small updates prior to passing the code along for comparison

updating contact and parproblems files
cleanup on finite difference test code. Making it easier for others to use
2024-07-29 16:34:06 -07:00 · 2024-07-15 10:47:54 -07:00 · 2024-07-15 08:08:00 -07:00 · 2024-07-14 20:43:48 -07:00 · 2024-06-17 13:15:59 -07:00
443 changed files with 71027 additions and 27327 deletions
@@ -0,0 +1,61 @@
+# Configuration for probot-stale - https://github.com/probot/stale
+
+# Number of days of inactivity before an Issue or Pull Request becomes stale
+daysUntilStale: 30
+
+# Number of days of inactivity before an Issue or Pull Request with the stale
+# label is closed.  Set to false to disable. If disabled, issues still need to
+# be closed manually, but will remain marked as stale.
+daysUntilClose: 7
+
+# Only issues or pull requests with all of these labels are check if stale.
+# Defaults to `[]` (disabled)
+onlyLabels: []
+
+# Issues or Pull Requests with these labels will never be considered stale. Set
+# to `[]` to disable
+exemptLabels:
+  - bug
+  - WIP
+  - ready-for-review
+  - in-review
+  - in-next
+
+# Set to true to ignore issues in a project (defaults to false)
+exemptProjects: false
+
+# Set to true to ignore issues in a milestone (defaults to false)
+exemptMilestones: false
+
+# Set to true to ignore issues with an assignee (defaults to false)
+exemptAssignees: false
+
+# Label to use when marking an issue as stale
+staleLabel: stale
+
+# Comment to post when marking an issue as stale. Set to `false` to disable
+markComment: >
+  :warning: This issue or PR has been automatically marked as stale because it has not
+  had any activity in the last month. *If no activity occurs in the next week, it will
+  be automatically closed.* Thank you for your contributions.
+
+# Comment to post when closing a stale issue. Set to `false` to disable
+closeComment: false
+
+# Limit the number of actions per hour, from 1-30. Default is 30
+limitPerRun: 30
+
+# Limit to only `issues` or `pulls`
+# only: issues
+
+# Optionally, specify configuration settings that are specific to just 'issues' or 'pulls':
+# pulls:
+#   daysUntilStale: 30
+#   markComment: >
+#     This pull request has been automatically marked as stale because it has not had
+#     recent activity. It will be closed if no further activity occurs. Thank you
+#     for your contributions.
+
+# issues:
+#   exemptLabels:
+#     - confirmed
@@ -1,31 +0,0 @@
-# This workflow warns and then closes issues and PRs that have had no activity for a specified amount of time.
-# For more information, see: https://github.com/actions/stale
-name: Mark stale issues and pull requests
-
-on:
-  workflow_dispatch:
-  schedule:
-  - cron: '0 0 * * *'
-
-jobs:
-  stale:
-
-    runs-on: ubuntu-latest
-    permissions:
-      issues: write
-      pull-requests: write
-      actions: write
-
-    steps:
-    - uses: actions/stale@v9
-      with:
-        repo-token: ${{ secrets.GITHUB_TOKEN }}
-        stale-issue-message: ':warning: This issue has been automatically marked as stale because it has not had any activity in the last month. *If no activity occurs in the next week, it will be automatically closed.* Thank you for your contributions.'
-        stale-pr-message: ':warning: This PR has been automatically marked as stale because it has not had any activity in the last month. *If no activity occurs in the next week, it will be automatically closed.* Thank you for your contributions.'
-        days-before-stale: 30
-        days-before-close: 7
-        stale-issue-label: 'stale'
-        stale-pr-label: 'stale'
-        operations-per-run: 500
-        exempt-issue-labels: "bug,WIP,ready-for-review,in-review,in-next"
-        exempt-pr-labels: "bug,WIP,ready-for-review,in-review,in-next"
@@ -1,31 +0,0 @@
-# Copyright (c) 2010-2024, Lawrence Livermore National Security, LLC. Produced
-# at the Lawrence Livermore National Laboratory. All Rights reserved. See files
-# LICENSE and NOTICE for details. LLNL-CODE-806117.
-#
-# This file is part of the MFEM library. For more information and source code
-# availability visit https://mfem.org.
-#
-# MFEM is free software; you can redistribute it and/or modify it under the
-# terms of the BSD-3 license. We welcome feedback and contributions, see file
-# CONTRIBUTING.md for details.
-
-name: "Trigger PyMFEM CI"
-
-on:
-  push:
-    branches:
-      - master
-
-jobs:
-  trigger-pymfem:
-    runs-on: ubuntu-latest
-    steps:
-    - name: Send POST request to trigger PyMFEM CI
-      run: |
-        curl -L \
-             -X POST \
-             -H "Accept: application/vnd.github+json" \
-             -H "Authorization: Bearer ${{ secrets.PYMFEM_CI_TOKEN }}" \
-             -H "X-GitHub-Api-Version: 2022-11-28" \
-             https://api.github.com/repos/mfem/pymfem/actions/workflows/build-and-test-dispatch.yml/dispatches \
-             -d '{"ref":"master", "inputs":{"test_options":"fast"}}'
@@ -15,9 +15,6 @@
 CMakeCache.txt
 CMakeFiles/

-# Clangd server cache
-*.cache*
-
 # Backup files
 *~

@@ -275,27 +272,16 @@ miniapps/navier/*_output

 miniapps/nurbs/nurbs_ex1
 miniapps/nurbs/nurbs_ex1p
-miniapps/nurbs/nurbs_ex3
-miniapps/nurbs/nurbs_ex5
 miniapps/nurbs/nurbs_ex11p
-miniapps/nurbs/nurbs_ex24
-miniapps/nurbs/nurbs_solenoidal
 miniapps/nurbs/nurbs_printfunc
 miniapps/nurbs/nurbs_patch_ex1
 miniapps/nurbs/nurbs_curveint
 miniapps/nurbs/refined.mesh
 miniapps/nurbs/mesh.*
-miniapps/nurbs/sol_?.gf
 miniapps/nurbs/sol.*
 miniapps/nurbs/mode_*
 miniapps/nurbs/Example1*
-miniapps/nurbs/Example3*
-miniapps/nurbs/Example5*
-miniapps/nurbs/Solenoidal*
-miniapps/nurbs/ParaView
 miniapps/nurbs/sin-fit.mesh
-miniapps/nurbs/ex5.mesh
-miniapps/nurbs/exsol.mesh
 miniapps/nurbs/CurveInt
 miniapps/nurbs/nurbs_naca_cmesh
 miniapps/nurbs/naca-cmesh.mesh
@@ -22,7 +22,7 @@ include:
 # the "needs" keyword and express the DAG of jobs for more efficiency.
 # - We use setup and setup_baseline phases to download content outside of mfem
 #   directory.
-# - Allocate/Release is where ruby resource are allocated/released once for all.
+# - Allocate/Release is where quartz resource are allocated/released once for all.
 # - Build and Test is where we build and MFEM for multiple toolchains.
 # - Baseline_checks gathers baseline-type test suites execution
 # - Baseline_publish, only available on master, allows to update baseline
@@ -53,7 +53,7 @@ variables:
  AUTOTEST_COMMIT: "YES"

 # Trigger subpipelines:
-ruby-build-and-test:
+quartz-build-and-test:
  stage: sub-pipelines
  variables:
    # Explicitly pass down values that we want to be able to set when triggering
@@ -61,10 +61,10 @@ ruby-build-and-test:
    AUTOTEST: "${AUTOTEST}"
    AUTOTEST_COMMIT: "${AUTOTEST_COMMIT}"
  trigger:
-    include: .gitlab/ruby-build-and-test.yml
+    include: .gitlab/quartz-build-and-test.yml
    strategy: depend

-ruby-baseline:
+quartz-baseline:
  stage: sub-pipelines
  variables:
    # Explicitly pass down values that we want to be able to set when triggering
@@ -73,7 +73,7 @@ ruby-baseline:
    AUTOTEST: "${AUTOTEST}"
    AUTOTEST_COMMIT: "${AUTOTEST_COMMIT}"
  trigger:
-    include: .gitlab/ruby-baseline.yml
+    include: .gitlab/quartz-baseline.yml
    strategy: depend

 lassen-build-and-test:
@@ -24,7 +24,7 @@ and `test type`.

 Machines typically include:

-* Ruby: 2nd Gen Intel Xeon (Cascade Lake)
+* Quartz: Intel bi-socket x86
 * Lassen: Power9 + Nvidia GPU
 * Corona: AMD GPU

@@ -76,13 +76,13 @@ with a spack spec of MFEM, within the limits permitted by the MFEM spack
 package.

 In any build-and-test sub-pipeline a job basically consists in defining the
-spack spec to use. Adding a job on ruby for example resumes to:
+spack spec to use. Adding a job on quartz for example resumes to:

 ```yaml
 <job_name>:
  variables:
    SPEC: "<spack_spec>"
-  extends: .build_and_test_on_ruby
+  extends: .build_and_test_on_quartz
 ```

 The remaining and non trivial work is to make sure this spec is working. To
@@ -24,7 +24,7 @@ variables:
 # TODO: add a clean-up mechanism
  BUILD_ROOT: ${USER_CI_TOP_DIR}/${CI_PROJECT_NAME}-${MACHINE_NAME}-pipeline-${CI_PIPELINE_ID}

-# On LLNL's ruby, there is only one allocation shared among jobs in order to
+# On LLNL's quartz, there is only one allocation shared among jobs in order to
 # save time and resource. This allocation has to be uniquely named so that we
 # are sure to retrieve it.
  ALLOC_NAME: ${CI_PROJECT_NAME}_ci_${CI_PIPELINE_ID}
@@ -9,17 +9,17 @@
 # terms of the BSD-3 license. We welcome feedback and contributions, see file
 # CONTRIBUTING.md for details.

-# GitLab pipelines configurations for the Ruby machine at LLNL
+# GitLab pipelines configurations for the Quartz machine at LLNL
 variables:
-  MACHINE_NAME: ruby
+  MACHINE_NAME: quartz

-.on_ruby:
+.on_quartz:
  tags:
    - shell
-    - ruby
+    - quartz
  rules:
-    # Don't run ruby jobs if...
-    - if: '$CI_COMMIT_BRANCH =~ /_qnone/ || $ON_RUBY == "OFF"'
+    # Don't run quartz jobs if...
+    - if: '$CI_COMMIT_BRANCH =~ /_qnone/ || $ON_QUARTZ == "OFF"'
      when: never
    # Don't run autotest update if...
    - if: '$CI_JOB_NAME =~ /report/ && $AUTOTEST != "YES"'
@@ -40,13 +40,13 @@ variables:
    - when: on_success

 # Spack helped builds
-# Generic ruby build job, extending build script
-.build_and_test_on_ruby:
-  extends: [.on_ruby]
+# Generic quartz build job, extending build script
+.build_and_test_on_quartz:
+  extends: [.on_quartz]
  stage: build_and_test
  script:
    # THREADS is used by 'tests/gitlab/build_and_test', run below
-    - export THREADS=16
+    - export THREADS=12
    - echo ${ALLOC_NAME}
    - export JOBID=$(squeue -h --name=${ALLOC_NAME} --format=%A)
    - echo ${JOBID}
@@ -18,7 +18,7 @@
 setup_baseline:
  tags:
    - shell
-    - ruby
+    - quartz
  stage: setup
  variables:
    GIT_STRATEGY: none
@@ -16,7 +16,7 @@
 setup:
  tags:
    - shell
-    - ruby
+    - quartz
  stage: setup
  variables:
    GIT_STRATEGY: none
@@ -19,8 +19,8 @@ stages:
  - cleanup
  - baseline_publish

-baselinecheck_mfem_intel_ruby:
-  extends: [.on_ruby]
+baselinecheck_mfem_intel_quartz:
+  extends: [.on_quartz]
  stage: baseline_check
  variables:
    # TPLS_DIR is used in .gitlab/scripts/baseline to provide the tpls location
@@ -32,7 +32,7 @@ baselinecheck_mfem_intel_ruby:
    - echo ${BUILD_ROOT}
    - echo ${TPLS_DIR}
    # Used by the tests in MFEM/tests:
-    - export MFEM_TEST_NP=48
+    - export MFEM_TEST_NP=32
    # The next script uses the following environment variables:
    # * BASELINE_TEST, SYS_TYPE, CI_PROJECT_DIR, ARTIFACTS_DIR,
    # * BUILD_ROOT, TPLS_DIR, MACHINE_NAME
@@ -44,16 +44,18 @@ baselinecheck_mfem_intel_ruby:
  allow_failure: true

 cleanup:
-  extends: .on_ruby
+  extends: .on_quartz
  stage: cleanup
  variables:
    GIT_STRATEGY: none
  script:
    - echo "BUILD_ROOT=${BUILD_ROOT}"
    - rm -rf "${BUILD_ROOT}" || true
+    - echo "CI_PROJECT_DIR=${CI_PROJECT_DIR}"
+    - make -C "${CI_PROJECT_DIR}" distclean

 report_baseline:
-  extends: [.on_ruby]
+  extends: [.on_quartz]
  stage: baseline_report
  script:
    - echo ${MACHINE_NAME}
@@ -113,8 +115,8 @@ report_baseline:
        exit $err
      ) 9> autotest.lock

-baselinepublish_mfem_ruby:
-  extends: [.on_ruby]
+baselinepublish_mfem_quartz:
+  extends: [.on_quartz]
  stage: baseline_publish
  rules:
    # - if: '$CI_COMMIT_BRANCH == "master" || $REBASELINE == "YES"'
@@ -129,5 +131,5 @@ baselinepublish_mfem_ruby:

 include:
  - local: .gitlab/configs/common.yml
-  - local: .gitlab/configs/ruby-config.yml
+  - local: .gitlab/configs/quartz-config.yml
  - local: .gitlab/configs/setup-baseline.yml
@@ -19,54 +19,54 @@ stages:
 allocate_resource:
  variables:
    GIT_STRATEGY: none
-  extends: .on_ruby
+  extends: .on_quartz
  stage: allocate_resource
  script:
    - echo ${ALLOC_NAME}
    - salloc --exclusive --nodes=1 --reservation=ci --time=60 --no-shell --job-name=${ALLOC_NAME}
  timeout: 6h

-# GitLab jobs for the Ruby machine at LLNL
+# GitLab jobs for the Quartz machine at LLNL
 debug_ser_gcc_10:
  variables:
    SPEC: "%gcc@10.3.1 +debug~mpi"
-  extends: .build_and_test_on_ruby
+  extends: .build_and_test_on_quartz

 debug_par_gcc_10:
  variables:
    SPEC: "%gcc@10.3.1 +debug+mpi"
-  extends: .build_and_test_on_ruby
+  extends: .build_and_test_on_quartz

 opt_ser_gcc_10:
  variables:
    SPEC: "%gcc@10.3.1 ~mpi"
-  extends: .build_and_test_on_ruby
+  extends: .build_and_test_on_quartz

 opt_par_gcc_10:
  variables:
    SPEC: "%gcc@10.3.1"
-  extends: .build_and_test_on_ruby
+  extends: .build_and_test_on_quartz

 opt_par_gcc_10_sundials:
  variables:
    SPEC: "%gcc@10.3.1 +sundials"
-  extends: .build_and_test_on_ruby
+  extends: .build_and_test_on_quartz

 opt_par_gcc_10_petsc:
  variables:
    SPEC: "%gcc@10.3.1 +petsc ^petsc+mumps~superlu-dist"
-  extends: .build_and_test_on_ruby
+  extends: .build_and_test_on_quartz

 opt_par_gcc_10_pumi:
  variables:
    SPEC: "%gcc@10.3.1 +pumi"
-  extends: .build_and_test_on_ruby
+  extends: .build_and_test_on_quartz

 # Release
 release_resource:
  variables:
    GIT_STRATEGY: none
-  extends: .on_ruby
+  extends: .on_quartz
  stage: release_resource_and_report
  script:
    - echo ${ALLOC_NAME}
@@ -78,17 +78,17 @@ release_resource:
 report_job_success:
  stage: release_resource_and_report
  extends:
-    - .on_ruby
+    - .on_quartz
    - .report_job_success

 report_job_failure:
  stage: release_resource_and_report
  extends:
-    - .on_ruby
+    - .on_quartz
    - .report_job_failure

 include:
  - local: .gitlab/configs/common.yml
-  - local: .gitlab/configs/ruby-config.yml
+  - local: .gitlab/configs/quartz-config.yml
  - local: .gitlab/configs/setup-build-and-test.yml
  - local: .gitlab/configs/report-build-and-test.yml
@@ -14,7 +14,7 @@
 # locals
 glob_err=${BASELINE_TEST}.err
 base=${BASELINE_TEST}-${SYS_TYPE}
-if [[ "${MACHINE_NAME}" == "ruby" ]]; then
+if [[ "${MACHINE_NAME}" == "quartz" ]]; then
  base="${BASELINE_TEST}-${MACHINE_NAME}"
 fi
 base_diff=${base}.diff
@@ -31,8 +31,8 @@ cd tests
 mkdir _${BASELINE_TEST} && cd _${BASELINE_TEST}

 # run
-if [[ "${MACHINE_NAME}" == "ruby" ]]; then
-  salloc --nodes=1 --exclusive --reservation=ci ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
+if [[ "${MACHINE_NAME}" == "quartz" || "${MACHINE_NAME}" == "ruby" ]]; then
+  salloc --nodes=1 --reservation=ci ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
 elif [[ ${MACHINE_NAME} == "corona" ]]; then
  salloc --nodes=1 -t 60 -p pbatch ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
 elif [[ ${MACHINE_NAME} == "lassen" ]]; then
@@ -41,11 +41,11 @@ else
  echo "Unknown machine: MACHINE_NAME=$MACHINE_NAME"
  exit 1
 fi
-status="$?"

 # post
 mkdir ${artifacts_path}

+status=0
 if [[ -f ${BASELINE_TEST}.out ]]; then
  cp ${BASELINE_TEST}.out ${artifacts_path}
 fi
@@ -11,7 +11,7 @@
 # terms of the BSD-3 license. We welcome feedback and contributions, see file
 # CONTRIBUTING.md for details.

-# There will be collision between corona and ruby baselines.
+# There will be collision between corona and quartz baselines.
 # Once the corresponding files have been generated, we can switch to machine
 # specific ref.
 ARTIFACT_PATH=${CI_PROJECT_DIR}/${ARTIFACTS_DIR}/baseline-${SYS_TYPE}
@@ -21,7 +21,7 @@ PATCH_FILE=${ARTIFACT_PATH}.patch
 FULL_FILE=${ARTIFACT_PATH}.out
 DIFF_FILE=${ARTIFACT_PATH}.diff

-# There will be collision between corona and ruby baselines.
+# There will be collision between corona and quartz baselines.
 # Once the corresponding files have been generated, we can switch to machine
 # specific ref.
 SAVED_NAME=baseline-${SYS_TYPE}.saved
@@ -10,87 +10,10 @@

 Version 4.7.1 (development)
 ===========================
- Refactored ALGOIM cut integration rules. The interface is unified with
-  the interface for moment based cut integration rules.
-
-Discretization improvements
---------------------------
- Added NURBS-based H(div) and H(curl) elements in 2D and 3D. Only on single
-  patch meshes. Only implemented for serial computations.
-
- Added support for boundary constraints to the hybridization class.
-
- Added support for external boundary submeshes with nonconformal mesh adaptation.
-
-Meshing improvements
--------------------
- The ExodusII reader now handles pyramid and wedge element types. Mixed meshes
-  are also supported.
-
-New and updated examples and miniapps
-------------------------------------
- Added miniapps to demonstrate the H(div) and H(curl) NURBS elements.

 - Added an MFEM example for the eikonal equation. This new solver is based on
  the proximal Galerkin method introduced by Keith and Surowiec.

- Added a command line option to all miniapps (`-p` or `--send-port`) for
-  specifying the GLVis server socket port (19916 by default).
-
-GPU computing
-------------
- Added support for GPU-accelerated batched linear algebra (using cuBLAS,
-  hipBLAS, MAGMA, or native MFEM functionality) through the BatchedLinAlg class.
-
- A new GPU kernel dispatch mechanism was introduced. Users can instantiate
-  specialized kernels for specific combinations of (for example) polynomial
-  degree and number of quadrature points using
-  `DiffusionIntegrator::AddSpecialization` and
-  `MassIntegrator::AddSpecialization` (this functionality may be added to more
-  integrators in the future).
-
- Calls to slower fallback kernels can be reported to `mfem::err` by setting
-  the environment variable `MFEM_REPORT_KERNELS` to any value other than `NO`
-  or by explicitly calling `KernelReporter::Enable`. Users can then add
-  specializations for these kernels to achieve higher performance.
-
- Element assembly kernels have been added for low-order refined to
-  high-order transfer operators. New kernels can be offloaded as device
-  kernels. Example usage may be found in lor-transfer.cpp under miniapps/tools.
-
-Miscellaneous
-------------
- Added support for SUNDIALS v7. See the section "API changes" for some small
-  changes related to this new version.
-
- Refactored the `ARKStepSolver` class (ARKODE interface) to use
-  `TimeDependentOperator::Mult` only when the associated ODE operator is
-  expressed in explicit form (i.e., `TimeDependentOperator::isExplicit()`),
-  otherwise `TimeDependentOperator::ExplicitMult` is used. A check has been
-  added to `ARKStepSolver` to verify that the associated ODE operator is not in
-  explicit form when a mass matrix solver is enabled via a call to either the
-  `UseMFEMMassLinearSolver` or `UseSundialsMassLinearSolver` methods. This is
-  because enabling a mass matrix solver assumes that F(u,k,t) = M k in the
-  associated ODE operator.
-
- Added support for custom interpolation procedure in FindPointsGSLIB.
-
-API changes
-----------
- API change: in class GridFunction, 'fec' was renamed to 'fec_owned'.
-
- API change: support for SUNDIALS v7:
-  * the SUNDIALS types `realtype` and `booleantype` are no longer defined by v7
-    and therefore MFEM now uses the new type names `sunrealtype` and
-    `sunbooleantype`, respectively, which MFEM defines when using SUNDIALS < v6
-    where these types were not defined.
-  * The SUNDIALS macro `SUNLS_SUCCESS` and some other `*_SUCCESS` macros were
-    removed and replaced by `SUN_SUCCESS` in v7, so to avoid tedious checks for
-    SUNDIALS versions, MFEM now defines and uses the constant `SUN_SUCCESS` when
-    using SUNDIALS < v7.
-  * The constants `SUN_PREC_*`, introduced by SUNDIALS v6 are now introduced by
-    MFEM when using SUNDIALS < v6 to avoid tedious version checks.
-

 Version 4.7, released on May 7, 2024
 ====================================
@@ -115,9 +38,6 @@ Meshing improvements

 - Added support for internal boundary elements in nonconforming meshes.

- Added ExodusII output capability. The writer can handle first-order (Pyramid5,
-  Wedge6, Hex8, Tet4) and second-order FE types (Pyramid14, Wedge18, Hex27, Tet10).
-
 - The ReadCubit Genesis mesh importer has been rewritten to improve readability.

 Discretization improvements
@@ -177,15 +97,6 @@ New and updated examples and miniapps
 - Added two new example codes: 38 and 39/39p described above. Substantially
  updated Example 18/18p.

- Added ODE solvers selection routines. This creates a uniformity across examples,
-  miniapps and other executables in regard to ODE(time-integrator) selection.
-
- Added new mechanism for retrieving and setting state vectors in ODE solvers.
-  This is relevant for AB/AM and gen-alpha solvers.
-
- Added ODEsolver/ODEsolver2 unit tests to verify order of convergence and
-  read/write functionality.
-
 Miscellaneous
 -------------
 - Updated the Doxygen documentation style, which now requires Doxygen version
@@ -146,9 +146,7 @@ if (MFEM_USE_CUDA)
   set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} ${CUDA_FLAGS}")
   find_package(CUDAToolkit REQUIRED)
   set(CUSPARSE_FOUND TRUE)
-   set(CUBLAS_FOUND TRUE)
   get_target_property(CUSPARSE_LIBRARIES CUDA::cusparse LOCATION)
-   get_target_property(CUBLAS_LIBRARIES CUDA::cublas LOCATION)
 endif()

 if (XSDK_ENABLE_C)
@@ -233,7 +231,6 @@ if (MFEM_USE_HIP)
    list(INSERT CMAKE_PREFIX_PATH 0 ${ROCM_PATH})
  endif()
  find_package(HIP REQUIRED)
-  find_package(HIPBLAS REQUIRED)
  find_package(HIPSPARSE REQUIRED)
 endif()

@@ -340,10 +337,7 @@ if (MFEM_USE_SUNDIALS)
  if (MFEM_USE_HIP)
    list(APPEND SUNDIALS_COMPONENTS NVector_Hip)
  endif()
-  # The Core component was added in SUNDIALS v7, so we treat it as optional in
-  # order to support older versions.
-  find_package(SUNDIALS REQUIRED ${SUNDIALS_COMPONENTS}
-    OPTIONAL_COMPONENTS Core)
+  find_package(SUNDIALS REQUIRED ${SUNDIALS_COMPONENTS})
 endif()

 # SuperLU_DIST can only be enabled in parallel
@@ -402,10 +396,6 @@ if (MFEM_USE_AMGX)
  find_package(AMGX REQUIRED)
 endif()

-if (MFEM_USE_MAGMA)
-  find_package(MAGMA REQUIRED)
-endif()
-
 if (MFEM_USE_CONDUIT)
    find_package(Conduit REQUIRED conduit relay blueprint)
 endif()
@@ -567,9 +557,8 @@ find_package(Threads REQUIRED)
 set(MFEM_TPLS OPENMP HYPRE LAPACK BLAS SuperLUDist STRUMPACK METIS SuiteSparse
    SUNDIALS PETSC SLEPC MUMPS AXOM FMS CONDUIT Ginkgo GNUTLS GSLIB
    NETCDF MPFR PUMI HIOP POSIXCLOCKS MFEMBacktrace ZLIB OCCA CEED RAJA UMPIRE
-    ADIOS2 MKL_CPARDISO MKL_PARDISO AMGX MAGMA CUSPARSE CUBLAS CALIPER CODIPACK
-    BENCHMARK PARELAG TRIBOL MPI_CXX HIP HIPBLAS HIPSPARSE MOONOLITH BLITZ
-    ALGOIM ENZYME)
+    ADIOS2 CUSPARSE MKL_CPARDISO MKL_PARDISO AMGX CALIPER CODIPACK
+    BENCHMARK PARELAG TRIBOL MPI_CXX HIP HIPSPARSE MOONOLITH BLITZ ALGOIM ENZYME)

 # Add all *_FOUND libraries in the variable TPL_LIBRARIES.
 set(TPL_LIBRARIES "")
@@ -684,7 +673,7 @@ if (NOT ("${PROJECT_SOURCE_DIR}" STREQUAL "${PROJECT_BINARY_DIR}"))
 #include \"${PROJECT_SOURCE_DIR}/${Header}\"
 ")

-    execute_process(COMMAND ${CMAKE_COMMAND} -E copy_if_different
+    execute_process(COMMAND ${CMAKE_COMMAND} -E copy_if_different 
      "${PROJECT_BINARY_DIR}/${Header}.tmp"
      "${PROJECT_BINARY_DIR}/${Header}"
    )
@@ -698,7 +687,7 @@ if (NOT ("${PROJECT_SOURCE_DIR}" STREQUAL "${PROJECT_BINARY_DIR}"))
 #include \"mfem/${Header}\"
 ")

-    execute_process(COMMAND ${CMAKE_COMMAND} -E copy_if_different
+    execute_process(COMMAND ${CMAKE_COMMAND} -E copy_if_different 
      "${PROJECT_BINARY_DIR}/InstallHeaders/${Header}.tmp"
      "${PROJECT_BINARY_DIR}/InstallHeaders/${Header}"
    )
@@ -273,13 +273,7 @@ Installation options:
   PREFIX  - Specify the installation directory. The library (libmfem.a) will be
             installed in $(PREFIX)/lib, the headers in $(PREFIX)/include, and
             the configuration makefile (config.mk) in $(PREFIX)/share/mfem.
-   INSTALL - Specify the install program, default = /usr/bin/install
-   INSTALL_DEF_PERM - Specify the default install permissions. This affects
-                      headers and configuration makefiles, default = 644
-   INSTALL_BIN_PERM - Specify the install permissions for binaries. This only
-                      affects the shared version of the library, default = 755
-   INSTALL_DIR_PERM - Specify the install permissions for directories and,
-                      on macOS/BSD, for symlinks as well, default = 755
+   INSTALL - Specify the install program, e.g /usr/bin/install

 MFEM library features/options (GNU make)
 ----------------------------------------
@@ -394,11 +388,6 @@ MFEM_USE_AMGX = YES/NO
   Allows the user to use SparseMatrices and HypreParMatrices to solve linear
   systems with the routines from the AmgX library.

-MFEM_USE_MAGMA = YES/NO
-   Enable MFEM functionality based on the MAGMA high-performance linear algebra
-   library. The MAGMA library provides a BLAS/LAPACK interface, with
-   implementations that have been optimized for Nvidia and AMD GPUs.
-
 MFEM_USE_GNUTLS = YES/NO
   Enable secure socket support in class socketstream, using the auxiliary
   GnuTLS_* classes, based on the GnuTLS library. This option may be useful in
@@ -502,14 +491,10 @@ MFEM_USE_CODIPACK = YES/NO
 MFEM_USE_ALGOIM = YES/NO
  Enable the usage of Algoim - a collection of high-order accurate numerical
  methods and C++ algorithms for working with implicitly-defined geometry and
-  level set methods, see https://algoim.github.io. MFEM provides interface to
-  Algoim v1. To check out the specific Algoim state use:
-  https://github.com/algoim/algoim
+  level set methods. The Algoim library requires the Blitz++ library. The MFEM
+  provides interface to Algoim v1. Thus, to check out the specific state use:
  git checkout 9c9ca0ef094d8ab0390ed36367a1151b459bbe0a
-  The Algoim library requires the Blitz++ library. To use the latest state of
-  Blitz++ that has been tested with MFEM, use:
-  https://github.com/blitzpp/blitz
-  git checkout f24a250a43dff88c31ad92916da828b7ea9a98b7
+  https://algoim.github.io

 MFEM_USE_ADFORWARD = YES/NO
  Enable forward mode for AD packages. This option is valid
@@ -714,11 +699,6 @@ The specific libraries and their options are:
  Options: AMGX_OPT, AMGX_LIB.
  Versions: AmgX >= 2.1, older versions may work too.

- MAGMA (optional), used with MFEM_USE_MAGMA = YES.
-  URL: https://icl.utk.edu/magma/
-  Options: MAGMA_OPT, MAGMA_LIB
-  Versions: MAGMA >= 2.8.0
-
 - GnuTLS (optional), used when MFEM_USE_GNUTLS = YES. On most Linux systems,
  GnuTLS is available as a development package, e.g. gnutls-devel. On Mac OS X,
  one can get the library through the Homebrew package manager (http://brew.sh).
@@ -37,7 +37,6 @@ set(MFEM_USE_MUMPS @MFEM_USE_MUMPS@)
 set(MFEM_USE_STRUMPACK @MFEM_USE_STRUMPACK@)
 set(MFEM_USE_GINKGO @MFEM_USE_GINKGO@)
 set(MFEM_USE_AMGX @MFEM_USE_AMGX@)
-set(MFEM_USE_MAGMA @MFEM_USE_MAGMA@)
 set(MFEM_USE_HIOP @MFEM_USE_HIOP@)
 set(MFEM_USE_GNUTLS @MFEM_USE_GNUTLS@)
 set(MFEM_USE_GSLIB @MFEM_USE_GSLIB@)
@@ -114,9 +114,6 @@
 // Enable MFEM functionality based on the AmgX library.
 #cmakedefine MFEM_USE_AMGX

-// Enable MFEM functionality based on the MAGMA library.
-#cmakedefine MFEM_USE_MAGMA
-
 // Enable secure socket streams based on the GNUTLS library.
 #cmakedefine MFEM_USE_GNUTLS

@@ -1,37 +0,0 @@
-# Copyright (c) 2010-2024, Lawrence Livermore National Security, LLC. Produced
-# at the Lawrence Livermore National Laboratory. All Rights reserved. See files
-# LICENSE and NOTICE for details. LLNL-CODE-806117.
-#
-# This file is part of the MFEM library. For more information and source code
-# availability visit https://mfem.org.
-#
-# MFEM is free software; you can redistribute it and/or modify it under the
-# terms of the BSD-3 license. We welcome feedback and contributions, see file
-# CONTRIBUTING.md for details.
-
-# Defines the following variables:
-#   - MAGMA_FOUND
-#   - MAGMA_LIBRARIES
-#   - MAGMA_INCLUDE_DIRS
-
-include(MfemCmakeUtilities)
-mfem_find_package(MAGMA MAGMA MAGMA_DIR "include" "magma.h" "lib" "magma"
-  "Paths to headers required by MAGMA." "Libraries required by MAGMA.")
-
-if (MAGMA_FOUND AND MFEM_USE_CUDA)
-  get_target_property(CUSPARSE_LIBRARIES CUDA::cusparse LOCATION)
-  get_target_property(CUBLAS_LIBRARIES CUDA::cublas LOCATION)
-  list(APPEND MAGMA_LIBRARIES ${CUSPARSE_LIBRARIES} ${CUBLAS_LIBRARIES})
-  set(MAGMA_LIBRARIES ${MAGMA_LIBRARIES} CACHE STRING
-      "MAGMA libraries + dependencies." FORCE)
-  message(STATUS "Updated MAGMA_LIBRARIES: ${MAGMA_LIBRARIES}")
-endif()
-
-if (MAGMA_FOUND AND MFEM_USE_HIP)
-  find_package(HIPBLAS REQUIRED)
-  find_package(HIPSPARSE REQUIRED)
-  list(APPEND MAGMA_LIBRARIES ${HIPBLAS_LIBRARIES} ${HIPSPARSE_LIBRARIES})
-  set(MAGMA_LIBRARIES ${MAGMA_LIBRARIES} CACHE STRING
-      "MAGMA libraries + dependencies." FORCE)
-  message(STATUS "Updated MAGMA_LIBRARIES: ${MAGMA_LIBRARIES}")
-endif()
@@ -31,5 +31,4 @@ mfem_find_package(SUNDIALS SUNDIALS SUNDIALS_DIR
  ADD_COMPONENT CVODE "include" cvode/cvode.h "lib" sundials_cvode
  ADD_COMPONENT CVODES "include" cvodes/cvodes.h "lib" sundials_cvodes
  ADD_COMPONENT ARKODE "include" arkode/arkode.h "lib" sundials_arkode
-  ADD_COMPONENT KINSOL "include" kinsol/kinsol.h "lib" sundials_kinsol
-  ADD_COMPONENT Core "include" sundials/sundials_core.h "lib" sundials_core)
+  ADD_COMPONENT KINSOL "include" kinsol/kinsol.h "lib" sundials_kinsol)
@@ -846,14 +846,14 @@ function(mfem_export_mk_files)
      MFEM_USE_ZLIB MFEM_USE_LIBUNWIND MFEM_USE_LAPACK MFEM_THREAD_SAFE
      MFEM_USE_LEGACY_OPENMP MFEM_USE_OPENMP MFEM_USE_MEMALLOC MFEM_USE_SUNDIALS
      MFEM_USE_SUITESPARSE MFEM_USE_SUPERLU MFEM_USE_SUPERLU5 MFEM_USE_MUMPS
-      MFEM_USE_STRUMPACK MFEM_USE_GINKGO MFEM_USE_AMGX MFEM_USE_MAGMA
-      MFEM_USE_GNUTLS MFEM_USE_NETCDF MFEM_USE_PETSC MFEM_USE_SLEPC
-      MFEM_USE_MPFR MFEM_USE_SIDRE MFEM_USE_FMS MFEM_USE_CONDUIT MFEM_USE_PUMI
-      MFEM_USE_HIOP MFEM_USE_GSLIB MFEM_USE_CUDA MFEM_USE_HIP MFEM_USE_RAJA
-      MFEM_USE_OCCA MFEM_USE_CEED MFEM_USE_CALIPER MFEM_USE_UMPIRE MFEM_USE_SIMD
-      MFEM_USE_ADIOS2 MFEM_USE_MKL_CPARDISO MFEM_USE_MKL_PARDISO
-      MFEM_USE_ADFORWARD MFEM_USE_CODIPACK MFEM_USE_BENCHMARK MFEM_USE_PARELAG
-      MFEM_USE_TRIBOL MFEM_USE_MOONOLITH MFEM_USE_ALGOIM MFEM_USE_ENZYME)
+      MFEM_USE_STRUMPACK MFEM_USE_GINKGO MFEM_USE_AMGX MFEM_USE_GNUTLS
+      MFEM_USE_NETCDF MFEM_USE_PETSC MFEM_USE_SLEPC MFEM_USE_MPFR MFEM_USE_SIDRE
+      MFEM_USE_FMS MFEM_USE_CONDUIT MFEM_USE_PUMI MFEM_USE_HIOP MFEM_USE_GSLIB
+      MFEM_USE_CUDA MFEM_USE_HIP MFEM_USE_RAJA MFEM_USE_OCCA MFEM_USE_CEED
+      MFEM_USE_CALIPER MFEM_USE_UMPIRE MFEM_USE_SIMD MFEM_USE_ADIOS2
+      MFEM_USE_MKL_CPARDISO MFEM_USE_MKL_PARDISO MFEM_USE_ADFORWARD
+      MFEM_USE_CODIPACK MFEM_USE_BENCHMARK MFEM_USE_PARELAG MFEM_USE_TRIBOL
+      MFEM_USE_MOONOLITH MFEM_USE_ALGOIM MFEM_USE_ENZYME)
  foreach(var ${CONFIG_MK_BOOL_VARS})
    if (${var})
      set(${var} YES)
@@ -114,9 +114,6 @@
 // Enable MFEM functionality based on the AmgX library.
 // #define MFEM_USE_AMGX

-// Enable MFEM functionality based on the MAGMA library.
-// #define MFEM_USE_MAGMA
-
 // Enable secure socket streams based on the GNUTLS library.
 // #define MFEM_USE_GNUTLS

@@ -38,7 +38,6 @@ MFEM_USE_MUMPS         = @MFEM_USE_MUMPS@
 MFEM_USE_STRUMPACK     = @MFEM_USE_STRUMPACK@
 MFEM_USE_GINKGO        = @MFEM_USE_GINKGO@
 MFEM_USE_AMGX          = @MFEM_USE_AMGX@
-MFEM_USE_MAGMA         = @MFEM_USE_MAGMA@
 MFEM_USE_GNUTLS        = @MFEM_USE_GNUTLS@
 MFEM_USE_NETCDF        = @MFEM_USE_NETCDF@
 MFEM_USE_PETSC         = @MFEM_USE_PETSC@
@@ -40,7 +40,6 @@ option(MFEM_USE_MUMPS "Enable MUMPS usage" OFF)
 option(MFEM_USE_STRUMPACK "Enable STRUMPACK usage" OFF)
 option(MFEM_USE_GINKGO "Enable Ginkgo usage" OFF)
 option(MFEM_USE_AMGX "Enable AmgX usage" OFF)
-option(MFEM_USE_MAGMA "Enable MAGMA usage" OFF)
 option(MFEM_USE_GNUTLS "Enable GNUTLS usage" OFF)
 option(MFEM_USE_GSLIB "Enable GSLIB usage" OFF)
 option(MFEM_USE_NETCDF "Enable NETCDF usage" OFF)
@@ -184,10 +183,6 @@ set(Ginkgo_DIR "${MFEM_DIR}/../ginkgo" CACHE PATH "Path to the Ginkgo library.")

 set(AMGX_DIR "${MFEM_DIR}/../amgx" CACHE PATH "Path to AmgX")

-set(MAGMA_DIR "${MFEM_DIR}/../magma" CACHE PATH "Path to MAGMA")
-set(MAGMA_REQUIRED_PACKAGES "BLAS" "LAPACK" CACHE STRING
-    "Additional packages required by MAGMA.")
-
 set(GNUTLS_DIR "" CACHE PATH "Path to the GnuTLS library.")

 set(GSLIB_DIR "" CACHE PATH "Path to the GSLIB library.")
@@ -264,7 +259,7 @@ set(PARELAG_LIBRARIES "${PARELAG_DIR}/build/src/libParELAG.a" CACHE STRING
    "The ParELAG library.")

 set(TRIBOL_DIR "${MFEM_DIR}/../tribol" CACHE PATH "Path to Tribol")
-set(Tribol_REQUIRED_PACKAGES "Axom/core/mint/slam/slic" CACHE STRING
+set(Tribol_REQUIRED_PACKAGES "Axom/core/mint/slam/slic" CACHE STRING 
    "Additional packages required by Tribol")

 set(BLAS_INCLUDE_DIRS "" CACHE STRING "Path to BLAS headers.")
@@ -95,10 +95,6 @@ else
   # Silence unused command line argument warnings when generating dependencies
   # with mpicxx and clang
   DEP_FLAGS := -Wno-unused-command-line-argument $(DEP_FLAGS)
-   # Silence "ignoring duplicate libraries" warnings on new (Xcode 15) linker
-   ifneq (,$(findstring PROJECT:dyld,$(shell ld -v 2>&1)))
-      LDFLAGS_INTERNAL = -Xlinker -no_warn_duplicate_libraries
-   endif
 endif

 # Set CXXFLAGS to overwrite the default selection of DEBUG_FLAGS/OPTIM_FLAGS
@@ -143,7 +139,6 @@ MFEM_USE_MUMPS         = NO
 MFEM_USE_STRUMPACK     = NO
 MFEM_USE_GINKGO        = NO
 MFEM_USE_AMGX          = NO
-MFEM_USE_MAGMA         = NO
 MFEM_USE_GNUTLS        = NO
 MFEM_USE_NETCDF        = NO
 MFEM_USE_PETSC         = NO
@@ -289,13 +284,6 @@ endif
 ifeq ($(MFEM_USE_HIP),YES)
   SUNDIALS_LIB += -lsundials_nvechip
 endif
-SUNDIALS_CORE_PAT = $(subst\
- @MFEM_DIR@,$(MFEM_DIR),$(SUNDIALS_DIR))/lib*/libsundials_core.*
-ifeq ($(MFEM_USE_SUNDIALS),YES)
-   ifneq ($(wildcard $(SUNDIALS_CORE_PAT)),)
-      SUNDIALS_LIB += -lsundials_core
-   endif
-endif
 # If SUNDIALS was built with KLU:
 # MFEM_USE_SUITESPARSE = YES

@@ -402,11 +390,6 @@ AMGX_DIR = @MFEM_DIR@/../amgx
 AMGX_OPT = -I$(AMGX_DIR)/include
 AMGX_LIB = -L$(AMGX_DIR)/lib -lamgx -lcusparse -lcusolver -lcublas -lnvToolsExt

-# MAGMA library configuration
-MAGMA_DIR = @MFEM_DIR@/../magma
-MAGMA_OPT = -I$(MAGMA_DIR)/include
-MAGMA_LIB = -L$(MAGMA_DIR)/lib -l:libmagma.a -lcublas -lcusparse $(LAPACK_LIB)
-
 # GnuTLS library configuration
 GNUTLS_OPT =
 GNUTLS_LIB = -lgnutls
@@ -514,11 +497,11 @@ GSLIB_LIB = -L$(GSLIB_DIR)/lib -lgs

 # CUDA library configuration
 CUDA_OPT =
-CUDA_LIB = -lcusparse -lcublas
+CUDA_LIB = -lcusparse

 # HIP library configuration
 HIP_OPT =
-HIP_LIB = -L$(HIP_DIR)/lib $(XLINKER)-rpath,$(HIP_DIR)/lib -lhipsparse -lhipblas
+HIP_LIB = -L$(HIP_DIR)/lib $(XLINKER)-rpath,$(HIP_DIR)/lib -lhipsparse

 # OCCA library configuration
 OCCA_DIR = @MFEM_DIR@/../occa
@@ -540,10 +523,8 @@ ifdef GOTCHA_DIR
 endif

 # BLITZ library configuration
-# BLITZ_DIR must be the custom installation folder (-DCMAKE_INSTALL_PREFIX).
-BLITZ_DIR = @MFEM_DIR@/../blitz/install
+BLITZ_DIR = @MFEM_DIR@/../blitz
 BLITZ_OPT = -I$(BLITZ_DIR)/include
-# On intel machines, use /lib64 instead of /lib.
 BLITZ_LIB = $(XLINKER)-rpath,$(BLITZ_DIR)/lib -L$(BLITZ_DIR)/lib -lblitz

