From jmwille at software.sandia.gov Thu Apr 20 20:06:56 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:06:58 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos Makefile.in aclocal.m4 configure Message-ID: <200604210206.k3L26u6G027327@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos In directory software.sandia.gov:/tmp/cvs-serv27187 Modified Files: Makefile.in aclocal.m4 configure Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:06:57 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:06:59 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example Makefile.in Message-ID: <200604210206.k3L26vEa027361@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example In directory software.sandia.gov:/tmp/cvs-serv27187/example Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:06:57 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:06:59 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/config acx_blas.m4 acx_lapack.m4 Message-ID: <200604210206.k3L26vot027343@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/config In directory software.sandia.gov:/tmp/cvs-serv27187/config Modified Files: acx_blas.m4 acx_lapack.m4 Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:06:58 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:01 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example/constExample Makefile.in Message-ID: <200604210206.k3L26weF027421@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example/constExample In directory software.sandia.gov:/tmp/cvs-serv27187/example/constExample Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:06:58 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:01 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example/basicExample Makefile.in Message-ID: <200604210206.k3L26vQ9027380@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example/basicExample In directory software.sandia.gov:/tmp/cvs-serv27187/example/basicExample Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:06:58 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:01 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example/classExample Makefile.in Message-ID: <200604210206.k3L26wv2027395@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example/classExample In directory software.sandia.gov:/tmp/cvs-serv27187/example/classExample Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:06:59 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:02 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example/epetra Makefile.in Message-ID: <200604210206.k3L26xss027438@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example/epetra In directory software.sandia.gov:/tmp/cvs-serv27187/example/epetra Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:07:00 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:02 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example/refCountPtrExample Makefile.in Message-ID: <200604210207.k3L270uT027471@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example/refCountPtrExample In directory software.sandia.gov:/tmp/cvs-serv27187/example/refCountPtrExample Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:07:00 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:02 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example/epetra/1Dfem Makefile.in Message-ID: <200604210207.k3L27043027454@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example/epetra/1Dfem In directory software.sandia.gov:/tmp/cvs-serv27187/example/epetra/1Dfem Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:07:00 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:02 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example/refCountPtrExample2 Makefile.in Message-ID: <200604210207.k3L2707C027495@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example/refCountPtrExample2 In directory software.sandia.gov:/tmp/cvs-serv27187/example/refCountPtrExample2 Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:07:02 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:04 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/example/verySimple Makefile.in Message-ID: <200604210207.k3L2721V027520@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/example/verySimple In directory software.sandia.gov:/tmp/cvs-serv27187/example/verySimple Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:07:02 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:06 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/src Makefile.in Message-ID: <200604210207.k3L272ve027537@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/src In directory software.sandia.gov:/tmp/cvs-serv27187/src Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:07:03 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:06 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/test Makefile.in Message-ID: <200604210207.k3L273eF027554@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/test In directory software.sandia.gov:/tmp/cvs-serv27187/test Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From jmwille at software.sandia.gov Thu Apr 20 20:07:03 2006 From: jmwille at software.sandia.gov (James M. Willenbring) Date: Thu Apr 20 20:07:10 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/test/unitTests Makefile.in Message-ID: <200604210207.k3L2732G027580@software.sandia.gov> Update of /space/CVS/Trilinos/packages/rythmos/test/unitTests In directory software.sandia.gov:/tmp/cvs-serv27187/test/unitTests Modified Files: Makefile.in Log Message: Bug 1409. Modified the blas and lapack search .m4 files to stop searching if a user specified blas or lapack library cannot be found. This only applies to blas or lapack libraries specified via --with-blas and --with-lapack. It is not reasonable to apply this to other options such as --with-libs and --with-ldflags. In the past if a user specified blas or lapack library was not found, the .m4 would continue to try to find blas or lapack in different places. This was confusing for users because they might have been trying to use an optimized blas library, but ended up with a different blas library. Unless a person looked through all of the configure output, this was not noticable. I tested this modification by checking it into Epetra for a few days before applying the change to other packages. The most recent bootstrap was done with the wrong version of automake so all Makefile.in's were modified by my bootstrap. From rabartl at software.sandia.gov Fri Apr 28 08:32:05 2006 From: rabartl at software.sandia.gov (Roscoe A. Bartlett) Date: Fri Apr 28 08:32:07 2006 Subject: [Rythmos-checkins] Trilinos/packages/rythmos/src Rythmos_Stepper_BackwardEuler.hpp Rythmos_Stepper_ExplicitRK.hpp Rythmos_Stepper_ExplicitTaylorPolynomial.hpp Rythmos_Stepper_ForwardEuler.hpp Rythmos_Stepper_ImplicitBDF.hpp Message-ID: <200604281432.k3SEW58i022765@software.sandia.gov> Update of /cvsroot/Trilinos/packages/rythmos/src In directory software.sandia.gov:/tmp/cvs-serv22505/packages/rythmos/src Modified Files: Rythmos_Stepper_BackwardEuler.hpp Rythmos_Stepper_ExplicitRK.hpp Rythmos_Stepper_ExplicitTaylorPolynomial.hpp Rythmos_Stepper_ForwardEuler.hpp Rythmos_Stepper_ImplicitBDF.hpp Log Message: I have done several things here: Changes to Thyra::ModelEvaluator and related software. 1) I Broke up the file Thyra_ModelEvaluator.hpp into two files Thyra_ModelEvaluatorBase.hpp (containing the non-templated base base class Thyra::ModelEvaluatorBase) and Thyra_ModelEvaluator.hpp (containing the templated base class Thyra::ModelEvaluator). 2) I removed all of the default function implementations from Thyra::ModelEvaluator and put them into a new base class (derived from Thyra::ModelEvaluator) called Thyra::StateFuncModelEvaluatorBase in thyra/src/support/nonlinear/model_evaluator/client_support/Thyra_StateFuncModelEvaluatorBase.hpp. I also moved all of the wrapper functions like calc_f() and calc_f_W() to a new file thyra/src/support/nonlinear/model_evaluator/client_support/Thyra_ModelEvaluatorHelpers.hpp. Now I think that the interface files Thyra_ModelEvaluator.hpp and Thyra_ModelEvaluatorBase.hpp are much cleaner and all extraneous non-interoperability code has been moved to support code directories. 3) From Thyra::ModelEvaluator I have replaced get_x_init(), get_p_init(l), and get_t_init() with a single function called getNominalValues() that returns an InArgs object. I did the same thing for upper and lower bounds. This significantly reduced the number of virtual functions in Thyra::ModelEvaluator, makes it easier to override the initial guess (or nominal values depending on how they are used), and make it easier to use in general (in my opinion). 4) From Thyra::ModelEvaluator I removed the create_DfDp_mv(), create_DgDx_mv() and create_DgDp_mv() functions and have replacement them with non-member helper functions in Thyra_ModelEvaluatorHelpers.hpp. The reason that the create_DfDp_mv(), create_DgDx_mv() and create_DgDp_mv() where part of the interface was to allow the ModelEvaluator subclass to attach a non-Euclidean scalar product to the domain (or range) space for these multi-vectors. However, this is not the right way to do this. What Thyra needs is to replace the createMembers(space,numMembers) with createMembers(space,domain). Our ANAs don't really handle non-Euclidean spaces well yet and this is up in the air somewhat. 5) I create a base class called Thyra::ModelEvaluatorDelegatorBase that makes it very easy to create and maintain Thyra::ModelEvaluator decorator classes. Now it is much easier to create a whole host of different types of decorators as described below. 6) I added several new concrete Decorator subclasses: - Thyra:::DefaultStateEliminationModelEvaluator: Given a Thyra::NonlinearSolverBase object, eliminates state variables x and the state equations f(x,) and computes the reduced functions g_j(x(),). I am using this to implement a black-box optimization approach. - Thyra::DefaultFiniteDifferenceModelEvaluator: Computes derivatives using finite differences. This is very simple right now just computing DgDp but it will become more sophisticated very quickly. - Thyra::DefaultNominalBoundsOverideModelEvaluator: Overrides the nominal values (initial guess) and/or the upper or lower bounds. I did not make any changes whatsoever to EpetraExt::ModelEvaluator and I do not think any such changes are needed. Thyra::ModelEvaluator and EpetraExt::ModelEvaluator have completely different purposes and it is appropriate that they be different. This checkin also includes the beginnings of a black-box finite-difference mode for the MOOCHO AdvDiffReact problem. There is still lots to do however. There are also other small changes but I do not want to list them all.