Merge branch 'master' into feature/partial_ess_bcs

Claude helped generate the docs / do the refactor as I was lazy...
However, the unit tests pass and the amount of common code between the two cases has now been reduced which is always a good thing...

.gitlab/jobs/dane.yml

@@ -85,8 +85,3 @@ opt_par_gcc_10_pumi:
   extends: .mfem_job_on_dane
   variables:
     SPEC: "%gcc@10.3.1 +pumi"
-
-opt_par_gcc_10_gslib:
-  extends: .mfem_job_on_dane
-  variables:
-    SPEC: "%gcc@10.3.1 +gslib"

.gitlab/jobs/matrix.yml

@@ -63,8 +63,3 @@ opt_mpi_cuda_hypre_cuda_gcc:
   extends: .mfem_job_on_matrix
   variables:
     SPEC: "%gcc@10.3.1 +mpi +cuda cuda_arch=90 ^hypre+cuda"
-
-opt_mpi_cuda_gcc_gslib:
-  extends: .mfem_job_on_matrix
-  variables:
-    SPEC: "%gcc@10.3.1 +mpi +cuda +gslib cuda_arch=90 ^hypre+cuda"

.gitlab/scripts/baseline

@@ -32,9 +32,9 @@ mkdir _${BASELINE_TEST} && cd _${BASELINE_TEST}
 
 # run
 if [[ "${MACHINE_NAME}" == "dane" ]]; then
-  srun --nodes=1 -t 60 --exclusive --reservation=ci ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
+  salloc --nodes=1 -t 60 --exclusive --reservation=ci ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
 elif [[ ${MACHINE_NAME} == "corona" ]]; then
-  srun --nodes=1 -t 60 -p pbatch ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
+  salloc --nodes=1 -t 60 -p pbatch ../runtest ../../mfem "${BASELINE_TEST} ${TPLS_DIR}"
 else
   echo "Unknown machine: MACHINE_NAME=$MACHINE_NAME"
   exit 1

CHANGELOG

@@ -8,10 +8,8 @@
                                https://mfem.org
 
 
-Version 4.9.1 (development)
-===========================
-
-- Policy for AI-assisted contribution added to CONTRIBUTING.md
+Version 4.10 (development)
+==========================
 
 Discretization improvements
 ---------------------------
@@ -25,6 +23,12 @@ Discretization improvements
   Tet rules (d=14-20): Chuluunbaatar et al., Comput. Math. Appl. 124:89-97,
   2022.
 
+
+Version 4.9.1 (development)
+===========================
+
+Discretization improvements
+---------------------------
 - Improved the gridfunction projection routines. Projections work for Scalar,
   Vector and VectorFE, also NURBS versions. Optionally different types of
   projections can be selected, default behaviour has not changed.

CONTRIBUTING.md

@@ -24,14 +24,6 @@ must be made under this license.
 Note also that MFEM has a [Code of Conduct](CODE_OF_CONDUCT.md). By participating
 in the MFEM community, you agree to abide by its rules.
 
-## AI Policy
-- Use of AI code generation in MFEM is allowed but must be disclosed, e.g. by
-  selecting the `AI-assisted` label on the PR.
-- By submitting a PR, the author acknowledges that they have reviewed and
-  understand the changes they are proposing.
-- PR authors are still responsible for correctness, licensing, and attribution
-  of all changes.
-
 If you plan on contributing to MFEM, consider reviewing the
 [issue tracker](https://github.com/mfem/mfem/issues) first to check if a thread
 already exists for your desired feature or the bug you ran into. Use a pull

examples/CMakeLists.txt

@@ -215,7 +215,7 @@ if (MFEM_ENABLE_TESTING)
     add_test(NAME ex1p_ceed_np=${MFEM_MPI_NP}
       COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
       ${MPIEXEC_PREFLAGS}
-      $<TARGET_FILE:ex1p> "-no-vis" "-d" "ceed-cpu" "-pa" "-a"
+      $<TARGET_FILE:ex1p> "-no-vis" "-d ceed-cpu" "-pa" "-a"
       ${MPIEXEC_POSTFLAGS})
   endif()
 endif()

examples/amgx/makefile

@@ -64,7 +64,7 @@ PARALLEL_NAME := Parallel AMGX example
 $(MFEM_LIB_FILE):
 	$(error The MFEM library is not build)
 
-clean: clean-build clean-exec
+clean: clean-build
 
 clean-build:
 	rm -f *.o *~ $(SEQ_EXAMPLES) $(PAR_EXAMPLES)

examples/caliper/makefile

@@ -64,12 +64,12 @@ ex1p-test-par: ex1p
 $(MFEM_LIB_FILE):
 	$(error The MFEM library is not built)
 
-clean: clean-build clean-exec
+clean: clean-build clean-exec $(SUBDIRS_CLEAN)
 
 clean-build:
 	rm -f *.o *~ $(SEQ_EXAMPLES) $(PAR_EXAMPLES)
 	rm -rf *.dSYM *.TVD.*breakpoints
 
 clean-exec:
-	@rm -f refined.mesh mesh.*
-	@rm -f sol.*
+	@rm -f refined.mesh displaced.mesh mesh.* ex5.mesh
+	@rm -f sphere_refined.* sol.* sol_u.* sol_p.* sol_r.* sol_i.*

examples/ginkgo/makefile

@@ -76,4 +76,4 @@ clean-build:
 	rm -rf *.dSYM *.TVD.*breakpoints
 
 clean-exec:
-	@rm -f refined.mesh sol.gf mesh.* sol.*
+	@rm -f refined.mesh sol.gf

examples/makefile

@@ -71,7 +71,6 @@ endif
 
 SUBDIRS_ALL = $(addsuffix /all,$(SUBDIRS))
 SUBDIRS_TEST = $(addsuffix /test,$(SUBDIRS))
-SUBDIRS_TEST_NOCLEAN = $(addsuffix /test-noclean,$(SUBDIRS))
 SUBDIRS_CLEAN = $(addsuffix /clean,$(SUBDIRS))
 SUBDIRS_TPRINT = $(addsuffix /test-print,$(SUBDIRS))
 
@@ -88,9 +87,8 @@ SUBDIRS_TPRINT = $(addsuffix /test-print,$(SUBDIRS))
 
 all: $(EXAMPLES) $(SUBDIRS_ALL)
 
-.PHONY: $(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_TEST_NOCLEAN) \
-   $(SUBDIRS_CLEAN) $(SUBDIRS_TPRINT)
-$(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_TEST_NOCLEAN) $(SUBDIRS_CLEAN):
+.PHONY: $(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_CLEAN) $(SUBDIRS_TPRINT)
+$(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_CLEAN):
 	$(MAKE) -C $(@D) $(@F)
 $(SUBDIRS_TPRINT):
 	@$(MAKE) -C $(@D) $(@F)
@@ -109,7 +107,6 @@ endif
 MFEM_TESTS = EXAMPLES
 include $(MFEM_TEST_MK)
 test: $(SUBDIRS_TEST)
-test-noclean: $(SUBDIRS_TEST_NOCLEAN)
 test-print: $(SUBDIRS_TPRINT)
 
 # Testing: Parallel vs. serial runs

fem/bounds.cpp

@@ -41,14 +41,9 @@ void PLBound::Setup(const int nb_i, const int ncp_i,
    tol = tol_i;
    lbound.SetSize(ncp, nb);
    ubound.SetSize(ncp, nb);
-   lbound_t.SetSize(nb, ncp);
-   ubound_t.SetSize(nb, ncp);
    nodes.SetSize(nb);
    weights.SetSize(nb);
    control_points.SetSize(ncp);
-   xhat.SetSize(nb);
-   what.SetSize(nb);
-   cphat.SetSize(ncp);
 
    auto scalenodes = [](const Vector &in, const real_t a, const real_t b) -> Vector
    {
@@ -95,10 +90,6 @@ void PLBound::Setup(const int nb_i, const int ncp_i,
       MFEM_ABORT("Unsupported interval points. Use [0,1].\n");
    }
    control_points = scalenodes(control_points, 0.0, 1.0); // rescale to [0,1]
-   for (int i = 0; i < ncp; i++)
-   {
-      cphat(i) = 2.0*control_points(i) - 1.0;
-   }
 
    Poly_1D::Basis &basis1d(poly1d.GetBasis(nb-1, b_type));
 
