation of Upsilon (norm of step SQUARED)
836 - updates to MSVS `v8-ifort` project files and addition of vc10
837 project files (using vc8-generated `IpoptFSS.dll`) (by Marcel Roelofs)
838 - minor bugfixes, include updates in BuildTools
840 ### 3.10.1 (2011-09-20)
842 - include updates in BuildTools, including new ThirdParty/Metis
843 (fix for URL to download Metis 4.0.3 release)
844 - linear solver loader prints error code when failing to load
845 library under Windows
846 - message on failure when reallocating memory in Mumps now includes
847 size of memory that was tried to be allocated
848 - added missing include of `cstdio/stdio.h` in `IpJournalist.hpp`
849 - minor fixes to build system
851 ### 3.10.0 (2011-06-20)
853 - move to new COIN-OR configuration and installation convention
854 - primal infeasibility output is now true infeasibility in original
859 ### 3.9.3 (2011-04-07)
861 - include updates in BuildTools, including new ThirdParty/Metis
862 (required to work with current metis release)
864 ### 3.9.2 (2010-12-22)
866 - converted from Common Public License to Eclipse Public License
867 - some bugfixes from BuildTools
869 ### 3.9.1 (2010-11-26)
871 - improved Hessian update for restoration phase
872 - added intermediate callback feature to C and Fortran interface
874 ### 3.9.0 (2010-11-05)
876 - switching to new BuildTools system
877 - added R interface (contributed by Jelmer Ypma)
878 - updates in AsNMPC (by Hans Pirnay)
882 ### 3.8.3 (2010-06-29)
884 - restated `SolveStatistics::TotalCPUTime` method for backward
887 ### 3.8.2 (2010-06-16)
889 - uses MUMPS version 4.9 and Lapack version 3.2.1
890 - added AsNMPC contribution made by Hans Pirnay
891 - enhanced MA57 options
892 - several bug fixes and minor additions
894 ### 3.8.1 (2009-10-30)
896 - Bugfix in NLP function evaluation timing measurement. The
897 time for the objective function gradient had been forgotten.
899 ### 3.8.0 (2009-10-30)
901 - Added MSVC solution with Intel Fortran compiler to generate DLLs
902 (contributed by Marcel Roelofs). To make this work, a lot of methods
903 in exported headers have been made virtual
904 - changed default convergence tolerance in restoration phase (now same
905 as regular tolerance)
906 - output is flushed after each iteration
910 ### 3.7.1 (2009-10-06)
912 - bugfix for square problems
913 - correct timing information (obj gradient was forgotten)
914 - flush output buffer after each iteration
915 - first code for iterative WSMP version (experimental and undocumented)
917 ### 3.7.0 (2009-07-16)
919 - a number of fixes (including those from 2009 COIN-OR Bug Squashing Party)
920 - changes in some exposed header files to provide access to internal
921 data structures for specific applications
925 ### 3.6.1 (2009-05-01)
927 - minor corrections in tutorial files
929 ### 3.6.0 (2009-04-29)
931 - new Matlab interface
932 - added new option to limit cpu time: `max_cpu_time`
933 - added ThirdParty directory for Metis to be used with MUMPS or MA57
934 - updated CUTEr Makefile to make it work with CUTEr2
935 - added files for a tutorial (including coding exercise)
939 ### 3.5.5 (2009-01-13)
941 - minor fixes regarding compilation
942 - undocumented version of inexact method
944 ### 3.5.4 (2008-09-29)
946 - changed to MUMPS version 4.8.3 in externals (Mumps developers
947 seem to have removed 4.8.1).
949 ### 3.5.3 (2008-09-19)
951 - changed back to MUMPS version 4.8.1 since there seem to be issues
954 ### 3.5.2 (2008-09-18)
956 - changed to latest version of MUMPS (4.8.2)
957 - some bugfixes (linear algebra objects, automatic problem scaling)
958 - made sure the subversion revision number is correct in all files
959 - allowed general additional data and cq in `IpData` and `IpCq`
961 ### 3.5.1 (2008-08-26)
963 - changed to latest version of MUMPS (4.8.1)
965 ### 3.5.0 (2008-08-25)
967 - added `ComputeRowAMax` and `ComputeColAMax` methods to Matrix base class
968 - changed externals for MUMPS to stable/1.1
969 - call `finalize_solution` in more failure cases
970 (this means that AMPL writes .sol file in more situations)
971 - added `IpTNLPReducer` as simple way to exclude constraints from problem
972 - several fixes, also from COIN-OR Bug Squashing Party 2008
976 ### 3.4.2 (2008-07-18)
979 - added wallclock time routine
980 - penalty function version does no longer crash if it
981 wants to go to restoration phase (not that this really helps
984 ### 3.4.1 (2008-05-30)
987 - deleted `v9` MSVC files again (since `v8` works fine for `v9`)
988 - print Ipopt version in default print level
989 - added option that allows to change name of options file
992 ### 3.4.0 (2008-04-25)
994 - added support to dynamically load HSL or Pardiso:
995 If Ipopt has been compiled without some HSL or Pardiso solver,
996 it can now load those solvers from a shared library at runtime
997 without recompilation. This will make distribution of binaries
998 easier. Does not work on all platforms yet.