 # ALGOIM library configuration
@@ -32,7 +32,7 @@ groups_serial=(
 '"examples"
   "Examples:"
   "examples"
-   "ex{,[1-9]}[0-9].cpp"'
+   "ex{,1,2,3}[0-9].cpp"'
 #   "ex1.cpp"'
 '"sundials"
   "SUNDIALS examples:"
@@ -58,10 +58,6 @@ groups_serial=(
   "HiOp examples:"
   "examples/hiop"
   "ex9.cpp"'
-'"moonolith"
-   "Moonolith examples:"
-   "examples/moonolith"
-   "ex1.cpp"'
 '"pumi"
   "PUMI examples:"
   "examples/pumi"
@@ -70,38 +66,25 @@ groups_serial=(
 '"meshing"
   "Meshing miniapps:"
   "miniapps/meshing"
-   "mobius-strip.cpp klein-bottle.cpp extruder.cpp toroid.cpp mesh-quality.cpp
-    polar-nc.cpp reflector.cpp shaper.cpp trimmer.cpp twist.cpp
+   "mobius-strip.cpp klein-bottle.cpp extruder.cpp toroid.cpp
    mesh-optimizer.cpp minimal-surface.cpp"'
 '"adjoint"
   "Adjoint miniapps:"
   "miniapps/adjoint"
   "cvsRoberts_ASAi_dns.cpp"'
-'"autodiff"
-   "Autodiff miniapps:"
-   "miniapps/autodiff"
-   "seq_example.cpp seq_test.cpp"' # 'seq_test.cpp' has no sample runs
-'"dpg"
-   "DPG miniapps:"
-   "miniapps/dpg"
-   "{acoustics,convection-diffusion,diffusion,maxwell}.cpp"'
 '"gslib"
   "GSLIB miniapps:"
   "miniapps/gslib"
   "field-diff.cpp field-interp.cpp findpts.cpp schwarz_ex1.cpp "'
-# todo: miniapps/mtop
 '"nurbs"
   "NURBS miniapps:"
   "miniapps/nurbs"
   "nurbs_ex1.cpp"'
-# todo: add other nurbs miniapps
-# todo: miniapps/solvers (serial)
 '"tools"
   "Tools miniapps:"
   "miniapps/tools"
   "convert-dc.cpp display-basis.cpp get-values.cpp load-dc.cpp
    lor-transfer.cpp"'
-# todo: add other tools miniapps
 '"toys"
   "Toys miniapps:"
   "miniapps/toys"
@@ -117,7 +100,7 @@ groups_parallel=(
 '"examples"
   "Examples:"
   "examples"
-   "ex{,[1-9]}[0-9]p.cpp"'
+   "ex{,1,2,3}[0-9]p.cpp"'
 #   "ex1p.cpp"'
 '"sundials"
   "SUNDIALS examples:"
@@ -143,10 +126,6 @@ groups_parallel=(
   "HiOp examples:"
   "examples/hiop"
   "ex9p.cpp"'
-'"moonolith"
-   "Moonolith examples:"
-   "examples/moonolith"
-   "ex{1,2}p.cpp"'
 '"pumi"
   "PUMI examples:"
   "examples/pumi"
@@ -159,41 +138,24 @@ groups_parallel=(
 '"meshing"
   "Meshing miniapps:"
   "miniapps/meshing"
-   "pmesh-optimizer.cpp pmesh-fitting.cpp pminimal-surface.cpp
-    fit-node-position.cpp"'
+   "pmesh-optimizer.cpp pmesh-fitting.cpp pminimal-surface.cpp"'
 '"electromagnetics"
   "Electromagnetics miniapps:"
   "miniapps/electromagnetics"
   "joule.cpp"'
-#   "{joule,maxwell,tesla,volta}.cpp"' # todo: multiline sample runs
+#   "{volta,tesla,joule}.cpp"' # todo: multiline sample runs
 '"adjoint"
   "Adjoint miniapps:"
   "miniapps/adjoint"
   "adjoint_advection_diffusion.cpp"'
-'"autodiff"
-   "Autodiff miniapps:"
-   "miniapps/autodiff"
-   "par_example.cpp"'
-'"dpg"
-   "DPG miniapps:"
-   "miniapps/dpg"
-   "p{acoustics,convection-diffusion,diffusion,maxwell}.cpp"'
 '"gslib"
   "GSLIB miniapps:"
   "miniapps/gslib"
   "pfindpts.cpp schwarz_ex1p.cpp"'
-'"hdiv-linear-solver"
-   "H(div) linear solver miniapps:"
-   "miniapps/hdiv-linear-solver"
-   "grad_div.cpp darcy.cpp"'
-# 'miniapps/hooke/hooke.cpp' has no sample runs
-# todo: miniapps/mtop
-# todo: miniapps/multidomain
 '"navier"
   "Navier miniapps:"
   "miniapps/navier"
   "navier_cht.cpp"'
-# todo: add other navier miniapps
 '"nurbs"
   "NURBS miniapps:"
   "miniapps/nurbs"
@@ -202,18 +164,14 @@ groups_parallel=(
   "Shifted miniapps:"
   "miniapps/shifted"
   "distance.cpp"'
-# todo: add other shifted miniapps
 '"solvers"
   "Solvers miniapps:"
   "miniapps/solvers"
   "block-solvers.cpp"'
-# todo: add other solvers miniapps
-# todo: miniapps/spde
 '"tools"
   "Tools miniapps:"
   "miniapps/tools"
-   "convert-dc.cpp get-values.cpp load-dc.cpp"'
-# todo: add other tools miniapps
+   "convert-cd.cpp get-values.cpp load-dc.cpp"'
 '"convergence"
   "Convergence tests:"
   "tests/convergence"
@@ -228,7 +186,7 @@ groups_all=(
 '"examples"
   "Examples:"
   "examples"
-   "ex\"{,[1-9]}[0-9]\"{,p}.cpp"'
+   "ex\"{,1,2,3}[0-9]\"{,p}.cpp"'
 '"sundials"
   "SUNDIALS examples:"
   "examples/sundials"
@@ -257,14 +215,10 @@ groups_all=(
   "HiOp examples:"
   "examples/hiop"
   "ex9.cpp ex9p.cpp"'
-'"moonolith"
-   "Moonolith examples:"
-   "examples/moonolith"
-   "ex1.cpp ex{1,2}p.cpp"'
 '"pumi"
   "PUMI examples:"
   "examples/pumi"
-   "ex1.cpp ex2.cpp ex1p.cpp ex6p.cpp"'
+   "ex1.cpp ex1p.cpp ex2.cpp ex6p.cpp"'
 '"superlu"
   "Superlu examples:"
   "examples/superlu"
@@ -272,67 +226,43 @@ groups_all=(
 '"meshing"
   "Meshing miniapps:"
   "miniapps/meshing"
-   "mobius-strip.cpp klein-bottle.cpp extruder.cpp toroid.cpp mesh-quality.cpp
-    polar-nc.cpp reflector.cpp shaper.cpp trimmer.cpp twist.cpp
-    {,p}mesh-optimizer.cpp pmesh-fitting.cpp {,p}minimal-surface.cpp
-    fit-node-position.cpp"'
+   "mobius-strip.cpp klein-bottle.cpp extruder.cpp toroid.cpp
+    {,p}mesh-optimizer.cpp pmesh-fitting.cpp {,p}minimal-surface.cpp"'
 '"electromagnetics"
   "Electromagnetics miniapps:"
   "miniapps/electromagnetics"
   "joule.cpp"'
-#   "{joule,maxwell,tesla,volta}.cpp"' # todo: multiline sample runs
+#   "{volta,tesla,joule}.cpp"' # todo: multiline sample runs
 '"adjoint"
   "Adjoint miniapps:"
   "miniapps/adjoint"
-   "cvsRoberts_ASAi_dns.cpp adjoint_advection_diffusion.cpp"'
-'"autodiff"
-   "Autodiff miniapps:"
-   "miniapps/autodiff"
-   "seq_example.cpp seq_test.cpp par_example.cpp"'
-# 'seq_test.cpp' has no sample runs
-'"dpg"
-   "DPG miniapps:"
-   "miniapps/dpg"
-   "{,p}{acoustics,convection-diffusion,diffusion,maxwell}.cpp"'
+   "adjoint_advection_diffusion.cpp cvsRoberts_ASAi_dns.cpp"'
 '"gslib"
   "GSLIB miniapps:"
   "miniapps/gslib"
   "field-diff.cpp field-interp.cpp findpts.cpp schwarz_ex1.cpp pfindpts.cpp
    schwarz_ex1p.cpp"'
-'"hdiv-linear-solver"
-   "H(div) linear solver miniapps:"
-   "miniapps/hdiv-linear-solver"
-   "grad_div.cpp darcy.cpp"'
-# 'miniapps/hooke/hooke.cpp' has no sample runs
-# todo: miniapps/mtop
-# todo: miniapps/multidomain
 '"navier"
   "Navier miniapps:"
   "miniapps/navier"
   "navier_cht.cpp"'
-# todo: add other navier miniapps
 '"nurbs"
   "NURBS miniapps:"
   "miniapps/nurbs"
   "nurbs_ex1.cpp nurbs_ex1p.cpp nurbs_ex11p.cpp"'
-# todo: add other nurbs miniapps
 '"shifted"
   "Shifted miniapps:"
   "miniapps/shifted"
   "distance.cpp"'
-# todo: add other shifted miniapps
 '"solvers"
   "Solvers miniapps:"
   "miniapps/solvers"
   "block-solvers.cpp"'
-# todo: add other solvers miniapps
-# todo: miniapps/spde
 '"tools"
   "Tools miniapps:"
   "miniapps/tools"
   "convert-dc.cpp display-basis.cpp get-values.cpp load-dc.cpp
    lor-transfer.cpp"'
-# todo: add other tools miniapps
 '"toys"
   "Toys miniapps:"
   "miniapps/toys"
@@ -456,7 +386,7 @@ function help_message()
      mfem_config [${mfem_config}]
         Set MFEM configuration options
      make [${make}], mpiexec [${mpiexec}], mpiexec_np [${mpiexec_np}]
-         Their values can also be set using the respective uppercase environment
+         Their values can also set using the respective uppercase environment
         variable
      mfem_build_dir [${mfem_build_dir}]
         Same as '-d': set this variable to something different from <mfem_dir>
@@ -18,9 +18,9 @@ elements
 boundary
 4
 1 1 0 1
-2 1 2 3
-3 1 3 0
-4 1 1 2
+1 1 2 3
+1 1 3 0
+1 1 1 2

 edges
 4
@@ -938,7 +938,6 @@ INPUT                  = @MFEM_SOURCE_DIR@/doc/CodeDocumentation.dox \
                         @MFEM_SOURCE_DIR@/config \
                         @MFEM_SOURCE_DIR@/general \
                         @MFEM_SOURCE_DIR@/linalg \
-                         @MFEM_SOURCE_DIR@/linalg/batched \
                         @MFEM_SOURCE_DIR@/linalg/simd \
                         @MFEM_SOURCE_DIR@/mesh \
                         @MFEM_SOURCE_DIR@/mesh/submesh \
@@ -1050,8 +1049,7 @@ RECURSIVE              = NO
 EXCLUDE                = @MFEM_SOURCE_DIR@/config/_config.hpp \
                         @MFEM_SOURCE_DIR@/config/get_hypre_version.cpp \
                         @MFEM_SOURCE_DIR@/general/tinyxml2.h \
-                         @MFEM_SOURCE_DIR@/general/tinyxml2.cpp \
-                         @MFEM_SOURCE_DIR@/linalg/lapack.hpp
+                         @MFEM_SOURCE_DIR@/general/tinyxml2.cpp

 # The EXCLUDE_SYMLINKS tag can be used to select whether or not files or
 # directories that are symbolic links (a Unix file system feature) are excluded
@@ -182,21 +182,6 @@ namespace mfem {
 *   <a class="el" href="examples_2superlu_2ex1p_8cpp_source.html">1p</a>,
 *   demonstrating the use of MFEM's \link superlu.hpp SuperLU integration\endlink.
 *
- * <H4>NURBS Examples</H4>
- * - Variants of Examples
- *  <a class="el" href="nurbs__ex1_8cpp_source.html">1</a>,
- *  <a class="el" href="nurbs__ex1p_8cpp_source.html">1p</a>,
- *  <a class="el" href="nurbs__ex3_8cpp_source.html">3</a>,
- *  <a class="el" href="nurbs__ex5_8cpp_source.html">5</a>,
- *  <a class="el" href="nurbs__ex11p_8cpp_source.html">11p</a>, and
- *  <a class="el" href="nurbs__ex24_8cpp_source.html">24</a>,
- *   demonstrating howto perform NURBS-based Isogeometric Analysis.
- * - Variant of Example <a class="el" href="nurbs__patch__ex1_8cpp_source.html">1</a>: demonstrates the use of patch integration
- * - <a class="el" href="nurbs__solenoidal_8cpp_source.html">NURBS Divergence-free</a>: solve a solenoidal vector projection with NURBS-based H(div) elements
- * - <a class="el" href="nurbs__curveint_8cpp_source.html">NURBS Interpolation</a>: NURBS interpolation of given geometry
- * - <a class="el" href="nurbs__naca__cmesh_8cpp_source.html">NURBS NACA Mesher</a>: generate NURBS based mesh around a NACA foil
- * - <a class="el" href="nurbs__printfunc_8cpp_source.html">NURBS Printer</a>: print the NURBS-basis
- *
 * <H3>Miniapps</H3>
 * - <a class="el" href="volta_8cpp_source.html">Volta</a>: simple electrostatics simulation code
 * - <a class="el" href="tesla_8cpp_source.html">Tesla</a>: simple magnetostatics simulation code
@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/amgx/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/caliper,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -3,14 +3,14 @@
 // Compile with: make ex10
 //
 // Sample runs:
-//    ex10 -m ../data/beam-quad.mesh -s 23 -r 2 -o 2 -dt 3
-//    ex10 -m ../data/beam-tri.mesh -s 23 -r 2 -o 2 -dt 3
-//    ex10 -m ../data/beam-hex.mesh -s 22 -r 1 -o 2 -dt 3
-//    ex10 -m ../data/beam-tet.mesh -s 22 -r 1 -o 2 -dt 3
-//    ex10 -m ../data/beam-wedge.mesh -s 22 -r 1 -o 2 -dt 3
-//    ex10 -m ../data/beam-quad.mesh -s 4 -r 2 -o 2 -dt 0.03 -vs 20
-//    ex10 -m ../data/beam-hex.mesh -s 4 -r 1 -o 2 -dt 0.05 -vs 20
-//    ex10 -m ../data/beam-quad-amr.mesh -s 23 -r 2 -o 2 -dt 3
+//    ex10 -m ../data/beam-quad.mesh -s 3 -r 2 -o 2 -dt 3
+//    ex10 -m ../data/beam-tri.mesh -s 3 -r 2 -o 2 -dt 3
+//    ex10 -m ../data/beam-hex.mesh -s 2 -r 1 -o 2 -dt 3
+//    ex10 -m ../data/beam-tet.mesh -s 2 -r 1 -o 2 -dt 3
+//    ex10 -m ../data/beam-wedge.mesh -s 2 -r 1 -o 2 -dt 3
+//    ex10 -m ../data/beam-quad.mesh -s 14 -r 2 -o 2 -dt 0.03 -vs 20
+//    ex10 -m ../data/beam-hex.mesh -s 14 -r 1 -o 2 -dt 0.05 -vs 20
+//    ex10 -m ../data/beam-quad-amr.mesh -s 3 -r 2 -o 2 -dt 3
 //
 // Description:  This examples solves a time dependent nonlinear elasticity
 //               problem of the form dv/dt = H(x) + S v, dx/dt = v, where H is a
@@ -87,16 +87,16 @@ public:
                        real_t visc, real_t mu, real_t K);

   /// Compute the right-hand side of the ODE system.
-   void Mult(const Vector &vx, Vector &dvx_dt) const override;
+   virtual void Mult(const Vector &vx, Vector &dvx_dt) const;
   /** Solve the Backward-Euler equation: k = f(x + dt*k, t), for the unknown k.
       This is the only requirement for high-order SDIRK implicit integration.*/
-   void ImplicitSolve(const real_t dt, const Vector &x, Vector &k) override;
+   virtual void ImplicitSolve(const real_t dt, const Vector &x, Vector &k);

   real_t ElasticEnergy(const Vector &x) const;
   real_t KineticEnergy(const Vector &v) const;
   void GetElasticEnergyDensity(const GridFunction &x, GridFunction &w) const;

-   ~HyperelasticOperator() override;
+   virtual ~HyperelasticOperator();
 };

 /** Nonlinear operator of the form:
@@ -120,12 +120,12 @@ public:
   void SetParameters(real_t dt_, const Vector *v_, const Vector *x_);

   /// Compute y = H(x + dt (v + dt k)) + M k + S (v + dt k).
-   void Mult(const Vector &k, Vector &y) const override;
+   virtual void Mult(const Vector &k, Vector &y) const;

   /// Compute J = M + dt S + dt^2 grad_H(x + dt (v + dt k)).
-   Operator &GetGradient(const Vector &k) const override;
+   virtual Operator &GetGradient(const Vector &k) const;

-   ~ReducedSystemOperator() override;
+   virtual ~ReducedSystemOperator();
 };


@@ -141,8 +141,8 @@ private:
 public:
   ElasticEnergyCoefficient(HyperelasticModel &m, const GridFunction &x_)
      : model(m), x(x_) { }
-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
-   ~ElasticEnergyCoefficient() override { }
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
+   virtual ~ElasticEnergyCoefficient() { }
 };

 void InitialDeformation(const Vector &x, Vector &y);
@@ -160,7 +160,7 @@ int main(int argc, char *argv[])
   const char *mesh_file = "../data/beam-quad.mesh";
   int ref_levels = 2;
   int order = 2;
-   int ode_solver_type = 23;
+   int ode_solver_type = 3;
   real_t t_final = 300.0;
   real_t dt = 3.0;
   real_t visc = 1e-2;
@@ -177,7 +177,11 @@ int main(int argc, char *argv[])
   args.AddOption(&order, "-o", "--order",
                  "Order (degree) of the finite elements.");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  ODESolver::Types.c_str());
+                  "ODE solver: 1 - Backward Euler, 2 - SDIRK2, 3 - SDIRK3,\n\t"
+                  "            11 - Forward Euler, 12 - RK2,\n\t"
+                  "            13 - RK3 SSP, 14 - RK4."
+                  "            22 - Implicit Midpoint Method,\n\t"
+                  "            23 - SDIRK23 (A-stable), 24 - SDIRK34");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -209,7 +213,28 @@ int main(int argc, char *argv[])
   // 3. Define the ODE solver used for time integration. Several implicit
   //    singly diagonal implicit Runge-Kutta (SDIRK) methods, as well as
   //    explicit Runge-Kutta methods are available.
-   unique_ptr<ODESolver> ode_solver = ODESolver::Select(ode_solver_type);
+   ODESolver *ode_solver;
+   switch (ode_solver_type)
+   {
+      // Implicit L-stable methods
+      case 1: ode_solver = new BackwardEulerSolver; break;
+      case 2: ode_solver = new SDIRK23Solver(2); break;
+      case 3: ode_solver = new SDIRK33Solver; break;
+      // Explicit methods
+      case 11: ode_solver = new ForwardEulerSolver; break;
+      case 12: ode_solver = new RK2Solver(0.5); break; // midpoint method
+      case 13: ode_solver = new RK3SSPSolver; break;
+      case 14: ode_solver = new RK4Solver; break;
+      case 15: ode_solver = new GeneralizedAlphaSolver(0.5); break;
+      // Implicit A-stable methods (not L-stable)
+      case 22: ode_solver = new ImplicitMidpointSolver; break;
+      case 23: ode_solver = new SDIRK23Solver; break;
+      case 24: ode_solver = new SDIRK34Solver; break;
+      default:
+         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         delete mesh;
+         return 3;
+   }

   // 4. Refine the mesh to increase the resolution. In this example we do
   //    'ref_levels' of uniform refinement, where 'ref_levels' is a
@@ -346,6 +371,7 @@ int main(int argc, char *argv[])
   }

   // 10. Free the used memory.
+   delete ode_solver;
   delete mesh;

   return 0;
@@ -3,14 +3,14 @@
 // Compile with: make ex10p
 //
 // Sample runs:
-//    mpirun -np 4 ex10p -m ../data/beam-quad.mesh -s 23 -rs 2 -dt 3
-//    mpirun -np 4 ex10p -m ../data/beam-tri.mesh -s 23 -rs 2 -dt 3
-//    mpirun -np 4 ex10p -m ../data/beam-hex.mesh -s 22 -rs 1 -dt 3
-//    mpirun -np 4 ex10p -m ../data/beam-tet.mesh -s 22 -rs 1 -dt 3
-//    mpirun -np 4 ex10p -m ../data/beam-wedge.mesh -s 22 -rs 1 -dt 3
-//    mpirun -np 4 ex10p -m ../data/beam-quad.mesh -s 4 -rs 2 -dt 0.03 -vs 20
-//    mpirun -np 4 ex10p -m ../data/beam-hex.mesh -s 4 -rs 1 -dt 0.05 -vs 20
-//    mpirun -np 4 ex10p -m ../data/beam-quad-amr.mesh -s 23 -rs 2 -dt 3
+//    mpirun -np 4 ex10p -m ../data/beam-quad.mesh -s 3 -rs 2 -dt 3
+//    mpirun -np 4 ex10p -m ../data/beam-tri.mesh -s 3 -rs 2 -dt 3
+//    mpirun -np 4 ex10p -m ../data/beam-hex.mesh -s 2 -rs 1 -dt 3
+//    mpirun -np 4 ex10p -m ../data/beam-tet.mesh -s 2 -rs 1 -dt 3
+//    mpirun -np 4 ex10p -m ../data/beam-wedge.mesh -s 2 -rs 1 -dt 3
+//    mpirun -np 4 ex10p -m ../data/beam-quad.mesh -s 14 -rs 2 -dt 0.03 -vs 20
+//    mpirun -np 4 ex10p -m ../data/beam-hex.mesh -s 14 -rs 1 -dt 0.05 -vs 20
+//    mpirun -np 4 ex10p -m ../data/beam-quad-amr.mesh -s 3 -rs 2 -dt 3
 //
 // Description:  This examples solves a time dependent nonlinear elasticity
 //               problem of the form dv/dt = H(x) + S v, dx/dt = v, where H is a
@@ -89,17 +89,17 @@ public:
                        real_t visc, real_t mu, real_t K);

   /// Compute the right-hand side of the ODE system.
-   void Mult(const Vector &vx, Vector &dvx_dt) const override;
+   virtual void Mult(const Vector &vx, Vector &dvx_dt) const;
   /** Solve the Backward-Euler equation: k = f(x + dt*k, t), for the unknown k.
       This is the only requirement for high-order SDIRK implicit integration.*/
-   void ImplicitSolve(const real_t dt, const Vector &x, Vector &k) override;
+   virtual void ImplicitSolve(const real_t dt, const Vector &x, Vector &k);

   real_t ElasticEnergy(const ParGridFunction &x) const;
   real_t KineticEnergy(const ParGridFunction &v) const;
   void GetElasticEnergyDensity(const ParGridFunction &x,
                                ParGridFunction &w) const;

-   ~HyperelasticOperator() override;
+   virtual ~HyperelasticOperator();
 };

 /** Nonlinear operator of the form:
@@ -125,12 +125,12 @@ public:
   void SetParameters(real_t dt_, const Vector *v_, const Vector *x_);

   /// Compute y = H(x + dt (v + dt k)) + M k + S (v + dt k).
-   void Mult(const Vector &k, Vector &y) const override;
+   virtual void Mult(const Vector &k, Vector &y) const;

   /// Compute J = M + dt S + dt^2 grad_H(x + dt (v + dt k)).
-   Operator &GetGradient(const Vector &k) const override;
+   virtual Operator &GetGradient(const Vector &k) const;

-   ~ReducedSystemOperator() override;
+   virtual ~ReducedSystemOperator();
 };


@@ -146,8 +146,8 @@ private:
 public:
   ElasticEnergyCoefficient(HyperelasticModel &m, const ParGridFunction &x_)
      : model(m), x(x_) { }
-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
-   ~ElasticEnergyCoefficient() override { }
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
+   virtual ~ElasticEnergyCoefficient() { }
 };

 void InitialDeformation(const Vector &x, Vector &y);
@@ -172,7 +172,7 @@ int main(int argc, char *argv[])
   int ser_ref_levels = 2;
   int par_ref_levels = 0;
   int order = 2;
-   int ode_solver_type = 23;
+   int ode_solver_type = 3;
   real_t t_final = 300.0;
   real_t dt = 3.0;
   real_t visc = 1e-2;
@@ -192,7 +192,11 @@ int main(int argc, char *argv[])
   args.AddOption(&order, "-o", "--order",
                  "Order (degree) of the finite elements.");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  ODESolver::Types.c_str());
+                  "ODE solver: 1 - Backward Euler, 2 - SDIRK2, 3 - SDIRK3,\n\t"
+                  "            11 - Forward Euler, 12 - RK2,\n\t"
+                  "            13 - RK3 SSP, 14 - RK4."
+                  "            22 - Implicit Midpoint Method,\n\t"
+                  "            23 - SDIRK23 (A-stable), 24 - SDIRK34");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -234,7 +238,31 @@ int main(int argc, char *argv[])
   // 4. Define the ODE solver used for time integration. Several implicit
   //    singly diagonal implicit Runge-Kutta (SDIRK) methods, as well as
   //    explicit Runge-Kutta methods are available.
-   unique_ptr<ODESolver> ode_solver = ODESolver::Select(ode_solver_type);
+   ODESolver *ode_solver;
+   switch (ode_solver_type)
+   {
+      // Implicit L-stable methods
+      case 1:  ode_solver = new BackwardEulerSolver; break;
+      case 2:  ode_solver = new SDIRK23Solver(2); break;
+      case 3:  ode_solver = new SDIRK33Solver; break;
+      // Explicit methods
+      case 11: ode_solver = new ForwardEulerSolver; break;
+      case 12: ode_solver = new RK2Solver(0.5); break; // midpoint method
+      case 13: ode_solver = new RK3SSPSolver; break;
+      case 14: ode_solver = new RK4Solver; break;
+      case 15: ode_solver = new GeneralizedAlphaSolver(0.5); break;
+      // Implicit A-stable methods (not L-stable)
+      case 22: ode_solver = new ImplicitMidpointSolver; break;
+      case 23: ode_solver = new SDIRK23Solver; break;
+      case 24: ode_solver = new SDIRK34Solver; break;
+      default:
+         if (myid == 0)
+         {
+            cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         }
+         delete mesh;
+         return 3;
+   }

   // 5. Refine the mesh in serial to increase the resolution. In this example
   //    we do 'ser_ref_levels' of uniform refinement, where 'ser_ref_levels' is
@@ -405,6 +433,7 @@ int main(int argc, char *argv[])
   }

   // 12. Free the used memory.
+   delete ode_solver;
   delete pmesh;

   return 0;
@@ -53,7 +53,7 @@ public:
      pmesh(pmesh_),
      pgf(pgf_) {}

-   void MonitorSolution(int i, real_t norm, const Vector &x, bool final) override
+   void MonitorSolution(int i, real_t norm, const Vector &x, bool final)
   {
      char vishost[] = "localhost";
      int  visport   = 19916;
@@ -5,10 +5,10 @@
 // Sample runs:  ex16
 //               ex16 -m ../data/inline-tri.mesh
 //               ex16 -m ../data/disc-nurbs.mesh -tf 2
-//               ex16 -s 21 -a 0.0 -k 1.0
-//               ex16 -s 22 -a 1.0 -k 0.0
-//               ex16 -s 23 -a 0.5 -k 0.5 -o 4
-//               ex16 -s 4 -dt 1.0e-4 -tf 4.0e-2 -vs 40
+//               ex16 -s 1 -a 0.0 -k 1.0
+//               ex16 -s 2 -a 1.0 -k 0.0
+//               ex16 -s 3 -a 0.5 -k 0.5 -o 4
+//               ex16 -s 14 -dt 1.0e-4 -tf 4.0e-2 -vs 40
 //               ex16 -m ../data/fichera-q2.mesh
 //               ex16 -m ../data/fichera-mixed.mesh
 //               ex16 -m ../data/escher.mesh
@@ -76,15 +76,15 @@ public:
   ConductionOperator(FiniteElementSpace &f, real_t alpha, real_t kappa,
                      const Vector &u);

-   void Mult(const Vector &u, Vector &du_dt) const override;
+   virtual void Mult(const Vector &u, Vector &du_dt) const;
   /** Solve the Backward-Euler equation: k = f(u + dt*k, t), for the unknown k.
       This is the only requirement for high-order SDIRK implicit integration.*/
-   void ImplicitSolve(const real_t dt, const Vector &u, Vector &k) override;
+   virtual void ImplicitSolve(const real_t dt, const Vector &u, Vector &k);

   /// Update the diffusion BilinearForm K using the given true-dof vector `u`.
   void SetParameters(const Vector &u);

-   ~ConductionOperator() override;
+   virtual ~ConductionOperator();
 };

 real_t InitialTemperature(const Vector &x);
@@ -95,13 +95,11 @@ int main(int argc, char *argv[])
   const char *mesh_file = "../data/star.mesh";
   int ref_levels = 2;
   int order = 2;
-
-   int ode_solver_type = 23;  // SDIRK33Solver
+   int ode_solver_type = 3;
   real_t t_final = 0.5;
   real_t dt = 1.0e-2;
   real_t alpha = 1.0e-2;
   real_t kappa = 0.5;
-
   bool visualization = true;
   bool visit = false;
   int vis_steps = 5;
@@ -117,7 +115,8 @@ int main(int argc, char *argv[])
   args.AddOption(&order, "-o", "--order",
                  "Order (degree) of the finite elements.");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  ODESolver::Types.c_str());
+                  "ODE solver: 1 - Backward Euler, 2 - SDIRK2, 3 - SDIRK3,\n\t"
+                  "\t   11 - Forward Euler, 12 - RK2, 13 - RK3 SSP, 14 - RK4.");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -150,7 +149,28 @@ int main(int argc, char *argv[])
   // 3. Define the ODE solver used for time integration. Several implicit
   //    singly diagonal implicit Runge-Kutta (SDIRK) methods, as well as
   //    explicit Runge-Kutta methods are available.
-   unique_ptr<ODESolver> ode_solver = ODESolver::Select(ode_solver_type);
+   ODESolver *ode_solver;
+   switch (ode_solver_type)
+   {
+      // Implicit L-stable methods
+      case 1:  ode_solver = new BackwardEulerSolver; break;
+      case 2:  ode_solver = new SDIRK23Solver(2); break;
+      case 3:  ode_solver = new SDIRK33Solver; break;
+      // Explicit methods
+      case 11: ode_solver = new ForwardEulerSolver; break;
+      case 12: ode_solver = new RK2Solver(0.5); break; // midpoint method
+      case 13: ode_solver = new RK3SSPSolver; break;
+      case 14: ode_solver = new RK4Solver; break;
+      case 15: ode_solver = new GeneralizedAlphaSolver(0.5); break;
+      // Implicit A-stable methods (not L-stable)
+      case 22: ode_solver = new ImplicitMidpointSolver; break;
+      case 23: ode_solver = new SDIRK23Solver; break;
+      case 24: ode_solver = new SDIRK34Solver; break;
+      default:
+         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         delete mesh;
+         return 3;
+   }

   // 4. Refine the mesh to increase the resolution. In this example we do
   //    'ref_levels' of uniform refinement, where 'ref_levels' is a
@@ -267,6 +287,7 @@ int main(int argc, char *argv[])
   }

   // 10. Free the used memory.
+   delete ode_solver;
   delete mesh;

   return 0;
@@ -5,10 +5,10 @@
 // Sample runs:  mpirun -np 4 ex16p
 //               mpirun -np 4 ex16p -m ../data/inline-tri.mesh
 //               mpirun -np 4 ex16p -m ../data/disc-nurbs.mesh -tf 2
-//               mpirun -np 4 ex16p -s 21 -a 0.0 -k 1.0
-//               mpirun -np 4 ex16p -s 22 -a 1.0 -k 0.0
-//               mpirun -np 8 ex16p -s 23 -a 0.5 -k 0.5 -o 4
-//               mpirun -np 4 ex16p -s 4 -dt 1.0e-4 -tf 4.0e-2 -vs 40
+//               mpirun -np 4 ex16p -s 1 -a 0.0 -k 1.0
+//               mpirun -np 4 ex16p -s 2 -a 1.0 -k 0.0
+//               mpirun -np 8 ex16p -s 3 -a 0.5 -k 0.5 -o 4
+//               mpirun -np 4 ex16p -s 14 -dt 1.0e-4 -tf 4.0e-2 -vs 40
 //               mpirun -np 16 ex16p -m ../data/fichera-q2.mesh
 //               mpirun -np 16 ex16p -m ../data/fichera-mixed.mesh
 //               mpirun -np 16 ex16p -m ../data/escher-p2.mesh
@@ -78,15 +78,15 @@ public:
   ConductionOperator(ParFiniteElementSpace &f, real_t alpha, real_t kappa,
                      const Vector &u);

-   void Mult(const Vector &u, Vector &du_dt) const override;
+   virtual void Mult(const Vector &u, Vector &du_dt) const;
   /** Solve the Backward-Euler equation: k = f(u + dt*k, t), for the unknown k.
       This is the only requirement for high-order SDIRK implicit integration.*/
-   void ImplicitSolve(const real_t dt, const Vector &u, Vector &k) override;
+   virtual void ImplicitSolve(const real_t dt, const Vector &u, Vector &k);

   /// Update the diffusion BilinearForm K using the given true-dof vector `u`.
   void SetParameters(const Vector &u);

-   ~ConductionOperator() override;
+   virtual ~ConductionOperator();
 };

 real_t InitialTemperature(const Vector &x);
@@ -104,13 +104,11 @@ int main(int argc, char *argv[])
   int ser_ref_levels = 2;
   int par_ref_levels = 1;
   int order = 2;
-
-   int ode_solver_type = 23;  // SDIRK33Solver
+   int ode_solver_type = 3;
   real_t t_final = 0.5;
   real_t dt = 1.0e-2;
   real_t alpha = 1.0e-2;
   real_t kappa = 0.5;
-
   bool visualization = true;
   bool visit = false;
   int vis_steps = 5;
@@ -129,7 +127,8 @@ int main(int argc, char *argv[])
   args.AddOption(&order, "-o", "--order",
                  "Order (degree) of the finite elements.");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  ODESolver::Types.c_str());
+                  "ODE solver: 1 - Backward Euler, 2 - SDIRK2, 3 - SDIRK3,\n\t"
+                  "\t   11 - Forward Euler, 12 - RK2, 13 - RK3 SSP, 14 - RK4.");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -170,7 +169,28 @@ int main(int argc, char *argv[])
   // 4. Define the ODE solver used for time integration. Several implicit
   //    singly diagonal implicit Runge-Kutta (SDIRK) methods, as well as
   //    explicit Runge-Kutta methods are available.
-   unique_ptr<ODESolver> ode_solver = ODESolver::Select(ode_solver_type);
+   ODESolver *ode_solver;
+   switch (ode_solver_type)
+   {
+      // Implicit L-stable methods
+      case 1:  ode_solver = new BackwardEulerSolver; break;
+      case 2:  ode_solver = new SDIRK23Solver(2); break;
+      case 3:  ode_solver = new SDIRK33Solver; break;
+      // Explicit methods
+      case 11: ode_solver = new ForwardEulerSolver; break;
+      case 12: ode_solver = new RK2Solver(0.5); break; // midpoint method
+      case 13: ode_solver = new RK3SSPSolver; break;
+      case 14: ode_solver = new RK4Solver; break;
+      case 15: ode_solver = new GeneralizedAlphaSolver(0.5); break;
+      // Implicit A-stable methods (not L-stable)
+      case 22: ode_solver = new ImplicitMidpointSolver; break;
+      case 23: ode_solver = new SDIRK23Solver; break;
+      case 24: ode_solver = new SDIRK34Solver; break;
+      default:
+         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         delete mesh;
+         return 3;
+   }

   // 5. Refine the mesh in serial to increase the resolution. In this example
   //    we do 'ser_ref_levels' of uniform refinement, where 'ser_ref_levels' is
@@ -356,6 +376,7 @@ int main(int argc, char *argv[])
   }

   // 12. Free the used memory.
+   delete ode_solver;
   delete pmesh;

   return 0;
@@ -69,7 +69,7 @@ public:
   void SetDisplacement(GridFunction &u_) { u = &u_; }
   void SetComponent(int i, int j) { si = i; sj = j; }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
 };

 // Simple GLVis visualization manager.
@@ -89,7 +89,7 @@ public:
   void NewWindow();
   void CloseConnection();
   void PositionWindow();
-   ~VisMan() override;
+   virtual ~VisMan();
 };

 // Manipulators for the GLVis visualization manager.
@@ -69,7 +69,7 @@ public:
   void SetDisplacement(GridFunction &u_) { u = &u_; }
   void SetComponent(int i, int j) { si = i; sj = j; }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
 };

 // Simple GLVis visualization manager.
@@ -89,7 +89,7 @@ public:
   void NewWindow();
   void CloseConnection();
   void PositionWindow();
-   ~VisMan() override;
+   virtual ~VisMan();
 };

 // Manipulators for the GLVis visualization manager.
@@ -90,7 +90,8 @@ int main(int argc, char *argv[])
   args.AddOption(&order, "-o", "--order",
                  "Order (degree) of the finite elements.");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  ODESolver::ExplicitTypes.c_str());
+                  "ODE solver: 1 - Forward Euler,\n\t"
+                  "            2 - RK2 SSP, 3 - RK3 SSP, 4 - RK4, 6 - RK6.");
   args.AddOption(&t_final, "-tf", "--t-final", "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
                  "Time step. Positive number skips CFL timestep calculation.");
@@ -124,7 +125,18 @@ int main(int argc, char *argv[])

   // 3. Define the ODE solver used for time integration. Several explicit
   //    Runge-Kutta methods are available.
-   unique_ptr<ODESolver> ode_solver = ODESolver::SelectExplicit(ode_solver_type);
+   ODESolver *ode_solver = NULL;
+   switch (ode_solver_type)
+   {
+      case 1: ode_solver = new ForwardEulerSolver; break;
+      case 2: ode_solver = new RK2Solver(1.0); break;
+      case 3: ode_solver = new RK3SSPSolver; break;
+      case 4: ode_solver = new RK4Solver; break;
+      case 6: ode_solver = new RK6Solver; break;
+      default:
+         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         return 3;
+   }

   // 4. Define the discontinuous DG finite element space of the given
   //    polynomial order on the refined mesh.
@@ -292,5 +304,8 @@ int main(int argc, char *argv[])
   const real_t error = sol.ComputeLpError(2, u0);
   cout << "Solution error: " << error << endl;

+   // Free the used memory.
+   delete ode_solver;
+
   return 0;
 }
@@ -99,7 +99,8 @@ int main(int argc, char *argv[])
   args.AddOption(&order, "-o", "--order",
                  "Order (degree) of the finite elements.");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  ODESolver::ExplicitTypes.c_str());
+                  "ODE solver: 1 - Forward Euler,\n\t"
+                  "            2 - RK2 SSP, 3 - RK3 SSP, 4 - RK4, 6 - RK6.");
   args.AddOption(&t_final, "-tf", "--t-final", "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
                  "Time step. Positive number skips CFL timestep calculation.");
@@ -147,7 +148,18 @@ int main(int argc, char *argv[])

   // 3. Define the ODE solver used for time integration. Several explicit
   //    Runge-Kutta methods are available.
-   unique_ptr<ODESolver> ode_solver = ODESolver::SelectExplicit(ode_solver_type);
+   ODESolver *ode_solver = NULL;
+   switch (ode_solver_type)
+   {
+      case 1: ode_solver = new ForwardEulerSolver; break;
+      case 2: ode_solver = new RK2Solver(1.0); break;
+      case 3: ode_solver = new RK3SSPSolver; break;
+      case 4: ode_solver = new RK4Solver; break;
+      case 6: ode_solver = new RK6Solver; break;
+      default:
+         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         return 3;
+   }

   // 4. Define the discontinuous DG finite element space of the given
   //    polynomial order on the refined mesh.
@@ -348,5 +360,8 @@ int main(int argc, char *argv[])
      cout << "Solution error: " << error << endl;
   }

+   // Free the used memory.
+   delete ode_solver;
+
   return 0;
 }
@@ -48,7 +48,7 @@ public:
      print_level = print_lvl;
   }

-   void MonitorResidual(int it, real_t norm, const Vector &r, bool final) override;
+   virtual void MonitorResidual(int it, real_t norm, const Vector &r, bool final);

 private:
   const std::string prefix;
@@ -116,10 +116,10 @@ public:
   JacobianPreconditioner(Array<FiniteElementSpace *> &fes,
                          SparseMatrix &mass, Array<int> &offsets);

-   void Mult(const Vector &k, Vector &y) const override;
-   void SetOperator(const Operator &op) override;
+   virtual void Mult(const Vector &k, Vector &y) const;
+   virtual void SetOperator(const Operator &op);

-   ~JacobianPreconditioner() override;
+   virtual ~JacobianPreconditioner();
 };

 // After spatial discretization, the rubber model can be written as:
@@ -161,13 +161,13 @@ public:
                  int iter, Coefficient &mu);

   // Required to use the native newton solver
-   Operator &GetGradient(const Vector &xp) const override;
-   void Mult(const Vector &k, Vector &y) const override;
+   virtual Operator &GetGradient(const Vector &xp) const;
+   virtual void Mult(const Vector &k, Vector &y) const;

   // Driver for the newton solver
   void Solve(Vector &xp) const;

-   ~RubberOperator() override;
+   virtual ~RubberOperator();
 };

 // Visualization driver
@@ -62,7 +62,7 @@ public:
 #endif
   }

-   void MonitorResidual(int it, real_t norm, const Vector &r, bool final) override;
+   virtual void MonitorResidual(int it, real_t norm, const Vector &r, bool final);

 private:
   const std::string prefix;
@@ -130,10 +130,10 @@ public:
   JacobianPreconditioner(Array<ParFiniteElementSpace *> &fes,
                          Operator &mass, Array<int> &offsets);

-   void Mult(const Vector &k, Vector &y) const override;
-   void SetOperator(const Operator &op) override;
+   virtual void Mult(const Vector &k, Vector &y) const;
+   virtual void SetOperator(const Operator &op);

-   ~JacobianPreconditioner() override;
+   virtual ~JacobianPreconditioner();
 };

 // After spatial discretization, the rubber model can be written as:
@@ -175,13 +175,13 @@ public:
                  int iter, Coefficient &mu);

   // Required to use the native newton solver
-   Operator &GetGradient(const Vector &xp) const override;
-   void Mult(const Vector &k, Vector &y) const override;
+   virtual Operator &GetGradient(const Vector &xp) const;
+   virtual void Mult(const Vector &k, Vector &y) const;

   // Driver for the newton solver
   void Solve(Vector &xp) const;

-   ~RubberOperator() override;
+   virtual ~RubberOperator();
 };

 // Visualization driver
@@ -79,14 +79,14 @@ class GradT : public Operator
 {
 public:
   GradT() : Operator(1) {}
-   void Mult(const Vector &x, Vector &y) const override { y.Set(1.0/m_, x); }
+   void Mult(const Vector &x, Vector &y) const { y.Set(1.0/m_, x); }
 };

 class NegGradV : public TimeDependentOperator
 {
 public:
   NegGradV() : TimeDependentOperator(1) {}
-   void Mult(const Vector &x, Vector &y) const override;
+   void Mult(const Vector &x, Vector &y) const;
 };

 int main(int argc, char *argv[])
@@ -84,14 +84,14 @@ class GradT : public Operator
 {
 public:
   GradT() : Operator(1) {}
-   void Mult(const Vector &x, Vector &y) const override { y.Set(1.0/m_, x); }
+   void Mult(const Vector &x, Vector &y) const { y.Set(1.0/m_, x); }
 };

 class NegGradV : public TimeDependentOperator
 {
 public:
   NegGradV() : TimeDependentOperator(1) {}
-   void Mult(const Vector &x, Vector &y) const override;
+   void Mult(const Vector &x, Vector &y) const;
 };

 int main(int argc, char *argv[])
@@ -44,7 +44,7 @@ protected:
   BilinearForm *M;
   BilinearForm *K;

-   SparseMatrix Mmat, Kmat;
+   SparseMatrix Mmat, Kmat, Kmat0;
   SparseMatrix *T; // T = M + dt K
   real_t current_dt;

@@ -61,20 +61,20 @@ public:
   WaveOperator(FiniteElementSpace &f, Array<int> &ess_bdr, real_t speed);

   using SecondOrderTimeDependentOperator::Mult;
-   void Mult(const Vector &u, const Vector &du_dt,
-             Vector &d2udt2) const override;
+   virtual void Mult(const Vector &u, const Vector &du_dt,
+                     Vector &d2udt2) const;

   /** Solve the Backward-Euler equation:
       d2udt2 = f(u + fac0*d2udt2,dudt + fac1*d2udt2, t),
       for the unknown d2udt2. */
   using SecondOrderTimeDependentOperator::ImplicitSolve;
-   void ImplicitSolve(const real_t fac0, const real_t fac1,
-                      const Vector &u, const Vector &dudt, Vector &d2udt2) override;
+   virtual void ImplicitSolve(const real_t fac0, const real_t fac1,
+                              const Vector &u, const Vector &dudt, Vector &d2udt2);

   ///
   void SetParameters(const Vector &u);

-   ~WaveOperator() override;
+   virtual ~WaveOperator();
 };


@@ -83,24 +83,25 @@ WaveOperator::WaveOperator(FiniteElementSpace &f,
   : SecondOrderTimeDependentOperator(f.GetTrueVSize(), (real_t) 0.0),
     fespace(f), M(NULL), K(NULL), T(NULL), current_dt(0.0), z(height)
 {
-   // Assemble Laplace matrix
+   const real_t rel_tol = 1e-8;
+
+   fespace.GetEssentialTrueDofs(ess_bdr, ess_tdof_list);
+
   c2 = new ConstantCoefficient(speed*speed);
+
   K = new BilinearForm(&fespace);
   K->AddDomainIntegrator(new DiffusionIntegrator(*c2));
   K->Assemble();

-   // Assemble Mass matrix
+   Array<int> dummy;
+   K->FormSystemMatrix(dummy, Kmat0);
+   K->FormSystemMatrix(ess_tdof_list, Kmat);
+
   M = new BilinearForm(&fespace);
   M->AddDomainIntegrator(new MassIntegrator());
   M->Assemble();
-
-   // Apply Bcs
-   fespace.GetEssentialTrueDofs(ess_bdr, ess_tdof_list);
-   K->FormSystemMatrix(ess_tdof_list, Kmat);
   M->FormSystemMatrix(ess_tdof_list, Mmat);

-   // Configure preconditioner
-   const real_t rel_tol = 1e-8;
   M_solver.iterative_mode = false;
   M_solver.SetRelTol(rel_tol);
   M_solver.SetAbsTol(0.0);
@@ -109,13 +110,14 @@ WaveOperator::WaveOperator(FiniteElementSpace &f,
   M_solver.SetPreconditioner(M_prec);
   M_solver.SetOperator(Mmat);

-   // Configure solver
   T_solver.iterative_mode = false;
   T_solver.SetRelTol(rel_tol);
   T_solver.SetAbsTol(0.0);
   T_solver.SetMaxIter(100);
   T_solver.SetPrintLevel(0);
   T_solver.SetPreconditioner(T_prec);
+
+   T = NULL;
 }

 void WaveOperator::Mult(const Vector &u, const Vector &du_dt,
@@ -124,11 +126,9 @@ void WaveOperator::Mult(const Vector &u, const Vector &du_dt,
   // Compute:
   //    d2udt2 = M^{-1}*-K(u)
   // for d2udt2
-   K->FullMult(u, z);
+   Kmat.Mult(u, z);
   z.Neg(); // z = -z
-   z.SetSubVector(ess_tdof_list, 0.0);
   M_solver.Mult(z, d2udt2);
-   d2udt2.SetSubVector(ess_tdof_list, 0.0);
 }

 void WaveOperator::ImplicitSolve(const real_t fac0, const real_t fac1,
@@ -142,11 +142,14 @@ void WaveOperator::ImplicitSolve(const real_t fac0, const real_t fac1,
      T = Add(1.0, Mmat, fac0, Kmat);
      T_solver.SetOperator(*T);
   }
-   K->FullMult(u, z);
+   Kmat0.Mult(u, z);
   z.Neg();
-   z.SetSubVector(ess_tdof_list, 0.0);
+
+   for (int i = 0; i < ess_tdof_list.Size(); i++)
+   {
+      z[ess_tdof_list[i]] = 0.0;
+   }
   T_solver.Mult(z, d2udt2);
-   d2udt2.SetSubVector(ess_tdof_list, 0.0);
 }

 void WaveOperator::SetParameters(const Vector &u)
@@ -201,7 +204,9 @@ int main(int argc, char *argv[])
   args.AddOption(&order, "-o", "--order",
                  "Order (degree) of the finite elements.");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  SecondOrderODESolver::Types.c_str());
+                  "ODE solver: [0--10] - GeneralizedAlpha(0.1 * s),\n\t"
+                  "\t   11 - Average Acceleration, 12 - Linear Acceleration\n"
+                  "\t   13 - CentralDifference, 14 - FoxGoodwin");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -236,7 +241,32 @@ int main(int argc, char *argv[])

   // 3. Define the ODE solver used for time integration. Several second order
   //    time integrators are available.
-   SecondOrderODESolver *ode_solver= SecondOrderODESolver::Select(ode_solver_type);
+   SecondOrderODESolver *ode_solver;
+   switch (ode_solver_type)
+   {
+      // Implicit methods
+      case 0: ode_solver = new GeneralizedAlpha2Solver(0.0); break;
+      case 1: ode_solver = new GeneralizedAlpha2Solver(0.1); break;
+      case 2: ode_solver = new GeneralizedAlpha2Solver(0.2); break;
+      case 3: ode_solver = new GeneralizedAlpha2Solver(0.3); break;
+      case 4: ode_solver = new GeneralizedAlpha2Solver(0.4); break;
+      case 5: ode_solver = new GeneralizedAlpha2Solver(0.5); break;
+      case 6: ode_solver = new GeneralizedAlpha2Solver(0.6); break;
+      case 7: ode_solver = new GeneralizedAlpha2Solver(0.7); break;
+      case 8: ode_solver = new GeneralizedAlpha2Solver(0.8); break;
+      case 9: ode_solver = new GeneralizedAlpha2Solver(0.9); break;
+      case 10: ode_solver = new GeneralizedAlpha2Solver(1.0); break;
+
+      case 11: ode_solver = new AverageAccelerationSolver(); break;
+      case 12: ode_solver = new LinearAccelerationSolver(); break;
+      case 13: ode_solver = new CentralDifferenceSolver(); break;
+      case 14: ode_solver = new FoxGoodwinSolver(); break;
+
+      default:
+         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         delete mesh;
+         return 3;
+   }

   // 4. Refine the mesh to increase the resolution. In this example we do
   //    'ref_levels' of uniform refinement, where 'ref_levels' is a
@@ -284,6 +314,7 @@ int main(int argc, char *argv[])
         ess_bdr = 0;
      }
   }
+
   WaveOperator oper(fespace, ess_bdr, speed);

   u_gf.SetFromTrueDofs(u);
@@ -103,8 +103,8 @@ public:

   using VectorCoefficient::Eval;

-   void Eval(Vector &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override
+   virtual void Eval(Vector &K, ElementTransformation &T,
+                     const IntegrationPoint &ip)
   {
      real_t x[3];
      Vector transip(x, 3);
@@ -102,8 +102,8 @@ public:

   using VectorCoefficient::Eval;

-   void Eval(Vector &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override
+   virtual void Eval(Vector &K, ElementTransformation &T,
+                     const IntegrationPoint &ip)
   {
      real_t x[3];
      Vector transip(x, 3);
@@ -58,7 +58,7 @@ public:
      }
   }

-   ~DiffusionMultigrid() override
+   virtual ~DiffusionMultigrid()
   {
      delete amg;
   }
@@ -53,7 +53,7 @@ public:
                                    real_t min_val_=-36)
      : u(&u_), obstacle(&obst_), min_val(min_val_) { }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
 };

 class ExponentialGridFunctionCoefficient : public Coefficient
@@ -69,7 +69,7 @@ public:
                                      real_t min_val_=0.0, real_t max_val_=1e6)
      : u(&u_), obstacle(&obst_), min_val(min_val_), max_val(max_val_) { }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
 };

 int main(int argc, char *argv[])
@@ -258,7 +258,7 @@ int main(int argc, char *argv[])
         MixedBilinearForm a10(&H1fes,&L2fes);
         a10.AddDomainIntegrator(new MixedScalarMassIntegrator());
         a10.Assemble();
-         a10.EliminateTrialEssentialBC(ess_bdr, x.GetBlock(0), rhs.GetBlock(1));
+         a10.EliminateTrialDofs(ess_bdr, x.GetBlock(0), rhs.GetBlock(1));
         a10.Finalize();
         SparseMatrix &A10 = a10.SpMat();

@@ -53,7 +53,7 @@ public:
                                    real_t min_val_=-36)
      : u(&u_), obstacle(&obst_), min_val(min_val_) { }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
 };

 class ExponentialGridFunctionCoefficient : public Coefficient
@@ -69,7 +69,7 @@ public:
                                      real_t min_val_=0.0, real_t max_val_=1e6)
      : u(&u_), obstacle(&obst_), min_val(min_val_), max_val(max_val_) { }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
 };

 int main(int argc, char *argv[])
@@ -52,8 +52,8 @@ public:
       fun(fun_) {}


-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip)
   {
      return fun(GridFunctionCoefficient::Eval(T, ip));
   }
@@ -83,8 +83,8 @@ public:
       OtherGridF_cf(OtherGridF),
       fun(fun_) {}