@@ -154,8 +145,6 @@ void PLBound::Setup(const int nb_i, const int ncp_i,
                lbound(j,i) = std::max(lbound(j,i),0_r);
             }
          }
-         lbound_t(i,j) = lbound(j,i);
-         ubound_t(i,j) = ubound(j,i);
       }
    }
 
@@ -187,11 +176,6 @@ void PLBound::Setup(const int nb_i, const int ncp_i,
          nodes(i) = irule.IntPoint(i).x;
       }
    }
-   for (int i = 0; i < nb; i++)
-   {
-      xhat(i) = 2.0*nodes(i) - 1.0;
-      what(i) = 2.0*weights(i);
-   }
 
    if (b_type == 2)
    {
@@ -771,4 +755,4 @@ void PLBound::Print(std::ostream &outp) const
    ubound.Print(outp);
 }
 
-}
+}

fem/bounds.hpp

@@ -13,7 +13,6 @@
 #define MFEM_BOUNDS
 
 #include "../config/config.hpp"
-#include "../general/forall.hpp"
 #include "fespace.hpp"
 
 namespace mfem
@@ -61,9 +60,7 @@ private:
    bool proj = true; // Use linear projection to compute bounds.
    real_t tol = 0.0; // offset bounds to avoid round-off errors
    Vector nodes, weights, control_points;
-   Vector xhat, what, cphat;
    DenseMatrix lbound, ubound; // ncp x nb matrices with bounds of all bases
-   DenseMatrix lbound_t, ubound_t; // nb x ncp transposes for device kernel
    // Some auxillary storage for computing the bounds with Bernstein
    DenseMatrix basisMatNodes; // Bernstein bases at equispaced nodes
    DenseMatrix basisMatInt;   // Bernstein bases at GLL nodes
@@ -116,10 +113,7 @@ public:
     *  @details This projection increases the computational cost but results in
     *  tighter bounds.
     */
-   void SetProjectionFlagForBounding(bool proj_)
-   {
-      proj = proj_;
-   }
+   void SetProjectionFlagForBounding(bool proj_) { proj = proj_; }
 
    /** @brief Compute piecewise linear bounds for the lexicographically-ordered
     *  nodal coefficients in @a coeff in 1D/2D/3D.
@@ -143,23 +137,9 @@ public:
    /// Get number of control points used to compute the bounds.
    int GetNControlPoints() const { return ncp; }
 
-   /// Get the underlying 1D basis type.
-   int GetBasisType() const { return b_type; }
-
    /// Get 1D control point locations (lexicographic order) in [0,1].
    const Vector &GetControlPoints() const { return control_points; }
 
-   /** @brief Compute element-wise bounds from a lexicographic E-vector.
-    *
-    *  @details The expected layout of @a e_vec is `ND x VDIM x NE`, where
-    *  `ND = nb^rdim`, `VDIM = fes_vdim`, and `NE` is the number of elements.
-    *  The output layout matches GridFunction::GetElementBounds:
-    *  `NE x active_vdim`, with the element index varying fastest.
-    */
-   void GetElementBoundsKernel(const int rdim, const int fes_vdim,
-                               const Vector &e_vec, Vector &lower,
-                               Vector &upper, const int vdim = 0) const;
-
    /** @brief Get lower and upper bounding matrix (ncp^dim x nb^dim)
     *
     *  @details The matrices can be used to compute the bounds at control points
@@ -203,600 +183,6 @@ private:
               const int cp_type_i, const real_t tol_i);
 };
 