1000 - ensured compilation of MSVS project files (`v8` and `v9`)
1001 - new special return code for square problems
1002 (`Feasible_Point_Found` returned if dual inf not small)
1003 - new initialization option for bound multipliers
1004 (see option `bound_mult_init_method`)
1005 - added simple penalty function line search option
1006 (`line_search_method=penalty`) - not guaranteed to converge, see
1007 Ipopt implementation paper (in MathProg)
1008 - some very basic method to approximate constraint Jacobian by
1009 finite differences (not efficient, but will hopefully be extended)
1013 ### 3.3.5 (2008-02-28)
1015 - corrected links for Ipopt mailing list
1016 - added missing `Makefile.in` for Matlab interface
1017 - the `addlibs*` files are now installed in `share/doc/coin/Ipopt`
1019 - updates in Matlab interface
1020 - bugfix for ticket #56
1022 ### 3.3.4 (2007-12-27)
1024 - headers are now installed in `include/coin` (no longer in `include/ipopt`)
1025 - default for `dual_inf_tol` is now 1 (instead of 1e-4)
1026 - In matlab interface, here the text from Peter Carbonetto:
1027 There have been several significant changes made to the MATLAB interface
1028 since the last release. The most important two changes are:
1029 1. Following the "warm start" feature of IPOPT, you may pass in initial
1030 estimates for the Lagrange multipliers.
1031 2. Callback routines for computing the objective, gradient (etc.) are now
1032 specified using function handles instead of strings.
1033 (Thanks to Marcus Brubaker at the University of Toronto for the initial suggestion.)
1035 ### 3.3.3 (2007-09-25)
1037 - minor changes, bug fixes
1039 ### 3.3.1 (2007-06-20)
1041 Synchronized with all changes in trunk; probably more than to be
1042 remembered. In the following a few:
1043 - support for Mumps linear solver (contributed by Damian Hocking)
1044 - `--print-options` flag for ipopt ASL solver executable to see all
1045 Ipopt options (available through `ipopt.opt` file)
1046 - added Matlab interface (contributed by Peter Carbonetto)
1047 - added support for `f2c` compiler to compiler Fortran code with
1049 - new MSVS support (now within MSVS project and also with `f2c`)
1050 - a number of small changes/bug fixes/improvements
1051 - small change in interface (e.g., `FinalizeSolution` method)
1055 ### 3.2.4 (2007-04-24)
1057 - updated download script for Blas to fit netlib's recent changes
1058 - using a more recent version of BuildTools
1060 ### 3.2.3 (2006-11-29)
1062 - updated download script for Lapack to fit to netlib's recent changes
1064 ### 3.2 r795 (2006-10-11)
1066 - Bugfix in L-BFGS update
1067 - fix in configure with detection of `sizeof(int*)` on Cygwin
1069 ### 3.2.1 (2006-07-14) - dev release number 764
1071 - Bugfix in least square multiplier estimate.
1072 It mainly showed up in LBFGS with restoration phase as seg fault
1074 ### 3.2.0 (2006-07-07) - dev release number 757
1076 - changed installation procedure and directory structure to
1077 conform with new COIN-OR convention
1081 ### 3.1.0 (2006-04-08) - dev release number 714
1083 Several bug-fixes, improvements and additions. In particular:
1084 - new quasi-Newton approximation using L-BFGS
1085 - interfaces to linear solver MA57, WSMP, Pardiso
1086 (MUMPS and TAUCS not yet completed)
1087 - derivative checker
1089 - configure supports compilation under Cygwin with native Windows compilers
1091 - user call-back method in TNLP
1095 ### 3.0.1 (2005-12-04)
1097 - Several corrections to Windows files
1098 - Fix termination if number of iterations is exceeded in restoration phase
1100 ### 3.0.0 (2005-08-26) - dev release number 510
1102 - First official release of the new C++ implementation of Ipopt.