-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip)
   {
      const real_t value1 = fun(GridFunctionCoefficient::Eval(T, ip));
      const real_t value2 = fun(OtherGridF_cf.Eval(T, ip));
@@ -108,7 +108,7 @@ public:
      : rho_filter(rho_filter_), min_val(min_val_), max_val(max_val_),
        exponent(exponent_) { }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip)
   {
      real_t val = rho_filter->GetValue(T, ip);
      real_t coeff = min_val + pow(val,exponent)*(max_val-min_val);
@@ -142,7 +142,7 @@ public:
      MFEM_ASSERT(rho_filter, "density field is not set");
   }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip)
   {
      real_t L = lambda->Eval(T, ip);
      real_t M = mu->Eval(T, ip);
@@ -176,8 +176,8 @@ public:

   using VectorCoefficient::Eval;

-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip)
   {
      Vector xx; xx.SetSize(T.GetDimension());
      T.Transform(ip,xx);
@@ -3,18 +3,18 @@
 // Compile with: make ex38
 //
 // Sample runs:
-// (since all sample runs require LAPACK or ALGOIM, the * symbol is used to
-//  exclude them from the automatically generated internal MFEM tests).
+// (since all sample runs require LAPACK, the * symbol is used to exclude them
+// from the automatically generated internal MFEM tests).
 //              * ex38
 //              * ex38 -i volumetric1d
 //              * ex38 -i surface2d
-//              * ex38 -i surface2d -o 4 -r 5 -m 1
+//              * ex38 -i surface2d -o 4 -r 5
 //              * ex38 -i volumetric2d
-//              * ex38 -i volumetric2d -o 4 -r 5 -m 1
+//              * ex38 -i volumetric2d -o 4 -r 5
 //              * ex38 -i surface3d
-//              * ex38 -i surface3d -o 3 -r 4 -m 1
+//              * ex38 -i surface3d -o 4 -r 5
 //              * ex38 -i volumetric3d
-//              * ex38 -i volumetric3d -o 3 -r 4 -m 1
+//              * ex38 -i volumetric3d -o 4 -r 5
 //
 // Description: This example code demonstrates the use of MFEM to integrate
 //              functions over implicit interfaces and subdomains bounded by
@@ -71,7 +71,7 @@ real_t integrand(const Vector& X)
   switch (itype)
   {
      case IntegrationType::Volumetric1D:
-         return pow(X(0), 2.);
+         return 1.;
      case IntegrationType::Surface2D:
         return 3. * pow(X(0), 2.) - pow(X(1), 2.);
      case IntegrationType::Volumetric2D:
@@ -91,7 +91,7 @@ real_t Surface()
   switch (itype)
   {
      case IntegrationType::Volumetric1D:
-         return .3025;
+         return 1.;
      case IntegrationType::Surface2D:
         return 2. * M_PI;
      case IntegrationType::Volumetric2D:
@@ -111,7 +111,7 @@ real_t Volume()
   switch (itype)
   {
      case IntegrationType::Volumetric1D:
-         return pow(.55, 3.) / 3.;
+         return .55;
      case IntegrationType::Surface2D:
         return NAN;
      case IntegrationType::Volumetric2D:
@@ -125,6 +125,7 @@ real_t Volume()
   }
 }

+#ifdef MFEM_USE_LAPACK
 /**
 @brief Class for surface IntegrationRule

@@ -134,14 +135,11 @@ real_t Volume()
 class SIntegrationRule : public IntegrationRule
 {
 protected:
-   /// method 0 is moments-based, 1 is Algoim.
-   int method, ir_order, ls_order;
-   Coefficient &level_set;
-   /// Space Dimension of the IntegrationRule
+   /// @brief Space Dimension of the IntegrationRule
   int dim;
-   /// Column-wise matrix of the quadtrature weights
+   /// @brief Column-wise matrix of the quadtrature weights
   DenseMatrix Weights;
-   /// Column-wise matrix of the transformation weights of the normal
+   /// @brief Column-wise matrix of the transformation weights of the normal
   DenseMatrix SurfaceWeights;

 public:
@@ -155,21 +153,15 @@ public:
    @param [in] lsOrder Polynomial degree for approx of level-set function
    @param [in] mesh Pointer to the mesh that is used
   */
-   SIntegrationRule(int method_, int Order,
-                    Coefficient& LvlSet, int lsOrder, Mesh* mesh)
-      : method(method_), ir_order(Order), ls_order(lsOrder),
-        level_set(LvlSet), dim(mesh->Dimension())
+   SIntegrationRule(int Order, Coefficient& LvlSet, int lsOrder, Mesh* mesh)
   {
-      // Nothing gets pre-computed for Algoim.
-      if (method == 1) { return; }
-
-#ifdef MFEM_USE_LAPACK
-      MomentFittingIntRules mf_ir(ir_order, level_set, ls_order);
+      dim = mesh->Dimension();

      IsoparametricTransformation Tr;
+      MomentFittingIntRules MFIRs(Order, LvlSet, lsOrder);
      mesh->GetElementTransformation(0, &Tr);
      IntegrationRule ir;
-      mf_ir.GetSurfaceIntegrationRule(Tr, ir);
+      MFIRs.GetSurfaceIntegrationRule(Tr, ir);
      if (dim >1)
      {
         Weights.SetSize(ir.GetNPoints(), mesh->GetNE());
@@ -180,7 +172,7 @@ public:
      }
      SurfaceWeights.SetSize(ir.GetNPoints(), mesh->GetNE());
      Vector w;
-      mf_ir.GetSurfaceWeights(Tr, ir, w);
+      MFIRs.GetSurfaceWeights(Tr, ir, w);
      SurfaceWeights.SetCol(0, w);
      SetSize(ir.GetNPoints());

@@ -206,8 +198,8 @@ public:
      for (int elem = 1; elem < mesh->GetNE(); elem++)
      {
         mesh->GetElementTransformation(elem, &Tr);
-         mf_ir.GetSurfaceIntegrationRule(Tr, ir);
-         mf_ir.GetSurfaceWeights(Tr, ir, w);
+         MFIRs.GetSurfaceIntegrationRule(Tr, ir);
+         MFIRs.GetSurfaceWeights(Tr, ir, w);
         SurfaceWeights.SetCol(elem, w);

         for (int ip = 0; ip < GetNPoints(); ip++)
@@ -223,48 +215,48 @@ public:
            }
         }
      }
-#else
-      MFEM_ABORT("Moment-fitting requires MFEM to be built with LAPACK!");
-#endif
   }

   /**
    @brief Set the weights for the given element and multiply them with the
    transformation of the interface
    */
-   void SetElementAndSurfaceWeight(ElementTransformation &Tr)
+   void SetElementinclSurfaceWeight(int Element)
   {
-      if (method == 1)
-      {
-#ifdef MFEM_USE_ALGOIM
-         AlgoimIntegrationRules a_ir(ir_order, level_set, ls_order);
-         a_ir.GetSurfaceIntegrationRule(Tr, *this);
-         Vector w;
-         a_ir.GetSurfaceWeights(Tr, *this, w);
-         for (int ip = 0; ip < GetNPoints(); ip++)
-         {
-            IntPoint(ip).weight *= w(ip);
-         }
-         return;
-#else
-         MFEM_ABORT("MFEM is not built with Algoim support!");
-#endif
-      }
-
      if (dim == 1)
      {
-         IntPoint(0).x = Weights(0, Tr.ElementNo);
-         IntPoint(0).weight = Weights(1, Tr.ElementNo);
+         IntegrationPoint &intp = IntPoint(0);
+         intp.x = Weights(0, Element);
+         intp.weight = Weights(1, Element);
+         cout << intp.x << " " << Element << endl;
      }
      else
-      {
         for (int ip = 0; ip < GetNPoints(); ip++)
         {
-            IntPoint(ip).weight = Weights(ip, Tr.ElementNo) *
-                                  SurfaceWeights(ip, Tr.ElementNo);
+            IntegrationPoint &intp = IntPoint(ip);
+            intp.weight = Weights(ip, Element) * SurfaceWeights(ip, Element);
         }
-      }
   }
+
+   /// @brief Set the weights for the given element
+   void SetElement(int Element)
+   {
+      if (dim == 1)
+      {
+         IntegrationPoint &intp = IntPoint(0);
+         intp.x = Weights(0, Element);
+         intp.weight = Weights(1, Element);
+      }
+      else
+         for (int ip = 0; ip < GetNPoints(); ip++)
+         {
+            IntegrationPoint &intp = IntPoint(ip);
+            intp.weight = Weights(ip, Element);
+         }
+   }
+
+   /// @brief Destructor of SIntegrationRule
+   ~SIntegrationRule() {}
 };

 /**
@@ -276,12 +268,9 @@ public:
 class CIntegrationRule : public IntegrationRule
 {
 protected:
-   /// method 0 is moments-based, 1 is Algoim.
-   int method, ir_order, ls_order;
-   Coefficient &level_set;
-   /// Space Dimension of the IntegrationRule
+   /// @brief Space Dimension of the IntegrationRule
   int dim;
-   /// Column-wise matrix of the quadtrature positions and weights.
+   /// @brief Column-wise matrix of the quadtrature weights
   DenseMatrix Weights;

 public:
@@ -295,21 +284,15 @@ public:
    @param [in] lsOrder Polynomial degree for approx of level-set function
    @param [in] mesh Pointer to the mesh that is used
   */
-   CIntegrationRule(int method_, int Order,
-                    Coefficient &LvlSet, int lsOrder, Mesh *mesh)
-      : method(method_), ir_order(Order), ls_order(lsOrder),
-        level_set(LvlSet), dim(mesh->Dimension())
+   CIntegrationRule(int Order, Coefficient& LvlSet, int lsOrder, Mesh* mesh)
   {
-      // Nothing gets pre-computed for Algoim.
-      if (method == 1) { return; }
-
-#ifdef MFEM_USE_LAPACK
-      MomentFittingIntRules mf_ir(ir_order, level_set, ls_order);
+      dim = mesh->Dimension();

      IsoparametricTransformation Tr;
+      MomentFittingIntRules MFIRs(Order, LvlSet, lsOrder);
      mesh->GetElementTransformation(0, &Tr);
      IntegrationRule ir;
-      mf_ir.GetVolumeIntegrationRule(Tr, ir);
+      MFIRs.GetVolumeIntegrationRule(Tr, ir);
      if (dim > 1)
      {
         Weights.SetSize(ir.GetNPoints(), mesh->GetNE());
@@ -341,9 +324,9 @@ public:
      for (int elem = 1; elem < mesh->GetNE(); elem++)
      {
         mesh->GetElementTransformation(elem, &Tr);
-         mf_ir.GetVolumeIntegrationRule(Tr, ir);
+         MFIRs.GetVolumeIntegrationRule(Tr, ir);

-         for (int ip = 0; ip < ir.GetNPoints(); ip++)
+         for (int ip = 0; ip < GetNPoints(); ip++)
         {
            if (dim > 1)
            {
@@ -356,39 +339,29 @@ public:
            }
         }
      }
-#else
-      MFEM_ABORT("Moment-fitting requires MFEM to be built with LAPACK!");
-#endif
   }

   /// @brief Set the weights for the given element
-   void SetElement(ElementTransformation &Tr)
+   void SetElement(int Element)
   {
-      if (method == 1)
-      {
-#ifdef MFEM_USE_ALGOIM
-         AlgoimIntegrationRules a_ir(ir_order, level_set, ls_order);
-         a_ir.GetVolumeIntegrationRule(Tr, *this);
-         return;
-#else
-         MFEM_ABORT("MFEM is not built with Algoim support!");
-#endif
-      }
-
-      for (int ip = 0; ip < GetNPoints(); ip++)
-      {
-         IntegrationPoint &intp = IntPoint(ip);
-         if (dim == 1)
+      if (dim == 1)
+         for (int ip = 0; ip < GetNPoints(); ip++)
         {
-            intp.x = Weights(2 * ip, Tr.ElementNo);
-            intp.weight = Weights(2 * ip + 1, Tr.ElementNo);
+            IntegrationPoint &intp = IntPoint(ip);
+            intp.x = Weights(2 * ip, Element);
+            intp.weight = Weights(2 * ip + 1, Element);
+         }
+      else
+         for (int ip = 0; ip < GetNPoints(); ip++)
+         {
+            IntegrationPoint &intp = IntPoint(ip);
+            intp.weight = Weights(ip, Element);
         }
-         else { intp.weight = Weights(ip, Tr.ElementNo); }
-      }
   }
+
+   /// @brief Destructor of CIntegrationRule
+   ~CIntegrationRule() {}
 };
-
-
 /**
 @brief Class for surface linearform integrator

@@ -435,9 +408,9 @@ public:
    @param [in] Tr transformation of finite element
    @param [out] elvect vector containing the
   */
-   void AssembleRHSElementVect(const FiniteElement &el,
-                               ElementTransformation &Tr,
-                               Vector &elvect) override
+   virtual void AssembleRHSElementVect(const FiniteElement &el,
+                                       ElementTransformation &Tr,
+                                       Vector &elvect) override
   {
      int dof = el.GetDof();
      shape.SetSize(dof);
@@ -445,7 +418,7 @@ public:
      elvect = 0.;

      // Update the surface integration rule for the current element
-      SIntRule->SetElementAndSurfaceWeight(Tr);
+      SIntRule->SetElementinclSurfaceWeight(Tr.ElementNo);

      for (int ip = 0; ip < SIntRule->GetNPoints(); ip++)
      {
@@ -455,8 +428,6 @@ public:
         add(elvect, SIntRule->IntPoint(ip).weight * val, shape, elvect);
      }
   }
-
-   using LinearFormIntegrator::AssembleRHSElementVect;
 };

 /**
@@ -505,9 +476,9 @@ public:
    @param [in] Tr transformation of finite element
    @param [out] elvect vector containing the
   */
-   void AssembleRHSElementVect(const FiniteElement &el,
-                               ElementTransformation &Tr,
-                               Vector &elvect) override
+   virtual void AssembleRHSElementVect(const FiniteElement &el,
+                                       ElementTransformation &Tr,
+                                       Vector &elvect) override
   {
      int dof = el.GetDof();
      shape.SetSize(dof);
@@ -515,7 +486,7 @@ public:
      elvect = 0.;

      // Update the subdomain integration rule
-      CIntRule->SetElement(Tr);
+      CIntRule->SetElement(Tr.ElementNo);

      for (int ip = 0; ip < CIntRule->GetNPoints(); ip++)
      {
@@ -526,17 +497,18 @@ public:
         add(elvect, CIntRule->IntPoint(ip).weight * val, shape, elvect);
      }
   }
-
-   using LinearFormIntegrator::AssembleRHSElementVect;
 };
+#endif // MFEM_USE_LAPACK

 int main(int argc, char *argv[])
 {
-#if defined(MFEM_USE_LAPACK) || defined(MFEM_USE_ALGOIM)
+#ifndef MFEM_USE_LAPACK
+   cout << "MFEM must be built with LAPACK for this example." << endl;
+   return MFEM_SKIP_RETURN_VALUE;
+#else
   // 1. Parse he command-line options.
   int ref_levels = 3;
   int order = 2;
-   int method = 0;
   const char *inttype = "surface2d";
   bool visualization = true;
   itype = IntegrationType::Surface2D;
@@ -544,8 +516,6 @@ int main(int argc, char *argv[])
   OptionsParser args(argc, argv);
   args.AddOption(&order, "-o", "--order", "Order of quadrature rule");
   args.AddOption(&ref_levels, "-r", "--refine", "Number of meh refinements");
-   args.AddOption(&method, "-m", "--method",
-                  "Cut integration method: 0 for moments-based, 1 for Algoim.");
   args.AddOption(&inttype, "-i", "--integrationtype",
                  "IntegrationType to demonstrate");
   args.AddOption(&visualization, "-vis", "--visualization", "-no-vis",
@@ -580,7 +550,7 @@ int main(int argc, char *argv[])
   }

   // 2. Construct and refine the mesh.
-   Mesh *mesh = nullptr;
+   Mesh *mesh;
   if (itype == IntegrationType::Volumetric1D)
   {
      mesh = new Mesh("../data/inline-segment.mesh");
@@ -628,14 +598,13 @@ int main(int argc, char *argv[])
   // 5. Define the necessary Integration rules on element 0.
   IsoparametricTransformation Tr;
   mesh->GetElementTransformation(0, &Tr);
-   SIntegrationRule* sir = new SIntegrationRule(method, order,
-                                                levelset, 2, mesh);
+   SIntegrationRule* sir = new SIntegrationRule(order, levelset, 2, mesh);
   CIntegrationRule* cir = NULL;
   if (itype == IntegrationType::Volumetric1D
       || itype == IntegrationType::Volumetric2D
       || itype == IntegrationType::Volumetric3D)
   {
-      cir = new CIntegrationRule(method, order, levelset, 2, mesh);
+      cir = new CIntegrationRule(order, levelset, 2, mesh);
   }

   // 6. Define and assemble the linear forms on the finite element space.
@@ -678,11 +647,11 @@ int main(int argc, char *argv[])
      cout << "Number of div free basis functions: " << nbasis << endl;
      cout << "Number of quadrature points:        " << ir.GetNPoints() << endl;
   }
-   cout << scientific << setprecision(10);
+   cout << scientific << setprecision(2);
   cout << "============================================" << endl;
   cout << "Computed value of surface integral: " << surface.Sum() << endl;
   cout << "True value of surface integral:     " << Surface() << endl;
-   cout << "Absolute Error (Surface):           ";
+   cout << "Absolute Error (Surface):            ";
   cout << abs(surface.Sum() - Surface()) << endl;
   cout << "Relative Error (Surface):           ";
   cout << abs(surface.Sum() - Surface()) / Surface() << endl;
@@ -693,7 +662,7 @@ int main(int argc, char *argv[])
      cout << "--------------------------------------------" << endl;
      cout << "Computed value of volume integral:  " << volume.Sum() << endl;
      cout << "True value of volume integral:      " << Volume() << endl;
-      cout << "Absolute Error (Volume):            ";
+      cout << "Absolute Error (Volume):             ";
      cout << abs(volume.Sum() - Volume()) << endl;
      cout << "Relative Error (Volume):            ";
      cout << abs(volume.Sum() - Volume()) / Volume() << endl;
@@ -722,8 +691,5 @@ int main(int argc, char *argv[])
   delete fespace;
   delete mesh;
   return EXIT_SUCCESS;
-#else
-   cout << "MFEM must be built with LAPACK or ALGOIM for this example." << endl;
-   return MFEM_SKIP_RETURN_VALUE;
-#endif // MFEM_USE_LAPACK
+#endif //MFEM_USE_LAPACK
 }
@@ -65,7 +65,6 @@ int main(int argc, char *argv[])
   bool static_cond = false;
   bool hybridization = false;
   bool pa = false;
-   bool ea = false;
   const char *device_config = "cpu";
   bool visualization = 1;

@@ -84,14 +83,18 @@ int main(int argc, char *argv[])
                  "--no-hybridization", "Enable hybridization.");
   args.AddOption(&pa, "-pa", "--partial-assembly", "-no-pa",
                  "--no-partial-assembly", "Enable Partial Assembly.");
-   args.AddOption(&ea, "-ea", "--element-assembly", "-no-ea",
-                  "--no-element-assembly", "Enable Element Assembly.");
   args.AddOption(&device_config, "-d", "--device",
                  "Device configuration string, see Device::Configure().");
   args.AddOption(&visualization, "-vis", "--visualization", "-no-vis",
                  "--no-visualization",
                  "Enable or disable GLVis visualization.");
-   args.ParseCheck();
+   args.Parse();
+   if (!args.Good())
+   {
+      args.PrintUsage(cout);
+      return 1;
+   }
+   args.PrintOptions(cout);
   kappa = freq * M_PI;

   // 2. Enable hardware devices such as GPUs, and programming models such as
@@ -163,7 +166,6 @@ int main(int argc, char *argv[])
   Coefficient *beta  = new ConstantCoefficient(1.0);
   BilinearForm *a = new BilinearForm(fespace);
   if (pa) { a->SetAssemblyLevel(AssemblyLevel::PARTIAL); }
-   if (ea) { a->SetAssemblyLevel(AssemblyLevel::ELEMENT); }
   a->AddDomainIntegrator(new DivDivIntegrator(*alpha));
   a->AddDomainIntegrator(new VectorFEMassIntegrator(*beta));

@@ -67,10 +67,8 @@ public:
   ZCoefficient(int vdim, GridFunction &psi_, real_t alpha_ = 1.0)
      : VectorCoefficient(vdim), psi(&psi_), alpha(alpha_) { }

-   using VectorCoefficient::Eval;
-
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   void SetAlpha(real_t alpha_) { alpha = alpha_; }
 };

@@ -84,8 +82,8 @@ public:
   DZCoefficient(int height, GridFunction &psi_, real_t alpha_ = 1.0)
      : MatrixCoefficient(height),  psi(&psi_), alpha(alpha_) { }

-   void Eval(DenseMatrix &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &K, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   void SetAlpha(real_t alpha_) { alpha = alpha_; }
 };

@@ -67,10 +67,8 @@ public:
   ZCoefficient(int vdim, ParGridFunction &psi_, real_t alpha_ = 1.0)
      : VectorCoefficient(vdim), psi(&psi_), alpha(alpha_) { }

-   using VectorCoefficient::Eval;
-
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   void SetAlpha(real_t alpha_) { alpha = alpha_; }
 };

@@ -84,8 +82,8 @@ public:
   DZCoefficient(int height, ParGridFunction &psi_, real_t alpha_ = 1.0)
      : MatrixCoefficient(height),  psi(&psi_), alpha(alpha_) { }

-   void Eval(DenseMatrix &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &K, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   void SetAlpha(real_t alpha_) { alpha = alpha_; }
 };

@@ -71,7 +71,6 @@ int main(int argc, char *argv[])
   bool static_cond = false;
   bool hybridization = false;
   bool pa = false;
-   bool ea = false;
   const char *device_config = "cpu";
   bool visualization = 1;

@@ -90,14 +89,24 @@ int main(int argc, char *argv[])
                  "--no-hybridization", "Enable hybridization.");
   args.AddOption(&pa, "-pa", "--partial-assembly", "-no-pa",
                  "--no-partial-assembly", "Enable Partial Assembly.");
-   args.AddOption(&ea, "-ea", "--element-assembly", "-no-ea",
-                  "--no-element-assembly", "Enable Element Assembly.");
   args.AddOption(&device_config, "-d", "--device",
                  "Device configuration string, see Device::Configure().");
   args.AddOption(&visualization, "-vis", "--visualization", "-no-vis",
                  "--no-visualization",
                  "Enable or disable GLVis visualization.");
-   args.ParseCheck();
+   args.Parse();
+   if (!args.Good())
+   {
+      if (myid == 0)
+      {
+         args.PrintUsage(cout);
+      }
+      return 1;
+   }
+   if (myid == 0)
+   {
+      args.PrintOptions(cout);
+   }
   kappa = freq * M_PI;

   // 3. Enable hardware devices such as GPUs, and programming models such as
@@ -185,7 +194,6 @@ int main(int argc, char *argv[])
   Coefficient *beta  = new ConstantCoefficient(1.0);
   ParBilinearForm *a = new ParBilinearForm(fespace);
   if (pa) { a->SetAssemblyLevel(AssemblyLevel::PARTIAL); }
-   if (ea) { a->SetAssemblyLevel(AssemblyLevel::ELEMENT); }
   a->AddDomainIntegrator(new DivDivIntegrator(*alpha));
   a->AddDomainIntegrator(new VectorFEMassIntegrator(*beta));

@@ -157,7 +157,7 @@ int main(int argc, char *argv[])
   MixedBilinearForm *B0 = new MixedBilinearForm(x0_space,test_space);
   B0->AddDomainIntegrator(new DiffusionIntegrator(one));
   B0->Assemble();
-   B0->EliminateTrialEssentialBC(ess_bdr, x.GetBlock(x0_var), F);
+   B0->EliminateTrialDofs(ess_bdr, x.GetBlock(x0_var), F);
   B0->Finalize();

   MixedBilinearForm *Bhat = new MixedBilinearForm(xhat_space,test_space);
@@ -9,7 +9,7 @@
 //    ex9 -m ../data/periodic-square.mesh -p 1 -r 2 -dt 0.005 -tf 9
 //    ex9 -m ../data/periodic-hexagon.mesh -p 1 -r 2 -dt 0.005 -tf 9
 //    ex9 -m ../data/amr-quad.mesh -p 1 -r 2 -dt 0.002 -tf 9
-//    ex9 -m ../data/amr-quad.mesh -p 1 -r 2 -dt 0.02 -s 23 -tf 9
+//    ex9 -m ../data/amr-quad.mesh -p 1 -r 2 -dt 0.02 -s 13 -tf 9
 //    ex9 -m ../data/star-q3.mesh -p 1 -r 2 -dt 0.005 -tf 9
 //    ex9 -m ../data/star-mixed.mesh -p 1 -r 2 -dt 0.005 -tf 9
 //    ex9 -m ../data/disc-nurbs.mesh -p 1 -r 3 -dt 0.005 -tf 9
@@ -104,12 +104,12 @@ public:
      }
   }

-   void SetOperator(const Operator &op) override
+   void SetOperator(const Operator &op)
   {
      linear_solver.SetOperator(op);
   }

-   void Mult(const Vector &x, Vector &y) const override
+   virtual void Mult(const Vector &x, Vector &y) const
   {
      linear_solver.Mult(x, y);
   }
@@ -134,10 +134,10 @@ private:
 public:
   FE_Evolution(BilinearForm &M_, BilinearForm &K_, const Vector &b_);

-   void Mult(const Vector &x, Vector &y) const override;
-   void ImplicitSolve(const real_t dt, const Vector &x, Vector &k) override;
+   virtual void Mult(const Vector &x, Vector &y) const;
+   virtual void ImplicitSolve(const real_t dt, const Vector &x, Vector &k);

-   ~FE_Evolution() override;
+   virtual ~FE_Evolution();
 };


@@ -182,7 +182,12 @@ int main(int argc, char *argv[])
   args.AddOption(&device_config, "-d", "--device",
                  "Device configuration string, see Device::Configure().");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  ODESolver::Types.c_str());
+                  "ODE solver: 1 - Forward Euler,\n\t"
+                  "            2 - RK2 SSP, 3 - RK3 SSP, 4 - RK4, 6 - RK6,\n\t"
+                  "            11 - Backward Euler,\n\t"
+                  "            12 - SDIRK23 (L-stable), 13 - SDIRK33,\n\t"
+                  "            22 - Implicit Midpoint Method,\n\t"
+                  "            23 - SDIRK23 (A-stable), 24 - SDIRK34");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -219,7 +224,28 @@ int main(int argc, char *argv[])

   // 3. Define the ODE solver used for time integration. Several explicit
   //    Runge-Kutta methods are available.
-   unique_ptr<ODESolver> ode_solver = ODESolver::Select(ode_solver_type);
+   ODESolver *ode_solver = NULL;
+   switch (ode_solver_type)
+   {
+      // Explicit methods
+      case 1: ode_solver = new ForwardEulerSolver; break;
+      case 2: ode_solver = new RK2Solver(1.0); break;
+      case 3: ode_solver = new RK3SSPSolver; break;
+      case 4: ode_solver = new RK4Solver; break;
+      case 6: ode_solver = new RK6Solver; break;
+      // Implicit (L-stable) methods
+      case 11: ode_solver = new BackwardEulerSolver; break;
+      case 12: ode_solver = new SDIRK23Solver(2); break;
+      case 13: ode_solver = new SDIRK33Solver; break;
+      // Implicit A-stable methods (not L-stable)
+      case 22: ode_solver = new ImplicitMidpointSolver; break;
+      case 23: ode_solver = new SDIRK23Solver; break;
+      case 24: ode_solver = new SDIRK34Solver; break;
+
+      default:
+         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         return 3;
+   }

   // 4. Refine the mesh to increase the resolution. In this example we do
   //    'ref_levels' of uniform refinement, where 'ref_levels' is a
@@ -414,6 +440,7 @@ int main(int argc, char *argv[])
   }

   // 10. Free the used memory.
+   delete ode_solver;
   delete pd;
   delete dc;

@@ -9,7 +9,7 @@
 //    mpirun -np 4 ex9p -m ../data/periodic-square.mesh -p 1 -dt 0.005 -tf 9
 //    mpirun -np 4 ex9p -m ../data/periodic-hexagon.mesh -p 1 -dt 0.005 -tf 9
 //    mpirun -np 4 ex9p -m ../data/amr-quad.mesh -p 1 -rp 1 -dt 0.002 -tf 9
-//    mpirun -np 4 ex9p -m ../data/amr-quad.mesh -p 1 -rp 1 -dt 0.02 -s 23 -tf 9
+//    mpirun -np 4 ex9p -m ../data/amr-quad.mesh -p 1 -rp 1 -dt 0.02 -s 13 -tf 9
 //    mpirun -np 4 ex9p -m ../data/star-q3.mesh -p 1 -rp 1 -dt 0.004 -tf 9
 //    mpirun -np 4 ex9p -m ../data/star-mixed.mesh -p 1 -rp 1 -dt 0.004 -tf 9
 //    mpirun -np 4 ex9p -m ../data/disc-nurbs.mesh -p 1 -rp 1 -dt 0.005 -tf 9
@@ -92,7 +92,7 @@ private:
 public:
   AIR_prec(int blocksize_) : AIR_solver(NULL), blocksize(blocksize_) { }

-   void SetOperator(const Operator &op) override
+   void SetOperator(const Operator &op)
   {
      width = op.Width();
      height = op.Height();
@@ -110,7 +110,7 @@ public:
      AIR_solver->SetMaxLevels(50);
   }

-   void Mult(const Vector &x, Vector &y) const override
+   virtual void Mult(const Vector &x, Vector &y) const
   {
      // Scale the rhs by block inverse and solve system
      HypreParVector z_s;
@@ -119,7 +119,7 @@ public:
      AIR_solver->Mult(z_s, y);
   }

-   ~AIR_prec() override
+   ~AIR_prec()
   {
      delete AIR_solver;
   }
@@ -185,17 +185,17 @@ public:
      }
   }

-   void SetOperator(const Operator &op) override
+   void SetOperator(const Operator &op)
   {
      linear_solver.SetOperator(op);
   }

-   void Mult(const Vector &x, Vector &y) const override
+   virtual void Mult(const Vector &x, Vector &y) const
   {
      linear_solver.Mult(x, y);
   }

-   ~DG_Solver() override
+   ~DG_Solver()
   {
      delete prec;
      delete A;
@@ -223,10 +223,10 @@ public:
   FE_Evolution(ParBilinearForm &M_, ParBilinearForm &K_, const Vector &b_,
                PrecType prec_type);

-   void Mult(const Vector &x, Vector &y) const override;
-   void ImplicitSolve(const real_t dt, const Vector &x, Vector &k) override;
+   virtual void Mult(const Vector &x, Vector &y) const;
+   virtual void ImplicitSolve(const real_t dt, const Vector &x, Vector &k);

-   ~FE_Evolution() override;
+   virtual ~FE_Evolution();
 };


@@ -285,7 +285,12 @@ int main(int argc, char *argv[])
   args.AddOption(&device_config, "-d", "--device",
                  "Device configuration string, see Device::Configure().");
   args.AddOption(&ode_solver_type, "-s", "--ode-solver",
-                  ODESolver::Types.c_str());
+                  "ODE solver: 1 - Forward Euler,\n\t"
+                  "            2 - RK2 SSP, 3 - RK3 SSP, 4 - RK4, 6 - RK6,\n\t"
+                  "            11 - Backward Euler,\n\t"
+                  "            12 - SDIRK23 (L-stable), 13 - SDIRK33,\n\t"
+                  "            22 - Implicit Midpoint Method,\n\t"
+                  "            23 - SDIRK23 (A-stable), 24 - SDIRK34");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -333,7 +338,31 @@ int main(int argc, char *argv[])

   // 4. Define the ODE solver used for time integration. Several explicit
   //    Runge-Kutta methods are available.
-   unique_ptr<ODESolver> ode_solver = ODESolver::Select(ode_solver_type);
+   ODESolver *ode_solver = NULL;
+   switch (ode_solver_type)
+   {
+      // Explicit methods
+      case 1: ode_solver = new ForwardEulerSolver; break;
+      case 2: ode_solver = new RK2Solver(1.0); break;
+      case 3: ode_solver = new RK3SSPSolver; break;
+      case 4: ode_solver = new RK4Solver; break;
+      case 6: ode_solver = new RK6Solver; break;
+      // Implicit (L-stable) methods
+      case 11: ode_solver = new BackwardEulerSolver; break;
+      case 12: ode_solver = new SDIRK23Solver(2); break;
+      case 13: ode_solver = new SDIRK33Solver; break;
+      // Implicit A-stable methods (not L-stable)
+      case 22: ode_solver = new ImplicitMidpointSolver; break;
+      case 23: ode_solver = new SDIRK23Solver; break;
+      case 24: ode_solver = new SDIRK34Solver; break;
+      default:
+         if (Mpi::Root())
+         {
+            cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+         }
+         delete mesh;
+         return 3;
+   }

   // 5. Refine the mesh in serial to increase the resolution. In this example
   //    we do 'ser_ref_levels' of uniform refinement, where 'ser_ref_levels' is
@@ -613,6 +642,7 @@ int main(int argc, char *argv[])
   delete m;
   delete fes;
   delete pmesh;
+   delete ode_solver;
   delete pd;
 #ifdef MFEM_USE_ADIOS2
   if (adios2)
@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/ginkgo/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -96,7 +96,6 @@ public:
   {
      Vector w_glob(width);
      pfes.Dof_TrueDof_Matrix()->MultTranspose(w, w_glob);
-      w_glob.HostReadWrite(); // read+write -> can use w_glob(i) (non-const)
      for (int i = 0; i < width; i++) { grad(0, i) = w_glob(i); }
   }

@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/hiop/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ..
 MFEM_BUILD_DIR ?= ..
-MFEM_INSTALL_DIR ?= ../mfem
 SRC = $(if $(MFEM_DIR:..=),$(MFEM_DIR)/examples/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/moonolith/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -206,7 +206,6 @@ int main(int argc, char *argv[])
   bool use_petsc = true;
   const char *petscrc_file = "";
   bool petsc_use_jfnk = false;
-   const char *device_config = "cpu";

   OptionsParser args(argc, argv);
   args.AddOption(&mesh_file, "-m", "--mesh",
@@ -244,8 +243,6 @@ int main(int argc, char *argv[])
   args.AddOption(&petsc_use_jfnk, "-jfnk", "--jfnk", "-no-jfnk",
                  "--no-jfnk",
                  "Use JFNK with user-defined preconditioner factory.");
-   args.AddOption(&device_config, "-d", "--device",
-                  "Device configuration string, see Device::Configure().");
   args.Parse();
   if (!args.Good())
   {
@@ -260,12 +257,7 @@ int main(int argc, char *argv[])
      args.PrintOptions(cout);
   }

-   // 2b. Enable hardware devices such as GPUs, and programming models such as
-   //    CUDA, OCCA, RAJA and OpenMP based on command line options.
-   Device device(device_config);
-   if (myid == 0) { device.Print(); }
-
-   // 2c. We initialize PETSc
+   // 2b. We initialize PETSc
   if (use_petsc)
   {
      MFEMInitializePetsc(NULL,NULL,petscrc_file,NULL);
@@ -67,7 +67,6 @@ int main(int argc, char *argv[])
   bool use_petsc = true;
   const char *petscrc_file = "";
   bool use_nonoverlapping = false;
-   const char *device_config = "cpu";

   OptionsParser args(argc, argv);
   args.AddOption(&mesh_file, "-m", "--mesh",
@@ -96,8 +95,6 @@ int main(int argc, char *argv[])
                  "-no-nonoverlapping", "--no-nonoverlapping",
                  "Use or not the block diagonal PETSc's matrix format "
                  "for non-overlapping domain decomposition.");
-   args.AddOption(&device_config, "-d", "--device",
-                  "Device configuration string, see Device::Configure().");
   args.Parse();
   if (!args.Good())
   {
@@ -112,12 +109,7 @@ int main(int argc, char *argv[])
      args.PrintOptions(cout);
   }

-   // 2b. Enable hardware devices such as GPUs, and programming models such as
-   //    CUDA, OCCA, RAJA and OpenMP based on command line options.
-   Device device(device_config);
-   if (myid == 0) { device.Print(); }
-
-   // 2c. We initialize PETSc
+   // 2b. We initialize PETSc
   if (use_petsc) { MFEMInitializePetsc(NULL,NULL,petscrc_file,NULL); }

   // 3. Read the (serial) mesh from the given mesh file on all processors.  We
@@ -61,7 +61,6 @@ int main(int argc, char *argv[])
   bool use_petsc = true;
   const char *petscrc_file = "";
   bool use_nonoverlapping = false;
-   const char *device_config = "cpu";

   OptionsParser args(argc, argv);
   args.AddOption(&mesh_file, "-m", "--mesh",
@@ -88,8 +87,6 @@ int main(int argc, char *argv[])
                  "-no-nonoverlapping", "--no-nonoverlapping",
                  "Use or not the block diagonal PETSc's matrix format "
                  "for non-overlapping domain decomposition.");
-   args.AddOption(&device_config, "-d", "--device",
-                  "Device configuration string, see Device::Configure().");
   args.Parse();
   if (!args.Good())
   {
@@ -103,15 +100,9 @@ int main(int argc, char *argv[])
   {
      args.PrintOptions(cout);
   }
-   kappa = freq * M_PI;
-
-   // 2b. Enable hardware devices such as GPUs, and programming models such as
-   //    CUDA, OCCA, RAJA and OpenMP based on command line options.
-   Device device(device_config);
-   if (myid == 0) { device.Print(); }
-
-   // 2c. We initialize PETSc
+   // 2b. We initialize PETSc
   if (use_petsc) { MFEMInitializePetsc(NULL,NULL,petscrc_file,NULL); }
+   kappa = freq * M_PI;

   // 3. Read the (serial) mesh from the given mesh file on all processors.  We
   //    can handle triangular, quadrilateral, tetrahedral, hexahedral, surface
@@ -58,7 +58,6 @@ int main(int argc, char *argv[])
   bool use_petsc = true;
   const char *petscrc_file = "";
   bool use_nonoverlapping = false;
-   const char *device_config = "cpu";

   OptionsParser args(argc, argv);
   args.AddOption(&mesh_file, "-m", "--mesh",
@@ -89,8 +88,6 @@ int main(int argc, char *argv[])
                  "-no-nonoverlapping", "--no-nonoverlapping",
                  "Use or not the block diagonal PETSc's matrix format "
                  "for non-overlapping domain decomposition.");
-   args.AddOption(&device_config, "-d", "--device",
-                  "Device configuration string, see Device::Configure().");
   args.Parse();
   if (!args.Good())
   {
@@ -104,15 +101,9 @@ int main(int argc, char *argv[])
   {
      args.PrintOptions(cout);
   }
-   kappa = freq * M_PI;
-
-   // 2b. Enable hardware devices such as GPUs, and programming models such as
-   //    CUDA, OCCA, RAJA and OpenMP based on command line options.
-   Device device(device_config);
-   if (myid == 0) { device.Print(); }
-
-   // 2c. We initialize PETSc
+   // 2b. We initialize PETSc
   if (use_petsc) { MFEMInitializePetsc(NULL,NULL,petscrc_file,NULL); }
+   kappa = freq * M_PI;

   // 3. Read the (serial) mesh from the given mesh file on all processors.  We
   //    can handle triangular, quadrilateral, tetrahedral, hexahedral, surface
@@ -59,8 +59,6 @@ int main(int argc, char *argv[])

   // 2. Parse command-line options.
   const char *mesh_file = "../../data/star.mesh";
-   int ser_ref_levels = -1;
-   int par_ref_levels = 2;
   int order = 1;
   bool par_format = false;
   bool visualization = 1;
@@ -68,22 +66,15 @@ int main(int argc, char *argv[])
   bool use_nonoverlapping = false;
   bool local_bdr_spec = false;
   const char *petscrc_file = "";
-   const char *device_config = "cpu";

   OptionsParser args(argc, argv);
   args.AddOption(&mesh_file, "-m", "--mesh",
                  "Mesh file to use.");
-   args.AddOption(&ser_ref_levels, "-rs", "--refine-serial",
-                  "Number of times to refine the mesh uniformly in serial.");
-   args.AddOption(&par_ref_levels, "-rp", "--refine-parallel",
-                  "Number of times to refine the mesh uniformly in parallel.");
   args.AddOption(&order, "-o", "--order",
                  "Finite element order (polynomial degree).");
   args.AddOption(&par_format, "-pf", "--parallel-format", "-sf",
                  "--serial-format",
                  "Format to use when saving the results for VisIt.");
-   args.AddOption(&device_config, "-d", "--device",
-                  "Device configuration string, see Device::Configure().");
   args.AddOption(&visualization, "-vis", "--visualization", "-no-vis",
                  "--no-visualization",
                  "Enable or disable GLVis visualization.");
@@ -112,13 +103,7 @@ int main(int argc, char *argv[])
   {
      args.PrintOptions(cout);
   }
-
-   // 2b. Enable hardware devices such as GPUs, and programming models such as
-   //    CUDA, OCCA, RAJA and OpenMP based on command line options.
-   Device device(device_config);
-   if (myid == 0) { device.Print(); }
-
-   // 2c. We initialize PETSc
+   // 2b. We initialize PETSc
   if (use_petsc) { MFEMInitializePetsc(NULL,NULL,petscrc_file,NULL); }

   // 3. Read the (serial) mesh from the given mesh file on all processors.  We
@@ -132,11 +117,9 @@ int main(int argc, char *argv[])
   //    'ref_levels' to be the largest number that gives a final mesh with no
   //    more than 10,000 elements.
   {
-      if (ser_ref_levels < 0)
-      {
-         ser_ref_levels = (int)floor(log(10000./mesh->GetNE())/log(2.)/dim);
-      }
-      for (int l = 0; l < ser_ref_levels; l++)
+      int ref_levels =
+         (int)floor(log(10000./mesh->GetNE())/log(2.)/dim);
+      for (int l = 0; l < ref_levels; l++)
      {
         mesh->UniformRefinement();
      }
@@ -148,6 +131,7 @@ int main(int argc, char *argv[])
   ParMesh *pmesh = new ParMesh(MPI_COMM_WORLD, *mesh);
   delete mesh;
   {
+      int par_ref_levels = 2;
      for (int l = 0; l < par_ref_levels; l++)
      {
         pmesh->UniformRefinement();
@@ -203,26 +187,21 @@ int main(int argc, char *argv[])

   // 9. Define the parallel grid function and parallel linear forms, solution
   //    vector and rhs.
-   MemoryType mt = device.GetMemoryType();
-   BlockVector x(block_offsets, mt), rhs(block_offsets, mt);
-   BlockVector trueX(block_trueOffsets, mt), trueRhs(block_trueOffsets, mt);
+   BlockVector x(block_offsets), rhs(block_offsets);
+   BlockVector trueX(block_trueOffsets), trueRhs(block_trueOffsets);

   ParLinearForm *fform(new ParLinearForm);
   fform->Update(R_space, rhs.GetBlock(0), 0);
   fform->AddDomainIntegrator(new VectorFEDomainLFIntegrator(fcoeff));
   fform->AddBoundaryIntegrator(new VectorFEBoundaryFluxLFIntegrator(fnatcoeff));
   fform->Assemble();
-   fform->SyncAliasMemory(rhs);
   fform->ParallelAssemble(trueRhs.GetBlock(0));
-   trueRhs.GetBlock(0).SyncAliasMemory(trueRhs);

   ParLinearForm *gform(new ParLinearForm);
   gform->Update(W_space, rhs.GetBlock(1), 0);
   gform->AddDomainIntegrator(new DomainLFIntegrator(gcoeff));
   gform->Assemble();
-   gform->SyncAliasMemory(rhs);
   gform->ParallelAssemble(trueRhs.GetBlock(1));
-   trueRhs.GetBlock(1).SyncAliasMemory(trueRhs);

   // 10. Assemble the finite element matrices for the Darcy operator
   //
@@ -53,7 +53,6 @@ int main(int argc, char *argv[])
   bool use_petsc = true;
   const char *petscrc_file = "";
   bool use_nonoverlapping = false;
-   const char *device_config = "cpu";

   OptionsParser args(argc, argv);
   args.AddOption(&mesh_file, "-m", "--mesh",
@@ -74,8 +73,6 @@ int main(int argc, char *argv[])
                  "-no-nonoverlapping", "--no-nonoverlapping",
                  "Use or not the block diagonal PETSc's matrix format "
                  "for non-overlapping domain decomposition.");
-   args.AddOption(&device_config, "-d", "--device",
-                  "Device configuration string, see Device::Configure().");
   args.Parse();
   if (!args.Good())
   {
@@ -89,13 +86,7 @@ int main(int argc, char *argv[])
   {
      args.PrintOptions(cout);
   }
-
-   // 2b. Enable hardware devices such as GPUs, and programming models such as
-   //    CUDA, OCCA, RAJA and OpenMP based on command line options.
-   Device device(device_config);
-   if (myid == 0) { device.Print(); }
-
-   // 2c. We initialize PETSc
+   // 2b. We initialize PETSc
   if (use_petsc) { MFEMInitializePetsc(NULL,NULL,petscrc_file,NULL); }

   // 3. Read the (serial) mesh from the given mesh file on all processors.  We
@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/petsc/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -66,6 +66,7 @@ int main(int argc, char *argv[])
 {
   // 1. Initialize MPI (required by PUMI) and HYPRE.
   Mpi::Init(argc, argv);
+   int num_procs = Mpi::WorldSize();
   int myid = Mpi::WorldRank();
   Hypre::Init();

@@ -80,6 +80,8 @@ int main(int argc, char *argv[])
 {
   // 1. Initialize MPI (required by PUMI) and HYPRE.
   Mpi::Init(argc, argv);
+   int num_proc = Mpi::WorldSize();
+   int myId = Mpi::WorldRank();
   Hypre::Init();

   // 2. Parse command-line options.
@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/pumi/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -31,21 +31,11 @@ include_directories(BEFORE ${PROJECT_BINARY_DIR})
 add_custom_target(test_sundials
  ${CMAKE_CTEST_COMMAND} -R sundials USES_TERMINAL)

-# Add one executable per cpp file, adding "sundials_" as prefix so the CMake
-# target is unique from those in the non-SUNDIALS examples. Also sets
-# "test_sundials" as a target that depends on the given SUNDIALS examples.
+# Add one executable per cpp file, adding "sundials_" as prefix. Sets
+# "test_sundials" as a target that depends on the given examples.
 set(PFX sundials_)
 add_mfem_examples(SUNDIALS_EXAMPLES_SRCS ${PFX} "" test_sundials)

-# Remove "sundials_" prefix from exectuable name for consistency with GNU build
-# system.
-foreach(SRC_FILE ${SUNDIALS_EXAMPLES_SRCS})
-  get_filename_component(SRC_FILENAME ${SRC_FILE} NAME)
-  string(REPLACE ".cpp" "" TARGET_NAME "${PFX}${SRC_FILENAME}")
-  string(REPLACE ${PFX} "" EXE_NAME ${TARGET_NAME})
-  set_target_properties(${TARGET_NAME} PROPERTIES OUTPUT_NAME ${EXE_NAME})
-endforeach()
-
 # Testing.
 # The SUNDIALS tests can be run separately using the target "test_sundials"
 # which builds the examples and runs:
@@ -61,10 +51,7 @@ if (MFEM_ENABLE_TESTING)
  set(EX10_COMMON_OPTS -m ../../data/beam-quad.mesh -o 2 -s 5 -dt 0.15 -tf 6 -vs 10)
  set(EX10_TEST_OPTS ${EX10_COMMON_OPTS}  -r  2)
  set(EX10P_TEST_OPTS ${EX10_COMMON_OPTS} -rp 1)
-  # Example 16: test ARKODE with implicit time stepping using mass form
-  set(EX16_COMMON_OPTS -s 15)
-  set(EX16_TEST_OPTS ${EX16_COMMON_OPTS})
-  set(EX16P_TEST_OPTS ${EX16_COMMON_OPTS})
+  # Example 16: use the default options

  # Add the tests: one test per source file.
  foreach(SRC_FILE ${SUNDIALS_EXAMPLES_SRCS})
@@ -1,17 +1,15 @@
 //                                MFEM Example 10
 //                             SUNDIALS Modification
 //
-// Compile with:
-//    make ex10             (GNU make)
-//    make sundials_ex10    (CMake)
+// Compile with: make ex10
 //
 // Sample runs:
 //    ex10 -m ../../data/beam-quad.mesh -r 2 -o 2 -s 12 -dt 0.15 -vs 10
 //    ex10 -m ../../data/beam-tri.mesh  -r 2 -o 2 -s 16 -dt 0.3  -vs 5
 //    ex10 -m ../../data/beam-hex.mesh  -r 1 -o 2 -s 12 -dt 0.2  -vs 5
-//    ex10 -m ../../data/beam-tri.mesh  -r 2 -o 2 -s  2 -dt 3 -nls 1
-//    ex10 -m ../../data/beam-quad.mesh -r 2 -o 2 -s  2 -dt 3 -nls 2
-//    ex10 -m ../../data/beam-hex.mesh  -r 1 -o 2 -s  2 -dt 3 -nls 4
+//    ex10 -m ../../data/beam-tri.mesh  -r 2 -o 2 -s  2 -dt 3 -nls kinsol
+//    ex10 -m ../../data/beam-quad.mesh -r 2 -o 2 -s  2 -dt 3 -nls kinsol
+//    ex10 -m ../../data/beam-hex.mesh  -r 1 -o 2 -s  2 -dt 3 -nls kinsol
 //    ex10 -m ../../data/beam-quad.mesh -r 2 -o 2 -s 14 -dt 0.15 -vs 10
 //    ex10 -m ../../data/beam-tri.mesh  -r 2 -o 2 -s 17 -dt 0.01 -vs 30
 //    ex10 -m ../../data/beam-hex.mesh  -r 1 -o 2 -s 14 -dt 0.15 -vs 10
@@ -99,11 +97,16 @@ protected:
   double saved_gamma; // saved gamma value from implicit setup

 public:
+   /// Solver type to use in the ImplicitSolve() method, used by SDIRK methods.
+   enum NonlinearSolverType
+   {
+      NEWTON = 0, ///< Use MFEM's plain NewtonSolver
+      KINSOL = 1  ///< Use SUNDIALS' KINSOL (through MFEM's class KINSolver)
+   };