-namespace internal
-{
-
-struct PLBoundDeviceData
-{
-   int nb;
-   int ncp;
-   const real_t *xhat;
-   const real_t *what;
-   const real_t *cphat;
-   const real_t *lbound;
-   const real_t *ubound;
-};
-
-template<int T_NB = 0, bool T_PROJ = true>
-inline void GetElementBoundsKernel1D(const PLBoundDeviceData &data,
-                                     const int fes_vdim,
-                                     const int ne,
-                                     const Vector &e_vec,
-                                     Vector &lower,
-                                     Vector &upper,
-                                     const int comp0,
-                                     const int ncomp)
-{
-   constexpr int GENERIC_MAX_ND = 32;
-   constexpr int MAX_ND = T_NB ? T_NB : GENERIC_MAX_ND;
-   constexpr int BLOCK_X = 2*MAX_ND;
-
-   const int nd = T_NB ? T_NB : data.nb;
-   MFEM_VERIFY(nd <= MAX_ND,
-               "Device element bounds kernel supports up to 32 "
-               "1D degrees of freedom.");
-
-   const auto E = Reshape(e_vec.Read(), nd, fes_vdim, ne);
-   auto L = Reshape(lower.Write(), ne, ncomp);
-   auto U = Reshape(upper.Write(), ne, ncomp);
-
-   mfem::forall_2D<BLOCK_X>(ne*ncomp, BLOCK_X, 1,
-                            [=] MFEM_HOST_DEVICE (int ec)
-   {
-      const int e = ec % ne;
-      const int c = ec / ne;
-      const int vc = comp0 + c;
-      const real_t *coeff = &E(0, vc, e);
-      const int tid = MFEM_THREAD_ID(x);
-
-      MFEM_SHARED real_t sproj[MAX_ND];
-      MFEM_SHARED real_t ssum0[MAX_ND];
-      MFEM_SHARED real_t ssum1[MAX_ND];
-      MFEM_SHARED real_t smin[BLOCK_X];
-      MFEM_SHARED real_t smax[BLOCK_X];
-      MFEM_SHARED real_t sa0;
-      MFEM_SHARED real_t sa1;
-
-      MFEM_FOREACH_THREAD(i, x, nd)
-      {
-         if constexpr (T_PROJ)
-         {
-            const real_t x = data.xhat[i];
-            const real_t w = data.what[i];
-            ssum0[i] = 0.5*coeff[i]*w;
-            ssum1[i] = 1.5*coeff[i]*w*x;
-         }
-         else
-         {
-            ssum0[i] = 0.0;
-            ssum1[i] = 0.0;
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      MFEM_FOREACH_THREAD(ii, x, 1)
-      {
-         sa0 = 0.0;
-         sa1 = 0.0;
-         for (int i = 0; i < nd; i++)
-         {
-            sa0 += ssum0[i];
-            sa1 += ssum1[i];
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      MFEM_FOREACH_THREAD(i, x, nd)
-      {
-         if constexpr (T_PROJ)
-         {
-            const real_t x = data.xhat[i];
-            sproj[i] = coeff[i] - sa0 - sa1*x;
-         }
-         else
-         {
-            sproj[i] = coeff[i];
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      real_t lower_local = HUGE_VAL;
-      real_t upper_local = -HUGE_VAL;
-      MFEM_FOREACH_THREAD(j, x, data.ncp)
-      {
-         real_t lo = 0.0;
-         real_t hi = 0.0;
-         if constexpr (T_PROJ)
-         {
-            const real_t xcp = data.cphat[j];
-            lo = sa0 + sa1*xcp;
-            hi = lo;
-         }
-
-         for (int i = 0; i < nd; i++)
-         {
-            const real_t val = sproj[i];
-            const real_t lv = data.lbound[j + i*data.ncp]*val;
-            const real_t uv = data.ubound[j + i*data.ncp]*val;
-            lo += lv < uv ? lv : uv;
-            hi += lv > uv ? lv : uv;
-         }
-         lower_local = lower_local < lo ? lower_local : lo;
-         upper_local = upper_local > hi ? upper_local : hi;
-      }
-
-      smin[tid] = lower_local;
-      smax[tid] = upper_local;
-      MFEM_SYNC_THREAD;
-
-      MFEM_FOREACH_THREAD(ii, x, 1)
-      {
-         real_t lower_ec = smin[0];
-         real_t upper_ec = smax[0];
-         const int nthreads = MFEM_THREAD_SIZE(x);
-         const int nactive = data.ncp < nthreads ? data.ncp : nthreads;
-         for (int t = 1; t < nactive; t++)
-         {
-            lower_ec = lower_ec < smin[t] ? lower_ec : smin[t];
-            upper_ec = upper_ec > smax[t] ? upper_ec : smax[t];
-         }
-         L(e, c) = lower_ec;
-         U(e, c) = upper_ec;
-      }
-   });
-}
-
-template<int T_NB = 0, int T_NCP = 0, bool T_PROJ = true>
-inline void GetElementBoundsKernel2D(const PLBoundDeviceData &data,
-                                     const int fes_vdim,
-                                     const int ne,
-                                     const Vector &e_vec,
-                                     Vector &lower,
-                                     Vector &upper,
-                                     const int comp0,
-                                     const int ncomp)
-{
-   constexpr int DEFAULT_MAX_NB = 8;
-   constexpr int DEFAULT_MAX_CP = 3*DEFAULT_MAX_NB;
-   constexpr int MAX_NB = T_NB ? T_NB : DEFAULT_MAX_NB;
-   constexpr int MAX_CP = T_NCP ? T_NCP : DEFAULT_MAX_CP;
-   constexpr int MAX_THREADS = MAX_CP*MAX_CP;
-
-   const int nb = data.nb;
-   const int ncp = data.ncp;
-   const int nd = nb*nb;
-   MFEM_VERIFY(nb <= MAX_NB,
-               "Device 2D element bounds kernel exceeds its compile-time "
-               "1D degree bound.");
-   MFEM_VERIFY(ncp <= MAX_CP,
-               "Device 2D element bounds kernel exceeds its compile-time "
-               "control-point bound.");
-   MFEM_VERIFY(ncp*ncp <= MAX_THREADS,
-               "Device 2D element bounds kernel exceeds its compile-time "
-               "thread-block bound.");
-
-   const auto E = Reshape(e_vec.Read(), nd, fes_vdim, ne);
-   auto L = Reshape(lower.Write(), ne, ncomp);
-   auto U = Reshape(upper.Write(), ne, ncomp);
-
-   mfem::forall_2D<MAX_THREADS>(ne*ncomp, ncp, ncp,
-                                [=] MFEM_HOST_DEVICE (int ec)
-   {
-      const int e = ec % ne;
-      const int c = ec / ne;
-      const int vc = comp0 + c;
-      const real_t *coeff = &E(0, vc, e);
-      const int tx = MFEM_THREAD_ID(x);
-      const int ty = MFEM_THREAD_ID(y);
-
-      MFEM_SHARED real_t sproj[MAX_NB*MAX_NB];
-      MFEM_SHARED real_t srow_min[MAX_NB*MAX_CP];
-      MFEM_SHARED real_t srow_max[MAX_NB*MAX_CP];
-      MFEM_SHARED real_t srow_a0[MAX_NB];
-      MFEM_SHARED real_t srow_a1[MAX_NB];
-      MFEM_SHARED real_t sa0[MAX_CP];
-      MFEM_SHARED real_t sa1[MAX_CP];
-      MFEM_SHARED real_t smin[MAX_THREADS];
-      MFEM_SHARED real_t smax[MAX_THREADS];
-
-      // Stage 1a: for each nodal row, form the per-node contributions to the
-      // row-wise linear fit used by the first 1D bounding solve.
-      MFEM_FOREACH_THREAD(jrow, y, nb)
-      {
-         const real_t *row_coeff = coeff + jrow*nb;
-         const int row_ncp_off = jrow*MAX_CP;
-         MFEM_FOREACH_THREAD(i, x, nb)
-         {
-            if constexpr (T_PROJ)
-            {
-               const real_t x = data.xhat[i];
-               const real_t w = data.what[i];
-               srow_min[row_ncp_off + i] = 0.5*row_coeff[i]*w;
-               srow_max[row_ncp_off + i] = 1.5*row_coeff[i]*w*x;
-            }
-            else
-            {
-               srow_min[row_ncp_off + i] = 0.0;
-               srow_max[row_ncp_off + i] = 0.0;
-            }
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      // Stage 1b: reduce the row-wise projection coefficients a0/a1.
-      if constexpr (T_PROJ)
-      {
-         MFEM_FOREACH_THREAD(jrow, y, nb)
-         {
-            const int row_ncp_off = jrow*MAX_CP;
-            real_t a0 = 0.0;
-            real_t a1 = 0.0;
-            MFEM_FOREACH_THREAD(ii, x, 1)
-            {
-               for (int i = 0; i < nb; i++)
-               {
-                  a0 += srow_min[row_ncp_off + i];
-                  a1 += srow_max[row_ncp_off + i];
-               }
-               srow_a0[jrow] = a0;
-               srow_a1[jrow] = a1;
-            }
-         }
-         MFEM_SYNC_THREAD;
-      }
-
-      // Stage 1c: subtract the row-wise linear fit once and cache the
-      // projected row coefficients for reuse across all x-control points.
-      MFEM_FOREACH_THREAD(jrow, y, nb)
-      {
-         const real_t *row_coeff = coeff + jrow*nb;
-         MFEM_FOREACH_THREAD(i, x, nb)
-         {
-            if constexpr (T_PROJ)
-            {
-               const real_t x = data.xhat[i];
-               sproj[jrow*MAX_NB + i] = row_coeff[i]
-                                        - srow_a0[jrow] - srow_a1[jrow]*x;
-            }
-            else
-            {
-               sproj[jrow*MAX_NB + i] = row_coeff[i];
-            }
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      // Stage 1d: solve the first 1D bounding problem along each nodal row and
-      // store bounds at every x-direction control point.
-      MFEM_FOREACH_THREAD(icp, x, ncp)
-      {
-         MFEM_FOREACH_THREAD(jrow, y, nb)
-         {
-            const int row_cp_off = jrow*ncp;
-            real_t lo = 0.0;
-            real_t hi = 0.0;
-            if constexpr (T_PROJ)
-            {
-               const real_t xcp = data.cphat[icp];
-               lo = srow_a0[jrow] + srow_a1[jrow]*xcp;
-               hi = lo;
-            }
-            for (int i = 0; i < nb; i++)
-            {
-               const real_t val = sproj[jrow*MAX_NB + i];
-               const real_t lv = data.lbound[icp + i*data.ncp]*val;
-               const real_t uv = data.ubound[icp + i*data.ncp]*val;
-               lo += lv < uv ? lv : uv;
-               hi += lv > uv ? lv : uv;
-            }
-            srow_min[row_cp_off + icp] = lo;
-            srow_max[row_cp_off + icp] = hi;
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      // Stage 2a: from the row bounds, form the per-row contributions to the
-      // second 1D projection solve in the y-direction.
-      MFEM_FOREACH_THREAD(icp, x, ncp)
-      {
-         MFEM_FOREACH_THREAD(jrow, y, nb)
-         {
-            const int row_cp_off = jrow*ncp;
-            if constexpr (T_PROJ)
-            {
-               const real_t x = data.xhat[jrow];
-               const real_t w = data.what[jrow];
-               const real_t t = 0.5*(srow_min[row_cp_off + icp] +
-                                     srow_max[row_cp_off + icp]);
-               smin[row_cp_off + icp] = 0.5*t*w;
-               smax[row_cp_off + icp] = 1.5*t*w*x;
-            }
-            else
-            {
-               smin[row_cp_off + icp] = 0.0;
-               smax[row_cp_off + icp] = 0.0;
-            }
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      // Stage 2b: reduce the y-direction projection coefficients for each