1106 ### no new release (2005-08-19)
1108 - corrected detection of BLAS libraries for SUN (make sure the example Makefiles work)
1109 - upgrade LICENSE file to CPL version 1.0 as retrieved from www.opensource.org
1111 ### 2.2.1e (2005-05-30)
1113 - fixed sign of multipliers returned to AMPL
1114 (bug reported by Rhoda Baker and Karsten Theissen)
1115 - switched to automake 1.9.5
1117 ### no new release (2005-01-07)
1119 - bugfix for the limited memory BFGS in case of square problems
1120 (bug reported by Wanhe Zhang and Ned Nedialkov)
1122 ### 2.2.1d (2004-10-05)
1124 - Added `outlev` as an option to the AMPL solver as a synonym for `iprint`
1125 - For `iprint` = 0, the output lines per iteration are now suppressed
1126 - corrected two bugs in `configure` script (test for size of `long` etc
1127 before Fortran libraries as added to `LIBS`; prevent cycling in
1129 - internally renamed subroutine `ERROR` to `OPTERROR` (`ERROR` had a
1130 name clash for some Fortran compiler)
1131 - avoid uninitialized variable in `update_b_lm.f`
1132 - minor correction in computation of residual in `get_step_full.F`
1133 - minor change for slack correction in `filter.F`
1135 ### 2.2.1c (2004-07-20)
1137 - corrected bug leading to very small `QTAU` in rare circumstances
1139 ### 2.2.1b (2004-05-21)
1141 - Make `DFILLINFACT` option available through AMPL interface
1142 - Now, later increase of memory requirement for Harwell solvers is also
1143 based on `DFILLINFACT`, instead of using a fixed values of 10.
1145 ### 2.2.1a (2004-05-13)
1147 - fix in `IPOPT/ipopt/mainloop.F`:
1148 The multipliers were not scaled back for low printlevel.
1150 ### no new release (2004-04-28)
1152 - fix in `IPOPT/AMPL_interface/ipoptAMPL.c`:
1153 Now the mulitpliers for the constraints are passed back to AMPL.
1154 - Added download scripts using `wget` to get ASL, BLAS, and LAPACK more easily.
1155 Thanks to Frank Wuebbeling for the hint.
1157 ### 2.2.1 (2004-04-25)
1159 - AMPL solver executable is now called `ipopt` (instead of `ipoptAMPL`).
1160 This fixed also problem with assigning IPOPT options from within
1161 AMPL. (reported by Karsten Theissen)
1162 - default value for number of iterations is now 10000 (instead of 1000)
1163 - new option: `IMAXCPUSEC` to be set to the maximum number of CPU seconds
1164 after which the algorithm should stop. The check is performed only at
1165 certain points in the algorithm, so that the executable might run longer
1167 The algorithm also stops (as before) when the file "STOP" is detected in
1168 the current directory. Finally, a call to `USER_REQUESTED_STOP` has
1169 been added if the preprocessor macro `USE_USER_REQUESTED_STOP` has been
1170 defined. If this `LOGICAL` function returns `.TRUE.`, the algorithm also
1171 stops. This feature was requested by David Ternet.
1172 - if IPOPT is run several time in a row, the counting of function
1173 evaluations is restarted after every new call of IPOPT.
1174 - in `get_step_full`, now check if the number of negative eigenvalues is LESS
1175 than then number constraints. If so, increase the pivot tolerance,
1176 and if that doesn't help, pretend that the system is singular.
1177 (this fixed a problem reported by Hans Mittelmann)
1178 - decrease default values for `DPIVTOLMAX`
1179 - suppress superflous leading zeros in iteration output. For long output
1180 (`ioutput=1`) include CPU time
1181 - a few changes regarding the inertia correction (`get_step_full.F`).
1182 This decreases CPU time significantly in a few cases.
1183 - corrected problem in C-files related to names for `struct`'s (some GNU
1184 compiler complained)
1185 - changed default options for GNU compilers (now `-O3 -funroll-loops`,
1186 no longer `-O2` and `-mieee-fp`)
1187 - no reference to MC35 in `resto_tron` if `HAVE_MC35` is not defined
1188 - `#error` preprocessor directive removed from `*.F` files, since not all
1189 compilers understand it (reported by Hans Mittelmann)
1190 - switch to automake 1.8.3
1192 ### 2.2.0 (2004-03-10)
1194 Many things have changed since the last official release.
1195 Here a few highlights:
1196 - easier installation procedure with autoconf
1197 - algorithm made more robust and efficient
1198 - new restoration phase for filter method (TRON no longer needed for
1199 full-space option anymore)