   HyperelasticOperator(FiniteElementSpace &f, Array<int> &ess_bdr,
                        double visc, double mu, double K,
-                        int kinsol_nls_type = -1, double kinsol_damping = 0.0,
-                        int kinsol_aa_n = 0);
+                        NonlinearSolverType nls_type);

   /// Compute the right-hand side of the ODE system.
   virtual void Mult(const Vector &vx, Vector &dvx_dt) const;
@@ -221,10 +224,8 @@ int main(int argc, char *argv[])
   double mu = 0.25;
   double K = 5.0;
   bool visualization = true;
-   int nonlinear_solver_type = 0;
+   const char *nls = "newton";
   int vis_steps = 1;
-   double kinsol_damping = 0.0;
-   int kinsol_aa_n = -1;

   // Relative and absolute tolerances for CVODE and ARKODE.
   const double reltol = 1e-1, abstol = 1e-1;
@@ -261,18 +262,9 @@ int main(int argc, char *argv[])
                  "15 - ARKODE implicit, approximate Jacobian,\n\t"
                  "16 - ARKODE implicit, specified Jacobian,\n\t"
                  "17 - ARKODE explicit, 4th order.");
-   args.AddOption(&nonlinear_solver_type, "-nls", "--nonlinear-solver",
-                  "Nonlinear system solver:\n\t"
-                  "0  - MFEM Newton method,\n\t"
-                  "1  - KINSOL Newton method,\n\t"
-                  "2  - KINSOL Newton method with globalization,\n\t"
-                  "3  - KINSOL fixed-point method (with or without AA),\n\t"
-                  "4  - KINSOL Picard method (with or without AA).");
-   args.AddOption(&kinsol_damping, "-damp", "--kinsol-damping",
-                  "Picard or Fixed-Point damping parameter (only valid with KINSOL): "
-                  "0 < d <= 1.0");
-   args.AddOption(&kinsol_aa_n, "-aan", "--anderson-subspace",
-                  "Anderson Acceleration subspace size (only valid with KINSOL)");
+   args.AddOption(&nls, "-nls", "--nonlinear-solver",
+                  "Nonlinear systems solver: "
+                  "\"newton\" (plain Newton) or \"kinsol\" (KINSOL).");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -303,32 +295,22 @@ int main(int argc, char *argv[])
      return 1;
   }

-   // check for valid nonlinear solver options
-   if (nonlinear_solver_type < 0 || nonlinear_solver_type > 4)
-   {
-      cout << "Unknown nonlinear solver type: " << nonlinear_solver_type << "\n";
-      return 1;
-   }
-   if (kinsol_damping > 0.0 &&
-       !(nonlinear_solver_type == 3 || nonlinear_solver_type == 4))
-   {
-      cout << "Only KINSOL fixed-point and Picard methods can use damping\n";
-      return 1;
-   }
-   if (kinsol_aa_n > 0 &&
-       !(nonlinear_solver_type == 3 || nonlinear_solver_type == 4))
-   {
-      cout << "Only KINSOL fixed-point and Picard methods can use AA\n";
-      return 1;
-   }
-
-
   // 2. Read the mesh from the given mesh file. We can handle triangular,
   //    quadrilateral, tetrahedral and hexahedral meshes with the same code.
   Mesh *mesh = new Mesh(mesh_file, 1, 1);
   int dim = mesh->Dimension();

-   // 3. Refine the mesh to increase the resolution. In this example we do
+   // 3. Setup the nonlinear solver
+   map<string,HyperelasticOperator::NonlinearSolverType> nls_map;
+   nls_map["newton"] = HyperelasticOperator::NEWTON;
+   nls_map["kinsol"] = HyperelasticOperator::KINSOL;
+   if (nls_map.find(nls) == nls_map.end())
+   {
+      cout << "Unknown type of nonlinear solver: " << nls << endl;
+      return 4;
+   }
+
+   // 4. Refine the mesh to increase the resolution. In this example we do
   //    'ref_levels' of uniform refinement, where 'ref_levels' is a
   //    command-line parameter.
   for (int lev = 0; lev < ref_levels; lev++)
@@ -336,7 +318,7 @@ int main(int argc, char *argv[])
      mesh->UniformRefinement();
   }

-   // 4. Define the vector finite element spaces representing the mesh
+   // 5. Define the vector finite element spaces representing the mesh
   //    deformation x, the velocity v, and the initial configuration, x_ref.
   //    Define also the elastic energy density, w, which is in a discontinuous
   //    higher-order space. Since x and v are integrated in time as a system,
@@ -364,7 +346,7 @@ int main(int argc, char *argv[])
   FiniteElementSpace w_fespace(mesh, &w_fec);
   GridFunction w(&w_fespace);

-   // 5. Set the initial conditions for v and x, and the boundary conditions on
+   // 6. Set the initial conditions for v and x, and the boundary conditions on
   //    a beam-like mesh (see description above).
   VectorFunctionCoefficient velo(dim, InitialVelocity);
   v.ProjectCoefficient(velo);
@@ -377,34 +359,9 @@ int main(int argc, char *argv[])
   ess_bdr = 0;
   ess_bdr[0] = 1; // boundary attribute 1 (index 0) is fixed

-   // 6. Initialize the hyperelastic operator, the GLVis visualization and print
+   // 7. Initialize the hyperelastic operator, the GLVis visualization and print
   //    the initial energies.
-   std::unique_ptr<HyperelasticOperator> oper;
-   if (nonlinear_solver_type == 0)
-      oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr, visc, mu,
-                                                    K);
-   else
-   {
-      switch (nonlinear_solver_type)
-      {
-         case 1:
-            oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr,
-                                                          visc, mu, K, KIN_NONE);
-            break;
-         case 2:
-            oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr,
-                                                          visc, mu, K, KIN_LINESEARCH);
-            break;
-         case 3:
-            oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr,
-                                                          visc, mu, K, KIN_FP, kinsol_damping, kinsol_aa_n);
-            break;
-         case 4:
-            oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr,
-                                                          visc, mu, K, KIN_PICARD, kinsol_damping, kinsol_aa_n);
-            break;
-      }
-   }
+   HyperelasticOperator oper(fespace, ess_bdr, visc, mu, K, nls_map[nls]);

   socketstream vis_v, vis_w;
   if (visualization)
@@ -418,23 +375,23 @@ int main(int argc, char *argv[])
      vis_w.open(vishost, visport);
      if (vis_w)
      {
-         oper->GetElasticEnergyDensity(x, w);
+         oper.GetElasticEnergyDensity(x, w);
         vis_w.precision(8);
         visualize(vis_w, mesh, &x, &w, "Elastic energy density", true);
      }
   }

-   double ee0 = oper->ElasticEnergy(x.GetTrueVector());
-   double ke0 = oper->KineticEnergy(v.GetTrueVector());
+   double ee0 = oper.ElasticEnergy(x.GetTrueVector());
+   double ke0 = oper.KineticEnergy(v.GetTrueVector());
   cout << "initial elastic energy (EE) = " << ee0 << endl;
   cout << "initial kinetic energy (KE) = " << ke0 << endl;
   cout << "initial   total energy (TE) = " << (ee0 + ke0) << endl;

-   // 7. Define the ODE solver used for time integration. Several implicit
+   // 8. Define the ODE solver used for time integration. Several implicit
   //    singly diagonal implicit Runge-Kutta (SDIRK) methods, as well as
   //    explicit Runge-Kutta methods are available.
   double t = 0.0;
-   oper->SetTime(t);
+   oper.SetTime(t);

   ODESolver *ode_solver = NULL;
   CVODESolver *cvode = NULL;
@@ -458,7 +415,7 @@ int main(int argc, char *argv[])
      case 11:
      case 12:
         cvode = new CVODESolver(CV_BDF);
-         cvode->Init(*oper);
+         cvode->Init(oper);
         cvode->SetSStolerances(reltol, abstol);
         CVodeSetEpsLin(cvode->GetMem(), cvode_eps_lin);
         cvode->SetMaxStep(dt);
@@ -471,7 +428,7 @@ int main(int argc, char *argv[])
      case 13:
      case 14:
         cvode = new CVODESolver(CV_ADAMS);
-         cvode->Init(*oper);
+         cvode->Init(oper);
         cvode->SetSStolerances(reltol, abstol);
         CVodeSetEpsLin(cvode->GetMem(), cvode_eps_lin);
         cvode->SetMaxStep(dt);
@@ -484,13 +441,9 @@ int main(int argc, char *argv[])
      case 15:
      case 16:
         arkode = new ARKStepSolver(ARKStepSolver::IMPLICIT);
-         arkode->Init(*oper);
+         arkode->Init(oper);
         arkode->SetSStolerances(reltol, abstol);
-#if MFEM_SUNDIALS_VERSION < 70100
         ARKStepSetNonlinConvCoef(arkode->GetMem(), arkode_eps_nonlin);
-#else
-         ARKodeSetNonlinConvCoef(arkode->GetMem(), arkode_eps_nonlin);
-#endif
         arkode->SetMaxStep(dt);
         if (ode_solver_type == 15)
         {
@@ -500,16 +453,16 @@ int main(int argc, char *argv[])
      // ARKStep Explicit methods
      case 17:
         arkode = new ARKStepSolver(ARKStepSolver::EXPLICIT);
-         arkode->Init(*oper);
+         arkode->Init(oper);
         arkode->SetSStolerances(reltol, abstol);
         arkode->SetMaxStep(dt);
         ode_solver = arkode; break;
   }

   // Initialize MFEM integrators, SUNDIALS integrators are initialized above
-   if (ode_solver_type < 11) { ode_solver->Init(*oper); }
+   if (ode_solver_type < 11) { ode_solver->Init(oper); }

-   // 8. Perform time-integration (looping over the time iterations, ti, with a
+   // 9. Perform time-integration (looping over the time iterations, ti, with a
   //    time-step dt).
   bool last_step = false;
   for (int ti = 1; !last_step; ti++)
@@ -522,8 +475,8 @@ int main(int argc, char *argv[])

      if (last_step || (ti % vis_steps) == 0)
      {
-         double ee = oper->ElasticEnergy(x.GetTrueVector());
-         double ke = oper->KineticEnergy(v.GetTrueVector());
+         double ee = oper.ElasticEnergy(x.GetTrueVector());
+         double ke = oper.KineticEnergy(v.GetTrueVector());

         cout << "step " << ti << ", t = " << t << ", EE = " << ee << ", KE = "
              << ke << ", ΔTE = " << (ee+ke)-(ee0+ke0) << endl;
@@ -537,14 +490,14 @@ int main(int argc, char *argv[])
            visualize(vis_v, mesh, &x, &v);
            if (vis_w)
            {
-               oper->GetElasticEnergyDensity(x, w);
+               oper.GetElasticEnergyDensity(x, w);
               visualize(vis_w, mesh, &x, &w);
            }
         }
      }
   }

-   // 9. Save the displaced mesh, the velocity and elastic energy.
+   // 10. Save the displaced mesh, the velocity and elastic energy.
   {
      v.SetFromTrueVector(); x.SetFromTrueVector();
      GridFunction *nodes = &x;
@@ -559,11 +512,11 @@ int main(int argc, char *argv[])
      v.Save(velo_ofs);
      ofstream ee_ofs("elastic_energy.sol");
      ee_ofs.precision(8);
-      oper->GetElasticEnergyDensity(x, w);
+      oper.GetElasticEnergyDensity(x, w);
      w.Save(ee_ofs);
   }

-   // 10. Free the used memory.
+   // 11. Free the used memory.
   delete ode_solver;
   delete mesh;

@@ -647,9 +600,7 @@ ReducedSystemOperator::~ReducedSystemOperator()
 HyperelasticOperator::HyperelasticOperator(FiniteElementSpace &f,
                                           Array<int> &ess_bdr, double visc,
                                           double mu, double K,
-                                           int kinsol_nls_type,
-                                           double kinsol_damping,
-                                           int kinsol_aa_n)
+                                           NonlinearSolverType nls_type)
   : TimeDependentOperator(2*f.GetTrueVSize(), 0.0), fespace(f),
     M(&fespace), S(&fespace), H(&fespace),
     viscosity(visc), z(height/2),
@@ -700,28 +651,15 @@ HyperelasticOperator::HyperelasticOperator(FiniteElementSpace &f,
   J_prec = NULL;
 #endif

-   if (kinsol_nls_type > 0)
+   if (nls_type == KINSOL)
   {
-      KINSolver *kinsolver = new KINSolver(kinsol_nls_type, true);
-      if (kinsol_nls_type != KIN_PICARD)
-      {
-         kinsolver->SetJFNK(true);
-         kinsolver->SetLSMaxIter(100);
-      }
-      if (kinsol_aa_n > 0)
-      {
-         kinsolver->EnableAndersonAcc(kinsol_aa_n);
-      }
+      KINSolver *kinsolver = new KINSolver(KIN_NONE, true);
      newton_solver = kinsolver;
      newton_solver->SetOperator(*reduced_oper);
      newton_solver->SetMaxIter(200);
      newton_solver->SetRelTol(rel_tol);
      newton_solver->SetPrintLevel(0);
      kinsolver->SetMaxSetupCalls(4);
-      if (kinsol_damping > 0.0)
-      {
-         kinsolver->SetDamping(kinsol_damping);
-      }
   }
   else
   {
@@ -1,17 +1,15 @@
 //                       MFEM Example 10 - Parallel Version
 //                             SUNDIALS Modification
 //
-// Compile with:
-//    make ex10p            (GNU make)
-//    make sundials_ex10p   (CMake)
+// Compile with: make ex10p
 //
 // Sample runs:
 //    mpirun -np 4 ex10p -m ../../data/beam-quad.mesh -rp 1 -o 2 -s 12 -dt 0.15 -vs 10
 //    mpirun -np 4 ex10p -m ../../data/beam-tri.mesh  -rp 1 -o 2 -s 16 -dt 0.25 -vs 10
 //    mpirun -np 4 ex10p -m ../../data/beam-hex.mesh  -rp 0 -o 2 -s 12 -dt 0.15 -vs 10
-//    mpirun -np 4 ex10p -m ../../data/beam-tri.mesh  -rp 1 -o 2 -s  2 -dt 3 -nls 1
-//    mpirun -np 4 ex10p -m ../../data/beam-quad.mesh -rp 1 -o 2 -s  2 -dt 3 -nls 2
-//    mpirun -np 4 ex10p -m ../../data/beam-hex.mesh  -rs 1 -o 2 -s  2 -dt 3 -nls 4
+//    mpirun -np 4 ex10p -m ../../data/beam-tri.mesh  -rp 1 -o 2 -s  2 -dt 3 -nls kinsol
+//    mpirun -np 4 ex10p -m ../../data/beam-quad.mesh -rp 1 -o 2 -s  2 -dt 3 -nls kinsol
+//    mpirun -np 4 ex10p -m ../../data/beam-hex.mesh  -rs 1 -o 2 -s  2 -dt 3 -nls kinsol
 //    mpirun -np 4 ex10p -m ../../data/beam-quad.mesh -rp 1 -o 2 -s 14 -dt 0.15 -vs 10
 //    mpirun -np 4 ex10p -m ../../data/beam-tri.mesh  -rp 1 -o 2 -s 17 -dt 5e-3 -vs 60
 //    mpirun -np 4 ex10p -m ../../data/beam-hex.mesh  -rp 0 -o 2 -s 14 -dt 0.15 -vs 10
@@ -101,11 +99,16 @@ protected:
   double saved_gamma; // saved gamma value from implicit setup

 public:
+   /// Solver type to use in the ImplicitSolve() method, used by SDIRK methods.
+   enum NonlinearSolverType
+   {
+      NEWTON = 0, ///< Use MFEM's plain NewtonSolver
+      KINSOL = 1  ///< Use SUNDIALS' KINSOL (through MFEM's class KINSolver)
+   };

   HyperelasticOperator(ParFiniteElementSpace &f, Array<int> &ess_bdr,
                        double visc, double mu, double K,
-                        int kinsol_nls_type = -1, double kinsol_damping = 0.0,
-                        int kinsol_aa_n = 0);
+                        NonlinearSolverType nls_type);

   /// Compute the right-hand side of the ODE system.
   virtual void Mult(const Vector &vx, Vector &dvx_dt) const;
@@ -230,10 +233,8 @@ int main(int argc, char *argv[])
   double mu = 0.25;
   double K = 5.0;
   bool visualization = true;
-   int nonlinear_solver_type = 0;
+   const char *nls = "newton";
   int vis_steps = 1;
-   double kinsol_damping = 0.0;
-   int kinsol_aa_n = -1;

   // Relative and absolute tolerances for CVODE and ARKODE.
   const double reltol = 1e-1, abstol = 1e-1;
@@ -272,18 +273,9 @@ int main(int argc, char *argv[])
                  "15 - ARKODE implicit, approximate Jacobian,\n\t"
                  "16 - ARKODE implicit, specified Jacobian,\n\t"
                  "17 - ARKODE explicit, 4th order.");
-   args.AddOption(&nonlinear_solver_type, "-nls", "--nonlinear-solver",
-                  "Nonlinear system solver:\n\t"
-                  "0  - MFEM Newton method,\n\t"
-                  "1  - KINSOL Newton method,\n\t"
-                  "2  - KINSOL Newton method with globalization,\n\t"
-                  "3  - KINSOL fixed-point method (with or without AA),\n\t"
-                  "4  - KINSOL Picard method (with or without AA).");
-   args.AddOption(&kinsol_damping, "-damp", "--kinsol-damping",
-                  "Picard or Fixed-Point damping parameter (only valid with KINSOL): "
-                  "0 < d <= 1.0");
-   args.AddOption(&kinsol_aa_n, "-aan", "--anderson-subspace",
-                  "Anderson Acceleration subspace size (only valid with KINSOL)");
+   args.AddOption(&nls, "-nls", "--nonlinear-solver",
+                  "Nonlinear systems solver: "
+                  "\"newton\" (plain Newton) or \"kinsol\" (KINSOL).");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -323,42 +315,27 @@ int main(int argc, char *argv[])
      return 1;
   }

-   // check for valid nonlinear solver options
-   if (nonlinear_solver_type < 0 || nonlinear_solver_type > 4)
-   {
-      if (myid == 0)
-      {
-         cout << "Unknown nonlinear solver type: " << nonlinear_solver_type
-              << "\n";
-      }
-      return 1;
-   }
-   if (kinsol_damping > 0.0 &&
-       !(nonlinear_solver_type == 3 || nonlinear_solver_type == 4))
-   {
-      if (myid == 0)
-      {
-         cout << "Only KINSOL fixed-point and Picard methods can use damping\n";
-      }
-      return 1;
-   }
-   if (kinsol_aa_n > 0 &&
-       !(nonlinear_solver_type == 3 || nonlinear_solver_type == 4))
-   {
-      if (myid == 0)
-      {
-         cout << "Only KINSOL fixed-point and Picard methods can use AA\n";
-      }
-      return 1;
-   }
-
   // 3. Read the serial mesh from the given mesh file on all processors. We can
   //    handle triangular, quadrilateral, tetrahedral and hexahedral meshes
   //    with the same code.
   Mesh *mesh = new Mesh(mesh_file, 1, 1);
   int dim = mesh->Dimension();

-   // 4. Refine the mesh in serial to increase the resolution. In this example
+   // 4. Nonlinear solver
+   map<string,HyperelasticOperator::NonlinearSolverType> nls_map;
+   nls_map["newton"] = HyperelasticOperator::NEWTON;
+   nls_map["kinsol"] = HyperelasticOperator::KINSOL;
+   if (nls_map.find(nls) == nls_map.end())
+   {
+      if (myid == 0)
+      {
+         cout << "Unknown type of nonlinear solver: " << nls << endl;
+      }
+      delete mesh;
+      return 4;
+   }
+
+   // 5. Refine the mesh in serial to increase the resolution. In this example
   //    we do 'ser_ref_levels' of uniform refinement, where 'ser_ref_levels' is
   //    a command-line parameter.
   for (int lev = 0; lev < ser_ref_levels; lev++)
@@ -366,7 +343,7 @@ int main(int argc, char *argv[])
      mesh->UniformRefinement();
   }

-   // 5. Define a parallel mesh by a partitioning of the serial mesh. Refine
+   // 6. Define a parallel mesh by a partitioning of the serial mesh. Refine
   //    this mesh further in parallel to increase the resolution. Once the
   //    parallel mesh is defined, the serial mesh can be deleted.
   ParMesh *pmesh = new ParMesh(MPI_COMM_WORLD, *mesh);
@@ -376,7 +353,7 @@ int main(int argc, char *argv[])
      pmesh->UniformRefinement();
   }

-   // 6. Define the parallel vector finite element spaces representing the mesh
+   // 7. Define the parallel vector finite element spaces representing the mesh
   //    deformation x_gf, the velocity v_gf, and the initial configuration,
   //    x_ref. Define also the elastic energy density, w_gf, which is in a
   //    discontinuous higher-order space. Since x and v are integrated in time
@@ -408,7 +385,7 @@ int main(int argc, char *argv[])
   ParFiniteElementSpace w_fespace(pmesh, &w_fec);
   ParGridFunction w_gf(&w_fespace);

-   // 7. Set the initial conditions for v_gf, x_gf and vx, and define the
+   // 8. Set the initial conditions for v_gf, x_gf and vx, and define the
   //    boundary conditions on a beam-like mesh (see description above).
   VectorFunctionCoefficient velo(dim, InitialVelocity);
   v_gf.ProjectCoefficient(velo);
@@ -423,38 +400,9 @@ int main(int argc, char *argv[])
   ess_bdr = 0;
   ess_bdr[0] = 1; // boundary attribute 1 (index 0) is fixed

-   // 8. Initialize the hyperelastic operator, the GLVis visualization and print
+   // 9. Initialize the hyperelastic operator, the GLVis visualization and print
   //    the initial energies.
-   std::unique_ptr<HyperelasticOperator> oper;
-   if (nonlinear_solver_type == 0)
-      oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr, visc, mu,
-                                                    K);
-   else
-   {
-      switch (nonlinear_solver_type)
-      {
-         case 1:
-            oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr,
-                                                          visc, mu, K, KIN_NONE);
-            break;
-         case 2:
-            oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr,
-                                                          visc, mu, K, KIN_LINESEARCH);
-            break;
-         case 3:
-            oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr,
-                                                          visc, mu, K, KIN_FP, kinsol_damping, kinsol_aa_n);
-            break;
-         case 4:
-            oper = std::make_unique<HyperelasticOperator>(fespace, ess_bdr,
-                                                          visc, mu, K, KIN_PICARD, kinsol_damping, kinsol_aa_n);
-            break;
-         default:
-            cout << "Unknown type of nonlinear solver: "
-                 << nonlinear_solver_type << endl;
-            return 4;
-      }
-   }
+   HyperelasticOperator oper(fespace, ess_bdr, visc, mu, K, nls_map[nls]);

   socketstream vis_v, vis_w;
   if (visualization)
@@ -470,14 +418,14 @@ int main(int argc, char *argv[])
      vis_w.open(vishost, visport);
      if (vis_w)
      {
-         oper->GetElasticEnergyDensity(x_gf, w_gf);
+         oper.GetElasticEnergyDensity(x_gf, w_gf);
         vis_w.precision(8);
         visualize(vis_w, pmesh, &x_gf, &w_gf, "Elastic energy density", true);
      }
   }

-   double ee0 = oper->ElasticEnergy(x_gf);
-   double ke0 = oper->KineticEnergy(v_gf);
+   double ee0 = oper.ElasticEnergy(x_gf);
+   double ke0 = oper.KineticEnergy(v_gf);
   if (myid == 0)
   {
      cout << "initial elastic energy (EE) = " << ee0 << endl;
@@ -485,11 +433,11 @@ int main(int argc, char *argv[])
      cout << "initial   total energy (TE) = " << (ee0 + ke0) << endl;
   }

-   // 9. Define the ODE solver used for time integration. Several implicit
-   //    singly diagonal implicit Runge-Kutta (SDIRK) methods, as well as
-   //    explicit Runge-Kutta methods are available.
+   // 10. Define the ODE solver used for time integration. Several implicit
+   //     singly diagonal implicit Runge-Kutta (SDIRK) methods, as well as
+   //     explicit Runge-Kutta methods are available.
   double t = 0.0;
-   oper->SetTime(t);
+   oper.SetTime(t);

   ODESolver *ode_solver = NULL;
   CVODESolver *cvode = NULL;
@@ -513,7 +461,7 @@ int main(int argc, char *argv[])
      case 11:
      case 12:
         cvode = new CVODESolver(MPI_COMM_WORLD, CV_BDF);
-         cvode->Init(*oper);
+         cvode->Init(oper);
         cvode->SetSStolerances(reltol, abstol);
         CVodeSetEpsLin(cvode->GetMem(), cvode_eps_lin);
         cvode->SetMaxStep(dt);
@@ -526,7 +474,7 @@ int main(int argc, char *argv[])
      case 13:
      case 14:
         cvode = new CVODESolver(MPI_COMM_WORLD, CV_ADAMS);
-         cvode->Init(*oper);
+         cvode->Init(oper);
         cvode->SetSStolerances(reltol, abstol);
         CVodeSetEpsLin(cvode->GetMem(), cvode_eps_lin);
         cvode->SetMaxStep(dt);
@@ -539,13 +487,9 @@ int main(int argc, char *argv[])
      case 15:
      case 16:
         arkode = new ARKStepSolver(MPI_COMM_WORLD, ARKStepSolver::IMPLICIT);
-         arkode->Init(*oper);
+         arkode->Init(oper);
         arkode->SetSStolerances(reltol, abstol);
-#if MFEM_SUNDIALS_VERSION < 70100
         ARKStepSetNonlinConvCoef(arkode->GetMem(), arkode_eps_nonlin);
-#else
-         ARKodeSetNonlinConvCoef(arkode->GetMem(), arkode_eps_nonlin);
-#endif
         arkode->SetMaxStep(dt);
         if (ode_solver_type == 15)
         {
@@ -555,16 +499,16 @@ int main(int argc, char *argv[])
      // ARKStep Explicit methods
      case 17:
         arkode = new ARKStepSolver(MPI_COMM_WORLD, ARKStepSolver::EXPLICIT);
-         arkode->Init(*oper);
+         arkode->Init(oper);
         arkode->SetSStolerances(reltol, abstol);
         arkode->SetMaxStep(dt);
         ode_solver = arkode; break;
   }

   // Initialize MFEM integrators, SUNDIALS integrators are initialized above
-   if (ode_solver_type < 11) { ode_solver->Init(*oper); }
+   if (ode_solver_type < 11) { ode_solver->Init(oper); }

-   // 10. Perform time-integration
+   // 11. Perform time-integration
   //     (looping over the time iterations, ti, with a time-step dt).
   bool last_step = false;
   for (int ti = 1; !last_step; ti++)
@@ -579,8 +523,8 @@ int main(int argc, char *argv[])
      {
         v_gf.SetFromTrueVector(); x_gf.SetFromTrueVector();

-         double ee = oper->ElasticEnergy(x_gf);
-         double ke = oper->KineticEnergy(v_gf);
+         double ee = oper.ElasticEnergy(x_gf);
+         double ke = oper.KineticEnergy(v_gf);

         if (myid == 0)
         {
@@ -596,14 +540,14 @@ int main(int argc, char *argv[])
            visualize(vis_v, pmesh, &x_gf, &v_gf);
            if (vis_w)
            {
-               oper->GetElasticEnergyDensity(x_gf, w_gf);
+               oper.GetElasticEnergyDensity(x_gf, w_gf);
               visualize(vis_w, pmesh, &x_gf, &w_gf);
            }
         }
      }
   }

-   // 11. Save the displaced mesh, the velocity and elastic energy.
+   // 12. Save the displaced mesh, the velocity and elastic energy.
   {
      v_gf.SetFromTrueVector(); x_gf.SetFromTrueVector();
      GridFunction *nodes = &x_gf;
@@ -624,11 +568,11 @@ int main(int argc, char *argv[])
      v_gf.Save(velo_ofs);
      ofstream ee_ofs(ee_name.str().c_str());
      ee_ofs.precision(8);
-      oper->GetElasticEnergyDensity(x_gf, w_gf);
+      oper.GetElasticEnergyDensity(x_gf, w_gf);
      w_gf.Save(ee_ofs);
   }

-   // 12. Free the used memory.
+   // 13. Free the used memory.
   delete ode_solver;
   delete pmesh;

@@ -718,10 +662,7 @@ ReducedSystemOperator::~ReducedSystemOperator()
 HyperelasticOperator::HyperelasticOperator(ParFiniteElementSpace &f,
                                           Array<int> &ess_bdr, double visc,
                                           double mu, double K,
-                                           int kinsol_nls_type,
-                                           double kinsol_damping,
-                                           int kinsol_aa_n)
-
+                                           NonlinearSolverType nls_type)
   : TimeDependentOperator(2*f.TrueVSize(), 0.0), fespace(f),
     M(&fespace), S(&fespace), H(&fespace),
     viscosity(visc), M_solver(f.GetComm()), z(height/2),
@@ -773,28 +714,17 @@ HyperelasticOperator::HyperelasticOperator(ParFiniteElementSpace &f,
   J_minres->SetPreconditioner(*J_prec);
   J_solver = J_minres;

-   if (kinsol_nls_type > 0)
+   if (nls_type == KINSOL)
   {
-      KINSolver *kinsolver = new KINSolver(f.GetComm(), kinsol_nls_type, true);
-      if (kinsol_nls_type != KIN_PICARD)
-      {
-         kinsolver->SetJFNK(true);
-         kinsolver->SetLSMaxIter(100);
-      }
-      if (kinsol_aa_n > 0)
-      {
-         kinsolver->EnableAndersonAcc(kinsol_aa_n);
-      }
+      KINSolver *kinsolver = new KINSolver(f.GetComm(), KIN_LINESEARCH, true);
+      kinsolver->SetJFNK(true);
+      kinsolver->SetLSMaxIter(100);
      newton_solver = kinsolver;
      newton_solver->SetOperator(*reduced_oper);
      newton_solver->SetMaxIter(200);
      newton_solver->SetRelTol(rel_tol);
-      newton_solver->SetPrintLevel(0);
+      newton_solver->SetPrintLevel(1);
      kinsolver->SetMaxSetupCalls(4);
-      if (kinsol_damping > 0.0)
-      {
-         kinsolver->SetDamping(kinsol_damping);
-      }
   }
   else
   {
@@ -1,21 +1,15 @@
 //                                MFEM Example 16
 //                             SUNDIALS Modification
 //
-// Compile with:
-//    make ex16             (GNU make)
-//    make sundials_ex16    (CMake)
+// Compile with: make ex16
 //
 // Sample runs:  ex16
 //               ex16 -m ../../data/inline-tri.mesh
 //               ex16 -m ../../data/disc-nurbs.mesh -tf 2
 //               ex16 -s 12 -a 0.0 -k 1.0
-//               ex16 -s 15 -a 0.0 -k 1.0
 //               ex16 -s 8 -a 1.0 -k 0.0 -dt 1e-4 -tf 5e-2 -vs 25
-//               ex16 -s 11 -a 1.0 -k 0.0 -dt 1e-4 -tf 5e-2 -vs 25
 //               ex16 -s 9 -a 0.5 -k 0.5 -o 4 -dt 1e-4 -tf 2e-2 -vs 25
-//               ex16 -s 12 -a 0.5 -k 0.5 -o 4 -dt 1e-4 -tf 2e-2 -vs 25
 //               ex16 -s 10 -dt 1.0e-4 -tf 4.0e-2 -vs 40
-//               ex16 -s 13 -dt 1.0e-4 -tf 4.0e-2 -vs 40
 //               ex16 -m ../../data/fichera-q2.mesh
 //               ex16 -m ../../data/escher.mesh
 //               ex16 -m ../../data/beam-tet.mesh -tf 10 -dt 0.1
@@ -43,102 +37,75 @@
 using namespace std;
 using namespace mfem;

-/** After spatial discretization, the conduction model is expressed as
+/** After spatial discretization, the conduction model can be written as:
 *
- *   M du/dt = - K(u) u
+ *     du/dt = M^{-1}(-Ku)
 *
 *  where u is the vector representing the temperature, M is the mass matrix,
- *  and K(u) is the diffusion operator with diffusivity depending on u:
+ *  and K is the diffusion operator with diffusivity depending on u:
 *  (\kappa + \alpha u).
 *
- *  Class ConductionOperatorOperator represents the above ODE operator in the
- *  general form F(u, k, t) = G(u, t) where
- *
- *    1. F(u, du/dt, t) = du/dt            (ODE is expressed in EXPLICIT form)
- *       G(u, t) = - inv(M) K(u) u
- *    2. F(u, du/dt, t) = M du/dt          (ODE is expressed in IMPLICIT form)
- *       G(u, t) = - K(u) u
+ *  Class ConductionOperator represents the right-hand side of the above ODE.
 */
 class ConductionOperator : public TimeDependentOperator
 {
+protected:
   FiniteElementSpace &fespace;
   Array<int> ess_tdof_list; // this list remains empty for pure Neumann b.c.

-   BilinearForm M;
-   SparseMatrix Mmat;
+   BilinearForm *M;
+   BilinearForm *K;

-   const real_t alpha, kappa;
-   std::unique_ptr<BilinearForm> K;
-   SparseMatrix Kmat;
-
-   std::unique_ptr<SparseMatrix> T; // T = M + gam K(u)
+   SparseMatrix Mmat, Kmat;
+   SparseMatrix *T; // T = M + dt K

   CGSolver M_solver; // Krylov solver for inverting the mass matrix M
   DSmoother M_prec;  // Preconditioner for the mass matrix M

-   CGSolver T_solver; // Implicit solver for T = M + gam K(u)
+   CGSolver T_solver; // Implicit solver for T = M + dt K
   DSmoother T_prec;  // Preconditioner for the implicit solver

+   double alpha, kappa;
+
   mutable Vector z; // auxiliary vector

 public:
+   ConductionOperator(FiniteElementSpace &f, double alpha, double kappa,
+                      const Vector &u);

-   ConductionOperator(FiniteElementSpace &f, const real_t alpha,
-                      const real_t kappa, const Vector &u,
-                      const Type &ode_expression_type);
+   virtual void Mult(const Vector &u, Vector &du_dt) const;

-   // Compute K(u_n) for use as an approximation in - K(u) u
-   void SetConductionTensor(const Vector &u);
+   /** Solve the Backward-Euler equation: k = f(u + dt*k, t), for the unknown k.
+       This is the only requirement for high-order SDIRK implicit integration.*/
+   virtual void ImplicitSolve(const double dt, const Vector &u, Vector &k);

-   /** Compute G(u, t) as defined in the IMPLICIT expression form of the ODE
-    operator, i.e., @a v = - K(u_n) @a u. Note that K(u_n) is an
-      approximation to K(u). */
-   void ExplicitMult(const Vector &u, Vector &v) const override;
+   /// Custom Jacobian system solver for the SUNDIALS time integrators.
+   /** For the ODE system represented by ConductionOperator

-   /** Solve for k in F(u, k, t) = G(u, t) for either EXPLICIT or IMPLICIT
-       expression forms of the ODE operator, i.e., @a k = - inv(M) K(u_n) @a u.
-       Note that K(u_n) is an approximation to K(u). */
-   void Mult(const Vector &u, Vector &k) const override;
+           M du/dt = -K(u),

-   /** Solve for k in F(u + gam*k, k, t) = G(u + gam*k, t) for either EXPLICIT
-       or IMPLICIT expression forms of the ODE operator, i.e.,
-       [ M + @a gam K(u_n) ] @a k = - K(u_n) @a u . Note that K(u_n) is an
-       approximation to K(u). */
-   void ImplicitSolve(const real_t gam, const Vector &u, Vector &k) override;
+       this class facilitates the solution of linear systems of the form

-   /** Setup to solve for dk in [dF/dk + gam*dF/du - gam*dG/du] dk = G - F for
-       either EXPLICIT or IMPLICIT expression forms of the ODE operator, i.e.,
-       [M - @a gam Jf(u)] dk = G - F, where Jf(u) is an approximation of the
-       Jacobian of -K(u) u. The approximation chosen here is Jf(u) = -K(u_n). */
-   int SUNImplicitSetup(const Vector &u, const Vector &fu, int jok, int *jcur,
-                        real_t gam) override;
+           (M + γK) y = M b,

-   /** Solve for @a dk in the system in SUNImplicitSetup to the given tolerance,
-       with the residual @a r providing either
-        1. @a r = G - F = inv(M) f(u) - k       (EXPLICIT expression form)
-        1. @a r = G - F = f(u) - M k            (IMPLICIT expression form)
-       */
-   int SUNImplicitSolve(const Vector &r, Vector &dk, real_t tol) override;
+       for given b, u (not used), and γ = GetTimeStep(). */

-   int SUNMassSetup() override;
+   /** Setup the system (M + dt K) x = M b. This method is used by the implicit
+       SUNDIALS solvers. */
+   virtual int SUNImplicitSetup(const Vector &x, const Vector &fx,
+                                int jok, int *jcur, double gamma);

-   int SUNMassSolve(const Vector &b, Vector &x, real_t tol) override;
+   /** Solve the system (M + dt K) x = M b. This method is used by the implicit
+       SUNDIALS solvers. */
+   virtual int SUNImplicitSolve(const Vector &b, Vector &x, double tol);

-   int SUNMassMult(const Vector &x, Vector &v) override;
+   /// Update the diffusion BilinearForm K using the given true-dof vector `u`.
+   void SetParameters(const Vector &u);
+
+   virtual ~ConductionOperator();
 };

-real_t InitialTemperature(const Vector &x)
-{
-   if (x.Norml2() < 0.5)
-   {
-      return 2.0;
-   }
-   else
-   {
-      return 1.0;
-   }
-}
-
+double InitialTemperature(const Vector &x);

 int main(int argc, char *argv[])
 {
@@ -150,16 +117,16 @@ int main(int argc, char *argv[])
   int ref_levels = 2;
   int order = 2;
   int ode_solver_type = 9; // CVODE implicit BDF
-   real_t t_final = 0.5;
-   real_t dt = 1.0e-2;
-   real_t alpha = 1.0e-2;
-   real_t kappa = 0.5;
+   double t_final = 0.5;
+   double dt = 1.0e-2;
+   double alpha = 1.0e-2;
+   double kappa = 0.5;
   bool visualization = true;
   bool visit = false;
   int vis_steps = 5;

   // Relative and absolute tolerances for CVODE and ARKODE.
-   const real_t reltol = 1e-4, abstol = 1e-4;
+   const double reltol = 1e-4, abstol = 1e-4;

   int precision = 8;
   cout.precision(precision);
@@ -184,10 +151,7 @@ int main(int argc, char *argv[])
                  "9  - CVODE (implicit BDF),\n\t"
                  "10 - ARKODE (default explicit),\n\t"
                  "11 - ARKODE (explicit Fehlberg-6-4-5),\n\t"
-                  "12 - ARKODE (default implicit),\n\t"
-                  "13 - ARKODE (default explicit with MFEM mass solve),\n\t"
-                  "14 - ARKODE (explicit Fehlberg-6-4-5 with MFEM mass solve),\n\t"
-                  "15 - ARKODE (default implicit with MFEM mass solve).");
+                  "12 - ARKODE (default impicit).");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -210,13 +174,16 @@ int main(int argc, char *argv[])
      args.PrintUsage(cout);
      return 1;
   }
+   if (ode_solver_type < 1 || ode_solver_type > 12)
+   {
+      cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+      return 3;
+   }
   args.PrintOptions(cout);

-   bool use_mass_solver = ode_solver_type >= 13;
-
   // 2. Read the mesh from the given mesh file. We can handle triangular,
   //    quadrilateral, tetrahedral and hexahedral meshes with the same code.
-   std::unique_ptr<Mesh> mesh(new Mesh(mesh_file, 1, 1));
+   Mesh *mesh = new Mesh(mesh_file, 1, 1);
   int dim = mesh->Dimension();

   // 3. Refine the mesh to increase the resolution. In this example we do
@@ -230,7 +197,7 @@ int main(int argc, char *argv[])
   // 4. Define the vector finite element space representing the current and the
   //    initial temperature, u_ref.
   H1_FECollection fe_coll(order, dim);
-   FiniteElementSpace fespace(mesh.get(), &fe_coll);
+   FiniteElementSpace fespace(mesh, &fe_coll);

   int fe_size = fespace.GetTrueVSize();
   cout << "Number of temperature unknowns: " << fe_size << endl;
@@ -244,17 +211,8 @@ int main(int argc, char *argv[])
   Vector u;
   u_gf.GetTrueDofs(u);

-   // 6. Initialize the conduction ODE operator and the visualization.
-   ConductionOperator::Type ode_expression_type;
-   if (use_mass_solver)
-   {
-      ode_expression_type = ConductionOperator::Type::IMPLICIT;
-   }
-   else
-   {
-      ode_expression_type = ConductionOperator::Type::EXPLICIT;
-   }
-   ConductionOperator oper(fespace, alpha, kappa, u, ode_expression_type);
+   // 6. Initialize the conduction operator and the visualization.
+   ConductionOperator oper(fespace, alpha, kappa, u);

   u_gf.SetFromTrueDofs(u);
   {
@@ -266,7 +224,7 @@ int main(int argc, char *argv[])
      u_gf.Save(osol);
   }

-   VisItDataCollection visit_dc("Example16", mesh.get());
+   VisItDataCollection visit_dc("Example16", mesh);
   visit_dc.RegisterField("temperature", &u_gf);
   if (visit)
   {
@@ -300,75 +258,52 @@ int main(int argc, char *argv[])
   }

   // 7. Define the ODE solver used for time integration.
-   real_t t = 0.0;
-   std::unique_ptr<ODESolver> ode_solver;
+   double t = 0.0;
+   ODESolver *ode_solver = NULL;
+   CVODESolver *cvode = NULL;
+   ARKStepSolver *arkode = NULL;
   switch (ode_solver_type)
   {
      // MFEM explicit methods
-      case 1: ode_solver = std::make_unique<ForwardEulerSolver>(); break;
-      case 2: ode_solver = std::make_unique<RK2Solver>(0.5); break; // midpoint method
-      case 3: ode_solver = std::make_unique<RK3SSPSolver>(); break;
-      case 4: ode_solver = std::make_unique<RK4Solver>(); break;
+      case 1: ode_solver = new ForwardEulerSolver; break;
+      case 2: ode_solver = new RK2Solver(0.5); break; // midpoint method
+      case 3: ode_solver = new RK3SSPSolver; break;
+      case 4: ode_solver = new RK4Solver; break;
      // MFEM implicit L-stable methods
-      case 5: ode_solver = std::make_unique<BackwardEulerSolver>(); break;
-      case 6: ode_solver = std::make_unique<SDIRK23Solver>(2); break;
-      case 7: ode_solver = std::make_unique<SDIRK33Solver>(); break;
+      case 5: ode_solver = new BackwardEulerSolver; break;
+      case 6: ode_solver = new SDIRK23Solver(2); break;
+      case 7: ode_solver = new SDIRK33Solver; break;
      // CVODE
      case 8:
-      case 9:
-      {
-         int cvode_solver_type;
-         if (ode_solver_type == 8)
-         {
-            cvode_solver_type = CV_ADAMS;
-         }
-         else
-         {
-            cvode_solver_type = CV_BDF;
-         }
-         std::unique_ptr<CVODESolver> cvode(new CVODESolver(cvode_solver_type));
+         cvode = new CVODESolver(CV_ADAMS);
         cvode->Init(oper);
         cvode->SetSStolerances(reltol, abstol);
         cvode->SetMaxStep(dt);
-         ode_solver = std::move(cvode);
-         break;
-      }
+         ode_solver = cvode; break;
+      case 9:
+         cvode = new CVODESolver(CV_BDF);
+         cvode->Init(oper);
+         cvode->SetSStolerances(reltol, abstol);
+         cvode->SetMaxStep(dt);
+         ode_solver = cvode; break;
      // ARKODE
      case 10:
      case 11:
-      case 12:
-      case 13:
-      case 14:
-      case 15:
-      {
-         ARKStepSolver::Type arkode_solver_type;
-         if (ode_solver_type == 12 || ode_solver_type == 15)
-         {
-            arkode_solver_type = ARKStepSolver::IMPLICIT;
-         }
-         else
-         {
-            arkode_solver_type = ARKStepSolver::EXPLICIT;
-         }
-         std::unique_ptr<ARKStepSolver> arkode(
-            new ARKStepSolver(arkode_solver_type));
+         arkode = new ARKStepSolver(ARKStepSolver::EXPLICIT);
         arkode->Init(oper);
         arkode->SetSStolerances(reltol, abstol);
         arkode->SetMaxStep(dt);
-         if (ode_solver_type == 11 || ode_solver_type == 14)
+         if (ode_solver_type == 11)
         {
            arkode->SetERKTableNum(ARKODE_FEHLBERG_13_7_8);
         }
-         if (use_mass_solver)
-         {
-            arkode->UseMFEMMassLinearSolver(SUNFALSE);
-         }
-         ode_solver = std::move(arkode);
-         break;
-      }
-      default:
-         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
-         return 3;
+         ode_solver = arkode; break;
+      case 12:
+         arkode = new ARKStepSolver(ARKStepSolver::IMPLICIT);
+         arkode->Init(oper);
+         arkode->SetSStolerances(reltol, abstol);
+         arkode->SetMaxStep(dt);
+         ode_solver = arkode; break;
   }

   // Initialize MFEM integrators, SUNDIALS integrators are initialized above
@@ -376,14 +311,8 @@ int main(int argc, char *argv[])

   // Since we want to update the diffusion coefficient after every time step,
   // we need to use the "one-step" mode of the SUNDIALS solvers.
-   if (CVODESolver* cvode = dynamic_cast<CVODESolver*>(ode_solver.get()))
-   {
-      cvode->SetStepMode(CV_ONE_STEP);
-   }
-   else if (ARKStepSolver* arkode = dynamic_cast<ARKStepSolver*>(ode_solver.get()))
-   {
-      arkode->SetStepMode(ARK_ONE_STEP);
-   }
+   if (cvode) { cvode->SetStepMode(CV_ONE_STEP); }
+   if (arkode) { arkode->SetStepMode(ARK_ONE_STEP); }

   // 8. Perform time-integration (looping over the time iterations, ti, with a
   //    time-step dt).
@@ -394,7 +323,7 @@ int main(int argc, char *argv[])
   bool last_step = false;
   for (int ti = 1; !last_step; ti++)
   {
-      real_t dt_real = min(dt, t_final - t);
+      double dt_real = min(dt, t_final - t);

      // Note that since we are using the "one-step" mode of the SUNDIALS
      // solvers, they will, generally, step over the final time and will not
@@ -408,14 +337,8 @@ int main(int argc, char *argv[])
      if (last_step || (ti % vis_steps) == 0)
      {
         cout << "step " << ti << ", t = " << t << endl;
-         if (CVODESolver* cvode = dynamic_cast<CVODESolver*>(ode_solver.get()))
-         {
-            cvode->PrintInfo();
-         }
-         else if (ARKStepSolver* arkode = dynamic_cast<ARKStepSolver*>(ode_solver.get()))
-         {
-            arkode->PrintInfo();
-         }
+         if (cvode) { cvode->PrintInfo(); }
+         if (arkode) { arkode->PrintInfo(); }

         u_gf.SetFromTrueDofs(u);
         if (visualization)
@@ -430,153 +353,137 @@ int main(int argc, char *argv[])
            visit_dc.Save();
         }
      }
-      oper.SetConductionTensor(u);
+      oper.SetParameters(u);
   }
   tic_toc.Stop();
   cout << "Done, " << tic_toc.RealTime() << "s." << endl;

   // 9. Save the final solution. This output can be viewed later using GLVis:
   //    "glvis -m ex16.mesh -g ex16-final.gf".
-   u_gf.Save("ex16-final.gf", precision);
+   {
+      ofstream osol("ex16-final.gf");
+      osol.precision(precision);
+      u_gf.Save(osol);
+   }
+
+   // 10. Free the used memory.
+   delete ode_solver;
+   delete mesh;

   return 0;
 }

-ConductionOperator::ConductionOperator(FiniteElementSpace &fes,
-                                       const real_t alpha, const real_t kappa,
-                                       const Vector &u,
-                                       const Type &ode_expression_type)
-   : TimeDependentOperator(fes.GetTrueVSize(), 0.0, ode_expression_type),
-     fespace(fes), M(&fespace), alpha(alpha), kappa(kappa), z(height)
+ConductionOperator::ConductionOperator(FiniteElementSpace &f, double al,
+                                       double kap, const Vector &u)
+   : TimeDependentOperator(f.GetTrueVSize(), 0.0), fespace(f), M(NULL), K(NULL),
+     T(NULL), z(height)
 {
-   // specify a relative tolerance for all solves with MFEM integrators
-   const real_t rel_tol = 1e-8;
+   const double rel_tol = 1e-8;