-      // x-control-point column.
-      MFEM_FOREACH_THREAD(jj, y, 1)
-      {
-         MFEM_FOREACH_THREAD(icp, x, ncp)
-         {
-            real_t a0 = 0.0;
-            real_t a1 = 0.0;
-            for (int jrow = 0; jrow < nb; jrow++)
-            {
-               a0 += smin[jrow*ncp + icp];
-               a1 += smax[jrow*ncp + icp];
-            }
-            sa0[icp] = a0;
-            sa1[icp] = a1;
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      // Stage 2c: subtract the y-direction linear fit from the intermediate
-      // row bounds so the final tensor-product bound uses the perturbation.
-      if constexpr (T_PROJ)
-      {
-         MFEM_FOREACH_THREAD(icp, x, ncp)
-         {
-            MFEM_FOREACH_THREAD(jrow, y, nb)
-            {
-               const int row_cp_off = jrow*ncp;
-               const real_t x = data.xhat[jrow];
-               const real_t t = sa0[icp] + sa1[icp]*x;
-               srow_min[row_cp_off + icp] -= t;
-               srow_max[row_cp_off + icp] -= t;
-            }
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      // Stage 3: each thread now owns one 2D control point (icp, kcp) and
-      // accumulates its final lower/upper bound from the row-bound data.
-      MFEM_FOREACH_THREAD(icp, x, ncp)
-      {
-         MFEM_FOREACH_THREAD(kcp, y, ncp)
-         {
-            real_t lo = 0.0;
-            real_t hi = 0.0;
-            if constexpr (T_PROJ)
-            {
-               const real_t xcp = data.cphat[kcp];
-               lo = sa0[icp] + sa1[icp]*xcp;
-               hi = lo;
-            }
-            for (int jrow = 0; jrow < nb; jrow++)
-            {
-               const real_t w0 = srow_min[jrow*ncp + icp];
-               const real_t w1 = srow_max[jrow*ncp + icp];
-               const real_t lb = data.lbound[kcp + jrow*data.ncp];
-               const real_t ub = data.ubound[kcp + jrow*data.ncp];
-               const real_t v0 = lb*w0;
-               const real_t v1 = ub*w0;
-               const real_t v2 = lb*w1;
-               const real_t v3 = ub*w1;
-               real_t vlo = v0 < v1 ? v0 : v1;
-               real_t vhi = v0 > v1 ? v0 : v1;
-               vlo = vlo < v2 ? vlo : v2;
-               vlo = vlo < v3 ? vlo : v3;
-               vhi = vhi > v2 ? vhi : v2;
-               vhi = vhi > v3 ? vhi : v3;
-               lo += vlo;
-               hi += vhi;
-            }
-            const int slot = kcp*ncp + icp;
-            smin[slot] = lo;
-            smax[slot] = hi;
-         }
-      }
-      MFEM_SYNC_THREAD;
-
-      const int lane = ty*ncp + tx;
-      const int nactive = ncp*ncp;
-      const int nthreads = MFEM_THREAD_SIZE(x)*MFEM_THREAD_SIZE(y);
-
-      // Reduce all 2D control-point bounds to one lower/upper pair per
-      // (element, component).
-      if (nthreads == 1)
-      {
-         if (tx == 0 && ty == 0)
-         {
-            real_t lower_ec = smin[0];
-            real_t upper_ec = smax[0];
-            for (int t = 1; t < nactive; t++)
-            {
-               lower_ec = lower_ec < smin[t] ? lower_ec : smin[t];
-               upper_ec = upper_ec > smax[t] ? upper_ec : smax[t];
-            }
-            L(e, c) = lower_ec;
-            U(e, c) = upper_ec;
-         }
-      }
-      else
-      {
-         for (int stride = (nactive + 1)/2; stride > 0;
-              stride = (stride + 1)/2)
-         {
-            if (lane < stride && lane + stride < nactive)
-            {
-               smin[lane] = smin[lane] < smin[lane + stride] ?
-                            smin[lane] : smin[lane + stride];
-               smax[lane] = smax[lane] > smax[lane + stride] ?
-                            smax[lane] : smax[lane + stride];
-            }
-            MFEM_SYNC_THREAD;
-            if (stride == 1) { break; }
-         }
-
-         if (lane == 0)
-         {
-            L(e, c) = smin[0];
-            U(e, c) = smax[0];
-         }
-      }
-   });
-}
-
-} // namespace internal
-
-inline void PLBound::GetElementBoundsKernel(const int rdim, const int fes_vdim,
-                                            const Vector &e_vec,
-                                            Vector &lower, Vector &upper,
-                                            const int vdim) const
-{
-   MFEM_VERIFY(b_type != BasisType::Positive,
-               "Bernstein device bounds are not implemented.");
-   if (rdim == 3)
-   {
-      MFEM_ABORT("Device element bounds kernel currently only supports 1D/2D.");
-   }
-   MFEM_VERIFY(rdim == 1 || rdim == 2, "Invalid element dimension.");
-   MFEM_VERIFY(vdim >= -1 && vdim <= fes_vdim, "Invalid vector component.");
-   const int nd = static_cast<int>(std::pow(nb, rdim));
-   const int ne = e_vec.Size()/(nd*fes_vdim);
-   const int ncomp = (vdim > 0) ? 1 : fes_vdim;
-
-   lower.SetSize(ne*ncomp, e_vec);
-   upper.SetSize(ne*ncomp, e_vec);
-   lower.UseDevice(true);
-   upper.UseDevice(true);
-
-   if (!proj)
-   {
-      MFEM_ABORT("Device element bounds kernel currently requires projection "
-                 "enabled.");
-   }
-
-   const real_t *dxhat = xhat.Read();
-   const real_t *dwhat = what.Read();
-   const real_t *dcphat = cphat.Read();
-   const real_t *dlbound = lbound.Read();
-   const real_t *dubound = ubound.Read();
-
-   internal::PLBoundDeviceData data
-   {
-      nb,
-      ncp,
-      dxhat,
-      dwhat,
-      dcphat,
-      dlbound,
-      dubound
-   };
-
-   const int comp0 = (vdim > 0) ? (vdim - 1) : 0;
-
-   if (rdim == 1)
-   {
-      switch (nb)
-      {
-         case 2: return internal::GetElementBoundsKernel1D<2, true>(data, fes_vdim, ne,
-                                                                     e_vec, lower, upper,
-                                                                     comp0, ncomp);
-         case 3: return internal::GetElementBoundsKernel1D<3, true>(data, fes_vdim, ne,
-                                                                     e_vec, lower, upper,
-                                                                     comp0, ncomp);
-         case 4: return internal::GetElementBoundsKernel1D<4, true>(data, fes_vdim, ne,
-                                                                     e_vec, lower, upper,
-                                                                     comp0, ncomp);
-         case 5: return internal::GetElementBoundsKernel1D<5, true>(data, fes_vdim, ne,
-                                                                     e_vec, lower, upper,
-                                                                     comp0, ncomp);
-         case 6: return internal::GetElementBoundsKernel1D<6, true>(data, fes_vdim, ne,
-                                                                     e_vec, lower, upper,
-                                                                     comp0, ncomp);
-         case 7: return internal::GetElementBoundsKernel1D<7, true>(data, fes_vdim, ne,
-                                                                     e_vec, lower, upper,
-                                                                     comp0, ncomp);
-         case 8: return internal::GetElementBoundsKernel1D<8, true>(data, fes_vdim, ne,
-                                                                     e_vec, lower, upper,
-                                                                     comp0, ncomp);
-         case 9: return internal::GetElementBoundsKernel1D<9, true>(data, fes_vdim, ne,
-                                                                     e_vec, lower, upper,
-                                                                     comp0, ncomp);
-         case 10: return internal::GetElementBoundsKernel1D<10, true>(data, fes_vdim, ne,
-                                                                       e_vec, lower, upper,
-                                                                       comp0, ncomp);
-         default: return internal::GetElementBoundsKernel1D<0, true>(data, fes_vdim, ne,
-                                                                      e_vec, lower, upper,
-                                                                      comp0, ncomp);
-      }
-   }
-#define MFEM_PLBOUND_2D_DISPATCH(NB, NCP) \
-   return internal::GetElementBoundsKernel2D<NB, NCP, true>(data, fes_vdim, ne, \
-                                                            e_vec, lower, upper, \
-                                                            comp0, ncomp)
-   switch (nb)
-   {
-      case 2:
-         switch (ncp)
-         {
-            case 4: MFEM_PLBOUND_2D_DISPATCH(2, 4);
-            case 6: MFEM_PLBOUND_2D_DISPATCH(2, 6);
-            case 8: MFEM_PLBOUND_2D_DISPATCH(2, 8);
-         }
-         break;
-      case 3:
-         switch (ncp)
-         {
-            case 6: MFEM_PLBOUND_2D_DISPATCH(3, 6);
-            case 9: MFEM_PLBOUND_2D_DISPATCH(3, 9);
-            case 12: MFEM_PLBOUND_2D_DISPATCH(3, 12);
-         }
-         break;
-      case 4:
-         switch (ncp)
-         {
-            case 8: MFEM_PLBOUND_2D_DISPATCH(4, 8);
-            case 12: MFEM_PLBOUND_2D_DISPATCH(4, 12);
-            case 16: MFEM_PLBOUND_2D_DISPATCH(4, 16);
-         }
-         break;
-      case 5:
-         switch (ncp)
-         {
-            case 10: MFEM_PLBOUND_2D_DISPATCH(5, 10);
-            case 15: MFEM_PLBOUND_2D_DISPATCH(5, 15);
-            case 20: MFEM_PLBOUND_2D_DISPATCH(5, 20);
-         }
-         break;
-      case 6:
-         switch (ncp)
-         {
-            case 12: MFEM_PLBOUND_2D_DISPATCH(6, 12);
-            case 18: MFEM_PLBOUND_2D_DISPATCH(6, 18);
-            case 24: MFEM_PLBOUND_2D_DISPATCH(6, 24);
-         }
-         break;
-      case 7:
-         switch (ncp)
-         {
-            case 14: MFEM_PLBOUND_2D_DISPATCH(7, 14);
-            case 21: MFEM_PLBOUND_2D_DISPATCH(7, 21);
-            case 28: MFEM_PLBOUND_2D_DISPATCH(7, 28);
-         }
-         break;
-      case 8:
-         switch (ncp)
-         {
-            case 16: MFEM_PLBOUND_2D_DISPATCH(8, 16);
-            case 24: MFEM_PLBOUND_2D_DISPATCH(8, 24);
-            case 32: MFEM_PLBOUND_2D_DISPATCH(8, 32);
-         }
-         break;
-   }
-#undef MFEM_PLBOUND_2D_DISPATCH
-   return internal::GetElementBoundsKernel2D<0, 0, true>(data, fes_vdim, ne,
-                                                         e_vec, lower, upper,
-                                                         comp0, ncomp);
-}
-
 } // namespace mfem
 