-   M.AddDomainIntegrator(new MassIntegrator());
-   M.Assemble();
-   M.FormSystemMatrix(ess_tdof_list, Mmat);
+   M = new BilinearForm(&fespace);
+   M->AddDomainIntegrator(new MassIntegrator());
+   M->Assemble();
+   M->FormSystemMatrix(ess_tdof_list, Mmat);

   M_solver.iterative_mode = false;
-   M_solver.SetRelTol(rel_tol); // will be overwritten with SUNDIALS integrators
+   M_solver.SetRelTol(rel_tol);
   M_solver.SetAbsTol(0.0);
   M_solver.SetMaxIter(50);
   M_solver.SetPrintLevel(0);
   M_solver.SetPreconditioner(M_prec);
   M_solver.SetOperator(Mmat);

+   alpha = al;
+   kappa = kap;
+
   T_solver.iterative_mode = false;
-   T_solver.SetRelTol(rel_tol); // will be overwritten with SUNDIALS integrators
+   T_solver.SetRelTol(rel_tol);
   T_solver.SetAbsTol(0.0);
   T_solver.SetMaxIter(100);
   T_solver.SetPrintLevel(0);
   T_solver.SetPreconditioner(T_prec);

-   SetConductionTensor(u);
+   SetParameters(u);
 }

-void ConductionOperator::SetConductionTensor(const Vector &u)
+void ConductionOperator::Mult(const Vector &u, Vector &du_dt) const
+{
+   // Compute:
+   //    du_dt = M^{-1}*-K(u)
+   // for du_dt
+   Kmat.Mult(u, z);
+   z.Neg(); // z = -z
+   M_solver.Mult(z, du_dt);
+}
+
+void ConductionOperator::ImplicitSolve(const double dt,
+                                       const Vector &u, Vector &du_dt)
+{
+   // Solve the equation:
+   //    du_dt = M^{-1}*[-K(u + dt*du_dt)]
+   // for du_dt
+   if (T) { delete T; }
+   T = Add(1.0, Mmat, dt, Kmat);
+   T_solver.SetOperator(*T);
+   Kmat.Mult(u, z);
+   z.Neg();
+   T_solver.Mult(z, du_dt);
+}
+
+void ConductionOperator::SetParameters(const Vector &u)
 {
-   // Compute K(u_n).
   GridFunction u_alpha_gf(&fespace);
   u_alpha_gf.SetFromTrueDofs(u);
   for (int i = 0; i < u_alpha_gf.Size(); i++)
   {
      u_alpha_gf(i) = kappa + alpha*u_alpha_gf(i);
   }
+
+   delete K;
+   K = new BilinearForm(&fespace);
+
   GridFunctionCoefficient u_coeff(&u_alpha_gf);

-   K = std::make_unique<BilinearForm>(&fespace);
   K->AddDomainIntegrator(new DiffusionIntegrator(u_coeff));
   K->Assemble();
   K->FormSystemMatrix(ess_tdof_list, Kmat);
 }

-void ConductionOperator::ExplicitMult(const Vector &u, Vector &v) const
+int ConductionOperator::SUNImplicitSetup(const Vector &x,
+                                         const Vector &fx, int jok, int *jcur,
+                                         double gamma)
 {
-   // Compute - K(u_n) u.
-   Kmat.Mult(u, v);
-   v.Neg();
-}
-
-void ConductionOperator::Mult(const Vector &u, Vector &k) const
-{
-   // Compute - inv(M) K(u_n) u.
-   ExplicitMult(u, z);
-   M_solver.Mult(z, k);
-}
-
-void ConductionOperator::ImplicitSolve(const real_t gam, const Vector &u,
-                                       Vector &k)
-{
-   // Solve for k in M k = - K(u_n) [u + gam*k].
-   ExplicitMult(u, z);
-   T = std::unique_ptr<SparseMatrix>(Add(1.0, Mmat, gam, Kmat));
+   // Setup the ODE Jacobian T = M + gamma K.
+   if (T) { delete T; }
+   T = Add(1.0, Mmat, gamma, Kmat);
   T_solver.SetOperator(*T);
-   T_solver.Mult(z, k);
+   *jcur = 1;
+   return (0);
 }

-int ConductionOperator::SUNImplicitSetup(const Vector &u, const Vector &fu,
-                                         int jok, int *jcur, real_t gam)
+int ConductionOperator::SUNImplicitSolve(const Vector &b, Vector &x, double tol)
 {
-   // Compute T = M + gamma K(u_n).
-   T = std::unique_ptr<SparseMatrix>(Add(1.0, Mmat, gam, Kmat));
-   T_solver.SetOperator(*T);
-   *jcur = SUNTRUE; // this should eventually only be set true if K(u) is used
-   return SUN_SUCCESS;
+   // Solve the system A x = z => (M - gamma K) x = M b.
+   Mmat.Mult(b, z);
+   T_solver.Mult(z, x);
+   return (0);
 }

-int ConductionOperator::SUNImplicitSolve(const Vector &r, Vector &dk,
-                                         real_t tol)
+ConductionOperator::~ConductionOperator()
 {
-   // Solve the system [M + gamma K(u_n)] dk = - K(u_n) u - M k.
-   // What value r is providing depends on the ODE expression form:
-   //   EXPLICIT form: r = -inv(M) K(u_n) u - k
-   //   IMPLICIT form: r = -K(u_n) u - M k
-   T_solver.SetRelTol(tol);
-   if (isExplicit())
+   delete T;
+   delete M;
+   delete K;
+}
+
+double InitialTemperature(const Vector &x)
+{
+   if (x.Norml2() < 0.5)
   {
-      Mmat.Mult(r, z);
-      T_solver.Mult(z, dk);
+      return 2.0;
   }
   else
   {
-      T_solver.Mult(r, dk);
-   }
-   if (T_solver.GetConverged())
-   {
-      return SUN_SUCCESS;
-   }
-   else
-   {
-      return SUNLS_CONV_FAIL;
+      return 1.0;
   }
 }
-
-int ConductionOperator::SUNMassSetup()
-{
-   // Do nothing b/c mass solver was setup in constructor.
-   return SUN_SUCCESS;
-}
-
-int ConductionOperator::SUNMassSolve(const Vector &b, Vector &x, real_t tol)
-{
-   // Solve the system M x = b.
-   M_solver.SetRelTol(tol);
-   M_solver.Mult(b, x);
-   if (M_solver.GetConverged())
-   {
-      return SUN_SUCCESS;
-   }
-   else
-   {
-      return SUNLS_CONV_FAIL;
-   }
-}
-
-int ConductionOperator::SUNMassMult(const Vector &x, Vector &v)
-{
-   // Compute M x.
-   Mmat.Mult(x, v);
-   return SUN_SUCCESS;
-}
-
@@ -1,22 +1,16 @@
 //                         MFEM Example 16 - Parallel Version
 //                              SUNDIALS Modification
 //
-// Compile with:
-//    make ex16p            (GNU make)
-//    make sundials_ex16p   (CMake)
+// Compile with: make ex16p
 //
 // Sample runs:
 //     mpirun -np 4 ex16p
 //     mpirun -np 4 ex16p -m ../../data/inline-tri.mesh
 //     mpirun -np 4 ex16p -m ../../data/disc-nurbs.mesh -tf 2
 //     mpirun -np 4 ex16p -s 12 -a 0.0 -k 1.0
-//     mpirun -np 4 ex16p -s 15 -a 0.0 -k 1.0
 //     mpirun -np 4 ex16p -s 8 -a 1.0 -k 0.0 -dt 4e-6 -tf 2e-2 -vs 50
-//     mpirun -np 4 ex16p -s 11 -a 1.0 -k 0.0 -dt 4e-6 -tf 2e-2 -vs 50
 //     mpirun -np 8 ex16p -s 9 -a 0.5 -k 0.5 -o 4 -dt 8e-6 -tf 2e-2 -vs 50
-//     mpirun -np 8 ex16p -s 12 -a 0.5 -k 0.5 -o 4 -dt 8e-6 -tf 2e-2 -vs 50
 //     mpirun -np 4 ex16p -s 10 -dt 2.0e-4 -tf 4.0e-2
-//     mpirun -np 4 ex16p -s 13 -dt 2.0e-4 -tf 4.0e-2
 //     mpirun -np 16 ex16p -m ../../data/fichera-q2.mesh
 //     mpirun -np 16 ex16p -m ../../data/escher-p2.mesh
 //     mpirun -np 8 ex16p -m ../../data/beam-tet.mesh -tf 10 -dt 0.1
@@ -44,102 +38,66 @@
 using namespace std;
 using namespace mfem;

-/** After spatial discretization, the conduction model is expressed as
+/** After spatial discretization, the conduction model can be written as:
 *
- *   M du/dt = - K(u) u
+ *     du/dt = M^{-1}(-Ku)
 *
 *  where u is the vector representing the temperature, M is the mass matrix,
- *  and K(u) is the diffusion operator with diffusivity depending on u:
+ *  and K is the diffusion operator with diffusivity depending on u:
 *  (\kappa + \alpha u).
 *
- *  Class ConductionOperatorOperator represents the above ODE operator in the
- *  general form F(u, k, t) = G(u, t) where either
- *
- *    1. F(u, du/dt, t) = du/dt            (ODE is expressed in EXPLICIT form)
- *       G(u, t) = - inv(M) K(u) u
- *    2. F(u, du/dt, t) = M du/dt          (ODE is expressed in IMPLICIT form)
- *       G(u, t) = - K(u) u
+ *  Class ConductionOperator represents the right-hand side of the above ODE.
 */
 class ConductionOperator : public TimeDependentOperator
 {
+protected:
   ParFiniteElementSpace &fespace;
   Array<int> ess_tdof_list; // this list remains empty for pure Neumann b.c.

-   ParBilinearForm M;
+   ParBilinearForm *M;
+   ParBilinearForm *K;
+
   HypreParMatrix Mmat;
-
-   const real_t alpha, kappa;
-   std::unique_ptr<BilinearForm> K;
   HypreParMatrix Kmat;
+   HypreParMatrix *T; // T = M + dt K
+   double current_dt;

-   std::unique_ptr<HypreParMatrix> T; // T = M + gam K(u)
+   CGSolver M_solver;    // Krylov solver for inverting the mass matrix M
+   HypreSmoother M_prec; // Preconditioner for the mass matrix M

-   CGSolver M_solver; // Krylov solver for inverting the mass matrix M
-   HypreSmoother M_prec;  // Preconditioner for the mass matrix M
+   CGSolver T_solver;    // Implicit solver for T = M + dt K
+   HypreSmoother T_prec; // Preconditioner for the implicit solver

-   CGSolver T_solver; // Implicit solver for T = M + gam K(u)
-   HypreSmoother T_prec;  // Preconditioner for the implicit solver
+   double alpha, kappa;

   mutable Vector z; // auxiliary vector

 public:
+   ConductionOperator(ParFiniteElementSpace &f, double alpha, double kappa,
+                      const Vector &u);

-   ConductionOperator(ParFiniteElementSpace &f, const real_t alpha,
-                      const real_t kappa, const Vector &u,
-                      const Type &ode_expression_type);
+   virtual void Mult(const Vector &u, Vector &du_dt) const;

-   // Compute K(u_n) for use as an approximation in - K(u) u
-   void SetConductionTensor(const Vector &u);
+   /** Solve the Backward-Euler equation: k = f(u + dt*k, t), for the unknown k.
+       This is the only requirement for high-order SDIRK implicit integration.*/
+   virtual void ImplicitSolve(const double dt, const Vector &u, Vector &k);

-   /** Compute G(u, t) as defined in the IMPLICIT expression form of the ODE
-       operator, i.e., @a v = - K(u_n) @a u. Note that K(u_n) is an
-       approximation to K(u). */
-   void ExplicitMult(const Vector &u, Vector &v) const override;
+   /** Setup the system (M + dt K) x = M b. This method is used by the implicit
+       SUNDIALS solvers. */
+   virtual int SUNImplicitSetup(const Vector &x, const Vector &fx,
+                                int jok, int *jcur, double gamma);

-   /** Solve for k in F(u, k, t) = G(u, t) for either EXPLICIT or IMPLICIT
-       expression forms of the ODE operator, i.e., @a k = - inv(M) K(u_n) @a u.
-       Note that K(u_n) is an approximation to K(u). */
-   void Mult(const Vector &u, Vector &k) const override;
+   /** Solve the system (M + dt K) x = M b. This method is used by the implicit
+       SUNDIALS solvers. */
+   virtual int SUNImplicitSolve(const Vector &b, Vector &x, double tol);

-   /** Solve for k in F(u + gam*k, k, t) = G(u + gam*k, t) for either EXPLICIT
-       or IMPLICIT expression forms of the ODE operator, i.e.,
-       [ M + @a gam K(u_n) ] @a k = - K(u_n) @a u . Note that K(u_n) is an
-       approximation to K(u). */
-   void ImplicitSolve(const real_t gam, const Vector &u, Vector &k) override;
+   /// Update the diffusion BilinearForm K using the given true-dof vector `u`.
+   void SetParameters(const Vector &u);

-   /** Setup to solve for dk in [dF/dk + gam*dF/du - gam*dG/du] dk = G - F for
-       either EXPLICIT or IMPLICIT expression forms of the ODE operator, i.e.,
-       [M - @a gam Jf(u)] dk = G - F, where Jf(u) is an approximation of the
-       Jacobian of -K(u) u. The approximation chosen here is Jf(u) = -K(u_n). */
-   int SUNImplicitSetup(const Vector &u, const Vector &fu, int jok, int *jcur,
-                        real_t gam) override;
-
-   /** Solve for @a dk in the system in SUNImplicitSetup to the given tolerance,
-       with the residual @a r providing either
-        1. @a r = G - F = inv(M) f(u) - k       (EXPLICIT expression form)
-        1. @a r = G - F = f(u) - M k            (IMPLICIT expression form)
-       */
-   int SUNImplicitSolve(const Vector &r, Vector &dk, real_t tol) override;
-
-   int SUNMassSetup() override;
-
-   int SUNMassSolve(const Vector &b, Vector &x, real_t tol) override;
-
-   int SUNMassMult(const Vector &x, Vector &v) override;
+   virtual ~ConductionOperator();
 };

-real_t InitialTemperature(const Vector &x)
-{
-   if (x.Norml2() < 0.5)
-   {
-      return 2.0;
-   }
-   else
-   {
-      return 1.0;
-   }
-}
-
+double InitialTemperature(const Vector &x);

 int main(int argc, char *argv[])
 {
@@ -156,16 +114,16 @@ int main(int argc, char *argv[])
   int par_ref_levels = 1;
   int order = 2;
   int ode_solver_type = 9; // CVODE implicit BDF
-   real_t t_final = 0.5;
-   real_t dt = 1.0e-2;
-   real_t alpha = 1.0e-2;
-   real_t kappa = 0.5;
+   double t_final = 0.5;
+   double dt = 1.0e-2;
+   double alpha = 1.0e-2;
+   double kappa = 0.5;
   bool visualization = true;
   bool visit = false;
   int vis_steps = 5;

   // Relative and absolute tolerances for CVODE and ARKODE.
-   const real_t reltol = 1e-4, abstol = 1e-4;
+   const double reltol = 1e-4, abstol = 1e-4;

   int precision = 8;
   cout.precision(precision);
@@ -192,10 +150,7 @@ int main(int argc, char *argv[])
                  "9  - CVODE (implicit BDF),\n\t"
                  "10 - ARKODE (default explicit),\n\t"
                  "11 - ARKODE (explicit Fehlberg-6-4-5),\n\t"
-                  "12 - ARKODE (default implicit),\n\t"
-                  "13 - ARKODE (default explicit with MFEM mass solve),\n\t"
-                  "14 - ARKODE (explicit Fehlberg-6-4-5 with MFEM mass solve),\n\t"
-                  "15 - ARKODE (default implicit with MFEM mass solve).");
+                  "12 - ARKODE (default impicit).");
   args.AddOption(&t_final, "-tf", "--t-final",
                  "Final time; start time is 0.");
   args.AddOption(&dt, "-dt", "--time-step",
@@ -219,33 +174,40 @@ int main(int argc, char *argv[])
      return 1;
   }

-   if (Mpi::Root())
+   if (myid == 0)
   {
      args.PrintOptions(cout);
   }

-   bool use_mass_solver = ode_solver_type >= 13;
+   // check for valid ODE solver option
+   if (ode_solver_type < 1 || ode_solver_type > 12)
+   {
+      if (myid == 0)
+      {
+         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
+      }
+      return 1;
+   }

-   // 3. Define a parallel mesh by a partitioning of a serial mesh. Read the
-   //    serial mesh from the given mesh file on all processors. We can
+   // 3. Read the serial mesh from the given mesh file on all processors. We can
   //    handle triangular, quadrilateral, tetrahedral and hexahedral meshes
   //    with the same code.
-   std::unique_ptr<ParMesh> pmesh;
+   Mesh *mesh = new Mesh(mesh_file, 1, 1);
+   int dim = mesh->Dimension();
+
+   // 4. Refine the mesh in serial to increase the resolution. In this example
+   //    we do 'ser_ref_levels' of uniform refinement, where 'ser_ref_levels' is
+   //    a command-line parameter.
+   for (int lev = 0; lev < ser_ref_levels; lev++)
   {
-      std::unique_ptr<Mesh> mesh(new Mesh(mesh_file, 1, 1));
-
-      // 4. Refine the mesh in serial to increase the resolution. In this example
-      //    we do 'ser_ref_levels' of uniform refinement, where 'ser_ref_levels' is
-      //    a command-line parameter.
-      for (int lev = 0; lev < ser_ref_levels; lev++)
-      {
-         mesh->UniformRefinement();
-      }
-
-      // 5. Refine this mesh further in parallel to increase the resolution.
-      //    Once the parallel mesh is defined, the serial mesh can be deleted.
-      pmesh = std::make_unique<ParMesh>(MPI_COMM_WORLD, *mesh);
+      mesh->UniformRefinement();
   }
+
+   // 5. Define a parallel mesh by a partitioning of the serial mesh. Refine
+   //    this mesh further in parallel to increase the resolution. Once the
+   //    parallel mesh is defined, the serial mesh can be deleted.
+   ParMesh *pmesh = new ParMesh(MPI_COMM_WORLD, *mesh);
+   delete mesh;
   for (int lev = 0; lev < par_ref_levels; lev++)
   {
      pmesh->UniformRefinement();
@@ -253,9 +215,8 @@ int main(int argc, char *argv[])

   // 6. Define the vector finite element space representing the current and the
   //    initial temperature, u_ref.
-   int dim = pmesh->Dimension();
   H1_FECollection fe_coll(order, dim);
-   ParFiniteElementSpace fespace(pmesh.get(), &fe_coll);
+   ParFiniteElementSpace fespace(pmesh, &fe_coll);

   int fe_size = fespace.GlobalTrueVSize();
   if (myid == 0)
@@ -272,17 +233,8 @@ int main(int argc, char *argv[])
   Vector u;
   u_gf.GetTrueDofs(u);

-   // 8. Initialize the conduction ODE operator and the visualization.
-   ConductionOperator::Type ode_expression_type;
-   if (use_mass_solver)
-   {
-      ode_expression_type = ConductionOperator::Type::IMPLICIT;
-   }
-   else
-   {
-      ode_expression_type = ConductionOperator::Type::EXPLICIT;
-   }
-   ConductionOperator oper(fespace, alpha, kappa, u, ode_expression_type);
+   // 8. Initialize the conduction operator and the VisIt visualization.
+   ConductionOperator oper(fespace, alpha, kappa, u);

   u_gf.SetFromTrueDofs(u);
   {
@@ -297,7 +249,7 @@ int main(int argc, char *argv[])
      u_gf.Save(osol);
   }

-   VisItDataCollection visit_dc("Example16-Parallel", pmesh.get());
+   VisItDataCollection visit_dc("Example16-Parallel", pmesh);
   visit_dc.RegisterField("temperature", &u_gf);
   if (visit)
   {
@@ -341,76 +293,52 @@ int main(int argc, char *argv[])
   }

   // 9. Define the ODE solver used for time integration.
-   real_t t = 0.0;
-   std::unique_ptr<ODESolver> ode_solver;
+   double t = 0.0;
+   ODESolver *ode_solver = NULL;
+   CVODESolver *cvode = NULL;
+   ARKStepSolver *arkode = NULL;
   switch (ode_solver_type)
   {
      // MFEM explicit methods
-      case 1: ode_solver = std::make_unique<ForwardEulerSolver>(); break;
-      case 2: ode_solver = std::make_unique<RK2Solver>(0.5); break; // midpoint method
-      case 3: ode_solver = std::make_unique<RK3SSPSolver>(); break;
-      case 4: ode_solver = std::make_unique<RK4Solver>(); break;
+      case 1: ode_solver = new ForwardEulerSolver; break;
+      case 2: ode_solver = new RK2Solver(0.5); break; // midpoint method
+      case 3: ode_solver = new RK3SSPSolver; break;
+      case 4: ode_solver = new RK4Solver; break;
      // MFEM implicit L-stable methods
-      case 5: ode_solver = std::make_unique<BackwardEulerSolver>(); break;
-      case 6: ode_solver = std::make_unique<SDIRK23Solver>(2); break;
-      case 7: ode_solver = std::make_unique<SDIRK33Solver>(); break;
+      case 5: ode_solver = new BackwardEulerSolver; break;
+      case 6: ode_solver = new SDIRK23Solver(2); break;
+      case 7: ode_solver = new SDIRK33Solver; break;
      // CVODE
      case 8:
-      case 9:
-      {
-         int cvode_solver_type;
-         if (ode_solver_type == 8)
-         {
-            cvode_solver_type = CV_ADAMS;
-         }
-         else
-         {
-            cvode_solver_type = CV_BDF;
-         }
-         std::unique_ptr<CVODESolver> cvode(
-            new CVODESolver(MPI_COMM_WORLD, cvode_solver_type));
+         cvode = new CVODESolver(MPI_COMM_WORLD, CV_ADAMS);
         cvode->Init(oper);
         cvode->SetSStolerances(reltol, abstol);
         cvode->SetMaxStep(dt);
-         ode_solver = std::move(cvode);
-         break;
-      }
+         ode_solver = cvode; break;
+      case 9:
+         cvode = new CVODESolver(MPI_COMM_WORLD, CV_BDF);
+         cvode->Init(oper);
+         cvode->SetSStolerances(reltol, abstol);
+         cvode->SetMaxStep(dt);
+         ode_solver = cvode; break;
      // ARKODE
      case 10:
      case 11:
-      case 12:
-      case 13:
-      case 14:
-      case 15:
-      {
-         ARKStepSolver::Type arkode_solver_type;
-         if (ode_solver_type == 12 || ode_solver_type == 15)
-         {
-            arkode_solver_type = ARKStepSolver::IMPLICIT;
-         }
-         else
-         {
-            arkode_solver_type = ARKStepSolver::EXPLICIT;
-         }
-         std::unique_ptr<ARKStepSolver> arkode(
-            new ARKStepSolver(MPI_COMM_WORLD, arkode_solver_type));
+         arkode = new ARKStepSolver(MPI_COMM_WORLD, ARKStepSolver::EXPLICIT);
         arkode->Init(oper);
         arkode->SetSStolerances(reltol, abstol);
         arkode->SetMaxStep(dt);
-         if (ode_solver_type == 11 || ode_solver_type == 14)
+         if (ode_solver_type == 11)
         {
            arkode->SetERKTableNum(ARKODE_FEHLBERG_13_7_8);
         }
-         if (use_mass_solver)
-         {
-            arkode->UseMFEMMassLinearSolver(SUNFALSE);
-         }
-         ode_solver = std::move(arkode);
-         break;
-      }
-      default:
-         cout << "Unknown ODE solver type: " << ode_solver_type << '\n';
-         return 3;
+         ode_solver = arkode; break;
+      case 12:
+         arkode = new ARKStepSolver(MPI_COMM_WORLD, ARKStepSolver::IMPLICIT);
+         arkode->Init(oper);
+         arkode->SetSStolerances(reltol, abstol);
+         arkode->SetMaxStep(dt);
+         ode_solver = arkode; break;
   }

   // Initialize MFEM integrators, SUNDIALS integrators are initialized above
@@ -418,18 +346,12 @@ int main(int argc, char *argv[])

   // Since we want to update the diffusion coefficient after every time step,
   // we need to use the "one-step" mode of the SUNDIALS solvers.
-   if (CVODESolver* cvode = dynamic_cast<CVODESolver*>(ode_solver.get()))
-   {
-      cvode->SetStepMode(CV_ONE_STEP);
-   }
-   else if (ARKStepSolver* arkode = dynamic_cast<ARKStepSolver*>(ode_solver.get()))
-   {
-      arkode->SetStepMode(ARK_ONE_STEP);
-   }
+   if (cvode) { cvode->SetStepMode(CV_ONE_STEP); }
+   if (arkode) { arkode->SetStepMode(ARK_ONE_STEP); }

   // 10. Perform time-integration (looping over the time iterations, ti, with a
   //     time-step dt).
-   if (Mpi::Root())
+   if (myid == 0)
   {
      cout << "Integrating the ODE ..." << endl;
   }
@@ -439,7 +361,7 @@ int main(int argc, char *argv[])
   bool last_step = false;
   for (int ti = 1; !last_step; ti++)
   {
-      real_t dt_real = min(dt, t_final - t);
+      double dt_real = min(dt, t_final - t);

      // Note that since we are using the "one-step" mode of the SUNDIALS
      // solvers, they will, generally, step over the final time and will not
@@ -455,14 +377,8 @@ int main(int argc, char *argv[])
         if (myid == 0)
         {
            cout << "step " << ti << ", t = " << t << endl;
-            if (CVODESolver* cvode = dynamic_cast<CVODESolver*>(ode_solver.get()))
-            {
-               cvode->PrintInfo();
-            }
-            else if (ARKStepSolver* arkode = dynamic_cast<ARKStepSolver*>(ode_solver.get()))
-            {
-               arkode->PrintInfo();
-            }
+            if (cvode) { cvode->PrintInfo(); }
+            if (arkode) { arkode->PrintInfo(); }
         }

         u_gf.SetFromTrueDofs(u);
@@ -479,38 +395,46 @@ int main(int argc, char *argv[])
            visit_dc.Save();
         }
      }
-      oper.SetConductionTensor(u);
+      oper.SetParameters(u);
   }
   tic_toc.Stop();
-   if (Mpi::Root())
+   if (myid == 0)
   {
      cout << "Done, " << tic_toc.RealTime() << "s." << endl;
   }

   // 11. Save the final solution in parallel. This output can be viewed later
   //     using GLVis: "glvis -np <np> -m ex16-mesh -g ex16-final".
-   u_gf.Save("ex16-final", precision);
+   {
+      ostringstream sol_name;
+      sol_name << "ex16-final." << setfill('0') << setw(6) << myid;
+      ofstream osol(sol_name.str().c_str());
+      osol.precision(precision);
+      u_gf.Save(osol);
+   }
+
+   // 12. Free the used memory.
+   delete ode_solver;
+   delete pmesh;

   return 0;
 }

-ConductionOperator::ConductionOperator(ParFiniteElementSpace &fes,
-                                       const real_t alpha, const real_t kappa,
-                                       const Vector &u,
-                                       const Type &ode_expression_type)
-   : TimeDependentOperator(fes.GetTrueVSize(), 0.0, ode_expression_type),
-     fespace(fes), M(&fespace), alpha(alpha), kappa(kappa),
-     M_solver(fes.GetComm()), T_solver(fes.GetComm()), z(height)
+ConductionOperator::ConductionOperator(ParFiniteElementSpace &f, double al,
+                                       double kap, const Vector &u)
+   : TimeDependentOperator(f.GetTrueVSize(), 0.0), fespace(f), M(NULL), K(NULL),
+     T(NULL),
+     M_solver(f.GetComm()), T_solver(f.GetComm()), z(height)
 {
-   // specify a relative tolerance for all solves with MFEM integrators
-   const real_t rel_tol = 1e-8;
+   const double rel_tol = 1e-8;

-   M.AddDomainIntegrator(new MassIntegrator());
-   M.Assemble(0); // keep zeros to keep sparsity pattern of M and K the same
-   M.FormSystemMatrix(ess_tdof_list, Mmat);
+   M = new ParBilinearForm(&fespace);
+   M->AddDomainIntegrator(new MassIntegrator());
+   M->Assemble(0); // keep sparsity pattern of M and K the same
+   M->FormSystemMatrix(ess_tdof_list, Mmat);

   M_solver.iterative_mode = false;
-   M_solver.SetRelTol(rel_tol); // will be overwritten with SUNDIALS integrators
+   M_solver.SetRelTol(rel_tol);
   M_solver.SetAbsTol(0.0);
   M_solver.SetMaxIter(100);
   M_solver.SetPrintLevel(0);
@@ -518,118 +442,97 @@ ConductionOperator::ConductionOperator(ParFiniteElementSpace &fes,
   M_solver.SetPreconditioner(M_prec);
   M_solver.SetOperator(Mmat);

+   alpha = al;
+   kappa = kap;
+
   T_solver.iterative_mode = false;
-   T_solver.SetRelTol(rel_tol); // will be overwritten with SUNDIALS integrators
+   T_solver.SetRelTol(rel_tol);
   T_solver.SetAbsTol(0.0);
   T_solver.SetMaxIter(100);
   T_solver.SetPrintLevel(0);
   T_solver.SetPreconditioner(T_prec);

-   SetConductionTensor(u);
+   SetParameters(u);
 }

-void ConductionOperator::SetConductionTensor(const Vector &u)
+void ConductionOperator::Mult(const Vector &u, Vector &du_dt) const
+{
+   // Compute:
+   //    du_dt = M^{-1}*-K(u)
+   // for du_dt
+   Kmat.Mult(u, z);
+   z.Neg(); // z = -z
+   M_solver.Mult(z, du_dt);
+}
+
+void ConductionOperator::ImplicitSolve(const double dt,
+                                       const Vector &u, Vector &du_dt)
+{
+   // Solve the equation:
+   //    du_dt = M^{-1}*[-K(u + dt*du_dt)]
+   // for du_dt
+   if (T) { delete T; }
+   T = Add(1.0, Mmat, dt, Kmat);
+   T_solver.SetOperator(*T);
+   Kmat.Mult(u, z);
+   z.Neg();
+   T_solver.Mult(z, du_dt);
+}
+
+int ConductionOperator::SUNImplicitSetup(const Vector &x,
+                                         const Vector &fx, int jok, int *jcur,
+                                         double gamma)
+{
+   // Setup the ODE Jacobian T = M + gamma K.
+   if (T) { delete T; }
+   T = Add(1.0, Mmat, gamma, Kmat);
+   T_solver.SetOperator(*T);
+   *jcur = 1;
+   return (0);
+}
+
+int ConductionOperator::SUNImplicitSolve(const Vector &b, Vector &x, double tol)
+{
+   // Solve the system A x = z => (M - gamma K) x = M b.
+   Mmat.Mult(b, z);
+   T_solver.Mult(z, x);
+   return (0);
+}
+
+void ConductionOperator::SetParameters(const Vector &u)
 {
-   // Compute K(u_n).
   ParGridFunction u_alpha_gf(&fespace);
   u_alpha_gf.SetFromTrueDofs(u);
   for (int i = 0; i < u_alpha_gf.Size(); i++)
   {
      u_alpha_gf(i) = kappa + alpha*u_alpha_gf(i);
   }
+
+   delete K;
+   K = new ParBilinearForm(&fespace);
+
   GridFunctionCoefficient u_coeff(&u_alpha_gf);

-   K = std::make_unique<ParBilinearForm>(&fespace);
   K->AddDomainIntegrator(new DiffusionIntegrator(u_coeff));
-   K->Assemble(0); // keep zeros to keep sparsity pattern of M and K the same
+   K->Assemble(0); // keep sparsity pattern of M and K the same
   K->FormSystemMatrix(ess_tdof_list, Kmat);
 }

-void ConductionOperator::ExplicitMult(const Vector &u, Vector &v) const
+ConductionOperator::~ConductionOperator()
 {
-   // Compute - K(u_n) u.
-   Kmat.Mult(u, v);
-   v.Neg();
+   delete T;
+   delete M;
+   delete K;
 }

-void ConductionOperator::Mult(const Vector &u, Vector &k) const
+double InitialTemperature(const Vector &x)
 {
-   // Compute - inv(M) K(u_n) u.
-   ExplicitMult(u, z);
-   M_solver.Mult(z, k);
-}
-
-void ConductionOperator::ImplicitSolve(const real_t gam, const Vector &u,
-                                       Vector &k)
-{
-   // Solve for k in M k = - K(u_n) [u + gam*k].
-   ExplicitMult(u, z);
-   T = std::unique_ptr<HypreParMatrix>(Add(1.0, Mmat, gam, Kmat));
-   T_solver.SetOperator(*T);
-   T_solver.Mult(z, k);
-}
-
-int ConductionOperator::SUNImplicitSetup(const Vector &u, const Vector &fu,
-                                         int jok, int *jcur, real_t gam)
-{
-   // Compute T = M + gamma K(u_n).
-   T = std::unique_ptr<HypreParMatrix>(Add(1.0, Mmat, gam, Kmat));
-   T_solver.SetOperator(*T);
-   *jcur = SUNTRUE; // this should eventually only be set true if K(u) is used
-   return SUN_SUCCESS;
-}
-
-int ConductionOperator::SUNImplicitSolve(const Vector &r, Vector &dk,
-                                         real_t tol)
-{
-   // Solve the system [M + gamma K(u_n)] dk = - K(u_n) u - M k.
-   // What value r is providing depends on the ODE expression form:
-   //   EXPLICIT form: r = -inv(M) K(u_n) u - k
-   //   IMPLICIT form: r = -K(u_n) u - M k
-   T_solver.SetRelTol(tol);
-   if (isExplicit())
+   if (x.Norml2() < 0.5)
   {
-      Mmat.Mult(r, z);
-      T_solver.Mult(z, dk);
+      return 2.0;
   }
   else
   {
-      T_solver.Mult(r, dk);
-   }
-   if (T_solver.GetConverged())
-   {
-      return SUN_SUCCESS;
-   }
-   else
-   {
-      return SUNLS_CONV_FAIL;
+      return 1.0;
   }
 }
-
-int ConductionOperator::SUNMassSetup()
-{
-   // Do nothing b/c mass solver was setup in constructor.
-   return SUN_SUCCESS;
-}
-
-int ConductionOperator::SUNMassSolve(const Vector &b, Vector &x, real_t tol)
-{
-   // Solve the system M x = b.
-   M_solver.SetRelTol(tol);
-   M_solver.Mult(b, x);
-   if (M_solver.GetConverged())
-   {
-      return SUN_SUCCESS;
-   }
-   else
-   {
-      return SUNLS_CONV_FAIL;
-   }
-}
-
-int ConductionOperator::SUNMassMult(const Vector &x, Vector &v)
-{
-   // Compute M x.
-   Mmat.Mult(x, v);
-   return SUN_SUCCESS;
-}
@@ -1,9 +1,7 @@
 //                                MFEM Example 9
 //                             SUNDIALS Modification
 //
-// Compile with:
-//    make ex9              (GNU make)
-//    make sundials_ex9     (CMake)
+// Compile with: make ex9
 //
 // Sample runs:
 //    ex9 -m ../../data/periodic-segment.mesh -p 0 -r 2 -s 7 -dt 0.005
@@ -1,9 +1,7 @@
 //                       MFEM Example 9 - Parallel Version
 //                             SUNDIALS Modification
 //
-// Compile with:
-//    make ex9p             (GNU make)
-//    make sundials_ex9p    (CMake)
+// Compile with: make ex9p
 //
 // Sample runs:
 //    mpirun -np 4 ex9p -m ../../data/periodic-segment.mesh -p 1 -rp 1 -s 7 -dt 0.0025
@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/sundials/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -99,12 +100,6 @@ ex10-test-seq: ex10
 	@$(call mfem-test,$<,, $(SERIAL_NAME),$(EX10_ARGS))
 ex10p-test-par: ex10p
 	@$(call mfem-test,$<, $(RUN_MPI), $(PARALLEL_NAME),$(EX10P_ARGS))
-# Example 16: test ARKODE with implicit time stepping using mass form
-EX16_COMMON_ARGS := -s 15
-ex16-test-seq: ex16
-	@$(call mfem-test,$<,, $(SERIAL_NAME),$(EX16_COMMON_ARGS))
-ex16p-test-par: ex16p
-	@$(call mfem-test,$<, $(RUN_MPI), $(PARALLEL_NAME),$(EX16_COMMON_ARGS))

 # Testing: "test" target and mfem-test* variables are defined in config/test.mk

@@ -12,10 +12,11 @@
 # Use the MFEM build directory
 MFEM_DIR ?= ../..
 MFEM_BUILD_DIR ?= ../..
-MFEM_INSTALL_DIR ?= ../../mfem
 SRC = $(if $(MFEM_DIR:../..=),$(MFEM_DIR)/examples/superlu/,)
-CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
-   $(wildcard $(MFEM_INSTALL_DIR)/share/mfem/config.mk))
+CONFIG_MK = $(MFEM_BUILD_DIR)/config/config.mk
+# Use the MFEM install directory
+# MFEM_INSTALL_DIR = ../../mfem
+# CONFIG_MK = $(MFEM_INSTALL_DIR)/share/mfem/config.mk

 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
@@ -35,7 +35,6 @@ set(SRCS
  integ/bilininteg_mass_ea.cpp
  integ/bilininteg_mixedcurl_pa.cpp
  integ/bilininteg_mixedvecgrad_pa.cpp
-  integ/bilininteg_trace_jump_ea.cpp
  integ/bilininteg_transpose_ea.cpp
  integ/bilininteg_vecdiffusion_mf.cpp
  integ/bilininteg_vecdiffusion_pa.cpp
@@ -47,7 +46,6 @@ set(SRCS
  integ/bilininteg_diffusion_kernels.cpp
  integ/bilininteg_elasticity_kernels.cpp
  integ/bilininteg_hcurl_kernels.cpp
-  integ/bilininteg_hdiv_ea.cpp
  integ/bilininteg_hdiv_kernels.cpp
  integ/bilininteg_hcurlhdiv_kernels.cpp
  integ/bilininteg_mass_kernels.cpp
@@ -82,7 +80,6 @@ set(SRCS
  geom.cpp
  gridfunc.cpp
  hybridization.cpp
-  hybridization_ext.cpp
  intrules.cpp
  intrules_cut.cpp
  ceed/interface/basis.cpp
@@ -115,6 +112,8 @@ set(SRCS
  qinterp/eval_by_vdim.cpp
  qinterp/grad_by_nodes.cpp
  qinterp/grad_by_vdim.cpp
+  qinterp/grad_phys_by_nodes.cpp
+  qinterp/grad_phys_by_vdim.cpp
  qspace.cpp
  quadinterpolator.cpp
  quadinterpolator_face.cpp
@@ -191,12 +190,8 @@ set(HDRS
  geom.hpp
  gridfunc.hpp
  hybridization.hpp
-  hybridization_ext.hpp
  intrules.hpp
  intrules_cut.hpp
-  kernel_dispatch.hpp
-  kernel_reporter.hpp
-  kernels.hpp
  ceed/interface/basis.hpp
  ceed/interface/integrator.hpp
  ceed/interface/interface.hpp
@@ -228,6 +223,7 @@ set(HDRS
  nonlinearform_ext.hpp
  nonlininteg.hpp
  qfunction.hpp
+  qinterp/dispatch.hpp
  qinterp/eval.hpp
  qinterp/grad.hpp
  qspace.hpp
@@ -71,11 +71,15 @@ BilinearForm::BilinearForm(FiniteElementSpace * f)
   sequence = f->GetSequence();
   mat = mat_e = NULL;
   extern_bfs = 0;
+   element_matrices = NULL;
+   static_cond = NULL;
+   hybridization = NULL;
   precompute_sparsity = 0;
   diag_policy = DIAG_KEEP;

   assembly = AssemblyLevel::LEGACY;
   batch = 1;
+   ext = NULL;
 }

 BilinearForm::BilinearForm (FiniteElementSpace * f, BilinearForm * bf, int ps)
@@ -85,11 +89,15 @@ BilinearForm::BilinearForm (FiniteElementSpace * f, BilinearForm * bf, int ps)
   sequence = f->GetSequence();
   mat_e = NULL;
   extern_bfs = 1;
+   element_matrices = NULL;
+   static_cond = NULL;
+   hybridization = NULL;
   precompute_sparsity = ps;
   diag_policy = DIAG_KEEP;

   assembly = AssemblyLevel::LEGACY;
   batch = 1;
+   ext = NULL;

   // Copy the pointers to the integrators
   domain_integs = bf->domain_integs;
@@ -119,16 +127,16 @@ void BilinearForm::SetAssemblyLevel(AssemblyLevel assembly_level)
         break;
      case AssemblyLevel::FULL:
         SetDiagonalPolicy( DIAG_ONE ); // Only diagonal policy supported on device
-         ext.reset(new FABilinearFormExtension(this));
+         ext = new FABilinearFormExtension(this);
         break;
      case AssemblyLevel::ELEMENT:
-         ext.reset(new EABilinearFormExtension(this));
+         ext = new EABilinearFormExtension(this);
         break;
      case AssemblyLevel::PARTIAL:
-         ext.reset(new PABilinearFormExtension(this));
+         ext = new PABilinearFormExtension(this);
         break;
      case AssemblyLevel::NONE:
-         ext.reset(new MFBilinearFormExtension(this));
+         ext = new MFBilinearFormExtension(this);
         break;
      default:
         MFEM_ABORT("BilinearForm: unknown assembly level");
@@ -137,13 +145,14 @@ void BilinearForm::SetAssemblyLevel(AssemblyLevel assembly_level)

 void BilinearForm::EnableStaticCondensation()
 {
+   delete static_cond;
   if (assembly != AssemblyLevel::LEGACY)
   {
-      static_cond.reset();
+      static_cond = NULL;
      MFEM_WARNING("Static condensation not supported for this assembly level");
      return;
   }
-   static_cond.reset(new StaticCondensation(fes));
+   static_cond = new StaticCondensation(fes);
   if (static_cond->ReducesTrueVSize())
   {
      bool symmetric = false;      // TODO
@@ -152,7 +161,8 @@ void BilinearForm::EnableStaticCondensation()
   }
   else
   {
-      static_cond.reset();
+      delete static_cond;
+      static_cond = NULL;
   }
 }

@@ -160,18 +170,15 @@ void BilinearForm::EnableHybridization(FiniteElementSpace *constr_space,
                                       BilinearFormIntegrator *constr_integ,
                                       const Array<int> &ess_tdof_list)
 {
-   if (assembly != AssemblyLevel::LEGACY && assembly != AssemblyLevel::ELEMENT)
+   delete hybridization;
+   if (assembly != AssemblyLevel::LEGACY)
   {
      delete constr_integ;
-      hybridization.reset();
+      hybridization = NULL;
      MFEM_WARNING("Hybridization not supported for this assembly level");
      return;
   }
-   hybridization.reset(new Hybridization(fes, constr_space));
-   if (assembly == AssemblyLevel::ELEMENT)
-   {
-      hybridization->EnableDeviceExecution();
-   }
+   hybridization = new Hybridization(fes, constr_space);
   hybridization->SetConstraintIntegrator(constr_integ);
   hybridization->Init(ess_tdof_list);
 }
@@ -224,8 +231,8 @@ void BilinearForm::Finalize (int skip_zeros)
      if (!static_cond) { mat->Finalize(skip_zeros); }
      if (mat_e) { mat_e->Finalize(skip_zeros); }
      if (static_cond) { static_cond->Finalize(); }
+      if (hybridization) { hybridization->Finalize(); }
   }
-   if (hybridization) { hybridization->Finalize(); }
 }

 void BilinearForm::AddDomainIntegrator(BilinearFormIntegrator *bfi)
@@ -273,7 +280,7 @@ void BilinearForm::AddBdrFaceIntegrator(BilinearFormIntegrator *bfi,
   boundary_face_integs_marker.Append(&bdr_marker);
 }

-void BilinearForm::ComputeElementMatrix(int i, DenseMatrix &elmat) const
+void BilinearForm::ComputeElementMatrix(int i, DenseMatrix &elmat)
 {
   if (element_matrices)
   {
@@ -282,10 +289,9 @@ void BilinearForm::ComputeElementMatrix(int i, DenseMatrix &elmat) const
      return;
   }

-   const FiniteElement &fe = *fes->GetFE(i);
-
   if (domain_integs.Size())
   {
+      const FiniteElement &fe = *fes->GetFE(i);
      ElementTransformation *eltrans = fes->GetElementTransformation(i);
      domain_integs[0]->AssembleElementMatrix(fe, *eltrans, elmat);
      for (int k = 1; k < domain_integs.Size(); k++)
@@ -296,18 +302,17 @@ void BilinearForm::ComputeElementMatrix(int i, DenseMatrix &elmat) const
   }
   else
   {
-      const int ndof = fe.GetDof() * fes->GetVDim();
-      elmat.SetSize(ndof);
+      fes->GetElementVDofs(i, vdofs);
+      elmat.SetSize(vdofs.Size());
      elmat = 0.0;
   }
 }

-void BilinearForm::ComputeBdrElementMatrix(int i, DenseMatrix &elmat) const
+void BilinearForm::ComputeBdrElementMatrix(int i, DenseMatrix &elmat)
 {
-   const FiniteElement &be = *fes->GetBE(i);
-
   if (boundary_integs.Size())
   {
+      const FiniteElement &be = *fes->GetBE(i);
      ElementTransformation *eltrans = fes->GetBdrElementTransformation(i);
      boundary_integs[0]->AssembleElementMatrix(be, *eltrans, elmat);
      for (int k = 1; k < boundary_integs.Size(); k++)
@@ -318,81 +323,8 @@ void BilinearForm::ComputeBdrElementMatrix(int i, DenseMatrix &elmat) const
   }
   else
   {
-      const int ndof = be.GetDof() * fes->GetVDim();
-      elmat.SetSize(ndof);
-      elmat = 0.0;
-   }
-}
-
-void BilinearForm::ComputeFaceMatrix(int i, DenseMatrix &elmat) const
-{
-   FaceElementTransformations *tr;
-   Mesh *mesh = fes -> GetMesh();
-   tr = mesh -> GetFaceElementTransformations (i);
-
-   const FiniteElement *fe1, *fe2;
-   fe1 = fes->GetFE(tr->Elem1No);
-   if (tr->Elem2No >= 0)
-   {
-      fe2 = fes->GetFE(tr->Elem2No);
-   }
-   else
-   {
-      // The fe2 object is really a dummy and not used on the
-      // boundaries, but we can't dereference a NULL pointer, and we don't
-      // want to actually make a fake element.
-      fe2 = fe1;
-   }
-
-   if (interior_face_integs.Size())
-   {
-      interior_face_integs[0] -> AssembleFaceMatrix (*fe1, *fe2, *tr, elmat);
-      for (int k = 1; k < interior_face_integs.Size(); k++)
-      {
-         interior_face_integs[k] -> AssembleFaceMatrix (*fe1, *fe2, *tr, elemmat);
-         elmat += elemmat;
-      }
-   }
-   else
-   {
-      int ndof = fe1->GetDof() * fes->GetVDim();
-      if (tr->Elem2No >= 0)
-      {
-         ndof += fe2->GetDof() * fes->GetVDim();
-      }
-
-      elmat.SetSize(ndof);
-      elmat = 0.0;
-   }
-}
-
-void BilinearForm::ComputeBdrFaceMatrix(int i, DenseMatrix &elmat) const
-{
-   FaceElementTransformations *tr;
-   Mesh *mesh = fes -> GetMesh();
-   tr = mesh -> GetBdrFaceTransformations (i);
-
-   const FiniteElement *fe1, *fe2;
-
-   fe1 = fes -> GetFE (tr -> Elem1No);
-   // The fe2 object is really a dummy and not used on the boundaries,
-   // but we can't dereference a NULL pointer, and we don't want to
-   // actually make a fake element.
-   fe2 = fe1;
-
-   if (boundary_face_integs.Size())
-   {
-      boundary_face_integs[0] -> AssembleFaceMatrix (*fe1, *fe2, *tr, elmat);
-      for (int k = 1; k < boundary_face_integs.Size(); k++)
-      {
-         boundary_face_integs[k] -> AssembleFaceMatrix (*fe1, *fe2, *tr, elemmat);
-         elmat += elemmat;
-      }
-   }
-   else
-   {
-      int ndof = fe1->GetDof() * fes->GetVDim();
-      elmat.SetSize(ndof);
+      fes->GetBdrElementVDofs(i, vdofs);
+      elmat.SetSize(vdofs.Size());
      elmat = 0.0;
   }
 }
@@ -458,10 +390,6 @@ void BilinearForm::Assemble(int skip_zeros)
   if (ext)
   {
      ext->Assemble();
-      if (hybridization)
-      {
-         hybridization->AssembleElementMatrices(GetElementMatrices());
-      }
      return;
   }