 #endif // MFEM_BOUNDS

fem/coefficient.hpp

@@ -52,9 +52,6 @@ public:
    /// Get the time for time dependent coefficients
    real_t GetTime() { return time; }
 
-   /// Returns dimension of the vector.
-   int GetVDim() { return 1; }
-
    /** @brief Evaluate the coefficient in the element described by @a T at the
        point @a ip. */
    /** @note When this method is called, the caller must make sure that the

fem/datacollection.cpp

@@ -492,8 +492,6 @@ void VisItDataCollection::SaveRootFile()
                            to_padded_string(cycle, pad_digits_cycle) +
                            ".mfem_root";
    std::ofstream root_file(root_name);
-   MFEM_VERIFY(root_file.is_open(),
-               "Failed to open ofstream " << root_name);
    root_file << GetVisItRootString();
    if (!root_file)
    {
@@ -979,10 +977,7 @@ void ParaViewDataCollection::Save()
    // Save the local part of the mesh and grid functions fields to the local
    // VTU file. Also save coefficient fields.
    {
-      std::string os_str = vtu_prefix + GenerateVTUFileName("proc", myid);
-      std::ofstream os(os_str);
-      MFEM_VERIFY(os.is_open(),
-                  "Failed to open ofstream " << os_str);
+      std::ofstream os(vtu_prefix + GenerateVTUFileName("proc", myid));
       os.precision(precision);
       SaveDataVTU(os, levels_of_detail);
    }
@@ -994,10 +989,7 @@ void ParaViewDataCollection::Save()
                   "QuadratureFunction output is not supported for "
                   "ParaViewDataCollection on domain boundary!");
       const std::string &field_name = qfield.first;
-      std::string os_str = vtu_prefix + GenerateVTUFileName(field_name, myid);
-      std::ofstream os(os_str);
-      MFEM_VERIFY(os.is_open(),
-                  "Failed to open ofstream " << os_str);
+      std::ofstream os(vtu_prefix + GenerateVTUFileName(field_name, myid));
       qfield.second->SaveVTU(os, pv_data_format, GetCompressionLevel(), field_name);
    }
 
@@ -1008,10 +1000,7 @@ void ParaViewDataCollection::Save()
    {
       // Create the main PVTU file
       {
-         std::string os_str = vtu_prefix + GeneratePVTUFileName("data");
-         std::ofstream pvtu_out(os_str);
-         MFEM_VERIFY(pvtu_out.is_open(),
-                     "Failed to open ofstream " << os_str);
+         std::ofstream pvtu_out(vtu_prefix + GeneratePVTUFileName("data"));
          WritePVTUHeader(pvtu_out);
 
          // Grid function fields and coefficient fields
@@ -1066,10 +1055,8 @@ void ParaViewDataCollection::Save()
          const std::string &q_field_name = q_field.first;
          std::string q_fname = GeneratePVTUPath() + "/"
                                + GeneratePVTUFileName(q_field_name);
-         std::string os_str = col_path + "/" + q_fname;
-         std::ofstream pvtu_out(os_str);
-         MFEM_VERIFY(pvtu_out.is_open(),
-                     "Failed to open ofstream " << os_str);
+
+         std::ofstream pvtu_out(col_path + "/" + q_fname);
          WritePVTUHeader(pvtu_out);
          int vec_dim = q_field.second->GetVDim();
          pvtu_out << "<PPointData>\n";

fem/dgmassinv_kernels.hpp

@@ -57,7 +57,7 @@ void DGMassApply(const int e,
       }
       else if (DIM == 3)
       {
-         SmemPAMassApply3D_Element<TD1D,TQ1D,NBZ,ACCUM>(e, NE, B, pa_data, x, y);
+         SmemPAMassApply3D_Element<TD1D,TQ1D,ACCUM>(e, NE, B, pa_data, x, y);
       }
       else
       {

fem/estimators.hpp

@@ -320,8 +320,8 @@ public:
     error estimation procedure where the flux averaging is replaced by a global
     L2 projection (requiring a mass matrix solve).
 
-    The required BilinearFormIntegrator must implement the method
-    ComputeElementFlux().
+    The required BilinearFormIntegrator must implement the methods
+    ComputeElementFlux() and ComputeFluxEnergy().
 
     Implemented for the parallel case only.
  */
@@ -357,8 +357,8 @@ protected:
 
 public:
    /** @brief Construct a new L2ZienkiewiczZhuEstimator object.
-       @param integ    This BilinearFormIntegrator must implement the method
-                       ComputeElementFlux().
+       @param integ    This BilinearFormIntegrator must implement the methods
+                       ComputeElementFlux() and ComputeFluxEnergy().
        @param sol      The solution field whose error is to be estimated.
        @param flux_fes The L2ZienkiewiczZhuEstimator assumes ownership of this
                        FiniteElementSpace and will call its Update() method when
@@ -382,8 +382,8 @@ public:
    { }
 
    /** @brief Construct a new L2ZienkiewiczZhuEstimator object.
-       @param integ    This BilinearFormIntegrator must implement the method
-                       ComputeElementFlux().
+       @param integ    This BilinearFormIntegrator must implement the methods
+                       ComputeElementFlux() and ComputeFluxEnergy().
        @param sol      The solution field whose error is to be estimated.
        @param flux_fes The L2ZienkiewiczZhuEstimator does NOT assume ownership
                        of this FiniteElementSpace; will call its Update() method

fem/fespace.cpp

@@ -547,11 +547,22 @@ void MarkDofs(const Array<int> &dofs, Array<int> &mark_array)
 
 void FiniteElementSpace::GetEssentialVDofs(const Array<int> &bdr_attr_is_ess,
                                            Array<int> &ess_vdofs,
-                                           int component) const
+                                           int component,
+                                           bool overwrite) const
 {
    Array<int> dofs;
-   ess_vdofs.SetSize(GetVSize());
-   ess_vdofs = 0;
+
+   if (overwrite)
+   {
+      ess_vdofs.SetSize(GetVSize());
+      ess_vdofs = 0;
+   }
+   else
+   {
+      MFEM_ASSERT(ess_vdofs.Size() == GetVSize(),
+                  "ess_vdofs size is not equal to FESpaces GetVSize().");
+   }
+
    for (int i = 0; i < GetNBE(); i++)
    {
       if (bdr_attr_is_ess[GetBdrAttribute(i)-1])
@@ -663,6 +674,54 @@ void FiniteElementSpace::GetEssentialTrueDofs(const Array<int> &bdr_attr_is_ess,
    MarkerToList(ess_tdofs, ess_tdof_list);
 }
 