@@ -831,19 +759,7 @@ void BilinearForm::FormLinearSystem(const Array<int> &ess_tdof_list, Vector &x,
 {
   if (ext)
   {
-      if (hybridization)
-      {
-         FormSystemMatrix(ess_tdof_list, A);
-         ConstrainedOperator A_constrained(this, ess_tdof_list);
-         A_constrained.EliminateRHS(x, b);
-         hybridization->ReduceRHS(b, B);
-         X.SetSize(B.Size());
-         X = 0.0;
-      }
-      else
-      {
-         ext->FormLinearSystem(ess_tdof_list, x, b, A, X, B, copy_interior);
-      }
+      ext->FormLinearSystem(ess_tdof_list, x, b, A, X, B, copy_interior);
      return;
   }
   const SparseMatrix *P = fes->GetConformingProlongation();
@@ -911,16 +827,7 @@ void BilinearForm::FormSystemMatrix(const Array<int> &ess_tdof_list,
 {
   if (ext)
   {
-      if (hybridization)
-      {
-         const int remove_zeros = 0;
-         Finalize(remove_zeros);
-         A.Reset(&hybridization->GetMatrix(), false);
-      }
-      else
-      {
-         ext->FormSystemMatrix(ess_tdof_list, A);
-      }
+      ext->FormSystemMatrix(ess_tdof_list, A);
      return;
   }

@@ -961,7 +868,7 @@ void BilinearForm::FormSystemMatrix(const Array<int> &ess_tdof_list,
 void BilinearForm::RecoverFEMSolution(const Vector &X,
                                      const Vector &b, Vector &x)
 {
-   if (ext && !hybridization)
+   if (ext)
   {
      ext->RecoverFEMSolution(X, b, x);
      return;
@@ -1018,26 +925,16 @@ void BilinearForm::RecoverFEMSolution(const Vector &X,

 void BilinearForm::ComputeElementMatrices()
 {
-   if (element_matrices) { return; }
-
-   if (auto *ea_ext = dynamic_cast<EABilinearFormExtension*>(ext.get()))
+   if (element_matrices || domain_integs.Size() == 0 || fes->GetNE() == 0)
   {
-      element_matrices.reset(new DenseTensor);
-      ea_ext->GetElementMatrices(*element_matrices, ElementDofOrdering::NATIVE, true);
-      return;
-   }
-
-   if (domain_integs.Size() == 0 || fes->GetNE() == 0)
-   {
-      element_matrices.reset(new DenseTensor);
      return;
   }

   int num_elements = fes->GetNE();
   int num_dofs_per_el = fes->GetFE(0)->GetDof() * fes->GetVDim();

-   element_matrices.reset(new DenseTensor(num_dofs_per_el, num_dofs_per_el,
-                                          num_elements));
+   element_matrices = new DenseTensor(num_dofs_per_el, num_dofs_per_el,
+                                      num_elements);

   DenseMatrix tmp;
   IsoparametricTransformation eltrans;
@@ -1068,12 +965,6 @@ void BilinearForm::ComputeElementMatrices()
   }
 }

-const DenseTensor &BilinearForm::GetElementMatrices()
-{
-   ComputeElementMatrices(); // Won't recompute if element_matrices exists
-   return *element_matrices;
-}
-
 void BilinearForm::EliminateEssentialBC(const Array<int> &bdr_attr_is_ess,
                                        const Vector &sol, Vector &rhs,
                                        DiagonalPolicy dpolicy)
@@ -1261,13 +1152,15 @@ void BilinearForm::Update(FiniteElementSpace *nfes)
   delete mat_e;
   mat_e = NULL;
   FreeElementMatrices();
-   static_cond.reset();
+   delete static_cond;
+   static_cond = NULL;

   if (full_update)
   {
      delete mat;
      mat = NULL;
-      hybridization.reset();
+      delete hybridization;
+      hybridization = NULL;
      sequence = fes->GetSequence();
   }
   else
@@ -1290,6 +1183,9 @@ BilinearForm::~BilinearForm()
 {
   delete mat_e;
   delete mat;
+   delete element_matrices;
+   delete static_cond;
+   delete hybridization;

   if (!extern_bfs)
   {
@@ -1301,6 +1197,8 @@ BilinearForm::~BilinearForm()
      for (k=0; k < boundary_face_integs.Size(); k++)
      { delete boundary_face_integs[k]; }
   }
+
+   delete ext;
 }


@@ -1327,6 +1225,7 @@ MixedBilinearForm::MixedBilinearForm (FiniteElementSpace *tr_fes,
   mat = NULL;
   mat_e = NULL;
   extern_bfs = 1;
+   ext = NULL;

   // Copy the pointers to the integrators
   domain_integs = mbf->domain_integs;
@@ -1356,22 +1255,22 @@ void MixedBilinearForm::SetAssemblyLevel(AssemblyLevel assembly_level)
      case AssemblyLevel::LEGACY:
         break;
      case AssemblyLevel::FULL:
-         // ext.reset(new FAMixedBilinearFormExtension(this));
+         // ext = new FAMixedBilinearFormExtension(this);
         // Use the original BilinearForm implementation for now
         break;
      case AssemblyLevel::ELEMENT:
-         MFEM_ABORT("Element assembly not supported yet... stay tuned!");
-         // ext.reset(new EAMixedBilinearFormExtension(this));
+         mfem_error("Element assembly not supported yet... stay tuned!");
+         // ext = new EAMixedBilinearFormExtension(this);
         break;
      case AssemblyLevel::PARTIAL:
-         ext.reset(new PAMixedBilinearFormExtension(this));
+         ext = new PAMixedBilinearFormExtension(this);
         break;
      case AssemblyLevel::NONE:
-         MFEM_ABORT("Matrix-free action not supported yet... stay tuned!");
-         // ext.reset(new MFMixedBilinearFormExtension(this));
+         mfem_error("Matrix-free action not supported yet... stay tuned!");
+         // ext = new MFMixedBilinearFormExtension(this);
         break;
      default:
-         MFEM_ABORT("Unknown assembly level");
+         mfem_error("Unknown assembly level");
   }
 }

@@ -1457,50 +1356,32 @@ void MixedBilinearForm::GetBlocks(Array2D<SparseMatrix *> &blocks) const
   mat->GetBlocks(blocks);
 }

-void MixedBilinearForm::AddDomainIntegrator(BilinearFormIntegrator *bfi)
+void MixedBilinearForm::AddDomainIntegrator (BilinearFormIntegrator * bfi)
 {
-   domain_integs.Append(bfi);
+   domain_integs.Append (bfi);
   domain_integs_marker.Append(NULL); // NULL marker means apply everywhere
 }

-void MixedBilinearForm::AddDomainIntegrator(BilinearFormIntegrator *bfi,
-                                            Array<int> &elem_marker)
+void MixedBilinearForm::AddDomainIntegrator (BilinearFormIntegrator * bfi,
+                                             Array<int> &elem_marker)
 {
-   domain_integs.Append(bfi);
+   domain_integs.Append (bfi);
   domain_integs_marker.Append(&elem_marker);
 }

-void MixedBilinearForm::AddBoundaryIntegrator(BilinearFormIntegrator *bfi)
+void MixedBilinearForm::AddBoundaryIntegrator (BilinearFormIntegrator * bfi)
 {
-   boundary_integs.Append(bfi);
+   boundary_integs.Append (bfi);
   boundary_integs_marker.Append(NULL); // NULL marker means apply everywhere
 }

-void MixedBilinearForm::AddBoundaryIntegrator(BilinearFormIntegrator *bfi,
-                                              Array<int> &bdr_marker)
+void MixedBilinearForm::AddBoundaryIntegrator (BilinearFormIntegrator * bfi,
+                                               Array<int> &bdr_marker)
 {
-   boundary_integs.Append(bfi);
+   boundary_integs.Append (bfi);
   boundary_integs_marker.Append(&bdr_marker);
 }

-void MixedBilinearForm::AddInteriorFaceIntegrator(BilinearFormIntegrator *bfi)
-{
-   interior_face_integs.Append(bfi);
-}
-
-void MixedBilinearForm::AddBdrFaceIntegrator(BilinearFormIntegrator *bfi)
-{
-   boundary_face_integs.Append(bfi);
-   boundary_face_integs_marker.Append(NULL); // NULL marker means apply everywhere
-}
-
-void MixedBilinearForm::AddBdrFaceIntegrator(BilinearFormIntegrator *bfi,
-                                             Array<int> &bdr_marker)
-{
-   boundary_face_integs.Append(bfi);
-   boundary_face_integs_marker.Append(&bdr_marker);
-}
-
 void MixedBilinearForm::AddTraceFaceIntegrator (BilinearFormIntegrator * bfi)
 {
   trace_face_integs.Append (bfi);
@@ -1633,108 +1514,6 @@ void MixedBilinearForm::Assemble(int skip_zeros)
      }
   }

-   if (interior_face_integs.Size())
-   {
-      FaceElementTransformations *ftr;
-      Array<int> trial_vdofs2, test_vdofs2;
-      const FiniteElement *trial_fe1, *trial_fe2, *test_fe1, *test_fe2;
-
-      int nfaces = mesh->GetNumFaces();
-      for (int i = 0; i < nfaces; i++)
-      {
-         ftr = mesh->GetInteriorFaceTransformations(i);
-         if (ftr != NULL)
-         {
-            trial_fes->GetElementVDofs(ftr->Elem1No, trial_vdofs);
-            test_fes->GetElementVDofs(ftr->Elem1No, test_vdofs);
-            trial_fe1 = trial_fes->GetFE(ftr->Elem1No);
-            test_fe1 = test_fes->GetFE(ftr->Elem1No);
-            if (ftr->Elem2No >= 0)
-            {
-               trial_fes->GetElementVDofs(ftr->Elem2No, trial_vdofs2);
-               test_fes->GetElementVDofs(ftr->Elem2No, test_vdofs2);
-               trial_vdofs.Append(trial_vdofs2);
-               test_vdofs.Append(test_vdofs2);
-               trial_fe2 = trial_fes->GetFE(ftr->Elem2No);
-               test_fe2 = test_fes->GetFE(ftr->Elem2No);
-            }
-            else
-            {
-               // The test_fe2 object is really a dummy and not used on the
-               // boundaries, but we can't dereference a NULL pointer, and we don't
-               // want to actually make a fake element.
-               trial_fe2 = trial_fe1;
-               test_fe2 = test_fe1;
-            }
-            for (int k = 0; k < interior_face_integs.Size(); k++)
-            {
-               interior_face_integs[k]->AssembleFaceMatrix(*trial_fe1, *test_fe1, *trial_fe2,
-                                                           *test_fe2,
-                                                           *ftr, elemmat);
-               mat->AddSubMatrix(test_vdofs, trial_vdofs, elemmat, skip_zeros);
-            }
-         }
-      }
-   }
-
-   if (boundary_face_integs.Size())
-   {
-      FaceElementTransformations *ftr;
-      Array<int> tr_vdofs2, te_vdofs2;
-      const FiniteElement *trial_fe1, *trial_fe2, *test_fe1, *test_fe2;
-
-      // Which boundary attributes need to be processed?
-      Array<int> bdr_attr_marker(mesh->bdr_attributes.Size() ?
-                                 mesh->bdr_attributes.Max() : 0);
-      bdr_attr_marker = 0;
-      for (int k = 0; k < boundary_face_integs.Size(); k++)
-      {
-         if (boundary_face_integs_marker[k] == NULL)
-         {
-            bdr_attr_marker = 1;
-            break;
-         }
-         Array<int> &bdr_marker = *boundary_face_integs_marker[k];
-         MFEM_ASSERT(bdr_marker.Size() == bdr_attr_marker.Size(),
-                     "invalid boundary marker for boundary face integrator #"
-                     << k << ", counting from zero");
-         for (int i = 0; i < bdr_attr_marker.Size(); i++)
-         {
-            bdr_attr_marker[i] |= bdr_marker[i];
-         }
-      }
-
-      for (int i = 0; i < trial_fes -> GetNBE(); i++)
-      {
-         const int bdr_attr = mesh->GetBdrAttribute(i);
-         if (bdr_attr_marker[bdr_attr-1] == 0) { continue; }
-
-         ftr = mesh -> GetBdrFaceTransformations (i);
-         if (ftr != NULL)
-         {
-            trial_fes->GetElementVDofs(ftr->Elem1No, trial_vdofs);
-            test_fes->GetElementVDofs(ftr->Elem1No, test_vdofs);
-            trial_fe1 = trial_fes->GetFE(ftr->Elem1No);
-            test_fe1 = test_fes->GetFE(ftr->Elem1No);
-            // The test_fe2 object is really a dummy and not used on the
-            // boundaries, but we can't dereference a NULL pointer, and we don't
-            // want to actually make a fake element.
-            trial_fe2 = trial_fe1;
-            test_fe2 = test_fe1;
-            for (int k = 0; k < boundary_face_integs.Size(); k++)
-            {
-               if (boundary_face_integs_marker[k] &&
-                   (*boundary_face_integs_marker[k])[bdr_attr-1] == 0) { continue; }
-
-               boundary_face_integs[k]->AssembleFaceMatrix(*trial_fe1, *test_fe1, *trial_fe2,
-                                                           *test_fe2,
-                                                           *ftr, elemmat);
-               mat->AddSubMatrix(test_vdofs, trial_vdofs, elemmat, skip_zeros);
-            }
-         }
-      }
-   }
-
   if (trace_face_integs.Size())
   {
      FaceElementTransformations *ftr;
@@ -1913,13 +1692,12 @@ void MixedBilinearForm::ConformingAssemble()
 }


-void MixedBilinearForm::ComputeElementMatrix(int i, DenseMatrix &elmat) const
+void MixedBilinearForm::ComputeElementMatrix(int i, DenseMatrix &elmat)
 {
-   const FiniteElement &trial_fe = *trial_fes->GetFE(i);
-   const FiniteElement &test_fe = *test_fes->GetFE(i);
-
   if (domain_integs.Size())
   {
+      const FiniteElement &trial_fe = *trial_fes->GetFE(i);
+      const FiniteElement &test_fe = *test_fes->GetFE(i);
      ElementTransformation *eltrans = test_fes->GetElementTransformation(i);
      domain_integs[0]->AssembleElementMatrix2(trial_fe, test_fe, *eltrans,
                                               elmat);
@@ -1932,21 +1710,19 @@ void MixedBilinearForm::ComputeElementMatrix(int i, DenseMatrix &elmat) const
   }
   else
   {
-      const int tr_dofs = trial_fe.GetDof() * trial_fes->GetVDim();
-      const int te_dofs = test_fe.GetDof() * test_fes->GetVDim();
-
-      elmat.SetSize(te_dofs, tr_dofs);
+      trial_fes->GetElementVDofs(i, trial_vdofs);
+      test_fes->GetElementVDofs(i, test_vdofs);
+      elmat.SetSize(test_vdofs.Size(), trial_vdofs.Size());
      elmat = 0.0;
   }
 }

-void MixedBilinearForm::ComputeBdrElementMatrix(int i, DenseMatrix &elmat) const
+void MixedBilinearForm::ComputeBdrElementMatrix(int i, DenseMatrix &elmat)
 {
-   const FiniteElement &trial_be = *trial_fes->GetBE(i);
-   const FiniteElement &test_be = *test_fes->GetBE(i);
-
   if (boundary_integs.Size())
   {
+      const FiniteElement &trial_be = *trial_fes->GetBE(i);
+      const FiniteElement &test_be = *test_fes->GetBE(i);
      ElementTransformation *eltrans = test_fes->GetBdrElementTransformation(i);
      boundary_integs[0]->AssembleElementMatrix2(trial_be, test_be, *eltrans,
                                                 elmat);
@@ -1959,191 +1735,9 @@ void MixedBilinearForm::ComputeBdrElementMatrix(int i, DenseMatrix &elmat) const
   }
   else
   {
-      const int tr_dofs = trial_be.GetDof() * trial_fes->GetVDim();
-      const int te_dofs = test_be.GetDof() * test_fes->GetVDim();
-
-      elmat.SetSize(te_dofs, tr_dofs);
-      elmat = 0.0;
-   }
-}
-
-void MixedBilinearForm::ComputeFaceMatrix(int i, DenseMatrix &elmat) const
-{
-   FaceElementTransformations *ftr;
-   Mesh *mesh = test_fes -> GetMesh();
-   ftr = mesh->GetFaceElementTransformations(i);
-   MFEM_ASSERT(ftr, "No associated face transformations.");
-
-   const FiniteElement *trial_fe1, *trial_fe2, *test_fe1, *test_fe2;
-
-   trial_fe1 = trial_fes->GetFE(ftr->Elem1No);
-   test_fe1 = test_fes->GetFE(ftr->Elem1No);
-   if (ftr->Elem2No >= 0)
-   {
-      trial_fe2 = trial_fes->GetFE(ftr->Elem2No);
-      test_fe2 = test_fes->GetFE(ftr->Elem2No);
-   }
-   else
-   {
-      // The test_fe2 object is really a dummy and not used on the
-      // boundaries, but we can't dereference a NULL pointer, and we don't
-      // want to actually make a fake element.
-      trial_fe2 = trial_fe1;
-      test_fe2 = test_fe1;
-   }
-
-   if (interior_face_integs.Size())
-   {
-      interior_face_integs[0]->AssembleFaceMatrix(*trial_fe1, *test_fe1, *trial_fe2,
-                                                  *test_fe2,
-                                                  *ftr, elmat);
-      for (int k = 1; k < interior_face_integs.Size(); k++)
-      {
-         interior_face_integs[k]->AssembleFaceMatrix(*trial_fe1, *test_fe1, *trial_fe2,
-                                                     *test_fe2,
-                                                     *ftr, elemmat);
-         elmat += elemmat;
-      }
-   }
-   else
-   {
-      int tr_dofs = trial_fe1->GetDof() * trial_fes->GetVDim();
-      int te_dofs = test_fe1->GetDof() * test_fes->GetVDim();
-      if (ftr->Elem2No >= 0)
-      {
-         tr_dofs += trial_fe2->GetDof() * trial_fes->GetVDim();
-         te_dofs += test_fe2->GetDof() * test_fes->GetVDim();
-      }
-
-      elmat.SetSize(te_dofs, tr_dofs);
-      elmat = 0.0;
-   }
-}
-
-void MixedBilinearForm::ComputeBdrFaceMatrix(int i, DenseMatrix &elmat) const
-{
-   FaceElementTransformations *ftr;
-   Mesh *mesh = test_fes -> GetMesh();
-   ftr = mesh->GetBdrFaceTransformations(i);
-   MFEM_ASSERT(ftr, "No associated boundary face.");
-
-   const FiniteElement *trial_fe1, *trial_fe2, *test_fe1, *test_fe2;
-
-   trial_fe1 = trial_fes->GetFE(ftr->Elem1No);
-   test_fe1 = test_fes->GetFE(ftr->Elem1No);
-   // The test_fe2 object is really a dummy and not used on the
-   // boundaries, but we can't dereference a NULL pointer, and we don't
-   // want to actually make a fake element.
-   trial_fe2 = trial_fe1;
-   test_fe2 = test_fe1;
-
-   if (boundary_face_integs.Size())
-   {
-      boundary_face_integs[0]->AssembleFaceMatrix(*trial_fe1, *test_fe1, *trial_fe2,
-                                                  *test_fe2,
-                                                  *ftr, elmat);
-      for (int k = 1; k < boundary_face_integs.Size(); k++)
-      {
-         boundary_face_integs[k]->AssembleFaceMatrix(*trial_fe1, *test_fe1, *trial_fe2,
-                                                     *test_fe2,
-                                                     *ftr, elemmat);
-         elmat += elemmat;
-      }
-   }
-   else
-   {
-      const int tr_dofs = trial_fe1->GetDof() * trial_fes->GetVDim();
-      const int te_dofs = test_fe1->GetDof() * test_fes->GetVDim();
-
-      elmat.SetSize(te_dofs, tr_dofs);
-      elmat = 0.0;
-   }
-}
-
-void MixedBilinearForm::ComputeTraceFaceMatrix(int i, DenseMatrix &elmat) const
-{
-   FaceElementTransformations *ftr;
-   Mesh *mesh = test_fes -> GetMesh();
-   ftr = mesh->GetFaceElementTransformations(i);
-   MFEM_ASSERT(ftr, "No associated face transformation.");
-
-   const FiniteElement *trial_face_fe, *test_fe1, *test_fe2;
-
-   trial_face_fe = trial_fes->GetFaceElement(i);
-   test_fe1 = test_fes->GetFE(ftr->Elem1No);
-   if (ftr->Elem2No >= 0)
-   {
-      test_fe2 = test_fes->GetFE(ftr->Elem2No);
-   }
-   else
-   {
-      // The test_fe2 object is really a dummy and not used on the
-      // boundaries, but we can't dereference a NULL pointer, and we don't
-      // want to actually make a fake element.
-      test_fe2 = test_fe1;
-   }
-
-   if (trace_face_integs.Size())
-   {
-      trace_face_integs[0]->AssembleFaceMatrix(*trial_face_fe, *test_fe1, *test_fe2,
-                                               *ftr, elmat);
-      for (int k = 1; k < trace_face_integs.Size(); k++)
-      {
-         trace_face_integs[k]->AssembleFaceMatrix(*trial_face_fe, *test_fe1, *test_fe2,
-                                                  *ftr, elemmat);
-         elmat += elemmat;
-      }
-   }
-   else
-   {
-      const int tr_face_dofs = trial_face_fe->GetDof() * trial_fes->GetVDim();
-      int te_dofs = test_fe1->GetDof() * test_fes->GetVDim();
-      if (ftr->Elem2No >= 0)
-      {
-         te_dofs += test_fe2->GetDof() * test_fes->GetVDim();
-      }
-
-      elmat.SetSize(te_dofs, tr_face_dofs);
-      elmat = 0.0;
-   }
-}
-
-void MixedBilinearForm::ComputeBdrTraceFaceMatrix(int i,
-                                                  DenseMatrix &elmat) const
-{
-   FaceElementTransformations *ftr;
-   Mesh *mesh = test_fes -> GetMesh();
-   ftr = mesh->GetBdrFaceTransformations(i);
-   MFEM_ASSERT(ftr, "No associated boundary face.");
-
-   const FiniteElement *trial_face_fe, *test_fe1, *test_fe2;
-   int iface = mesh->GetBdrElementFaceIndex(i);
-   trial_face_fe = trial_fes->GetFaceElement(iface);
-   test_fe1 = test_fes->GetFE(ftr->Elem1No);
-   // The test_fe2 object is really a dummy and not used on the
-   // boundaries, but we can't dereference a NULL pointer, and we don't
-   // want to actually make a fake element.
-   test_fe2 = test_fe1;
-
-   if (boundary_trace_face_integs.Size())
-   {
-      boundary_trace_face_integs[0]->AssembleFaceMatrix(*trial_face_fe, *test_fe1,
-                                                        *test_fe2,
-                                                        *ftr, elmat);
-      for (int k = 1; k < boundary_trace_face_integs.Size(); k++)
-      {
-         boundary_trace_face_integs[k]->AssembleFaceMatrix(*trial_face_fe, *test_fe1,
-                                                           *test_fe2,
-                                                           *ftr, elemmat);
-         elmat += elemmat;
-      }
-   }
-   else
-   {
-      const int tr_face_dofs = trial_face_fe->GetDof() * trial_fes->GetVDim();
-      int te_dofs = test_fe1->GetDof() * test_fes->GetVDim();
-
-      elmat.SetSize(te_dofs, tr_face_dofs);
+      trial_fes->GetBdrElementVDofs(i, trial_vdofs);
+      test_fes->GetBdrElementVDofs(i, test_vdofs);
+      elmat.SetSize(test_vdofs.Size(), trial_vdofs.Size());
      elmat = 0.0;
   }
 }
@@ -2186,59 +1780,36 @@ void MixedBilinearForm::AssembleBdrElementMatrix(
   mat->AddSubMatrix(test_vdofs_, trial_vdofs_, elmat, skip_zeros);
 }

-void MixedBilinearForm::EliminateTrialEssentialBC(
+void MixedBilinearForm::EliminateTrialDofs (
   const Array<int> &bdr_attr_is_ess, const Vector &sol, Vector &rhs )
 {
-   Array<int> trial_ess_dofs;
-   trial_fes->GetEssentialVDofs(bdr_attr_is_ess, trial_ess_dofs);
-   mat->EliminateCols(trial_ess_dofs, &sol, &rhs);
+   int i, j, k;
+   Array<int> tr_vdofs, cols_marker (trial_fes -> GetVSize());
+
+   cols_marker = 0;
+   for (i = 0; i < trial_fes -> GetNBE(); i++)
+      if (bdr_attr_is_ess[trial_fes -> GetBdrAttribute (i)-1])
+      {
+         trial_fes -> GetBdrElementVDofs (i, tr_vdofs);
+         for (j = 0; j < tr_vdofs.Size(); j++)
+         {
+            if ( (k = tr_vdofs[j]) < 0 )
+            {
+               k = -1-k;
+            }
+            cols_marker[k] = 1;
+         }
+      }
+   mat -> EliminateCols (cols_marker, &sol, &rhs);
 }

-void MixedBilinearForm::EliminateTrialEssentialBC(const Array<int>
-                                                  &bdr_attr_is_ess)
-{
-   Array<int> trial_ess_dofs;
-   trial_fes->GetEssentialVDofs(bdr_attr_is_ess, trial_ess_dofs);
-   mat->EliminateCols(trial_ess_dofs);
-}
-
-void MixedBilinearForm::EliminateTrialVDofs(const Array<int> &trial_vdofs_,
-                                            const Vector &sol, Vector &rhs)
-{
-   Array<int> trial_vdofs_marker;
-   FiniteElementSpace::ListToMarker(trial_vdofs_, mat->Width(),
-                                    trial_vdofs_marker);
-   mat->EliminateCols(trial_vdofs_marker, &sol, &rhs);
-}
-
-void MixedBilinearForm::EliminateTrialVDofs(const Array<int> &trial_vdofs_)
-{
-   if (mat_e == NULL)
-   {
-      mat_e = new SparseMatrix(mat->Height(), mat->Width());
-   }
-
-   Array<int> trial_vdofs_marker;
-   FiniteElementSpace::ListToMarker(trial_vdofs_, mat->Width(),
-                                    trial_vdofs_marker);
-   mat->EliminateCols(trial_vdofs_marker, *mat_e);
-   mat_e->Finalize();
-}
-
-void MixedBilinearForm::EliminateTrialVDofsInRHS(const Array<int> &trial_vdofs_,
-                                                 const Vector &x, Vector &b)
-{
-   mat_e->AddMult(x, b, -1.);
-}
-
-void MixedBilinearForm::EliminateEssentialBCFromTrialDofs(
+void MixedBilinearForm::EliminateEssentialBCFromTrialDofs (
   const Array<int> &marked_vdofs, const Vector &sol, Vector &rhs)
 {
-   mat->EliminateCols(marked_vdofs, &sol, &rhs);
+   mat -> EliminateCols (marked_vdofs, &sol, &rhs);
 }

-void MixedBilinearForm::EliminateTestEssentialBC(const Array<int>
-                                                 &bdr_attr_is_ess)
+void MixedBilinearForm::EliminateTestDofs (const Array<int> &bdr_attr_is_ess)
 {
   int i, j, k;
   Array<int> te_vdofs;
@@ -2258,14 +1829,6 @@ void MixedBilinearForm::EliminateTestEssentialBC(const Array<int>
      }
 }

-void MixedBilinearForm::EliminateTestVDofs(const Array<int> &test_vdofs_)
-{
-   for (int i=0; i<test_vdofs_.Size(); ++i)
-   {
-      mat->EliminateRow(test_vdofs_[i]);
-   }
-}
-
 void MixedBilinearForm::FormRectangularSystemMatrix(
   const Array<int> &trial_tdof_list,
   const Array<int> &test_tdof_list,
@@ -2302,9 +1865,20 @@ void MixedBilinearForm::FormRectangularSystemMatrix(
      mat = m;
   }

-   EliminateTrialVDofs(trial_tdof_list);
-   EliminateTestVDofs(test_tdof_list);
+   Array<int> ess_trial_tdof_marker, ess_test_tdof_marker;
+   FiniteElementSpace::ListToMarker(trial_tdof_list, trial_fes->GetTrueVSize(),
+                                    ess_trial_tdof_marker);
+   FiniteElementSpace::ListToMarker(test_tdof_list, test_fes->GetTrueVSize(),
+                                    ess_test_tdof_marker);

+   mat_e = new SparseMatrix(mat->Height(), mat->Width());
+   mat->EliminateCols(ess_trial_tdof_marker, *mat_e);
+
+   for (int i=0; i<test_tdof_list.Size(); ++i)
+   {
+      mat->EliminateRow(test_tdof_list[i]);
+   }
+   mat_e->Finalize();
   A.Reset(mat, false);
 }

@@ -2333,7 +1907,7 @@ void MixedBilinearForm::FormRectangularLinearSystem(
                                  A); // Set A = mat_e
   }
   // Eliminate essential BCs with B -= Ab xb
-   EliminateTrialVDofsInRHS(trial_tdof_list, X, B);
+   mat_e->AddMult(X, B, -1.0);

   B.SetSubVector(test_tdof_list, 0.0);
 }
@@ -2359,15 +1933,12 @@ MixedBilinearForm::~MixedBilinearForm()
      for (i = 0; i < domain_integs.Size(); i++) { delete domain_integs[i]; }
      for (i = 0; i < boundary_integs.Size(); i++)
      { delete boundary_integs[i]; }
-      for (i = 0; i < interior_face_integs.Size(); i++)
-      { delete interior_face_integs[i]; }
-      for (i = 0; i < boundary_face_integs.Size(); i++)
-      { delete boundary_face_integs[i]; }
      for (i = 0; i < trace_face_integs.Size(); i++)
      { delete trace_face_integs[i]; }
      for (i = 0; i < boundary_trace_face_integs.Size(); i++)
      { delete boundary_trace_face_integs[i]; }
   }
+   delete ext;
 }

 void DiscreteLinearOperator::SetAssemblyLevel(AssemblyLevel assembly_level)
@@ -2384,16 +1955,16 @@ void DiscreteLinearOperator::SetAssemblyLevel(AssemblyLevel assembly_level)
         // Use the original implementation for now
         break;
      case AssemblyLevel::ELEMENT:
-         MFEM_ABORT("Element assembly not supported yet... stay tuned!");
+         mfem_error("Element assembly not supported yet... stay tuned!");
         break;
      case AssemblyLevel::PARTIAL:
-         ext.reset(new PADiscreteLinearOperatorExtension(this));
+         ext = new PADiscreteLinearOperatorExtension(this);
         break;
      case AssemblyLevel::NONE:
-         MFEM_ABORT("Matrix-free action not supported yet... stay tuned!");
+         mfem_error("Matrix-free action not supported yet... stay tuned!");
         break;
      default:
-         MFEM_ABORT("Unknown assembly level");
+         mfem_error("Unknown assembly level");
   }
 }

@@ -83,7 +83,7 @@ protected:
   /** @brief Extension for supporting Full Assembly (FA),
       Element Assembly (EA),Partial Assembly (PA),
       or Matrix Free assembly (MF). */
-   std::unique_ptr<BilinearFormExtension> ext;
+   BilinearFormExtension *ext;

   /** Indicates if the sparse matrix is sorted after assembly when using
       Full Assembly (FA). */
@@ -119,13 +119,13 @@ protected:
   Array<BilinearFormIntegrator*> boundary_face_integs;
   Array<Array<int>*> boundary_face_integs_marker; ///< Entries are not owned.

-   mutable DenseMatrix elemmat;
-   mutable Array<int>  vdofs;
+   DenseMatrix elemmat;
+   Array<int>  vdofs;

-   std::unique_ptr<DenseTensor> element_matrices;
+   DenseTensor *element_matrices; ///< Owned.

-   std::unique_ptr<StaticCondensation> static_cond;
-   std::unique_ptr<Hybridization> hybridization;
+   StaticCondensation *static_cond; ///< Owned.
+   Hybridization *hybridization; ///< Owned.

   /** @brief This data member allows one to specify what should be done to the
       diagonal matrix entries and corresponding RHS values upon elimination of
@@ -148,11 +148,13 @@ protected:
   BilinearForm() : Matrix (0)
   {
      fes = NULL; sequence = -1;
-      mat = mat_e = NULL; extern_bfs = 0;
+      mat = mat_e = NULL; extern_bfs = 0; element_matrices = NULL;
+      static_cond = NULL; hybridization = NULL;
      precompute_sparsity = 0;
      diag_policy = DIAG_KEEP;
      assembly = AssemblyLevel::LEGACY;
      batch = 1;
+      ext = NULL;
   }

 private:
@@ -212,7 +214,7 @@ public:
   /// Returns the assembly level
   AssemblyLevel GetAssemblyLevel() const { return assembly; }

-   Hybridization *GetHybridization() const { return hybridization.get(); }
+   Hybridization *GetHybridization() const { return hybridization; }

   /** @brief Enable the use of static condensation. For details see the
       description for class StaticCondensation in fem/staticcond.hpp This
@@ -222,7 +224,7 @@ public:

   /** @brief Check if static condensation was actually enabled by a previous
       call to EnableStaticCondensation(). */
-   bool StaticCondensationIsEnabled() const { return static_cond.get() != nullptr; }
+   bool StaticCondensationIsEnabled() const { return static_cond; }

   /// Return the trace FE space associated with static condensation.
   FiniteElementSpace *SCFESpace() const
@@ -292,13 +294,13 @@ public:
   const real_t &operator()(int i, int j) { return (*mat)(i,j); }

   /// Returns a reference to: $ M_{ij} $
-   real_t &Elem(int i, int j) override;
+   virtual real_t &Elem(int i, int j);

   /// Returns constant reference to: $ M_{ij} $
-   const real_t &Elem(int i, int j) const override;
+   virtual const real_t &Elem(int i, int j) const;

   /// Matrix vector multiplication:  $ y = M x $
-   void Mult(const Vector &x, Vector &y) const override;
+   virtual void Mult(const Vector &x, Vector &y) const;

   /** @brief Matrix vector multiplication with the original uneliminated
       matrix.  The original matrix is $ M + M_e $ so we have:
@@ -307,7 +309,7 @@ public:
   { mat->Mult(x, y); mat_e->AddMult(x, y); }

   /// Add the matrix vector multiple to a vector:  $ y += a M x $
-   void AddMult(const Vector &x, Vector &y, const real_t a = 1.0) const override
+   virtual void AddMult(const Vector &x, Vector &y, const real_t a = 1.0) const
   { mat -> AddMult (x, y, a); }

   /** @brief Add the original uneliminated matrix vector multiple to a vector.
@@ -317,8 +319,8 @@ public:
   { mat->AddMult(x, y); mat_e->AddMult(x, y); }

   /// Add the matrix transpose vector multiplication:  $ y += a M^T x $
-   void AddMultTranspose(const Vector & x, Vector & y,
-                         const real_t a = 1.0) const override
+   virtual void AddMultTranspose(const Vector & x, Vector & y,
+                                 const real_t a = 1.0) const
   { mat->AddMultTranspose(x, y, a); }

   /** @brief Add the original uneliminated matrix transpose vector
@@ -328,7 +330,7 @@ public:
   { mat->AddMultTranspose(x, y); mat_e->AddMultTranspose(x, y); }

   /// Matrix transpose vector multiplication:  $ y = M^T x $
-   void MultTranspose(const Vector & x, Vector & y) const override;
+   virtual void MultTranspose(const Vector & x, Vector & y) const;

   /// Compute $ y^T M x $
   real_t InnerProduct(const Vector &x, const Vector &y) const
@@ -336,13 +338,13 @@ public:

   /** @brief Returns a pointer to (approximation) of the matrix inverse:
       $ M^{-1} $ (currently returns NULL) */
-   MatrixInverse *Inverse() const override;
+   virtual MatrixInverse *Inverse() const;

   /** @brief Finalizes the matrix initialization if the ::AssemblyLevel is
       AssemblyLevel::LEGACY.
       The matrix that gets finalized is different if you are using static
       condensation or hybridization.*/
-   void Finalize(int skip_zeros = 1) override;
+   virtual void Finalize(int skip_zeros = 1);

   /** @brief Returns a const reference to the sparse matrix:  $ M $
    *
@@ -456,18 +458,18 @@ public:
       conforming prolongation, and |.| denotes the entry-wise absolute value.
       In general, this is just an approximation of the exact diagonal for this
       case. */
-   void AssembleDiagonal(Vector &diag) const override;
+   virtual void AssembleDiagonal(Vector &diag) const;

   /// Get the finite element space prolongation operator.
-   const Operator *GetProlongation() const override
+   virtual const Operator *GetProlongation() const
   { return fes->GetConformingProlongation(); }

   /// Get the finite element space restriction operator
-   const Operator *GetRestriction() const override
+   virtual const Operator *GetRestriction() const
   { return fes->GetConformingRestriction(); }

   /// Get the output finite element space prolongation matrix
-   const Operator *GetOutputProlongation() const override
+   virtual const Operator *GetOutputProlongation() const
   { return GetProlongation(); }

   /** @brief Returns the output fe space restriction matrix, transposed
@@ -475,11 +477,11 @@ public:
       Logically, this is the transpose of GetOutputRestriction, but in
       practice it is convenient to have it in transposed form for
       construction of RAP operators in matrix-free methods. */
-   const Operator *GetOutputRestrictionTranspose() const override
+   virtual const Operator *GetOutputRestrictionTranspose() const
   { return fes->GetRestrictionTransposeOperator(); }

   /// Get the output finite element space restriction matrix
-   const Operator *GetOutputRestriction() const override
+   virtual const Operator *GetOutputRestriction() const
   { return GetRestriction(); }

   /// Compute serial RAP operator and store it in @a A as a SparseMatrix.
@@ -564,41 +566,24 @@ public:
       FormLinearSystem() method to recover the solution as a GridFunction-size
       vector in @a x. Use the same arguments as in the FormLinearSystem() call.
   */
-   void RecoverFEMSolution(const Vector &X, const Vector &b,
-                           Vector &x) override;
+   virtual void RecoverFEMSolution(const Vector &X, const Vector &b, Vector &x);

-   /// @brief Compute and store internally all element matrices.
-   ///
-   /// If AssemblyLevel::ELEMENT is selected with SetAssemblyLeve(), this will
-   /// use effecient (device-accelerated) assembly of the element matrices.
+   /// Compute and store internally all element matrices.
   void ComputeElementMatrices();

   /// Free the memory used by the element matrices.
-   void FreeElementMatrices() { element_matrices.reset(); }
-
-   /// @brief Return a DenseTensor containing the assembled element matrices.
-   ///
-   /// If AssemblyLevel::ELEMENT is selected with SetAssemblyLeve(), this will
-   /// use effecient (device-accelerated) assembly of the element matrices.
-   const DenseTensor &GetElementMatrices();
+   void FreeElementMatrices()
+   { delete element_matrices; element_matrices = NULL; }

   /// Compute the element matrix of the given element
   /** The element matrix is computed by calling the domain integrators
       or the one stored internally by a prior call of ComputeElementMatrices()
       is returned when available.
   */
-   void ComputeElementMatrix(int i, DenseMatrix &elmat) const;
+   void ComputeElementMatrix(int i, DenseMatrix &elmat);

   /// Compute the boundary element matrix of the given boundary element
-   /** @note The boundary attribute markers of the integrators are ignored. */
-   void ComputeBdrElementMatrix(int i, DenseMatrix &elmat) const;
-
-   /// Compute the face matrix of the given face element
-   void ComputeFaceMatrix(int i, DenseMatrix &elmat) const;
-
-   /// Compute the boundary face matrix of the given boundary element
-   /** @note The boundary attribute markers of the integrators are ignored. */
-   void ComputeBdrFaceMatrix(int i, DenseMatrix &elmat) const;
+   void ComputeBdrElementMatrix(int i, DenseMatrix &elmat);

   /// Assemble the given element matrix
   /** The element matrix @a elmat is assembled for the element @a i, i.e.
@@ -759,7 +744,7 @@ protected:

   /** Extension for supporting Full Assembly (FA), Element Assembly (EA),
       Partial Assembly (PA), or Matrix Free assembly (MF). */
-   std::unique_ptr<MixedBilinearFormExtension> ext;
+   MixedBilinearFormExtension *ext;

   /** @brief Indicates the BilinearFormIntegrator%s stored in
       MixedBilinearForm#domain_integs, MixedBilinearForm#boundary_integs,
@@ -778,14 +763,6 @@ protected:
   /// Entries are not owned.
   Array<Array<int>*> boundary_integs_marker;

-   /// Interior face integrators.
-   Array<BilinearFormIntegrator*> interior_face_integs;
-
-   /// Boundary face integrators.
-   Array<BilinearFormIntegrator*> boundary_face_integs;
-   /// Entries are not owned.
-   Array<Array<int>*> boundary_face_integs_marker;
-
   /// Trace face (skeleton) integrators.
   Array<BilinearFormIntegrator*> trace_face_integs;

@@ -794,8 +771,8 @@ protected:
   /// Entries are not owned.
   Array<Array<int>*> boundary_trace_face_integs_marker;

-   mutable DenseMatrix elemmat;
-   mutable Array<int>  trial_vdofs, test_vdofs;
+   DenseMatrix elemmat;
+   Array<int>  trial_vdofs, test_vdofs;

 private:
   /// Copy construction is not supported; body is undefined.
@@ -826,32 +803,32 @@ public:
                     MixedBilinearForm *mbf);

   /// Returns a reference to: $ M_{ij} $
-   real_t &Elem(int i, int j) override;
+   virtual real_t &Elem(int i, int j);

   /// Returns a reference to: $ M_{ij} $
-   const real_t &Elem(int i, int j) const override;
+   virtual const real_t &Elem(int i, int j) const;

   /// Matrix multiplication: $ y = M x $
-   void Mult(const Vector & x, Vector & y) const override;
+   virtual void Mult(const Vector & x, Vector & y) const;

   /// Add the matrix vector multiple to a vector:  $ y += a M x $
-   void AddMult(const Vector & x, Vector & y,
-                const real_t a = 1.0) const override;
+   virtual void AddMult(const Vector & x, Vector & y,
+                        const real_t a = 1.0) const;

   /// Matrix transpose vector multiplication:  $ y = M^T x $
-   void MultTranspose(const Vector & x, Vector & y) const override;
+   virtual void MultTranspose(const Vector & x, Vector & y) const;

   /// Add the matrix transpose vector multiplication:  $ y += a M^T x $
-   void AddMultTranspose(const Vector & x, Vector & y,
-                         const real_t a = 1.0) const override;
+   virtual void AddMultTranspose(const Vector & x, Vector & y,
+                                 const real_t a = 1.0) const;

   /** @brief Returns a pointer to (approximation) of the matrix inverse:
       $ M^{-1} $ (currently unimplemented and returns NULL)*/
-   MatrixInverse *Inverse() const override;
+   virtual MatrixInverse *Inverse() const;

   /** @brief Finalizes the matrix initialization if the ::AssemblyLevel is
       AssemblyLevel::LEGACY.*/
-   void Finalize(int skip_zeros = 1) override;
+   virtual void Finalize(int skip_zeros = 1);

   /** @brief Extract the associated matrix as SparseMatrix blocks. The number
       of block rows and columns is given by the vector dimensions (vdim) of the
@@ -862,37 +839,15 @@ public:
   /** This will segfault if the usual sparse mat is not defined
       like when static condensation is being used or AllocMat() has
       not yet been called. */
-   const SparseMatrix &SpMat() const
-   {
-      MFEM_VERIFY(mat, "mat is NULL and can't be dereferenced");
-      return *mat;
-   }
+   const SparseMatrix &SpMat() const { return *mat; }

   /// Returns a reference to the sparse matrix:  $ M $
-   SparseMatrix &SpMat()
-   {
-      MFEM_VERIFY(mat, "mat is NULL and can't be dereferenced");
-      return *mat;
-   }
+   SparseMatrix &SpMat() { return *mat; }

   /**  @brief Nullifies the internal matrix $ M $ and returns a pointer
        to it.  Used for transferring ownership. */
   SparseMatrix *LoseMat() { SparseMatrix *tmp = mat; mat = NULL; return tmp; }

-   /// Returns a const reference to the sparse matrix of eliminated b.c.:  $ M_e $
-   const SparseMatrix &SpMatElim() const
-   {
-      MFEM_VERIFY(mat_e, "mat_e is NULL and can't be dereferenced");
-      return *mat_e;
-   }
-
-   /// Returns a reference to the sparse matrix of eliminated b.c.:  $ M_e $
-   SparseMatrix &SpMatElim()
-   {
-      MFEM_VERIFY(mat_e, "mat_e is NULL and can't be dereferenced");
-      return *mat_e;
-   }
-
   /// Adds a domain integrator. Assumes ownership of @a bfi.
   void AddDomainIntegrator(BilinearFormIntegrator *bfi);

@@ -907,16 +862,6 @@ public:
   void AddBoundaryIntegrator(BilinearFormIntegrator * bfi,
                              Array<int> &bdr_marker);

-   /// Adds an interior face integrator. Assumes ownership of @a bfi.
-   void AddInteriorFaceIntegrator(BilinearFormIntegrator *bfi);
-
-   /// Adds a boundary face integrator. Assumes ownership of @a bfi.
-   void AddBdrFaceIntegrator(BilinearFormIntegrator *bfi);
-
-   /// Adds a boundary face integrator. Assumes ownership of @a bfi.
-   void AddBdrFaceIntegrator(BilinearFormIntegrator *bfi,
-                             Array<int> &bdr_marker);
-
   /** @brief Add a trace face integrator. Assumes ownership of @a bfi.

       This type of integrator assembles terms over all faces of the mesh using
@@ -947,16 +892,6 @@ public:
       corresponding pointer (to Array<int>) will be NULL. */
   Array<Array<int>*> *GetBBFI_Marker() { return &boundary_integs_marker; }

-   /// Access all integrators added with AddInteriorFaceIntegrator().
-   Array<BilinearFormIntegrator*> *GetFBFI() { return &interior_face_integs; }
-
-   /// Access all integrators added with AddBdrFaceIntegrator().
-   Array<BilinearFormIntegrator*> *GetBFBFI() { return &boundary_face_integs; }
-   /** @brief Access all boundary markers added with AddBdrFaceIntegrator().
-       If no marker was specified when the integrator was added, the
-       corresponding pointer (to Array<int>) will be NULL. */
-   Array<Array<int>*> *GetBFBFI_Marker() { return &boundary_face_integs_marker; }
-
   /// Access all integrators added with AddTraceFaceIntegrator().
   Array<BilinearFormIntegrator*> *GetTFBFI() { return &trace_face_integs; }

@@ -985,19 +920,19 @@ public:
   void AssembleDiagonal_ADAt(const Vector &D, Vector &diag) const;

   /// Get the input finite element space prolongation matrix
-   const Operator *GetProlongation() const override
+   virtual const Operator *GetProlongation() const
   { return trial_fes->GetProlongationMatrix(); }

   /// Get the input finite element space restriction matrix
-   const Operator *GetRestriction() const override
+   virtual const Operator *GetRestriction() const
   { return trial_fes->GetRestrictionMatrix(); }

   /// Get the test finite element space prolongation matrix
-   const Operator *GetOutputProlongation() const override
+   virtual const Operator *GetOutputProlongation() const
   { return test_fes->GetProlongationMatrix(); }

   /// Get the test finite element space restriction matrix
-   const Operator *GetOutputRestriction() const override
+   virtual const Operator *GetOutputRestriction() const
   { return test_fes->GetRestrictionMatrix(); }

   /** @brief For partially conforming trial and/or test FE spaces, complete the
@@ -1009,25 +944,10 @@ public:
   void ConformingAssemble();

   /// Compute the element matrix of the given element
-   void ComputeElementMatrix(int i, DenseMatrix &elmat) const;
+   void ComputeElementMatrix(int i, DenseMatrix &elmat);

   /// Compute the boundary element matrix of the given boundary element
-   /** @note The boundary attribute markers of the integrators are ignored. */
-   void ComputeBdrElementMatrix(int i, DenseMatrix &elmat) const;
-
-   /// Compute the trace face matrix of the given face element
-   void ComputeTraceFaceMatrix(int i, DenseMatrix &elmat) const;
-
-   /// Compute the boundary trace face matrix of the given boundary element
-   /** @note The boundary attribute markers of the integrators are ignored. */
-   void ComputeBdrTraceFaceMatrix(int i, DenseMatrix &elmat) const;
-
-   /// Compute the face matrix of the given face element
-   void ComputeFaceMatrix(int i, DenseMatrix &elmat) const;
-
-   /// Compute the boundary face matrix of the given boundary element
-   /** @note The boundary attribute markers of the integrators are ignored. */
-   void ComputeBdrFaceMatrix(int i, DenseMatrix &elmat) const;
+   void ComputeBdrElementMatrix(int i, DenseMatrix &elmat);

   /// Assemble the given element matrix
   /** The element matrix @a elmat is assembled for the element @a i, i.e.
@@ -1069,61 +989,24 @@ public:
                                 Array<int> &test_vdofs,
                                 int skip_zeros = 1);

-   /// Eliminate essential boundary trial DOFs from the system.
+   /// Eliminate essential boundary DOFs from the columns of the system.
   /** The array @a bdr_attr_is_ess marks boundary attributes that constitute
-       the essential part of the boundary. */
-   void EliminateTrialEssentialBC(const Array<int> &bdr_attr_is_ess,
-                                  const Vector &sol, Vector &rhs);
+       the essential part of the boundary.  All entries in the columns will be
+       set to 0.0 through elimination.*/
+   void EliminateTrialDofs(const Array<int> &bdr_attr_is_ess,
+                           const Vector &sol, Vector &rhs);

-   /// Eliminate essential boundary trial DOFs from the system matrix.
-   /** The array @a bdr_attr_is_ess marks boundary attributes that constitute
-       the essential part of the boundary. */
-   void EliminateTrialEssentialBC(const Array<int> &bdr_attr_is_ess);
-
-   /// (DEPRECATED) Eliminate essential boundary trial DOFs from the system.
-   /** @see EliminateTrialEssentialBC() */
-   MFEM_DEPRECATED void EliminateTrialDofs(const Array<int> &bdr_attr_is_ess,
-                                           const Vector &sol, Vector &rhs)
-   { EliminateTrialEssentialBC(bdr_attr_is_ess, sol, rhs); }
-
-   /// Eliminate the given trial @a vdofs. NOTE: here, @a vdofs is a list of DOFs.
-   /** In this case the eliminations are applied to the internal $ M $
-       and @a rhs without storing the elimination matrix $ M_e $. */
-   void EliminateTrialVDofs(const Array<int> &vdofs, const Vector &sol,
-                            Vector &rhs);
-
-   /// Eliminate the given trial @a vdofs, storing the eliminated part internally in $ M_e $.
-   /** This method works in conjunction with EliminateTrialVDofsInRHS() and allows
-       elimination of boundary conditions in multiple right-hand sides. In this
-       method, @a vdofs is a list of DOFs. */
-   void EliminateTrialVDofs(const Array<int> &vdofs);
-
-   /** @brief Use the stored eliminated part of the matrix (see
-       EliminateTrialVDofs(const Array<int> &)) to modify the r.h.s.
-       @a b; @a vdofs is a list of DOFs (non-directional, i.e. >= 0). */
-   void EliminateTrialVDofsInRHS(const Array<int> &vdofs, const Vector &x,
-                                 Vector &b);
-
-   /** @brief Similar to
-      EliminateTrialVDofs(const Array<int> &, const Vector &, Vector &)
-      but here @a ess_dofs is a marker (boolean) array on all vector-dofs
-      (@a ess_dofs[i] < 0 is true). */
+   /// Eliminate the list of DOFs from the columns of the system.
+   /** @a marked_vdofs is the of colunm numbers that will be eliminated.  All
+       entries in the columns will be set to 0.0 through elimination.*/
   void EliminateEssentialBCFromTrialDofs(const Array<int> &marked_vdofs,
                                          const Vector &sol, Vector &rhs);

-   /// Eliminate essential boundary test DOFs from the system matrix.
+   /// Eliminate essential boundary DOFs from the rows of the system.
   /** The array @a bdr_attr_is_ess marks boundary attributes that constitute
-       the essential part of the boundary. */
-   void EliminateTestEssentialBC(const Array<int> &bdr_attr_is_ess);
-
-   /// (DEPRECATED) Eliminate essential boundary test DOFs from the system.
-   /** @see EliminateTestEssentialBC() */
-   MFEM_DEPRECATED virtual void EliminateTestDofs(const Array<int>
-                                                  &bdr_attr_is_ess)
-   { EliminateTestEssentialBC(bdr_attr_is_ess); }
-
-   /// Eliminate the given test @a vdofs. NOTE: here, @a vdofs is a list of DOFs.
-   void EliminateTestVDofs(const Array<int> &vdofs);
+       the essential part of the boundary.  All entries in the rows will be
+       set to 0.0 through elimination.*/
+   virtual void EliminateTestDofs(const Array<int> &bdr_attr_is_ess);

   /** @brief Return in @a A that is column-constrained.