+void FiniteElementSpace::GetEssentialVDofsFromComponent(
+   const Array<int> &bdr_attr_is_ess,
+   const Array2D<bool> &component,
+   Array<int> &ess_vdofs) const
+{
+   MFEM_ASSERT(component.NumCols() == vdim,
+               "Number of columns of component was not equal to FESpace vdim");
+   MFEM_ASSERT(component.NumRows() == bdr_attr_is_ess.Size(),
+               "Number of rows of component was not equal to bdr_attr_is_ess.Size()");
+
+   Array<int> bdr_attr_is_ess_single_comp;
+   bdr_attr_is_ess_single_comp.SetSize(bdr_attr_is_ess.Size());
+
+   for (int i = 0; i < vdim; i++)
+   {
+      const bool overwrite = (i == 0);
+      bdr_attr_is_ess_single_comp = 0;
+      for (int j = 0; j < bdr_attr_is_ess.Size(); j++)
+      {
+         if (bdr_attr_is_ess[j] && component(j, i))
+         {
+            bdr_attr_is_ess_single_comp[j] = bdr_attr_is_ess[j];
+         }
+      }
+      GetEssentialVDofs(bdr_attr_is_ess_single_comp, ess_vdofs, i, overwrite);
+   }
+}
+
+void FiniteElementSpace::GetEssentialTrueDofs(const Array<int> &bdr_attr_is_ess,
+                                              Array<int> &ess_tdof_list,
+                                              const Array2D<bool> &component)
+{
+   Array<int> ess_vdofs, ess_tdofs;
+
+   GetEssentialVDofsFromComponent(bdr_attr_is_ess, component, ess_vdofs);
+
+   const SparseMatrix *R = GetConformingRestriction();
+   if (!R)
+   {
+      ess_tdofs.MakeRef(ess_vdofs);
+   }
+   else
+   {
+      R->BooleanMult(ess_vdofs, ess_tdofs);
+   }
+   MarkerToList(ess_tdofs, ess_tdof_list);
+}
+
 void FiniteElementSpace::GetBoundaryTrueDofs(Array<int> &boundary_dofs,
                                              int component)
 {

fem/fespace.hpp

@@ -595,6 +595,29 @@ protected:
    virtual void CopyProlongationAndRestriction(const FiniteElementSpace &fes,
                                                const Array<int> *perm);
 
+   /** @brief Helper function to mark essential VDofs based on a component
+       matrix specifying which vector components are essential on each boundary
+       attribute.
+
+       This method loops over all vector dimensions (vdim) and calls
+       GetEssentialVDofs() for each component where the corresponding entry
+       in the @a component matrix is true.
+
+       @param[in] bdr_attr_is_ess Array marking which boundary attributes are
+                                  essential (1 for essential, 0 otherwise).
+       @param[in] component 2D boolean array of size (num_bdr_attributes x vdim)
+                           indicating which vector components are essential for
+                           each boundary attribute. component(j,i) == true means
+                           component i is essential on boundary attribute j.
+       @param[out] ess_vdofs Marker array for essential VDofs. On exit,
+                            ess_vdofs[i] != 0 if VDof i is essential.
+
+       @note This is a helper method used by GetEssentialTrueDofs() when
+             component-wise boundary conditions are specified. */
+   void GetEssentialVDofsFromComponent(const Array<int> &bdr_attr_is_ess,
+                                       const Array2D<bool> &component,
+                                       Array<int> &ess_vdofs) const;
+
 public:
 
 
@@ -1356,11 +1379,17 @@ public:
 
    /** @brief Mark degrees of freedom associated with boundary elements with
        the specified boundary attributes (marked in 'bdr_attr_is_ess').
+
        For spaces with 'vdim' > 1, the 'component' parameter can be used
-       to restricts the marked vDOFs to the specified component. */
+       to restricts the marked vDOFs to the specified component.
+       If overwrite is set to false then values in ess_vdofs are preserved
+       and not reset which allows the accumulation of multiple DOFs into a single array.
+       However, the assumption here is that ess_vdofs is set to
+       the correct size already.*/
    virtual void GetEssentialVDofs(const Array<int> &bdr_attr_is_ess,
                                   Array<int> &ess_vdofs,
-                                  int component = -1) const;
+                                  int component = -1,
+                                  bool overwrite = true) const;
 
    /** @brief Get a list of essential true dofs, ess_tdof_list, corresponding to the
        boundary attributes marked in the array bdr_attr_is_ess.
@@ -1370,6 +1399,20 @@ public:
                                      Array<int> &ess_tdof_list,
                                      int component = -1) const;
 
+
+   /** @brief Get a list of essential true dofs, ess_tdof_list, corresponding to the
+       boundary attributes marked in the array bdr_attr_is_ess.
+
+       For spaces with 'vdim' > 1, the 'component' array can be used
+       to restricts the marked tDOFs per boundary to the specified components.
+       If vdim > 1 then one can specify per boundary attribute which components
+       on a boundary are essential by assigning a value of true to its location
+       in the component array.
+       The component has dimensions number of boundary attributes x vdim. */
+   virtual void GetEssentialTrueDofs(const Array<int> &bdr_attr_is_ess,
+                                     Array<int> &ess_tdof_list,
+                                     const Array2D<bool> &component);
+
    /** @brief Get a list of all boundary true dofs, @a boundary_dofs. For spaces
        with 'vdim' > 1, the 'component' parameter can be used to restricts the
        marked tDOFs to the specified component. Equivalent to

fem/gridfunc.cpp

@@ -3030,14 +3030,10 @@ void GridFunction::ProjectCoefficient(Coefficient *coeff[])
    }
 }
 
-void GridFunction::ProjectDiscCoefficient(
-   std::variant<Coefficient*, VectorCoefficient*> coeff, Array<int> &dof_attr)
+void GridFunction::ProjectDiscCoefficient(VectorCoefficient &coeff,
+                                          Array<int> &dof_attr)
 {
-   std::visit([&](auto* c)
-   {
-      MFEM_VERIFY(VectorDim() == c->GetVDim(), "coeff vdim != VectorDim()");
-   }, coeff);
-
+   MFEM_VERIFY(VectorDim() == coeff.GetVDim(), "coeff vdim != VectorDim()");
    Array<int> vdofs;
    Vector vals;
 
@@ -3051,10 +3047,7 @@ void GridFunction::ProjectDiscCoefficient(
    {
       fes->GetElementVDofs(i, vdofs);
       vals.SetSize(vdofs.Size());
-      std::visit([&](auto* c)
-      {
-         fes->GetFE(i)->Project(*c, *fes->GetElementTransformation(i), vals);
-      }, coeff);
+      fes->GetFE(i)->Project(coeff, *fes->GetElementTransformation(i), vals);
 
       // the values in shared dofs are determined from the element with maximal
       // attribute
@@ -3070,6 +3063,13 @@ void GridFunction::ProjectDiscCoefficient(
    }
 }
 
+void GridFunction::ProjectDiscCoefficient(VectorCoefficient &coeff)
+{
+   MFEM_VERIFY(VectorDim() == coeff.GetVDim(), "coeff vdim != VectorDim()");
+   Array<int> dof_attr;
+   ProjectDiscCoefficient(coeff, dof_attr);
+}
+
 void GridFunction::ProjectDiscCoefficient(Coefficient &coeff, AvgType type)
 {
    // Harmonic  (x1 ... xn) = [ (1/x1 + ... + 1/xn) / n ]^-1.
@@ -5252,30 +5252,6 @@ void GridFunction::GetElementBounds(const PLBound &plb,
                                     Vector &lower, Vector &upper,
                                     const int vdim) const
 {
-   if (UseDevice() && Device::Allows(Backend::DEVICE_MASK) &&
-       plb.GetBasisType() != BasisType::Positive &&
-       UsesTensorBasis(*fes))
-   {
-      const FiniteElement &fe = *fes->GetTypicalFE();
-      const int rdim = fe.GetDim();
-      const int fes_dim = fes->GetVDim();
-      const int nel = fes->GetNE();
-      const int nd = fe.GetDof();
-
-      Vector e_vec(nd*fes_dim*nel, Device::GetDeviceMemoryType());
-      e_vec.UseDevice(true);
-      const ElementRestrictionOperator *elem_restr =
-         fes->GetElementRestriction(ElementDofOrdering::LEXICOGRAPHIC);
-      MFEM_VERIFY(elem_restr != nullptr,
-                  "Element restriction is required for device bounds.");
-      elem_restr->Mult(*this, e_vec);
-
-      plb.GetElementBoundsKernel(rdim, fes_dim, e_vec, lower, upper, vdim);
-      lower.HostRead();
-      upper.HostRead();
-      return;
-   }
-
    int nel  = fes->GetNE();
    int fes_dim  = fes->GetVDim();
    lower.SetSize(nel*(vdim > 0 ? 1 :fes_dim));

fem/gridfunc.hpp

@@ -23,7 +23,6 @@
 #include <limits>
 #include <ostream>
 #include <string>
-#include <variant>
 
 namespace mfem
 {
@@ -80,18 +79,10 @@ protected:
                         bool wcoef,
                         int subdomain);
 
-   /** @brief Project a discontinuous (vector) coefficient as a grid function on
-       a continuous finite element space. Return in dof_attr the maximal
-       attribute of the elements containing each degree of freedom. */
-   virtual void ProjectDiscCoefficient(
-      std::variant<Coefficient*, VectorCoefficient*> coeff, Array<int> &dof_attr);
-
-   /** @brief Project a discontinuous (vector) coefficient as a grid function on
-       a continuous finite element space. The values in shared dofs are
-       determined from the element with maximal attribute. */
-   virtual void ProjectDiscCoefficient(
-      std::variant<Coefficient*, VectorCoefficient*> coeff)
-   { Array<int> dof_attr; ProjectDiscCoefficient(coeff, dof_attr); };
+   /** Project a discontinuous vector coefficient in a continuous space and
+       return in dof_attr the maximal attribute of the elements containing each
+       degree of freedom. */
+   void ProjectDiscCoefficient(VectorCoefficient &coeff, Array<int> &dof_attr);
 
    /** Helper function for ProjectCoefficientElementL2 */
    void ProjectCoefficientElementL2_(Coefficient &coeff, Vector &sol, Vector &Va);
@@ -524,17 +515,10 @@ public:
        but using an array of scalar coefficients for each component. */
    void ProjectCoefficient(Coefficient *coeff[]);
 
-   /** @brief Project a discontinuous coefficient as a grid function on
-       a continuous finite element space. The values in shared dofs are
-       determined from the element with maximal attribute. */
-   virtual void ProjectDiscCoefficient(Coefficient &coeff)
-   { ProjectDiscCoefficient(&coeff); }
-
    /** @brief Project a discontinuous vector coefficient as a grid function on
        a continuous finite element space. The values in shared dofs are
        determined from the element with maximal attribute. */
-   virtual void ProjectDiscCoefficient(VectorCoefficient &coeff)
-   { ProjectDiscCoefficient(&coeff); }
+   virtual void ProjectDiscCoefficient(VectorCoefficient &coeff);
 
    enum AvgType {ARITHMETIC, HARMONIC};
    /** @brief Projects a discontinuous coefficient so that the values in shared

fem/gslib.cpp

@@ -106,7 +106,9 @@ FindPointsGSLIB::FindPointsGSLIB()
    gsl_comm = new gslib::comm;
    cr       = new gslib::crystal;
 #ifdef MFEM_USE_MPI
-   if (!Mpi::IsInitialized()) { Mpi::Init(); }
+   int initialized = 0;
+   MPI_Initialized(&initialized);
+   if (!initialized) { MPI_Init(NULL, NULL); }
    MPI_Comm comm = MPI_COMM_WORLD;
    comm_init(gsl_comm, comm);
 #else
@@ -488,7 +490,7 @@ void FindPointsGSLIB::FindPointsOnDevice(const Vector &point_pos,
    }
    DEV.find_device = true;
 
-   const unsigned int id = gsl_comm->id, np = gsl_comm->np;
+   const int id = gsl_comm->id, np = gsl_comm->np;
 
    gsl_mfem_ref.SetSize(points_cnt * dim);
    gsl_mfem_elem.SetSize(points_cnt);
@@ -650,7 +652,7 @@ void FindPointsGSLIB::FindPointsOnDevice(const Vector &point_pos,
          {
             const int pp = hash_offset[i];
             /* don't send back to where it just came from */
-            if (static_cast<unsigned>(pp) == p->proc)
+            if (pp == p->proc)
             {
                continue;
             }
@@ -1066,7 +1068,7 @@ void FindPointsGSLIB::InterpolateOnDevice(const Vector &field_in_evec,
          sarray_transfer(struct evalOutPt_t, &outpt, proc, 1, cr);
 
          opt = (evalOutPt_t *)outpt.ptr;
-         for (size_t index = 0; index < outpt.n; index++)
+         for (int index = 0; index < outpt.n; index++)
          {
             int idx = ordering == Ordering::byNODES ?
                       opt->index + i*points_cnt :
@@ -1411,7 +1413,7 @@ void FindPointsGSLIB::SetupSplitMeshesAndIntegrationRules(const int order)
 {
    MFEM_VERIFY(mesh, "Setup FindPointsGSLIB with mesh first.");
    const int dof1D = order+1;
-   dim = mesh->Dimension();
+   const int dim = mesh->Dimension();
 
    SetupSplitMeshes();
    if (dim == 2)
@@ -2252,8 +2254,7 @@ void FindPointsGSLIB::DistributeInterpolatedValues(const Vector &int_vals,
       sarray_transfer(struct out_pt, outpt, proc, 1, cr);
 
       // Store received data
-      MFEM_VERIFY(outpt->n == static_cast<size_t>(points_cnt),
-                  "Incompatible size. Number of points "
+      MFEM_VERIFY(outpt->n == points_cnt, "Incompatible size. Number of points "
                   "received does not match the number of points originally "
                   "found using FindPoints.");
 
@@ -2622,7 +2623,9 @@ GSOPGSLIB::GSOPGSLIB(Array<long long> &ids)
    gsl_comm = new gslib::comm;
    cr       = new gslib::crystal;
 #ifdef MFEM_USE_MPI
-   if (!Mpi::IsInitialized()) { Mpi::Init(); }
+   int initialized;
+   MPI_Initialized(&initialized);
+   if (!initialized) { MPI_Init(NULL, NULL); }
    MPI_Comm comm = MPI_COMM_WORLD;
    comm_init(gsl_comm, comm);
 #else

fem/gslib/findpts_local_2.cpp

@@ -254,7 +254,7 @@ get_edge(const double *elx[2], const double *wtend, int ei,
       edge.dxdn[d] = workspace + (2 + d) * pN; //dxdn and dydn at DOFs along edge
    }
 
-   if (static_cast<unsigned>(side_init) != (1u << ei))
+   if (side_init != (1u << ei))
    {
 #define ELX(d, j, k) elx[d][j + k * pN] // assumes lexicographic ordering
       for (int d = 0; d < 2; ++d)

fem/gslib/findpts_local_3.cpp

@@ -294,7 +294,7 @@ get_face(const double *elx[3], const double *wtend, int fi, double *workspace,
       face.dxdn[d] = workspace+(3+d)*p_Nfr;
    }
 
-   if (static_cast<unsigned>(side_init) != (1u << fi))
+   if (side_init != (1u << fi))
    {
       const int e_stride[3] = {1, pN, pN*pN};
 #define ELX(d, j, k, l) elx[d][j*e_stride[d1]+k*e_stride[d2]+l*e_stride[dn]]
@@ -342,7 +342,7 @@ get_edge(const double *elx[3], const double *wtend, int ei, double *workspace,
 
    if (jidx >= 3*pN) { return edge; }
 
-   if (static_cast<unsigned>(side_init) != (64u << ei))
+   if (side_init != (64u << ei))
    {
       const int e_stride[3] = {1, pN, pN*pN};
 #define ELX(d, j, k, l) elx[d][j*e_stride[de]+k*e_stride[dn1]+l*e_stride[dn2]]

fem/integ/bilininteg_mass_kernels.hpp

@@ -181,12 +181,6 @@ constexpr int NBZ(int D1D)
 {
    return ipow(2, D(D1D) >= 0 ? D(D1D) : 0);
 }
-constexpr int NBZ3D(int MDQ)
-{
-   return MDQ > 0 ? std::min<int>(
-             (128 + MDQ * MDQ * MDQ - 1) / (MDQ * MDQ * MDQ), 64)
-          : 1;
-}
 }
 
 // Shared memory PA Mass Diagonal 2D kernel
@@ -810,23 +804,19 @@ void PAMassApply3D_Element(const int e,
    }
 }
 