@@ -1279,7 +1162,7 @@ public:

   /** @brief Get the output finite element space restriction matrix in
       transposed form. */
-   const Operator *GetOutputRestrictionTranspose() const override
+   virtual const Operator *GetOutputRestrictionTranspose() const
   { return test_fes->GetRestrictionTransposeOperator(); }
 };

@@ -16,7 +16,6 @@
 #include "bilinearform.hpp"
 #include "pbilinearform.hpp"
 #include "pgridfunc.hpp"
-#include "fe/face_map_utils.hpp"
 #include "ceed/interface/util.hpp"

 namespace mfem
@@ -865,139 +864,54 @@ void EABilinearFormExtension::Assemble()
   ne = trial_fes->GetMesh()->GetNE();
   elemDofs = trial_fes->GetFE(0)->GetDof();

-   Vector ea_data_tmp;
+   ea_data.SetSize(ne*elemDofs*elemDofs, Device::GetMemoryType());
+   ea_data.UseDevice(true);

-   auto add_with_markers = [&](const Vector &ea_1, Vector &ea_2, const int ne_,
-                               const Array<int> &markers, const Array<int> &attrs,
-                               const bool add)
+   Array<BilinearFormIntegrator*> &integrators = *a->GetDBFI();
+   const int integratorCount = integrators.Size();
+   if ( integratorCount == 0 )
   {
-      if (ne_ == 0) { return; }
-      const int sz = ea_1.Size() / ne_;
-      const int *d_m = markers.Read();
-      const int *d_a = attrs.Read();
-      const auto d_ea_1 = Reshape(ea_1.Read(), sz, ne_);
-      auto d_ea_2 = Reshape(add ? ea_2.ReadWrite() : ea_2.Write(), sz, ne_);
-
-      mfem::forall(sz*ne_, [=] MFEM_HOST_DEVICE (int idx)
-      {
-         const int i = idx % sz;
-         const int e = idx / sz;
-         const real_t val = d_m[d_a[e] - 1] ? d_ea_1(i, e) : 0.0;
-         if (add)
-         {
-            d_ea_2(i, e) += val;
-         }
-         else
-         {
-            d_ea_2(i, e) = val;
-         }
-      });
-   };
-
+      ea_data = 0.0;
+   }
+   for (int i = 0; i < integratorCount; ++i)
   {
-      ea_data.SetSize(ne*elemDofs*elemDofs);
-      ea_data.UseDevice(true);
-      Array<BilinearFormIntegrator*> &integrators = *a->GetDBFI();
-      Array<Array<int>*> &markers_array = *a->GetDBFI_Marker();
-
-      if (integrators.Size() == 0) { ea_data = 0.0; }
-
-      for (int i = 0; i < integrators.Size(); ++i)
-      {
-         const bool add = (i > 0);
-         const Array<int> *markers = markers_array[i];
-         if (markers == nullptr)
-         {
-            integrators[i]->AssembleEA(*a->FESpace(), ea_data, add);
-         }
-         else
-         {
-            ea_data_tmp.SetSize(ea_data.Size());
-            integrators[i]->AssembleEA(*a->FESpace(), ea_data_tmp, false);
-            add_with_markers(ea_data_tmp, ea_data, ne, *markers,
-                             elem_attributes, add);
-         }
-      }
+      integrators[i]->AssembleEA(*a->FESpace(), ea_data, i);
   }

   faceDofs = trial_fes ->
              GetTraceElement(0, trial_fes->GetMesh()->GetFaceGeometry(0)) ->
              GetDof();

+   MFEM_VERIFY(a->GetBBFI()->Size() == 0,
+               "Element assembly does not support AddBoundaryIntegrator yet.");
+
+   Array<BilinearFormIntegrator*> &intFaceIntegrators = *a->GetFBFI();
+   const int intFaceIntegratorCount = intFaceIntegrators.Size();
+   if (intFaceIntegratorCount>0)
   {
-      Array<BilinearFormIntegrator*> &bdr_integs = *a->GetBBFI();
-      Array<Array<int>*> &markers_array = *a->GetBBFI_Marker();
-      const int n_bdr_integs = bdr_integs.Size();
-      if (n_bdr_integs > 0)
-      {
-         nf_bdr = trial_fes->GetNFbyType(FaceType::Boundary);
-         ea_data_bdr.SetSize(nf_bdr*faceDofs*faceDofs);
-      }
-      for (int i = 0; i < n_bdr_integs; ++i)
-      {
-         const bool add = (i > 0);
-         const Array<int> *markers = markers_array[i];
-         if (markers == nullptr)
-         {
-            bdr_integs[i]->AssembleEABoundary(*a->FESpace(), ea_data_bdr, add);
-         }
-         else
-         {
-            ea_data_tmp.SetSize(ea_data_bdr.Size());
-            bdr_integs[i]->AssembleEABoundary(*a->FESpace(), ea_data_tmp, add);
-            add_with_markers(ea_data_tmp, ea_data_bdr, nf_bdr, *markers,
-                             bdr_attributes, add);
-         }
-      }
+      nf_int = trial_fes->GetNFbyType(FaceType::Interior);
+      ea_data_int.SetSize(2*nf_int*faceDofs*faceDofs, Device::GetMemoryType());
+      ea_data_ext.SetSize(2*nf_int*faceDofs*faceDofs, Device::GetMemoryType());
+   }
+   for (int i = 0; i < intFaceIntegratorCount; ++i)
+   {
+      intFaceIntegrators[i]->AssembleEAInteriorFaces(*a->FESpace(),
+                                                     ea_data_int,
+                                                     ea_data_ext,
+                                                     i);
   }

+   Array<BilinearFormIntegrator*> &bdrFaceIntegrators = *a->GetBFBFI();
+   const int boundFaceIntegratorCount = bdrFaceIntegrators.Size();
+   if (boundFaceIntegratorCount>0)
   {
-      Array<BilinearFormIntegrator*> &intFaceIntegrators = *a->GetFBFI();
-      const int intFaceIntegratorCount = intFaceIntegrators.Size();
-      if (intFaceIntegratorCount>0)
-      {
-         nf_int = trial_fes->GetNFbyType(FaceType::Interior);
-         ea_data_int.SetSize(2*nf_int*faceDofs*faceDofs);
-         ea_data_ext.SetSize(2*nf_int*faceDofs*faceDofs);
-      }
-      for (int i = 0; i < intFaceIntegratorCount; ++i)
-      {
-         const bool add = (i > 0);
-         intFaceIntegrators[i]->AssembleEAInteriorFaces(*a->FESpace(),
-                                                        ea_data_int,
-                                                        ea_data_ext,
-                                                        add);
-      }
+      nf_bdr = trial_fes->GetNFbyType(FaceType::Boundary);
+      ea_data_bdr.SetSize(nf_bdr*faceDofs*faceDofs, Device::GetMemoryType());
+      ea_data_bdr = 0.0;
   }
-
+   for (int i = 0; i < boundFaceIntegratorCount; ++i)
   {
-      Array<BilinearFormIntegrator*> &bdr_face_integs = *a->GetBFBFI();
-      Array<Array<int>*> &markers_array = *a->GetBFBFI_Marker();
-      const int n_bdr_face_integs = bdr_face_integs.Size();
-      if (n_bdr_face_integs > 0)
-      {
-         nf_bdr = trial_fes->GetNFbyType(FaceType::Boundary);
-         ea_data_bdr.SetSize(nf_bdr*faceDofs*faceDofs);
-      }
-      for (int i = 0; i < n_bdr_face_integs; ++i)
-      {
-         const bool add = (i > 0);
-         const Array<int> *markers = markers_array[i];
-         if (markers == nullptr)
-         {
-            bdr_face_integs[i]->AssembleEABoundaryFaces(
-               *a->FESpace(), ea_data_bdr, add);
-         }
-         else
-         {
-            ea_data_tmp.SetSize(ea_data_bdr.Size());
-            bdr_face_integs[i]->AssembleEABoundaryFaces(*a->FESpace(),
-                                                        ea_data_tmp,
-                                                        add);
-            add_with_markers(ea_data_tmp, ea_data_bdr, nf_bdr, *markers,
-                             bdr_attributes, add);
-         }
-      }
+      bdrFaceIntegrators[i]->AssembleEABoundaryFaces(*a->FESpace(),ea_data_bdr,i);
   }

   if (factorize_face_terms && int_face_restrict_lex)
@@ -1109,29 +1023,34 @@ void EABilinearFormExtension::Mult(const Vector &x, Vector &y) const
   }

   // Treatment of boundary faces
-   if (!factorize_face_terms && bdr_face_restrict_lex && ea_data_bdr.Size() > 0)
+   Array<BilinearFormIntegrator*> &bdrFaceIntegrators = *a->GetBFBFI();
+   const int bFISz = bdrFaceIntegrators.Size();
+   if (!factorize_face_terms && bdr_face_restrict_lex && bFISz>0)
   {
      // Apply the Boundary Face Restriction
      bdr_face_restrict_lex->Mult(x, bdr_face_X);
-      bdr_face_Y = 0.0;
-      // Apply the boundary face matrices
-      const int NDOFS = faceDofs;
-      auto X = Reshape(bdr_face_X.Read(), NDOFS, nf_bdr);
-      auto Y = Reshape(bdr_face_Y.ReadWrite(), NDOFS, nf_bdr);
-      auto A = Reshape(ea_data_bdr.Read(), NDOFS, NDOFS, nf_bdr);
-      mfem::forall(nf_bdr*NDOFS, [=] MFEM_HOST_DEVICE (int glob_j)
+      if (bdr_face_X.Size()>0)
      {
-         const int f = glob_j/NDOFS;
-         const int j = glob_j%NDOFS;
-         real_t res = 0.0;
-         for (int i = 0; i < NDOFS; i++)
+         bdr_face_Y = 0.0;
+         // Apply the boundary face matrices
+         const int NDOFS = faceDofs;
+         auto X = Reshape(bdr_face_X.Read(), NDOFS, nf_bdr);
+         auto Y = Reshape(bdr_face_Y.ReadWrite(), NDOFS, nf_bdr);
+         auto A = Reshape(ea_data_bdr.Read(), NDOFS, NDOFS, nf_bdr);
+         mfem::forall(nf_bdr*NDOFS, [=] MFEM_HOST_DEVICE (int glob_j)
         {
-            res += A(i, j, f)*X(i, f);
-         }
-         Y(j, f) += res;
-      });
-      // Apply the Boundary Face Restriction transposed
-      bdr_face_restrict_lex->AddMultTransposeInPlace(bdr_face_Y, y);
+            const int f = glob_j/NDOFS;
+            const int j = glob_j%NDOFS;
+            real_t res = 0.0;
+            for (int i = 0; i < NDOFS; i++)
+            {
+               res += A(i, j, f)*X(i, f);
+            }
+            Y(j, f) += res;
+         });
+         // Apply the Boundary Face Restriction transposed
+         bdr_face_restrict_lex->AddMultTransposeInPlace(bdr_face_Y, y);
+      }
   }
 }

@@ -1232,163 +1151,34 @@ void EABilinearFormExtension::MultTranspose(const Vector &x, Vector &y) const
   }

   // Treatment of boundary faces
-   if (!factorize_face_terms && bdr_face_restrict_lex && ea_data_bdr.Size() > 0)
+   Array<BilinearFormIntegrator*> &bdrFaceIntegrators = *a->GetBFBFI();
+   const int bFISz = bdrFaceIntegrators.Size();
+   if (!factorize_face_terms && bdr_face_restrict_lex && bFISz>0)
   {
      // Apply the Boundary Face Restriction
      bdr_face_restrict_lex->Mult(x, bdr_face_X);
-      bdr_face_Y = 0.0;
-      // Apply the boundary face matrices transposed
-      const int NDOFS = faceDofs;
-      auto X = Reshape(bdr_face_X.Read(), NDOFS, nf_bdr);
-      auto Y = Reshape(bdr_face_Y.ReadWrite(), NDOFS, nf_bdr);
-      auto A = Reshape(ea_data_bdr.Read(), NDOFS, NDOFS, nf_bdr);
-      mfem::forall(nf_bdr*NDOFS, [=] MFEM_HOST_DEVICE (int glob_j)
+      if (bdr_face_X.Size()>0)
      {
-         const int f = glob_j/NDOFS;
-         const int j = glob_j%NDOFS;
-         real_t res = 0.0;
-         for (int i = 0; i < NDOFS; i++)
+         bdr_face_Y = 0.0;
+         // Apply the boundary face matrices transposed
+         const int NDOFS = faceDofs;
+         auto X = Reshape(bdr_face_X.Read(), NDOFS, nf_bdr);
+         auto Y = Reshape(bdr_face_Y.ReadWrite(), NDOFS, nf_bdr);
+         auto A = Reshape(ea_data_bdr.Read(), NDOFS, NDOFS, nf_bdr);
+         mfem::forall(nf_bdr*NDOFS, [=] MFEM_HOST_DEVICE (int glob_j)
         {
-            res += A(j, i, f)*X(i, f);
-         }
-         Y(j, f) += res;
-      });
-      // Apply the Boundary Face Restriction transposed
-      bdr_face_restrict_lex->AddMultTransposeInPlace(bdr_face_Y, y);
-   }
-}
-
-void EABilinearFormExtension::GetElementMatrices(
-   DenseTensor &element_matrices, ElementDofOrdering ordering, bool add_bdr)
-{
-   // Ensure the EA data is assembled
-   if (ea_data.Size() == 0) { Assemble(); }
-
-   const int ndofs = elemDofs;
-   element_matrices.SetSize(ndofs, ndofs, ne);
-   const int N = element_matrices.TotalSize();
-
-   const auto d_ea_data = Reshape(ea_data.Read(), ndofs, ndofs, ne);
-   auto d_element_matrices = Reshape(element_matrices.Write(),
-                                     ndofs, ndofs,
-                                     ne);
-
-   const int *d_dof_map = nullptr;
-   Array<int> dof_map;
-   if (ordering == ElementDofOrdering::NATIVE)
-   {
-      const TensorBasisElement* tbe =
-         dynamic_cast<const TensorBasisElement*>(trial_fes->GetFE(0));
-      if (tbe)
-      {
-         // Deep copy to avoid issues with host device (see similar comment in
-         // HybridizationExtension::ConstructC).
-         dof_map = tbe->GetDofMap();
-         d_dof_map = dof_map.Read();
-      }
-   }
-
-   if (d_dof_map)
-   {
-      // Reordering required
-      mfem::forall(N, [=] MFEM_HOST_DEVICE (int idx)
-      {
-         const int e = idx / ndofs / ndofs;
-         const int i = idx % ndofs;
-         const int j = (idx / ndofs) % ndofs;
-         const int ii_s = d_dof_map[i];
-         const int ii = (ii_s >= 0) ? ii_s : -1 - ii_s;
-         const int s_i = (ii_s >= 0) ? 1 : -1;
-         const int jj_s = d_dof_map[j];
-         const int jj = (jj_s >= 0) ? jj_s : -1 - jj_s;
-         const int s_j = (jj_s >= 0) ? 1 : -1;
-         d_element_matrices(ii, jj, e) = s_i*s_j*d_ea_data(j, i, e);
-      });
-   }
-   else
-   {
-      // No reordering required
-      mfem::forall(N, [=] MFEM_HOST_DEVICE (int idx)
-      {
-         const int e = idx / ndofs / ndofs;
-         const int i = idx % ndofs;
-         const int j = (idx / ndofs) % ndofs;
-         d_element_matrices(i, j, e) = d_ea_data(j, i, e);
-      });
-   }
-
-   if (add_bdr && ea_data_bdr.Size() > 0)
-   {
-      const int ndof_face = faceDofs;
-      const auto d_ea_bdr = Reshape(ea_data_bdr.Read(),
-                                    ndof_face, ndof_face, nf_bdr);
-
-      // Get all the local face maps (mapping from lexicographic face index to
-      // lexicographic volume index, depending on the local face index).
-      const Mesh &mesh = *trial_fes->GetMesh();
-      const int dim = mesh.Dimension();
-      const int n_faces_per_el = 2*dim; // assuming tensor product
-      Array<int> face_maps(ndof_face * n_faces_per_el);
-      for (int lf_i = 0; lf_i < n_faces_per_el; ++lf_i)
-      {
-         Array<int> face_map(ndof_face);
-         trial_fes->GetFE(0)->GetFaceMap(lf_i, face_map);
-         for (int i = 0; i < ndof_face; ++i)
-         {
-            face_maps[i + lf_i*ndof_face] = face_map[i];
-         }
-      }
-
-      Array<int> face_info(nf_bdr * 2);
-      {
-         int fidx = 0;
-         for (int f = 0; f < mesh.GetNumFaces(); ++f)
-         {
-            Mesh::FaceInformation finfo = mesh.GetFaceInformation(f);
-            if (!finfo.IsBoundary()) { continue; }
-            face_info[0 + fidx*2] = finfo.element[0].local_face_id;
-            face_info[1 + fidx*2] = finfo.element[0].index;
-            fidx++;
-         }
-      }
-
-      const auto d_face_maps = Reshape(face_maps.Read(), ndof_face, n_faces_per_el);
-      const auto d_face_info = Reshape(face_info.Read(), 2, nf_bdr);
-
-      const bool reorder = (ordering == ElementDofOrdering::NATIVE);
-
-      mfem::forall_2D(nf_bdr, ndof_face, ndof_face, [=] MFEM_HOST_DEVICE (int f)
-      {
-         const int lf_i = d_face_info(0, f);
-         const int e = d_face_info(1, f);
-         // Loop over face indices in "native ordering"
-         MFEM_FOREACH_THREAD(i_lex_face, x, ndof_face)
-         {
-            // Convert from lexicographic face DOF to volume DOF
-            const int i_lex = d_face_maps(i_lex_face, lf_i);
-
-            const int ii_s = d_dof_map[i_lex];
-            const int ii = (ii_s >= 0) ? ii_s : -1 - ii_s;
-
-            const int i = reorder ? ii : i_lex;
-            const int s_i = (ii_s < 0 && reorder) ? -1 : 1;
-
-            MFEM_FOREACH_THREAD(j_lex_face, y, ndof_face)
+            const int f = glob_j/NDOFS;
+            const int j = glob_j%NDOFS;
+            real_t res = 0.0;
+            for (int i = 0; i < NDOFS; i++)
            {
-               // Convert from lexicographic face DOF to volume DOF
-               const int j_lex = d_face_maps(j_lex_face, lf_i);
-
-               const int jj_s = d_dof_map[j_lex];
-               const int jj = (jj_s >= 0) ? jj_s : -1 - jj_s;
-
-               const int j = reorder ? jj : j_lex;
-               const int s_j = (jj_s < 0 && reorder) ? -1 : 1;
-
-               AtomicAdd(d_element_matrices(i, j, e),
-                         s_i*s_j*d_ea_bdr(i_lex_face, j_lex_face, f));
+               res += A(j, i, f)*X(i, f);
            }
-         }
-      });
+            Y(j, f) += res;
+         });
+         // Apply the Boundary Face Restriction transposed
+         bdr_face_restrict_lex->AddMultTransposeInPlace(bdr_face_Y, y);
+      }
   }
 }

@@ -37,19 +37,19 @@ protected:
 public:
   BilinearFormExtension(BilinearForm *form);

-   MemoryClass GetMemoryClass() const override
+   virtual MemoryClass GetMemoryClass() const
   { return Device::GetDeviceMemoryClass(); }

   /// Get the finite element space prolongation matrix
-   const Operator *GetProlongation() const override;
+   virtual const Operator *GetProlongation() const;

   /// Get the finite element space restriction matrix
-   const Operator *GetRestriction() const override;
+   virtual const Operator *GetRestriction() const;

   /// Assemble at the level given for the BilinearFormExtension subclass
   virtual void Assemble() = 0;

-   void AssembleDiagonal(Vector &diag) const override
+   virtual void AssembleDiagonal(Vector &diag) const
   {
      MFEM_ABORT("AssembleDiagonal not implemented for this assembly level!");
   }
@@ -83,17 +83,16 @@ protected:
 public:
   PABilinearFormExtension(BilinearForm*);

-   void Assemble() override;
-   void AssembleDiagonal(Vector &diag) const override;
-   void FormSystemMatrix(const Array<int> &ess_tdof_list,
-                         OperatorHandle &A) override;
+   void Assemble();
+   void AssembleDiagonal(Vector &diag) const;
+   void FormSystemMatrix(const Array<int> &ess_tdof_list, OperatorHandle &A);
   void FormLinearSystem(const Array<int> &ess_tdof_list,
                         Vector &x, Vector &b,
                         OperatorHandle &A, Vector &X, Vector &B,
-                         int copy_interior = 0) override;
-   void Mult(const Vector &x, Vector &y) const override;
-   void MultTranspose(const Vector &x, Vector &y) const override;
-   void Update() override;
+                         int copy_interior = 0);
+   void Mult(const Vector &x, Vector &y) const;
+   void MultTranspose(const Vector &x, Vector &y) const;
+   void Update();

 protected:
   void SetupRestrictionOperators(const L2FaceValues m);
@@ -151,20 +150,9 @@ protected:
 public:
   EABilinearFormExtension(BilinearForm *form);

-   void Assemble() override;
-   void Mult(const Vector &x, Vector &y) const override;
-   void MultTranspose(const Vector &x, Vector &y) const override;
-
-   /// @brief Populates @a element_matrices with the element matrices.
-   ///
-   /// The element matrices are converted from row-major (how they are stored in
-   /// @a ea_data) to column-major format.
-   ///
-   /// If @a ordering is ElementDofOrdering::NATIVE, then the matrices are
-   /// reordered from the lexicographic ordering used internally.
-   void GetElementMatrices(DenseTensor &element_matrices,
-                           ElementDofOrdering ordering,
-                           bool add_bdr);
+   void Assemble();
+   void Mult(const Vector &x, Vector &y) const;
+   void MultTranspose(const Vector &x, Vector &y) const;
 };

 /// Data and methods for fully-assembled bilinear forms
@@ -177,19 +165,18 @@ private:
 public:
   FABilinearFormExtension(BilinearForm *form);

-   void Assemble() override;
+   void Assemble();
   void RAP(OperatorHandle &A);
   /** @note Always does `DIAG_ONE` policy to be consistent with
       `Operator::FormConstrainedSystemOperator`. */
   void EliminateBC(const Array<int> &ess_dofs, OperatorHandle &A);
-   void FormSystemMatrix(const Array<int> &ess_tdof_list,
-                         OperatorHandle &A) override;
+   void FormSystemMatrix(const Array<int> &ess_tdof_list, OperatorHandle &A);
   void FormLinearSystem(const Array<int> &ess_tdof_list,
                         Vector &x, Vector &b,
                         OperatorHandle &A, Vector &X, Vector &B,
-                         int copy_interior = 0) override;
-   void Mult(const Vector &x, Vector &y) const override;
-   void MultTranspose(const Vector &x, Vector &y) const override;
+                         int copy_interior = 0);
+   void Mult(const Vector &x, Vector &y) const;
+   void MultTranspose(const Vector &x, Vector &y) const;

   /** DGMult and DGMultTranspose use the extended L-vector to perform the
       computation. */
@@ -212,17 +199,16 @@ protected:
 public:
   MFBilinearFormExtension(BilinearForm *form);

-   void Assemble() override;
-   void AssembleDiagonal(Vector &diag) const override;
-   void FormSystemMatrix(const Array<int> &ess_tdof_list,
-                         OperatorHandle &A) override;
+   void Assemble();
+   void AssembleDiagonal(Vector &diag) const;
+   void FormSystemMatrix(const Array<int> &ess_tdof_list, OperatorHandle &A);
   void FormLinearSystem(const Array<int> &ess_tdof_list,
                         Vector &x, Vector &b,
                         OperatorHandle &A, Vector &X, Vector &B,
-                         int copy_interior = 0) override;
-   void Mult(const Vector &x, Vector &y) const override;
-   void MultTranspose(const Vector &x, Vector &y) const override;
-   void Update() override;
+                         int copy_interior = 0);
+   void Mult(const Vector &x, Vector &y) const;
+   void MultTranspose(const Vector &x, Vector &y) const;
+   void Update();
 };

 /// Class extending the MixedBilinearForm class to support different AssemblyLevels.
@@ -239,20 +225,20 @@ protected:
 public:
   MixedBilinearFormExtension(MixedBilinearForm *form);

-   MemoryClass GetMemoryClass() const override
+   virtual MemoryClass GetMemoryClass() const
   { return Device::GetMemoryClass(); }

   /// Get the finite element space prolongation matrix
-   const Operator *GetProlongation() const override;
+   virtual const Operator *GetProlongation() const;

   /// Get the finite element space restriction matrix
-   const Operator *GetRestriction() const override;
+   virtual const Operator *GetRestriction() const;

   /// Get the output finite element space restriction matrix
-   const Operator *GetOutputProlongation() const override;
+   virtual const Operator *GetOutputProlongation() const;

   /// Get the output finite element space restriction matrix
-   const Operator *GetOutputRestriction() const override;
+   virtual const Operator *GetOutputRestriction() const;

   virtual void Assemble() = 0;
   virtual void FormRectangularSystemOperator(const Array<int> &trial_tdof_list,
@@ -287,7 +273,7 @@ public:
   PAMixedBilinearFormExtension(MixedBilinearForm *form);

   /// Partial assembly of all internal integrators
-   void Assemble() override;
+   void Assemble();
   /**
      @brief Setup OperatorHandle A to contain constrained linear operator

@@ -297,7 +283,7 @@ public:
   */
   void FormRectangularSystemOperator(const Array<int> &trial_tdof_list,
                                      const Array<int> &test_tdof_list,
-                                      OperatorHandle &A) override;
+                                      OperatorHandle &A);
   /**
      Setup OperatorHandle A to contain constrained linear operator and
      eliminate columns corresponding to essential dofs from system,
@@ -306,21 +292,20 @@ public:
   void FormRectangularLinearSystem(const Array<int> &trial_tdof_list,
                                    const Array<int> &test_tdof_list,
                                    Vector &x, Vector &b,
-                                    OperatorHandle &A, Vector &X, Vector &B) override;
+                                    OperatorHandle &A, Vector &X, Vector &B);
   /// y = A*x
-   void Mult(const Vector &x, Vector &y) const override;
+   void Mult(const Vector &x, Vector &y) const;
   /// y += c*A*x
-   void AddMult(const Vector &x, Vector &y, const real_t c=1.0) const override;
+   void AddMult(const Vector &x, Vector &y, const real_t c=1.0) const;
   /// y = A^T*x
-   void MultTranspose(const Vector &x, Vector &y) const override;
+   void MultTranspose(const Vector &x, Vector &y) const;
   /// y += c*A^T*x
-   void AddMultTranspose(const Vector &x, Vector &y,
-                         const real_t c=1.0) const override;
+   void AddMultTranspose(const Vector &x, Vector &y, const real_t c=1.0) const;
   /// Assemble the diagonal of ADA^T for a diagonal vector D.
-   void AssembleDiagonal_ADAt(const Vector &D, Vector &diag) const override;
+   void AssembleDiagonal_ADAt(const Vector &D, Vector &diag) const;

   /// Update internals for when a new MixedBilinearForm is given to this class
-   void Update() override;
+   void Update();
 };


@@ -337,17 +322,16 @@ public:
   PADiscreteLinearOperatorExtension(DiscreteLinearOperator *linop);

   /// Partial assembly of all internal integrators
-   void Assemble() override;
+   void Assemble();

-   void AddMult(const Vector &x, Vector &y, const real_t c=1.0) const override;
+   void AddMult(const Vector &x, Vector &y, const real_t c=1.0) const;

-   void AddMultTranspose(const Vector &x, Vector &y,
-                         const real_t c=1.0) const override;
+   void AddMultTranspose(const Vector &x, Vector &y, const real_t c=1.0) const;

   void FormRectangularSystemOperator(const Array<int>&, const Array<int>&,
-                                      OperatorHandle& A) override;
+                                      OperatorHandle& A);

-   const Operator * GetOutputRestrictionTranspose() const override;
+   const Operator * GetOutputRestrictionTranspose() const;

 private:
   Vector test_multiplicity;
@@ -72,14 +72,6 @@ void BilinearFormIntegrator::AssembleEA(const FiniteElementSpace &fes,
              "   is not implemented for this class.");
 }

-void BilinearFormIntegrator::AssembleEABoundary(const FiniteElementSpace &fes,
-                                                Vector &emat,
-                                                const bool add)
-{
-   MFEM_ABORT("BilinearFormIntegrator::AssembleEABoundary(...)\n"
-              "   is not implemented for this class.");
-}
-
 void BilinearFormIntegrator::AssembleEAInteriorFaces(const FiniteElementSpace
                                                     &fes,
                                                     Vector &ea_data_int,
@@ -90,16 +82,6 @@ void BilinearFormIntegrator::AssembleEAInteriorFaces(const FiniteElementSpace
              "   is not implemented for this class.");
 }

-void BilinearFormIntegrator::AssembleEAInteriorFaces(
-   const FiniteElementSpace &trial_fes,
-   const FiniteElementSpace &test_fes,
-   Vector &emat,
-   const bool add)
-{
-   MFEM_ABORT("BilinearFormIntegrator::AssembleEAInteriorFaces(...)\n"
-              "   is not implemented for this class.");
-}
-
 void BilinearFormIntegrator::AssembleEABoundaryFaces(const FiniteElementSpace
                                                     &fes,
                                                     Vector &ea_data_bdr,
@@ -188,16 +170,6 @@ void BilinearFormIntegrator::AssembleFaceMatrix(
              "   is not implemented for this class.");
 }

-void BilinearFormIntegrator::AssembleFaceMatrix(
-   const FiniteElement &trial_fe1, const FiniteElement &test_fe1,
-   const FiniteElement &trial_fe2, const FiniteElement &test_fe2,
-   FaceElementTransformations &Trans,
-   DenseMatrix &elmat)
-{
-   MFEM_ABORT("AssembleFaceMatrix (mixed form) is not implemented for this"
-              " Integrator class.");
-}
-
 void BilinearFormIntegrator::AssembleFaceMatrix(
   const FiniteElement &trial_face_fe, const FiniteElement &test_fe1,
   const FiniteElement &test_fe2, FaceElementTransformations &Trans,
@@ -251,38 +223,28 @@ void TransposeIntegrator::SetIntRule(const IntegrationRule *ir)
   bfi->SetIntRule(ir);
 }

-void TransposeIntegrator::AssembleElementMatrix(
+void TransposeIntegrator::AssembleElementMatrix (
   const FiniteElement &el, ElementTransformation &Trans, DenseMatrix &elmat)
 {
-   bfi->AssembleElementMatrix(el, Trans, bfi_elmat);
+   bfi -> AssembleElementMatrix (el, Trans, bfi_elmat);
   // elmat = bfi_elmat^t
   elmat.Transpose (bfi_elmat);
 }

-void TransposeIntegrator::AssembleElementMatrix2(
+void TransposeIntegrator::AssembleElementMatrix2 (
   const FiniteElement &trial_fe, const FiniteElement &test_fe,
   ElementTransformation &Trans, DenseMatrix &elmat)
 {
-   bfi->AssembleElementMatrix2(test_fe, trial_fe, Trans, bfi_elmat);
+   bfi -> AssembleElementMatrix2 (test_fe, trial_fe, Trans, bfi_elmat);
   // elmat = bfi_elmat^t
   elmat.Transpose (bfi_elmat);
 }

-void TransposeIntegrator::AssembleFaceMatrix(
+void TransposeIntegrator::AssembleFaceMatrix (
   const FiniteElement &el1, const FiniteElement &el2,
   FaceElementTransformations &Trans, DenseMatrix &elmat)
 {
-   bfi->AssembleFaceMatrix(el1, el2, Trans, bfi_elmat);
-   // elmat = bfi_elmat^t
-   elmat.Transpose (bfi_elmat);
-}
-
-void TransposeIntegrator::AssembleFaceMatrix(
-   const FiniteElement &tr_el1, const FiniteElement &te_el1,
-   const FiniteElement &tr_el2, const FiniteElement &te_el2,
-   FaceElementTransformations &Trans, DenseMatrix &elmat)
-{
-   bfi->AssembleFaceMatrix(te_el1, tr_el1, te_el2, tr_el2, Trans, bfi_elmat);
+   bfi -> AssembleFaceMatrix (el1, el2, Trans, bfi_elmat);
   // elmat = bfi_elmat^t
   elmat.Transpose (bfi_elmat);
 }
@@ -873,34 +835,6 @@ const IntegrationRule &GradientIntegrator::GetRule(const FiniteElement
 }


-DiffusionIntegrator::DiffusionIntegrator(const IntegrationRule *ir)
-   : BilinearFormIntegrator(ir),
-     Q(nullptr), VQ(nullptr), MQ(nullptr), maps(nullptr), geom(nullptr)
-{
-   static Kernels kernels;
-}
-
-DiffusionIntegrator::DiffusionIntegrator(Coefficient &q,
-                                         const IntegrationRule *ir)
-   : DiffusionIntegrator(ir)
-{
-   Q = &q;
-}
-
-DiffusionIntegrator::DiffusionIntegrator(VectorCoefficient &q,
-                                         const IntegrationRule *ir)
-   : DiffusionIntegrator(ir)
-{
-   VQ = &q;
-}
-
-DiffusionIntegrator::DiffusionIntegrator(MatrixCoefficient &q,
-                                         const IntegrationRule *ir)
-   : DiffusionIntegrator(ir)
-{
-   MQ = &q;
-}
-
 void DiffusionIntegrator::AssembleElementMatrix
 ( const FiniteElement &el, ElementTransformation &Trans,
  DenseMatrix &elmat )
@@ -1288,8 +1222,7 @@ real_t DiffusionIntegrator::ComputeFluxEnergy
   for (int i = 0; i < ir->GetNPoints(); i++)
   {
      const IntegrationPoint &ip = ir->IntPoint(i);
-      Trans.SetIntPoint(&ip);
-      fluxelem.CalcPhysShape(Trans, shape);
+      fluxelem.CalcShape(ip, shape);

      pointflux = 0.0;
      for (int k = 0; k < spaceDim; k++)
@@ -1300,6 +1233,7 @@ real_t DiffusionIntegrator::ComputeFluxEnergy
         }
      }

+      Trans.SetIntPoint(&ip);
      real_t w = Trans.Weight() * ip.weight;

      if (MQ)
@@ -1356,17 +1290,6 @@ const IntegrationRule &DiffusionIntegrator::GetRule(
   return IntRules.Get(trial_fe.GetGeomType(), order);
 }

-MassIntegrator::MassIntegrator(const IntegrationRule *ir)
-   : BilinearFormIntegrator(ir), Q(nullptr), maps(nullptr), geom(nullptr)
-{
-   static Kernels kernels;
-}
-
-MassIntegrator::MassIntegrator(Coefficient &q, const IntegrationRule *ir)
-   : MassIntegrator(ir)
-{
-   Q = &q;
-}

 void MassIntegrator::AssembleElementMatrix
 ( const FiniteElement &el, ElementTransformation &Trans,
@@ -1487,7 +1410,9 @@ void BoundaryMassIntegrator::AssembleFaceMatrix(
      // Set the integration point in the face and the neighboring element
      Trans.SetAllIntPoints(&ip);

-      el1.CalcPhysShape(*Trans.Elem1, shape);
+      // Access the neighboring element's integration point
+      const IntegrationPoint &eip = Trans.GetElement1IntPoint();
+      el1.CalcShape(eip, shape);

      w = Trans.Weight() * ip.weight;
      if (Q)
@@ -1657,9 +1582,9 @@ void VectorMassIntegrator::AssembleElementMatrix
   for (int s = 0; s < ir->GetNPoints(); s++)
   {
      const IntegrationPoint &ip = ir->IntPoint(s);
-      Trans.SetIntPoint (&ip);
-      el.CalcPhysShape(Trans, shape);
+      el.CalcShape(ip, shape);

+      Trans.SetIntPoint (&ip);
      norm = ip.weight * Trans.Weight();

      MultVVt(shape, partelmat);
@@ -1741,10 +1666,10 @@ void VectorMassIntegrator::AssembleElementMatrix2(
   for (int s = 0; s < ir->GetNPoints(); s++)
   {
      const IntegrationPoint &ip = ir->IntPoint(s);
-      Trans.SetIntPoint(&ip);
-      trial_fe.CalcPhysShape(Trans, shape);
-      test_fe.CalcPhysShape(Trans, te_shape);
+      trial_fe.CalcShape(ip, shape);
+      test_fe.CalcShape(ip, te_shape);

+      Trans.SetIntPoint(&ip);
      norm = ip.weight * Trans.Weight();

      MultVWt(te_shape, shape, partelmat);
@@ -1972,12 +1897,12 @@ void VectorFECurlIntegrator::AssembleElementMatrix2(
         if ( trial_fe.GetMapType() == mfem::FiniteElement::H_CURL )
         {
            trial_fe.CalcCurlShape(ip, curlshapeTrial_dFT);
-            test_fe.CalcPhysShape(Trans, shapeTest);
+            test_fe.CalcShape(ip, shapeTest);
         }
         else
         {
            test_fe.CalcCurlShape(ip, curlshapeTrial_dFT);
-            trial_fe.CalcPhysShape(Trans, shapeTest);
+            trial_fe.CalcShape(ip, shapeTest);
         }
      }

@@ -2000,89 +1925,6 @@ void VectorFECurlIntegrator::AssembleElementMatrix2(
   }
 }

-void VectorFEBoundaryFluxIntegrator::AssembleElementMatrix(
-   const FiniteElement &el, ElementTransformation &Tr,
-   DenseMatrix &elmat)
-{
-   int nd = el.GetDof();
-   real_t w;
-
-#ifdef MFEM_THREAD_SAFE
-   Vector shape;
-#endif
-   elmat.SetSize(nd);
-   shape.SetSize(nd);
-
-   const IntegrationRule *ir = IntRule;
-   if (ir == NULL)
-   {
-      int intorder = 2*el.GetOrder() + Tr.OrderW();  // <----------
-      ir = &IntRules.Get(el.GetGeomType(), intorder);
-   }
-
-   elmat = 0.0;
-   for (int i = 0; i < ir->GetNPoints(); i++)
-   {
-      const IntegrationPoint &ip = ir->IntPoint(i);
-      el.CalcShape(ip, shape);
-
-      Tr.SetIntPoint (&ip);
-      w = ip.weight / Tr.Weight();
-
-      if (Q)
-      {
-         w *= Q->Eval(Tr, ip);
-      }
-
-      AddMult_a_VVt(w, shape, elmat);
-   }
-}
-
-void VectorFEBoundaryFluxIntegrator::AssembleElementMatrix2(
-   const FiniteElement &trial_fe,
-   const FiniteElement &test_fe,
-   ElementTransformation &Tr,
-   DenseMatrix &elmat)
-{
-   int tr_nd = trial_fe.GetDof();
-   int te_nd = test_fe.GetDof();
-   real_t w;
-
-#ifdef MFEM_THREAD_SAFE
-   Vector shape, te_shape;
-#endif
-   elmat.SetSize(te_nd, tr_nd);
-   shape.SetSize(tr_nd);
-   te_shape.SetSize(te_nd);
-
-   const IntegrationRule *ir = IntRule;
-   if (ir == NULL)
-   {
-      int order = trial_fe.GetOrder() + test_fe.GetOrder() + Tr.OrderW();
-
-      ir = &IntRules.Get(trial_fe.GetGeomType(), order);
-   }
-
-   elmat = 0.0;
-   for (int i = 0; i < ir->GetNPoints(); i++)
-   {
-      const IntegrationPoint &ip = ir->IntPoint(i);
-      trial_fe.CalcShape(ip, shape);
-      test_fe.CalcShape(ip, te_shape);
-
-      Tr.SetIntPoint (&ip);
-      w = ip.weight / Tr.Weight();
-
-      if (Q)
-      {
-         w *= Q->Eval(Tr, ip);
-      }
-
-      te_shape *= w;
-      AddMultVWt(te_shape, shape, elmat);
-   }
-}
-
 void DerivativeIntegrator::AssembleElementMatrix2 (
   const FiniteElement &trial_fe,
   const FiniteElement &test_fe,
@@ -2139,7 +1981,7 @@ void DerivativeIntegrator::AssembleElementMatrix2 (
      det = Trans.Weight();
      Mult (dshape, invdfdx, dshapedxt);

-      test_fe.CalcPhysShape(Trans, shape);
+      test_fe.CalcShape(ip, shape);

      for (l = 0; l < trial_nd; l++)
      {
@@ -2724,7 +2566,7 @@ void VectorFEMassIntegrator::AssembleElementMatrix2(
         Trans.SetIntPoint (&ip);

         trial_fe.CalcVShape(Trans, trial_vshape);
-         test_fe.CalcPhysShape(Trans, shape);
+         test_fe.CalcShape(ip, shape);

         w = ip.weight * Trans.Weight();
         if (DQ)
@@ -2884,11 +2726,11 @@ void VectorDivergenceIntegrator::AssembleElementMatrix2(
   for (int i = 0; i < ir -> GetNPoints(); i++)
   {
      const IntegrationPoint &ip = ir->IntPoint(i);
-      Trans.SetIntPoint (&ip);

      trial_fe.CalcDShape (ip, dshape);
-      test_fe.CalcPhysShape (Trans, shape);
+      test_fe.CalcShape (ip, shape);

+      Trans.SetIntPoint (&ip);
      CalcAdjugate(Trans.Jacobian(), Jadj);

      Mult (dshape, Jadj, gshape);
@@ -3389,11 +3231,11 @@ real_t ElasticityIntegrator::ComputeFluxEnergy(const FiniteElement &fluxelem,
   for (int i = 0; i < ir->GetNPoints(); i++)
   {
      const IntegrationPoint &ip = ir->IntPoint(i);
-      Trans.SetIntPoint(&ip);
-      fluxelem.CalcPhysShape(Trans, shape);
+      fluxelem.CalcShape(ip, shape);

      flux_mat.MultTranspose(shape, pointstress);

+      Trans.SetIntPoint(&ip);
      real_t w = Trans.Weight() * ip.weight;

      M = mu->Eval(Trans, ip);
@@ -3500,7 +3342,7 @@ void DGTraceIntegrator::AssembleFaceMatrix(const FiniteElement &el1,
      const IntegrationPoint &eip1 = Trans.GetElement1IntPoint();
      const IntegrationPoint &eip2 = Trans.GetElement2IntPoint();

-      el1.CalcPhysShape(*Trans.Elem1, shape1);
+      el1.CalcShape(eip1, shape1);

      u->Eval(vu, *Trans.Elem1, eip1);

@@ -3547,7 +3389,7 @@ void DGTraceIntegrator::AssembleFaceMatrix(const FiniteElement &el1,

      if (ndof2)
      {
-         el2.CalcPhysShape(*Trans.Elem2, shape2);
+         el2.CalcShape(eip2, shape2);

         if (w != 0.0)
            for (int i = 0; i < ndof2; i++)
@@ -3575,150 +3417,6 @@ void DGTraceIntegrator::AssembleFaceMatrix(const FiniteElement &el1,
   }
 }

-void DGTraceIntegrator::AssembleFaceMatrix(const FiniteElement &trial_fe1,
-                                           const FiniteElement &test_fe1,
-                                           const FiniteElement &trial_fe2,
-                                           const FiniteElement &test_fe2,
-                                           FaceElementTransformations &Trans,
-                                           DenseMatrix &elmat)
-{
-   int tr_ndof1, te_ndof1, tr_ndof2, te_ndof2;
-
-   real_t un, a, b, w;
-
-   dim = test_fe1.GetDim();
-   tr_ndof1 = trial_fe1.GetDof();
-   te_ndof1 = test_fe1.GetDof();
-   Vector vu(dim), nor(dim);
-
-   if (Trans.Elem2No >= 0)
-   {
-      tr_ndof2 = trial_fe2.GetDof();
-      te_ndof2 = test_fe2.GetDof();
-   }
-   else
-   {
-      tr_ndof2 = 0;
-      te_ndof2 = 0;
-   }
-
-   tr_shape1.SetSize(tr_ndof1);
-   te_shape1.SetSize(te_ndof1);
-   tr_shape2.SetSize(tr_ndof2);
-   te_shape2.SetSize(te_ndof2);
-   elmat.SetSize(te_ndof1 + te_ndof2, tr_ndof1 + tr_ndof2);
-   elmat = 0.0;
-
-   const IntegrationRule *ir = IntRule;
-   if (ir == NULL)
-   {
-      int order;
-      // Assuming order(u)==order(mesh)
-      if (Trans.Elem2No >= 0)
-         order = (min(Trans.Elem1->OrderW(), Trans.Elem2->OrderW()) +
-                  max(trial_fe1.GetOrder(), trial_fe2.GetOrder()) +
-                  max(test_fe1.GetOrder(), test_fe2.GetOrder()));
-      else
-      {
-         order = Trans.Elem1->OrderW() + trial_fe1.GetOrder() + test_fe1.GetOrder();
-      }
-      if (trial_fe1.Space() == FunctionSpace::Pk)
-      {
-         order++;
-      }
-      ir = &IntRules.Get(Trans.FaceGeom, order);
-   }
-
-   for (int p = 0; p < ir->GetNPoints(); p++)
-   {
-      const IntegrationPoint &ip = ir->IntPoint(p);
-      IntegrationPoint eip1, eip2;
-      Trans.Loc1.Transform(ip, eip1);
-      Trans.Elem1->SetIntPoint(&eip1);
-      if (tr_ndof2 && te_ndof2)
-      {
-         Trans.Loc2.Transform(ip, eip2);
-         Trans.Elem2->SetIntPoint(&eip2);
-      }
-      trial_fe1.CalcPhysShape(*Trans.Elem1, tr_shape1);
-      test_fe1.CalcPhysShape(*Trans.Elem1, te_shape1);
-
-      Trans.Face->SetIntPoint(&ip);
-
-      u->Eval(vu, *Trans.Elem1, eip1);
-
-      if (dim == 1)
-      {
-         nor(0) = 2*eip1.x - 1.0;
-      }
-      else
-      {
-         CalcOrtho(Trans.Face->Jacobian(), nor);
-      }
-
-      un = vu * nor;
-      a = 0.5 * alpha * un;
-      b = beta * fabs(un);
-      // note: if |alpha/2|==|beta| then |a|==|b|, i.e. (a==b) or (a==-b)
-      //       and therefore two blocks in the element matrix contribution
-      //       (from the current quadrature point) are 0
-
-      if (rho)
-      {
-         real_t rho_p;
-         if (un >= 0.0 && tr_ndof2 && te_ndof2)
-         {
-            Trans.Elem2->SetIntPoint(&eip2);
-            rho_p = rho->Eval(*Trans.Elem2, eip2);
-         }
-         else
-         {
-            rho_p = rho->Eval(*Trans.Elem1, eip1);
-         }
-         a *= rho_p;
-         b *= rho_p;
-      }
-
-      w = ip.weight * (a+b);
-      if (w != 0.0)
-      {
-         for (int i = 0; i < te_ndof1; i++)
-            for (int j = 0; j < tr_ndof1; j++)
-            {
-               elmat(i, j) += w * te_shape1(i) * tr_shape1(j);
-            }
-      }
-
-      if (tr_ndof2 && te_ndof2)
-      {
-         trial_fe2.CalcPhysShape(*Trans.Elem2, tr_shape2);
-         test_fe2.CalcPhysShape(*Trans.Elem2, te_shape2);
-
-         if (w != 0.0)
-            for (int i = 0; i < te_ndof2; i++)
-               for (int j = 0; j < tr_ndof1; j++)
-               {
-                  elmat(te_ndof1+i, j) -= w * te_shape2(i) * tr_shape1(j);
-               }
-
-         w = ip.weight * (b-a);
-         if (w != 0.0)
-         {
-            for (int i = 0; i < te_ndof2; i++)
-               for (int j = 0; j < tr_ndof2; j++)
-               {
-                  elmat(te_ndof1+i, tr_ndof1+j) += w * te_shape2(i) * tr_shape2(j);
-               }
-
-            for (int i = 0; i < te_ndof1; i++)
-               for (int j = 0; j < tr_ndof2; j++)
-               {
-                  elmat(i, tr_ndof1+j) -= w * te_shape1(i) * tr_shape2(j);
-               }
-         }
-      }
-   }
-}

 const IntegrationRule &DGTraceIntegrator::GetRule(
   Geometry::Type geom, int order, FaceElementTransformations &T)
@@ -4241,14 +3939,19 @@ void TraceJumpIntegrator::AssembleFaceMatrix(
      // Set the integration point in the face and the neighboring elements
      Trans.SetAllIntPoints(&ip);

+      // Access the neighboring elements' integration points
+      // Note: eip2 will only contain valid data if Elem2 exists
+      const IntegrationPoint &eip1 = Trans.GetElement1IntPoint();
+      const IntegrationPoint &eip2 = Trans.GetElement2IntPoint();
+
      // Trace finite element shape function
      trial_face_fe.CalcShape(ip, face_shape);
      // Side 1 finite element shape function
-      test_fe1.CalcPhysShape(*Trans.Elem1, shape1);
+      test_fe1.CalcShape(eip1, shape1);
      if (ndof2)
      {
         // Side 2 finite element shape function
-         test_fe2.CalcPhysShape(*Trans.Elem2, shape2);
+         test_fe2.CalcShape(eip2, shape2);
      }
      w = ip.weight;
      if (trial_face_fe.GetMapType() == FiniteElement::VALUE)
@@ -4611,8 +4314,8 @@ struct ShapeCoefficient : public VectorCoefficient
      : VectorCoefficient(fe_.GetDof()), Q(q), fe(fe_) { }

   using VectorCoefficient::Eval;
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip)
   {
      V.SetSize(vdim);
      fe.CalcPhysShape(T, V);
@@ -4654,8 +4357,8 @@ ScalarVectorProductInterpolator::AssembleElementMatrix2(
      VShapeCoefficient(Coefficient &q, const FiniteElement &fe_, int sdim)
         : MatrixCoefficient(fe_.GetDof(), sdim), Q(q), fe(fe_) { }

-      void Eval(DenseMatrix &M, ElementTransformation &T,
-                const IntegrationPoint &ip) override
+      virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                        const IntegrationPoint &ip)
      {
         M.SetSize(height, width);
         fe.CalcPhysVShape(T, M);
@@ -4691,8 +4394,8 @@ VectorScalarProductInterpolator::AssembleElementMatrix2(
         : MatrixCoefficient(fe_.GetDof(), vq.GetVDim()), VQ(vq), fe(fe_),
           vc(width), shape(height) { }

-      void Eval(DenseMatrix &M, ElementTransformation &T,
-                const IntegrationPoint &ip) override
+      virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                        const IntegrationPoint &ip)
      {
         M.SetSize(height, width);
         VQ.Eval(vc, T, ip);
@@ -4731,8 +4434,8 @@ ScalarCrossProductInterpolator::AssembleElementMatrix2(
           vshape(vdim, vq.GetVDim()), vc(vq.GetVDim()) { }

      using VectorCoefficient::Eval;
-      void Eval(Vector &V, ElementTransformation &T,
-                const IntegrationPoint &ip) override
+      virtual void Eval(Vector &V, ElementTransformation &T,
+                        const IntegrationPoint &ip)
      {
         V.SetSize(vdim);
         VQ.Eval(vc, T, ip);
@@ -4775,8 +4478,8 @@ VectorCrossProductInterpolator::AssembleElementMatrix2(
         MFEM_ASSERT(width == 3, "");
      }

-      void Eval(DenseMatrix &M, ElementTransformation &T,
-                const IntegrationPoint &ip) override
+      virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                        const IntegrationPoint &ip)
      {
         M.SetSize(height, width);
         VQ.Eval(vc, T, ip);
@@ -4824,8 +4527,8 @@ struct VDotVShapeCoefficient : public VectorCoefficient
        vshape(vdim, vq.GetVDim()), vc(vq.GetVDim()) { }

   using VectorCoefficient::Eval;
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip)
   {
      V.SetSize(vdim);
      VQ.Eval(vc, T, ip);
@@ -90,12 +90,12 @@ public:
   explicit ConstantCoefficient(real_t c = 1.0) { constant=c; }

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip)
   { return (constant); }

   /// Fill the QuadratureFunction @a qf with the constant value.
-   void Project(QuadratureFunction &qf) override;
+   void Project(QuadratureFunction &qf);
 };

 /** @brief A piecewise constant coefficient with the constants keyed
@@ -130,8 +130,8 @@ public:
   int GetNConst() { return constants.Size(); }

   /// Evaluate the coefficient.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /** @brief A piecewise coefficient with the pieces keyed off the element
@@ -195,7 +195,7 @@ public:
   { InitMap(attr, coefs); }

   /// Set the time for time dependent coefficients
-   void SetTime(real_t t) override;
+   virtual void SetTime(real_t t);

   /// Replace a set of coefficients
   void UpdateCoefficients(const Array<int> & attr,
@@ -211,8 +211,8 @@ public:
   { pieces.erase(attr); }

   /// Evaluate the coefficient.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// A general function coefficient
@@ -254,8 +254,8 @@ public:
   }

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// A common base class for returning individual components of the domain's
@@ -271,8 +271,8 @@ protected:

 public:
   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Scalar coefficient which returns the x-component of the evaluation point
@@ -307,8 +307,8 @@ public:
   CylindricalRadialCoefficient() : transip(3) {}

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Scalar coefficient which returns the angular position or azimuth (often
@@ -323,8 +323,8 @@ public:
   CylindricalAzimuthalCoefficient() : transip(3) {}

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Scalar coefficient which returns the height or altitude of
@@ -342,8 +342,8 @@ public:
   SphericalRadialCoefficient() : transip(3) {}

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Scalar coefficient which returns the azimuthal angle (often denoted by phi)
@@ -357,8 +357,8 @@ public:
   SphericalAzimuthalCoefficient() : transip(3) {}

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Scalar coefficient which returns the polar angle (often denoted by theta)
@@ -372,8 +372,8 @@ public:
   SphericalPolarCoefficient() : transip(3) {}

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 class GridFunction;
@@ -399,15 +399,15 @@ public:
   const GridFunction * GetGridFunction() const { return GridF; }

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);

   /// @brief Fill the QuadratureFunction @a qf by evaluating the coefficient at
   /// the quadrature points.
   ///
   /// This function uses the efficient QuadratureFunction::ProjectGridFunction
   /// to fill the QuadratureFunction.
-   void Project(QuadratureFunction &qf) override;
+   virtual void Project(QuadratureFunction &qf);
 };


@@ -433,10 +433,10 @@ public:
      : Q1(q1), Q2(q2), Transform2(std::move(F)) { Transform1 = 0; }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);
 };

 /** @brief Delta function coefficient optionally multiplied by a weight
@@ -488,7 +488,7 @@ public:
   }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Set the center location of the delta function.
   void SetDeltaCenter(const Vector& center);
@@ -534,7 +534,7 @@ public:
   virtual real_t EvalDelta(ElementTransformation &T, const IntegrationPoint &ip);
   /** @brief A DeltaFunction cannot be evaluated. Calling this method will
       cause an MFEM error, terminating the application. */
-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip)
   { mfem_error("DeltaCoefficient::Eval"); return 0.; }
   virtual ~DeltaCoefficient() { delete weight; }
 };
@@ -555,10 +555,10 @@ public:
   { c = &c_; attr.Copy(active_attr); }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip)
   { return active_attr[T.Attribute-1] ? c->Eval(T, ip, GetTime()) : 0.0; }
 };

@@ -628,8 +628,8 @@ public:
   using VectorCoefficient::Eval;

   ///  Evaluate the vector coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override { V = vec; }
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip) { V = vec; }

   /// Return a reference to the constant vector in this class.
   const Vector& GetVec() const { return vec; }
@@ -698,7 +698,7 @@ public:
      : VectorCoefficient(vd) { InitMap(attr, coefs); }

   /// Set the time for time dependent coefficients
-   void SetTime(real_t t) override;
+   virtual void SetTime(real_t t);

   /// Replace a set of coefficients
   void UpdateCoefficients(const Array<int> & attr,
@@ -713,8 +713,8 @@ public:
   { pieces.erase(attr); }

   /// Evaluate the coefficient.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   using VectorCoefficient::Eval;
 };

@@ -728,8 +728,8 @@ public:

   using VectorCoefficient::Eval;
   /// Evaluate the vector coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   virtual ~PositionVectorCoefficient() { }
 };
@@ -765,8 +765,8 @@ public:

   using VectorCoefficient::Eval;
   /// Evaluate the vector coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   virtual ~VectorFunctionCoefficient() { }
 };
@@ -787,7 +787,7 @@ public:
   explicit VectorArrayCoefficient(int dim);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Returns i'th coefficient.
   Coefficient* GetCoeff(int i) { return Coeff[i]; }
@@ -798,12 +798,6 @@ public:
   /// Sets coefficient in the vector.
   void Set(int i, Coefficient *c, bool own=true);

-   /// Set ownership of the i'th coefficient
-   void SetOwnership(int i, bool own) { ownCoeff[i] = own; }
-
-   /// Get ownership of the i'th coefficient
-   bool GetOwnership(int i) const { return ownCoeff[i]; }
-
   /// Evaluates i'th component of the vector of coefficients and returns the
   /// value.
   real_t Eval(int i, ElementTransformation &T, const IntegrationPoint &ip)
@@ -812,8 +806,8 @@ public:
   using VectorCoefficient::Eval;
   /** @brief Evaluate the coefficient. Each element of vector V comes from the
       associated array of scalar coefficients. */
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   /// Destroys vector coefficient.
   virtual ~VectorArrayCoefficient();
@@ -842,21 +836,21 @@ public:
   const GridFunction * GetGridFunction() const { return GridFunc; }

   /// Evaluate the vector coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   /** @brief Evaluate the vector coefficients at all of the locations in the
       integration rule and write the vectors into the columns of matrix @a
       M. */
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationRule &ir) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationRule &ir);

   /// @brief Fill the QuadratureFunction @a qf by evaluating the coefficient at
   /// the quadrature points.
   ///
   /// This function uses the efficient QuadratureFunction::ProjectGridFunction
   /// to fill the QuadratureFunction.
-   void Project(QuadratureFunction &qf) override;
+   virtual void Project(QuadratureFunction &qf);

   virtual ~VectorGridFunctionCoefficient() { }
 };
@@ -880,14 +874,14 @@ public:
   const GridFunction * GetGridFunction() const { return GridFunc; }

   /// Evaluate the gradient vector coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   /** @brief Evaluate the gradient vector coefficient at all of the locations
       in the integration rule and write the vectors into columns of matrix @a
       M. */
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationRule &ir) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationRule &ir);

   virtual ~GradientGridFunctionCoefficient() { }
 };
@@ -911,8 +905,8 @@ public:

   using VectorCoefficient::Eval;
   /// Evaluate the vector curl coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   virtual ~CurlGridFunctionCoefficient() { }
 };
@@ -935,8 +929,8 @@ public:
   const GridFunction * GetGridFunction() const { return GridFunc; }

   /// Evaluate the scalar divergence coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);

   virtual ~DivergenceGridFunctionCoefficient() { }
 };
@@ -979,7 +973,7 @@ public:
      : VectorCoefficient(dir_.Size()), dir(dir_), d(x,y,z,s) { }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Replace the associated DeltaCoefficient with a new DeltaCoefficient.
   /** The new DeltaCoefficient cannot have a specified weight Coefficient, i.e.
@@ -1004,8 +998,8 @@ public:
   using VectorCoefficient::Eval;
   /** @brief A VectorDeltaFunction cannot be evaluated. Calling this method
       will cause an MFEM error, terminating the application. */
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip)
   { mfem_error("VectorDeltaCoefficient::Eval"); }
   virtual ~VectorDeltaCoefficient() { }
 };
@@ -1027,17 +1021,17 @@ public:
   { c = &vc; attr.Copy(active_attr); }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Evaluate the vector coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   /** @brief Evaluate the vector coefficient at all of the locations in the
       integration rule and write the vectors into the columns of matrix @a
       M. */
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationRule &ir) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationRule &ir);
 };

 typedef VectorCoefficient DiagonalMatrixCoefficient;
@@ -1119,8 +1113,8 @@ public:
      : MatrixCoefficient(m.Height(), m.Width()), mat(m) { }
   using MatrixCoefficient::Eval;
   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override { M = mat; }
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip) { M = mat; }
   /// Return a reference to the constant matrix.
   const DenseMatrix& GetMatrix() { return mat; }
 };
@@ -1213,7 +1207,7 @@ public:
      : MatrixCoefficient(h, w, symm) { InitMap(attr, coefs); }

   /// Set the time for time dependent coefficients
-   void SetTime(real_t t) override;
+   virtual void SetTime(real_t t);

   /// Replace a set of coefficients
   void UpdateCoefficients(const Array<int> & attr,
@@ -1228,8 +1222,8 @@ public:
   { pieces.erase(attr); }

   /// Evaluate the coefficient.
-   void Eval(DenseMatrix &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &K, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /** @brief A matrix coefficient with an optional scalar coefficient multiplier
@@ -1286,16 +1280,16 @@ public:
   { }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &K, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   /// (DEPRECATED) Evaluate the symmetric matrix coefficient at @a ip.
   /** @deprecated Use Eval() instead. */
-   void EvalSymmetric(Vector &K, ElementTransformation &T,
-                      const IntegrationPoint &ip) override;
+   virtual void EvalSymmetric(Vector &K, ElementTransformation &T,
+                              const IntegrationPoint &ip);

   virtual ~MatrixFunctionCoefficient() { }
 };
@@ -1316,7 +1310,7 @@ public:
   explicit MatrixArrayCoefficient (int dim);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Get the coefficient located at (i,j) in the matrix.
   Coefficient* GetCoeff (int i, int j) { return Coeff[i*width+j]; }
@@ -1326,12 +1320,6 @@ public:
       can be overridden with the @a own parameter. */
   void Set(int i, int j, Coefficient * c, bool own=true);

-   /// Set ownership of the coefficient at (i,j) in the matrix
-   void SetOwnership(int i, int j, bool own) { ownCoeff[i*width+j] = own; }
-
-   /// Get ownership of the coefficient at (i,j) in the matrix
-   bool GetOwnership(int i, int j) const { return ownCoeff[i*width+j]; }
-
   using MatrixCoefficient::Eval;

   /// Evaluate coefficient located at (i,j) in the matrix using integration
@@ -1340,8 +1328,8 @@ public:
   { return Coeff[i*width+j] ? Coeff[i*width+j] -> Eval(T, ip, GetTime()) : 0.0; }

   /// Evaluate the matrix coefficient @a ip.
-   void Eval(DenseMatrix &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &K, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   virtual ~MatrixArrayCoefficient();
 };
@@ -1372,12 +1360,6 @@ public:
       can be overridden with the @a own parameter. */
   void Set(int i, VectorCoefficient * c, bool own=true);

-   /// Set ownership of the i'th coefficient
-   void SetOwnership(int i, bool own) { ownCoeff[i] = own; }
-
-   /// Get ownership of the i'th coefficient
-   bool GetOwnership(int i) const { return ownCoeff[i]; }
-
   using MatrixCoefficient::Eval;

   /// Evaluate coefficient located at the i-th row of the matrix using integration
@@ -1410,11 +1392,11 @@ public:
   { c = &mc; attr.Copy(active_attr); }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &K, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /// Coefficients based on sums, products, or other functions of coefficients.
@@ -1443,7 +1425,7 @@ public:
      : aConst(0.0), a(&A), b(&B), alpha(alpha_), beta(beta_) { }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the first term in the linear combination as a constant
   void SetAConst(real_t A) { a = NULL; aConst = A; }
@@ -1471,8 +1453,8 @@ public:
   real_t GetBeta() const { return beta; }

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip)
   {
      return alpha * ((a == NULL ) ? aConst : a->Eval(T, ip) )
             + beta * b->Eval(T, ip);
@@ -1520,8 +1502,8 @@ public:
       @note When this method is called, the caller must make sure that the
       IntegrationPoint associated with @a T is the same as @a ip. This can be
       achieved by calling T.SetIntPoint(&ip). */
-   void Eval(DenseMatrix &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &K, ElementTransformation &T,
+                     const IntegrationPoint &ip);


   /// @deprecated Return a reference to the internal matrix used when evaluating this coefficient as a DenseMatrix.
@@ -1543,8 +1525,8 @@ public:
      : SymmetricMatrixCoefficient(m.Height()), mat(m) { }
   using SymmetricMatrixCoefficient::Eval;
   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseSymmetricMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override { M = mat; }
+   virtual void Eval(DenseSymmetricMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip) { M = mat; }

   /// Return a reference to the constant matrix.
   const DenseSymmetricMatrix& GetMatrix() { return mat; }
@@ -1594,12 +1576,12 @@ public:
   { }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   using SymmetricMatrixCoefficient::Eval;
   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseSymmetricMatrix &K, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseSymmetricMatrix &K, ElementTransformation &T,
+                     const IntegrationPoint &ip);

   virtual ~SymmetricMatrixFunctionCoefficient() { }
 };
@@ -1624,7 +1606,7 @@ public:
      : aConst(0.0), a(&A), b(&B) { }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the first term in the product as a constant
   void SetAConst(real_t A) { a = NULL; aConst = A; }
@@ -1642,8 +1624,8 @@ public:
   Coefficient * GetBCoef() const { return b; }

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip)
   { return ((a == NULL ) ? aConst : a->Eval(T, ip) ) * b->Eval(T, ip); }
 };

@@ -1672,7 +1654,7 @@ public:
      : aConst(0.0), bConst(B), a(&A), b(NULL) { }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the numerator in the ratio as a constant
   void SetAConst(real_t A) { a = NULL; aConst = A; }
@@ -1695,8 +1677,8 @@ public:
   Coefficient * GetBCoef() const { return b; }

   /// Evaluate the coefficient
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip)
   {
      real_t den = (b == NULL ) ? bConst : b->Eval(T, ip);
      MFEM_ASSERT(den != 0.0, "Division by zero in RatioCoefficient");
@@ -1718,7 +1700,7 @@ public:
      : a(&A), p(p_) { }

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the base coefficient
   void SetACoef(Coefficient &A) { a = &A; }
@@ -1731,8 +1713,8 @@ public:
   real_t GetExponent() const { return p; }

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip)
   { return pow(a->Eval(T, ip), p); }
 };

@@ -1751,7 +1733,7 @@ public:
   InnerProductCoefficient(VectorCoefficient &A, VectorCoefficient &B);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the first vector in the inner product
   void SetACoef(VectorCoefficient &A) { a = &A; }
@@ -1764,8 +1746,8 @@ public:
   VectorCoefficient * GetBCoef() const { return b; }

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Scalar coefficient defined as a cross product of two vectors in the xy-plane.
@@ -1783,7 +1765,7 @@ public:
   VectorRotProductCoefficient(VectorCoefficient &A, VectorCoefficient &B);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the first vector in the product
   void SetACoef(VectorCoefficient &A) { a = &A; }
@@ -1796,8 +1778,8 @@ public:
   VectorCoefficient * GetBCoef() const { return b; }

   /// Evaluate the coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Scalar coefficient defined as the determinant of a matrix coefficient
@@ -1813,7 +1795,7 @@ public:
   DeterminantCoefficient(MatrixCoefficient &A);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the matrix coefficient
   void SetACoef(MatrixCoefficient &A) { a = &A; }
@@ -1821,8 +1803,8 @@ public:
   MatrixCoefficient * GetACoef() const { return a; }

   /// Evaluate the determinant coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Scalar coefficient defined as the trace of a matrix coefficient
@@ -1838,7 +1820,7 @@ public:
   TraceCoefficient(MatrixCoefficient &A);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the matrix coefficient
   void SetACoef(MatrixCoefficient &A) { a = &A; }
@@ -1846,8 +1828,8 @@ public:
   MatrixCoefficient * GetACoef() const { return a; }

   /// Evaluate the trace coefficient at @a ip.
-   real_t Eval(ElementTransformation &T,
-               const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T,
+                       const IntegrationPoint &ip);
 };

 /// Vector coefficient defined as the linear combination of two vectors
@@ -1884,7 +1866,7 @@ public:
                        Coefficient &alpha_, Coefficient &beta_);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the first vector coefficient
   void SetACoef(VectorCoefficient &A_) { ACoef = &A_; }
@@ -1927,8 +1909,8 @@ public:
   real_t GetBeta() const { return beta; }

   /// Evaluate the coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   using VectorCoefficient::Eval;
 };

@@ -1948,7 +1930,7 @@ public:
   ScalarVectorProductCoefficient(Coefficient &A, VectorCoefficient &B);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the scalar factor as a constant
   void SetAConst(real_t A) { a = NULL; aConst = A; }
@@ -1966,8 +1948,8 @@ public:
   VectorCoefficient * GetBCoef() const { return b; }

   /// Evaluate the coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   using VectorCoefficient::Eval;
 };

@@ -1989,7 +1971,7 @@ public:
   NormalizedVectorCoefficient(VectorCoefficient &A, real_t tol = 1e-6);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the vector coefficient
   void SetACoef(VectorCoefficient &A) { a = &A; }
@@ -1997,8 +1979,8 @@ public:
   VectorCoefficient * GetACoef() const { return a; }

   /// Evaluate the coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   using VectorCoefficient::Eval;
 };

@@ -2017,7 +1999,7 @@ public:
   VectorCrossProductCoefficient(VectorCoefficient &A, VectorCoefficient &B);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the first term in the product
   void SetACoef(VectorCoefficient &A) { a = &A; }
@@ -2030,8 +2012,8 @@ public:
   VectorCoefficient * GetBCoef() const { return b; }

   /// Evaluate the coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   using VectorCoefficient::Eval;
 };

@@ -2051,7 +2033,7 @@ public:
   MatrixVectorProductCoefficient(MatrixCoefficient &A, VectorCoefficient &B);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the matrix coefficient
   void SetACoef(MatrixCoefficient &A) { a = &A; }
@@ -2064,8 +2046,8 @@ public:
   VectorCoefficient * GetBCoef() const { return b; }

   /// Evaluate the vector coefficient at @a ip.
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);
   using VectorCoefficient::Eval;
 };

@@ -2084,8 +2066,8 @@ public:
      : MatrixCoefficient(d, d), dim(d) { }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /// Matrix coefficient defined as the linear combination of two matrices
@@ -2106,7 +2088,7 @@ public:
                        real_t alpha_ = 1.0, real_t beta_ = 1.0);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the first matrix coefficient
   void SetACoef(MatrixCoefficient &A) { a = &A; }
@@ -2129,8 +2111,8 @@ public:
   real_t GetBeta() const { return beta; }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /// Matrix coefficient defined as the product of two matrices
@@ -2158,8 +2140,8 @@ public:
   MatrixCoefficient * GetBCoef() const { return b; }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /** @brief Matrix coefficient defined as a product of a scalar coefficient and a
@@ -2179,7 +2161,7 @@ public:
   ScalarMatrixProductCoefficient(Coefficient &A, MatrixCoefficient &B);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the scalar factor as a constant
   void SetAConst(real_t A) { a = NULL; aConst = A; }
@@ -2197,8 +2179,8 @@ public:
   MatrixCoefficient * GetBCoef() const { return b; }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /// Matrix coefficient defined as the transpose of a matrix coefficient
@@ -2212,7 +2194,7 @@ public:
   TransposeMatrixCoefficient(MatrixCoefficient &A);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the matrix coefficient
   void SetACoef(MatrixCoefficient &A) { a = &A; }
@@ -2220,8 +2202,8 @@ public:
   MatrixCoefficient * GetACoef() const { return a; }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /// Matrix coefficient defined as the inverse of a matrix coefficient.
@@ -2235,7 +2217,7 @@ public:
   InverseMatrixCoefficient(MatrixCoefficient &A);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the matrix coefficient
   void SetACoef(MatrixCoefficient &A) { a = &A; }
@@ -2243,8 +2225,8 @@ public:
   MatrixCoefficient * GetACoef() const { return a; }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /// Matrix coefficient defined as the exponential of a matrix coefficient.
@@ -2258,7 +2240,7 @@ public:
   ExponentialMatrixCoefficient(MatrixCoefficient &A);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the matrix coefficient
   void SetACoef(MatrixCoefficient &A) { a = &A; }
@@ -2266,8 +2248,8 @@ public:
   MatrixCoefficient * GetACoef() const { return a; }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /// Matrix coefficient defined as the outer product of two vector coefficients.
@@ -2285,7 +2267,7 @@ public:
   OuterProductCoefficient(VectorCoefficient &A, VectorCoefficient &B);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the first vector in the outer product
   void SetACoef(VectorCoefficient &A) { a = &A; }
@@ -2298,8 +2280,8 @@ public:
   VectorCoefficient * GetBCoef() const { return b; }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };

 /** @brief Matrix coefficient defined as -a k x k x, for a vector k and scalar a
@@ -2323,7 +2305,7 @@ public:
   CrossCrossCoefficient(Coefficient &A, VectorCoefficient &K);

   /// Set the time for internally stored coefficients
-   void SetTime(real_t t) override;
+   void SetTime(real_t t);

   /// Reset the scalar factor as a constant
   void SetAConst(real_t A) { a = NULL; aConst = A; }
@@ -2341,8 +2323,8 @@ public:
   VectorCoefficient * GetKCoef() const { return k; }

   /// Evaluate the matrix coefficient at @a ip.
-   void Eval(DenseMatrix &M, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(DenseMatrix &M, ElementTransformation &T,
+                     const IntegrationPoint &ip);
 };
 ///@}

@@ -2367,10 +2349,10 @@ public:
   const QuadratureFunction& GetQuadFunction() const { return QuadF; }

   using VectorCoefficient::Eval;
-   void Eval(Vector &V, ElementTransformation &T,
-             const IntegrationPoint &ip) override;
+   virtual void Eval(Vector &V, ElementTransformation &T,
+                     const IntegrationPoint &ip);

-   void Project(QuadratureFunction &qf) override;
+   virtual void Project(QuadratureFunction &qf);

   virtual ~VectorQuadratureFunctionCoefficient() { }
 };
@@ -2389,9 +2371,9 @@ public:

   const QuadratureFunction& GetQuadFunction() const { return QuadF; }

-   real_t Eval(ElementTransformation &T, const IntegrationPoint &ip) override;
+   virtual real_t Eval(ElementTransformation &T, const IntegrationPoint &ip);

-   void Project(QuadratureFunction &qf) override;
+   virtual void Project(QuadratureFunction &qf);

   virtual ~QuadratureFunctionCoefficient() { }
 };
@@ -1245,7 +1245,7 @@ ParSesquilinearForm::FormLinearSystem(const Array<int> &ess_tdof_list,
         hypre_ParCSRMatrix *Aih = *Ah;
         Ah->HypreReadWrite();
         const int *d_ess_tdof_list =
-            ess_tdof_list.GetMemory().Read(GetHypreForallMemoryClass(), n);
+            ess_tdof_list.GetMemory().Read(GetHypreMemoryClass(), n);
         HYPRE_Int *d_diag_i = Aih->diag->i;
         real_t *d_diag_data = Aih->diag->data;
         mfem::hypre_forall(n, [=] MFEM_HOST_DEVICE (int k)
@@ -943,7 +943,6 @@ void ParaViewDataCollection::Save()
            pvtu_out << "<PDataArray type=\"" << GetDataTypeString()
                     << "\" Name=\"" << field_it.first
                     << "\" NumberOfComponents=\"" << vec_dim << "\" "
-                     << VTKComponentLabels(vec_dim) << " "
                     << "format=\"" << GetDataFormatString() << "\" />\n";
         }
         pvtu_out << "</PPointData>\n";
@@ -978,7 +977,6 @@ void ParaViewDataCollection::Save()
         pvtu_out << "<PDataArray type=\"" << GetDataTypeString()
                  << "\" Name=\"" << q_field_name
                  << "\" NumberOfComponents=\"" << vec_dim << "\" "
-                  << VTKComponentLabels(vec_dim) << " "
                  << "format=\"" << GetDataFormatString() << "\" />\n";
         pvtu_out << "</PPointData>\n";
         WritePVTUFooter(pvtu_out, q_field_name);
@@ -1071,9 +1069,8 @@ void ParaViewDataCollection::SaveGFieldVTU(std::ostream &os, int ref_,
   int vec_dim = it->second->VectorDim();
   os << "<DataArray type=\"" << GetDataTypeString()
      << "\" Name=\"" << it->first
-      << "\" NumberOfComponents=\"" << vec_dim << "\" "
-      << VTKComponentLabels(vec_dim) << " "
-      << "format=\"" << GetDataFormatString() << "\" >" << '\n';
+      << "\" NumberOfComponents=\"" << vec_dim << "\""
+      << " format=\"" << GetDataFormatString() << "\" >" << '\n';
   if (vec_dim == 1)
   {
      // scalar data
@@ -454,28 +454,28 @@ public:
 #endif

   /// Set/change the mesh associated with the collection
-   void SetMesh(Mesh *new_mesh) override;
+   virtual void SetMesh(Mesh *new_mesh) override;

 #ifdef MFEM_USE_MPI
   /// Set/change the mesh associated with the collection.
-   void SetMesh(MPI_Comm comm, Mesh *new_mesh) override;
+   virtual void SetMesh(MPI_Comm comm, Mesh *new_mesh) override;
 #endif

   /// Add a grid function to the collection and update the root file
-   void RegisterField(const std::string& field_name,
-                      GridFunction *gf) override;
+   virtual void RegisterField(const std::string& field_name,
+                              GridFunction *gf) override;

   /// Add a quadrature function to the collection and update the root file.
   /** Visualization of quadrature function is not supported in VisIt(3.12).
       A patch has been sent to VisIt developers in June 2020. */
-   void RegisterQField(const std::string& q_field_name,
-                       QuadratureFunction *qf) override;
+   virtual void RegisterQField(const std::string& q_field_name,
+                               QuadratureFunction *qf) override;

   /// Set the number of digits used for both the cycle and the MPI rank
   /// @note VisIt seems to require 6 pad digits for the MPI rank. Therefore,
   /// this function uses this default value. This behavior can be overridden
   /// by calling SetPadDigitsCycle() and SetPadDigitsRank() instead.
-   void SetPadDigits(int digits) override
+   virtual void SetPadDigits(int digits) override
   { pad_digits_cycle=digits; pad_digits_rank=6; }

   /// Set VisIt parameter: default levels of detail for the MultiresControl
@@ -489,13 +489,13 @@ public:
   void DeleteAll();

   /// Save the collection and a VisIt root file
-   void Save() override;
+   virtual void Save() override;

   /// Save a VisIt root file for the collection
   void SaveRootFile();

   /// Load the collection based on its VisIt data (described in its root file)
-   void Load(int cycle_ = 0) override;
+   virtual void Load(int cycle_ = 0) override;

   /// We will delete the mesh and fields if we own them
   virtual ~VisItDataCollection() {}
@@ -546,7 +546,7 @@ public:

   /// Save the collection - the directory name is constructed based on the
   /// cycle value
-   void Save() override;
+   virtual void Save() override;

   /// Set the data format for the ParaView output files. Possible options are
   /// VTKFormat::ASCII, VTKFormat::BINARY, and VTKFormat::BINARY32.
@@ -590,7 +590,7 @@ public:
   void UseRestartMode(bool restart_mode_);

   /// Load the collection - not implemented in the ParaView writer
-   void Load(int cycle_ = 0) override;
+   virtual void Load(int cycle_ = 0) override;
 };

 }
@@ -52,15 +52,6 @@ protected:
   const DenseMatrix &EvalTransAdjugateJ();
   const DenseMatrix &EvalInverseJ();

-   /// @name Tolerance used for point comparisons
-   ///@{
-#ifdef MFEM_USE_DOUBLE
-   static constexpr real_t tol_0 = 1e-15;
-#elif defined(MFEM_USE_SINGLE)
-   static constexpr real_t tol_0 = 1e-7;
-#endif
-   ///@}
-
 public:

   /** This enumeration declares the values stored in
@@ -185,7 +176,7 @@ public:
       returned. This method is not 100 percent reliable for non-linear
       transformations. */
   virtual int TransformBack(const Vector &pt, IntegrationPoint &ip,
-                             const real_t phys_tol = tol_0) = 0;
+                             const real_t phys_tol = 1e-15) = 0;

   virtual ~ElementTransformation() { }
 };
@@ -290,15 +281,9 @@ public:
        rel_qpts_order(-1),
        solver_type(NewtonElementProject),
        max_iter(16),
-#ifdef MFEM_USE_DOUBLE
        ref_tol(1e-15),
        phys_rtol(1e-15),
        ip_tol(1e-8),
-#elif defined(MFEM_USE_SINGLE)
-        ref_tol(1e-7),
-        phys_rtol(1e-7),
-        ip_tol(1e-4),
-#endif
        print_level(-1)
   { }

@@ -385,10 +370,10 @@ private:

   /** @brief Evaluate the Jacobian of the transformation at the IntPoint and
       store it in dFdx. */
-   const DenseMatrix &EvalJacobian() override;
+   virtual const DenseMatrix &EvalJacobian();
   // Evaluate the Hessian of the transformation at the IntPoint and store it
   // in d2Fdx2.
-   const DenseMatrix &EvalHessian() override;
+   virtual const DenseMatrix &EvalHessian();

 public:
   IsoparametricTransformation() : FElem(NULL) {}
@@ -430,32 +415,32 @@ public:

   /** @brief Transform integration point from reference coordinates to
       physical coordinates and store them in the vector. */
-   void Transform(const IntegrationPoint &, Vector &) override;
+   virtual void Transform(const IntegrationPoint &, Vector &);

   /** @brief Transform all the integration points from the integration rule
       from reference coordinates to physical
      coordinates and store them as column vectors in the matrix. */
-   void Transform(const IntegrationRule &, DenseMatrix &) override;
+   virtual void Transform(const IntegrationRule &, DenseMatrix &);

   /** @brief Transform all the integration points from the column vectors
       of @a matrix from reference coordinates to physical
       coordinates and store them as column vectors in @a result. */
-   void Transform(const DenseMatrix &matrix, DenseMatrix &result) override;
+   virtual void Transform(const DenseMatrix &matrix, DenseMatrix &result);

   /// Return the order of the current element we are using for the transformation.
-   int Order() const override { return FElem->GetOrder(); }
+   virtual int Order() const { return FElem->GetOrder(); }

   /// Return the order of the elements of the Jacobian of the transformation.
-   int OrderJ() const override;
+   virtual int OrderJ() const;

   /** @brief Return the order of the determinant of the Jacobian (weight)
       of the transformation. */
-   int OrderW() const override;
+   virtual int OrderW() const;

   /// Return the order of $ adj(J)^T \nabla fi $
-   int OrderGrad(const FiniteElement *fe) const override;
+   virtual int OrderGrad(const FiniteElement *fe) const;

-   int GetSpaceDim() const override { return PointMat.Height(); }
+   virtual int GetSpaceDim() const { return PointMat.Height(); }

   /** @brief Transform a point @a pt from physical space to a point @a ip in
       reference space and optionally can set a solver tolerance using @a phys_tol. */
@@ -463,8 +448,8 @@ public:
       point in physical space. If the inversion fails a non-zero value is
       returned. This method is not 100 percent reliable for non-linear
       transformations. */
-   int TransformBack (const Vector & v, IntegrationPoint & ip,
-                      const real_t phys_rel_tol = tol_0) override
+   virtual int TransformBack(const Vector & v, IntegrationPoint & ip,
+                             const real_t phys_rel_tol = 1e-15)
   {
      InverseElementTransformation inv_tr(this);
      inv_tr.SetPhysicalRelTol(phys_rel_tol);
@@ -604,9 +589,9 @@ public:
       has been configured. */
   const IntegrationPoint &GetElement2IntPoint() { return eip2; }

-   void Transform(const IntegrationPoint &, Vector &) override;
-   void Transform(const IntegrationRule &, DenseMatrix &) override;
-   void Transform(const DenseMatrix &matrix, DenseMatrix &result) override;
+   virtual void Transform(const IntegrationPoint &, Vector &);
+   virtual void Transform(const IntegrationRule &, DenseMatrix &);
+   virtual void Transform(const DenseMatrix &matrix, DenseMatrix &result);

   ElementTransformation & GetElement1Transformation();
   ElementTransformation & GetElement2Transformation();
@@ -172,10 +172,10 @@ public:
   void SetFluxAveraging(int fa) { flux_averaging = fa; }

   /// Return the total error from the last error estimate.
-   real_t GetTotalError() const override { return total_error; }
+   virtual real_t GetTotalError() const override { return total_error; }

   /// Get a Vector with all element errors.
-   const Vector &GetLocalErrors() override
+   virtual const Vector &GetLocalErrors() override
   {
      if (MeshIsModified()) { ComputeEstimates(); }
      return error_estimates;
@@ -184,14 +184,14 @@ public:
   /** @brief Get an Array<int> with anisotropic flags for all mesh elements.
       Return an empty array when anisotropic estimates are not available or
       enabled. */
-   const Array<int> &GetAnisotropicFlags() override
+   virtual const Array<int> &GetAnisotropicFlags() override
   {
      if (MeshIsModified()) { ComputeEstimates(); }
      return aniso_flags;
   }

   /// Reset the error estimator.
-   void Reset() override { current_sequence = -1; }
+   virtual void Reset() override { current_sequence = -1; }

   /** @brief Destroy a ZienkiewiczZhuEstimator object. Destroys, if owned, the
       FiniteElementSpace, flux_space. */
@@ -298,17 +298,17 @@ public:
   }

   /// Return the total error from the last error estimate.
-   real_t GetTotalError() const override { return total_error; }
+   virtual real_t GetTotalError() const override { return total_error; }

   /// Get a Vector with all element errors.
-   const Vector &GetLocalErrors() override
+   virtual const Vector &GetLocalErrors() override
   {
      if (MeshIsModified()) { ComputeEstimates(); }
      return error_estimates;
   }

   /// Reset the error estimator.
-   void Reset() override { current_sequence = -1; }
+   virtual void Reset() override { current_sequence = -1; }

   virtual ~LSZienkiewiczZhuEstimator() { }
 };
@@ -411,17 +411,17 @@ public:
   void SetLocalErrorNormP(int p) { local_norm_p = p; }

   /// Return the total error from the last error estimate.
-   real_t GetTotalError() const override { return total_error; }
+   virtual real_t GetTotalError() const override { return total_error; }

   /// Get a Vector with all element errors.
-   const Vector &GetLocalErrors() override
+   virtual const Vector &GetLocalErrors() override
   {
      if (MeshIsModified()) { ComputeEstimates(); }
      return error_estimates;
   }

   /// Reset the error estimator.
-   void Reset() override { current_sequence = -1; }
+   virtual void Reset() override { current_sequence = -1; }

   /** @brief Destroy a L2ZienkiewiczZhuEstimator object. Destroys, if owned,
       the FiniteElementSpace, flux_space. */
@@ -505,10 +505,10 @@ public:
   void SetCoef(VectorCoefficient &A) { vcoef = &A; }

   /// Reset the error estimator.
-   void Reset() override { current_sequence = -1; }
+   virtual void Reset() override { current_sequence = -1; }

   /// Get a Vector with all element errors.
-   const Vector &GetLocalErrors() override
+   virtual const Vector &GetLocalErrors() override
   {
      if (MeshIsModified()) { ComputeEstimates(); }
      return error_estimates;
@@ -661,7 +661,7 @@ public:
   /// Reset the error estimator.
   void Reset() override { current_sequence = -1; };

-   real_t GetTotalError() const override { return total_error; }
+   virtual real_t GetTotalError() const override { return total_error; }

   /** @brief Change the method to compute hₑ on a per-element basis.
       @param compute_element_coefficient_
@@ -69,16 +69,16 @@ inline int ToLexOrdering2D(const int face_id, const int size1d, const int i)
 }

 /// @brief Given a face DOF index on a shared face, ordered lexicographically
-/// relative to element the element (where the local face is face_id), and
-/// return the corresponding face DOF index ordered lexicographically relative
-/// to the face itself.
+/// relative to element 1, return the corresponding face DOF index ordered
+/// lexicographically relative to element 2.
 MFEM_HOST_DEVICE
-inline int PermuteFace2D(const int face_id, const int orientation,
-                         const int size1d, const int index)
+inline int PermuteFace2D(const int face_id1, const int face_id2,
+                         const int orientation, const int size1d,
+                         const int index)
 {
   int new_index;
   // Convert from element 1 lex ordering to native ordering
-   if (face_id == 2 || face_id == 3)
+   if (face_id1 == 2 || face_id1 == 3)
   {
      new_index = size1d-1-index;
   }
@@ -91,18 +91,7 @@ inline int PermuteFace2D(const int face_id, const int orientation,
   {
      new_index = size1d-1-new_index;
   }
-   return new_index;
-}
-
-/// @brief Given a face DOF index on a shared face, ordered lexicographically
-/// relative to element 1, return the corresponding face DOF index ordered
-/// lexicographically relative to element 2.
-MFEM_HOST_DEVICE
-inline int PermuteFace2D(const int face_id1, const int face_id2,
-                         const int orientation, const int size1d,
-                         const int index)
-{
-   const int new_index = PermuteFace2D(face_id1, orientation, size1d, index);
+   // Covert to element 2 lex ordering
   return ToLexOrdering2D(face_id2, size1d, new_index);
 }

@@ -127,22 +116,26 @@ inline int ToLexOrdering3D(const int face_id, const int size1d, const int i,
   }
 }

-/// @brief Given the index of a face DOF in lexicographic ordering relative the
-/// element (where the local face id is @a face_id), permute the index so that
-/// it is lexicographically ordered relative to the face itself.
+/// @brief Given the index of a face DOF in lexicographic ordering relative
+/// element 1, permute the index so that it is lexicographically ordered
+/// relative to element 2.
+///
+/// The given face corresponds to local face index @a face_id1 relative to
+/// element 1, and @a face_id2 (with @a orientation) relative to element 2.
 MFEM_HOST_DEVICE
-inline int PermuteFace3D(const int face_id, const int orientation,
+inline int PermuteFace3D(const int face_id1, const int face_id2,
+                         const int orientation,
                         const int size1d, const int index)
 {
   int i=0, j=0, new_i=0, new_j=0;
   i = index%size1d;
   j = index/size1d;
   // Convert from lex ordering
-   if (face_id==3 || face_id==4)
+   if (face_id1==3 || face_id1==4)
   {
      i = size1d-1-i;
   }
-   else if (face_id==0)
+   else if (face_id1==0)
   {
      j = size1d-1-j;
   }
@@ -182,23 +175,6 @@ inline int PermuteFace3D(const int face_id, const int orientation,
         new_j = (size1d-1-j);
         break;
   }
-   return new_i + new_j*size1d;
-}
-
-/// @brief Given the index of a face DOF in lexicographic ordering relative
-/// element 1, permute the index so that it is lexicographically ordered
-/// relative to element 2.
-///
-/// The given face corresponds to local face index @a face_id1 relative to
-/// element 1, and @a face_id2 (with @a orientation) relative to element 2.
-MFEM_HOST_DEVICE
-inline int PermuteFace3D(const int face_id1, const int face_id2,
-                         const int orientation,
-                         const int size1d, const int index)
-{
-   const int new_index = PermuteFace3D(face_id1, orientation, size1d, index);
-   const int new_i = new_index%size1d;
-   const int new_j = new_index/size1d;
   return ToLexOrdering3D(face_id2, size1d, new_i, new_j);
 }

@@ -394,32 +394,7 @@ public:
   /// Get a const reference to the nodes of the element
   const IntegrationRule & GetNodes() const { return Nodes; }

-   /** @brief Evaluate the Hessians of all shape functions of a scalar finite
-       element in reference space at the given point @a ip. */
-   /** Each row of the result DenseMatrix @a Hessian contains upper triangular
-       part of the Hessian of one shape function.
-       The order in 2D is {u_xx, u_xy, u_yy}.
-       The size (#dof x (#dim (#dim+1)/2) of @a Hessian must be set in advance.*/
-   virtual void CalcHessian(const IntegrationPoint &ip,
-                            DenseMatrix &Hessian) const;
-
-   /** @brief Evaluate the Hessian of all shape functions of a scalar finite
-       element in physical space at the given point @a ip. */
-   /** The size (#dof, #dim*(#dim+1)/2) of @a Hessian must be set in advance. */
-   void CalcPhysHessian(ElementTransformation &Trans,
-                        DenseMatrix& Hessian) const;
-
-   /** @brief Evaluate the Laplacian of all shape functions of a scalar finite
-       element in physical space at the given point @a ip. */
-   /** The size (#dof) of @a Laplacian must be set in advance. */
-   void CalcPhysLaplacian(ElementTransformation &Trans,
-                          Vector& Laplacian) const;
-
-   /** @brief Evaluate the Laplacian of all shape functions of a scalar finite
-       element in physical space at the given point @a ip. */
-   /** The size (#dof) of @a Laplacian must be set in advance. */
-   void CalcPhysLinLaplacian(ElementTransformation &Trans,
-                             Vector& Laplacian) const;
+   // virtual functions for finite elements on vector spaces

   /** @brief Evaluate the values of all shape functions of a *vector* finite
       element in reference space at the given point @a ip. */
@@ -479,6 +454,30 @@ public:
   */
   virtual void GetFaceDofs(int face, int **dofs, int *ndofs) const;

+   /** @brief Evaluate the Hessians of all shape functions of a scalar finite
+       element in reference space at the given point @a ip. */
+   /** Each row of the result DenseMatrix @a Hessian contains upper triangular
+       part of the Hessian of one shape function.
+       The order in 2D is {u_xx, u_xy, u_yy}.
+       The size (#dof x (#dim (#dim+1)/2) of @a Hessian must be set in advance.*/
+   virtual void CalcHessian(const IntegrationPoint &ip,
+                            DenseMatrix &Hessian) const;
+
+   /** @brief Evaluate the Hessian of all shape functions of a scalar finite
+       element in reference space at the given point @a ip. */
+   /** The size (#dof, #dim*(#dim+1)/2) of @a Hessian must be set in advance. */
+   virtual void CalcPhysHessian(ElementTransformation &Trans,
+                                DenseMatrix& Hessian) const;
+
+   /** @brief Evaluate the Laplacian of all shape functions of a scalar finite
+       element in reference space at the given point @a ip. */
+   /** The size (#dof) of @a Laplacian must be set in advance. */
+   virtual void CalcPhysLaplacian(ElementTransformation &Trans,
+                                  Vector& Laplacian) const;
+
+   virtual void CalcPhysLinLaplacian(ElementTransformation &Trans,
+                                     Vector& Laplacian) const;
+
   /** @brief Return the local interpolation matrix @a I (Dof x Dof) where the
       fine element is the image of the base geometry under the given
       transformation. */
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
tuckerh	f10bc713a4	small updates prior to passing the code along for comparison	2024-07-29 16:34:06 -07:00
tuckerh	96c110dab4	updating contact and parproblems files	2024-07-15 10:47:54 -07:00
tuckerh	a417272578	cleanup on finite difference test code. Making it easier for others to use	2024-07-15 08:08:00 -07:00
tuckerh	7423f8c998	adding finite difference test	2024-07-14 20:43:48 -07:00
tuckerh	db8d1f6cd4	initial commit, pushing updates from my modifications from contact-dev branch	2024-06-17 13:15:59 -07:00