-template <int T_D1D, int T_Q1D, int TBATCH, bool ACCUMULATE = true>
-MFEM_HOST_DEVICE inline void
-SmemPAMassApply3D_Element(const int e, const int NE, const real_t *b_,
-                          const real_t *d_, const real_t *x_, real_t *y_,
-                          int d1d = 0, int q1d = 0)
+template<int T_D1D, int T_Q1D, bool ACCUMULATE = true>
+MFEM_HOST_DEVICE inline
+void SmemPAMassApply3D_Element(const int e,
+                               const int NE,
+                               const real_t *b_,
+                               const real_t *d_,
+                               const real_t *x_,
+                               real_t *y_,
+                               const int d1d = 0,
+                               const int q1d = 0)
 {
-   static_assert(TBATCH > 0, "TBATCH must be positive");
-#if defined(__CUDA_ARCH__) || defined(__HIP_DEVICE_COMPILE__)
-   constexpr int tbatch = TBATCH;
-   const int tidz = MFEM_THREAD_ID(z);
-#else
-   // host always batch size 1
-   constexpr int tbatch = 1;
-   constexpr int tidz = 0;
-#endif
-   const int D1D = T_D1D ? T_D1D : d1d;
-   const int Q1D = T_Q1D ? T_Q1D : q1d;
+   constexpr int D1D = T_D1D ? T_D1D : d1d;
+   constexpr int Q1D = T_Q1D ? T_Q1D : q1d;
    constexpr int MQ1 = T_Q1D ? T_Q1D : DofQuadLimits::MAX_Q1D;
    constexpr int MD1 = T_D1D ? T_D1D : DofQuadLimits::MAX_D1D;
    constexpr int MDQ = (MQ1 > MD1) ? MQ1 : MD1;
@@ -839,37 +829,33 @@ SmemPAMassApply3D_Element(const int e, const int NE, const real_t *b_,
    MFEM_SHARED real_t sDQ[MQ1*MD1];
    real_t (*B)[MD1] = (real_t (*)[MD1]) sDQ;
    real_t (*Bt)[MQ1] = (real_t (*)[MQ1]) sDQ;
-   MFEM_SHARED real_t sm0[tbatch][MDQ*MDQ*MDQ];
-   MFEM_SHARED real_t sm1[tbatch][MDQ*MDQ*MDQ];
-   real_t (*X)[MD1][MD1]   = (real_t (*)[MD1][MD1]) (sm0+tidz);
-   real_t (*DDQ)[MD1][MQ1] = (real_t (*)[MD1][MQ1]) (sm1+tidz);
-   real_t (*DQQ)[MQ1][MQ1] = (real_t (*)[MQ1][MQ1]) (sm0+tidz);
-   real_t (*QQQ)[MQ1][MQ1] = (real_t (*)[MQ1][MQ1]) (sm1+tidz);
-   real_t (*QQD)[MQ1][MD1] = (real_t (*)[MQ1][MD1]) (sm0+tidz);
-   real_t (*QDD)[MD1][MD1] = (real_t (*)[MD1][MD1]) (sm1+tidz);
-   MFEM_FOREACH_THREAD(dy, y, D1D)
+   MFEM_SHARED real_t sm0[MDQ*MDQ*MDQ];
+   MFEM_SHARED real_t sm1[MDQ*MDQ*MDQ];
+   real_t (*X)[MD1][MD1]   = (real_t (*)[MD1][MD1]) sm0;
+   real_t (*DDQ)[MD1][MQ1] = (real_t (*)[MD1][MQ1]) sm1;
+   real_t (*DQQ)[MQ1][MQ1] = (real_t (*)[MQ1][MQ1]) sm0;
+   real_t (*QQQ)[MQ1][MQ1] = (real_t (*)[MQ1][MQ1]) sm1;
+   real_t (*QQD)[MQ1][MD1] = (real_t (*)[MQ1][MD1]) sm0;
+   real_t (*QDD)[MD1][MD1] = (real_t (*)[MD1][MD1]) sm1;
+   MFEM_FOREACH_THREAD(dy,y,D1D)
    {
-      MFEM_FOREACH_THREAD(dx, x, D1D)
+      MFEM_FOREACH_THREAD(dx,x,D1D)
       {
          MFEM_UNROLL(MD1)
          for (int dz = 0; dz < D1D; ++dz)
          {
-            X[dz][dy][dx] = x(dx, dy, dz, e);
+            X[dz][dy][dx] = x(dx,dy,dz,e);
          }
       }
-      MFEM_FOREACH_THREAD(dx, x, Q1D) { B[dx][dy] = b(dx, dy); }
-   }
-   if (tidz == 0)
-   {
-      MFEM_FOREACH_THREAD(dy, y, D1D)
+      MFEM_FOREACH_THREAD(dx,x,Q1D)
       {
-         MFEM_FOREACH_THREAD(dx, x, Q1D) { B[dx][dy] = b(dx, dy); }
+         B[dx][dy] = b(dx,dy);
       }
    }
    MFEM_SYNC_THREAD;
-   MFEM_FOREACH_THREAD(dy, y, D1D)
+   MFEM_FOREACH_THREAD(dy,y,D1D)
    {
-      MFEM_FOREACH_THREAD(qx, x, Q1D)
+      MFEM_FOREACH_THREAD(qx,x,Q1D)
       {
          real_t u[D1D];
          MFEM_UNROLL(MD1)
@@ -894,9 +880,9 @@ SmemPAMassApply3D_Element(const int e, const int NE, const real_t *b_,
       }
    }
    MFEM_SYNC_THREAD;
-   MFEM_FOREACH_THREAD(qy, y, Q1D)
+   MFEM_FOREACH_THREAD(qy,y,Q1D)
    {
-      MFEM_FOREACH_THREAD(qx, x, Q1D)
+      MFEM_FOREACH_THREAD(qx,x,Q1D)
       {
          real_t u[D1D];
          MFEM_UNROLL(MD1)
@@ -921,9 +907,9 @@ SmemPAMassApply3D_Element(const int e, const int NE, const real_t *b_,
       }
    }
    MFEM_SYNC_THREAD;
-   MFEM_FOREACH_THREAD(qy, y, Q1D)
+   MFEM_FOREACH_THREAD(qy,y,Q1D)
    {
-      MFEM_FOREACH_THREAD(qx, x, Q1D)
+      MFEM_FOREACH_THREAD(qx,x,Q1D)
       {
          real_t u[Q1D];
          MFEM_UNROLL(MQ1)
@@ -943,22 +929,22 @@ SmemPAMassApply3D_Element(const int e, const int NE, const real_t *b_,
          MFEM_UNROLL(MQ1)
          for (int qz = 0; qz < Q1D; qz++)
          {
-            QQQ[qz][qy][qx] = u[qz] * d(qx, qy, qz, e);
+            QQQ[qz][qy][qx] = u[qz] * d(qx,qy,qz,e);
          }
       }
    }
    MFEM_SYNC_THREAD;
-   if (tidz == 0)
+   MFEM_FOREACH_THREAD(di,y,D1D)
    {
-      MFEM_FOREACH_THREAD(di, y, D1D)
+      MFEM_FOREACH_THREAD(q,x,Q1D)
       {
-         MFEM_FOREACH_THREAD(q, x, Q1D) { Bt[di][q] = b(q, di); }
+         Bt[di][q] = b(q,di);
       }
    }
    MFEM_SYNC_THREAD;
-   MFEM_FOREACH_THREAD(qy, y, Q1D)
+   MFEM_FOREACH_THREAD(qy,y,Q1D)
    {
-      MFEM_FOREACH_THREAD(dx, x, D1D)
+      MFEM_FOREACH_THREAD(dx,x,D1D)
       {
          real_t u[Q1D];
          MFEM_UNROLL(MQ1)
@@ -983,9 +969,9 @@ SmemPAMassApply3D_Element(const int e, const int NE, const real_t *b_,
       }
    }
    MFEM_SYNC_THREAD;
-   MFEM_FOREACH_THREAD(dy, y, D1D)
+   MFEM_FOREACH_THREAD(dy,y,D1D)
    {
-      MFEM_FOREACH_THREAD(dx, x, D1D)
+      MFEM_FOREACH_THREAD(dx,x,D1D)
       {
          real_t u[Q1D];
          MFEM_UNROLL(MQ1)
@@ -1010,9 +996,9 @@ SmemPAMassApply3D_Element(const int e, const int NE, const real_t *b_,
       }
    }
    MFEM_SYNC_THREAD;
-   MFEM_FOREACH_THREAD(dy, y, D1D)
+   MFEM_FOREACH_THREAD(dy,y,D1D)
    {
-      MFEM_FOREACH_THREAD(dx, x, D1D)
+      MFEM_FOREACH_THREAD(dx,x,D1D)
       {
          real_t u[D1D];
          MFEM_UNROLL(MD1)
@@ -1034,11 +1020,11 @@ SmemPAMassApply3D_Element(const int e, const int NE, const real_t *b_,
          {
             if (ACCUMULATE)
             {
-               y(dx, dy, dz, e) += u[dz];
+               y(dx,dy,dz,e) += u[dz];
             }
             else
             {
-               y(dx, dy, dz, e) = u[dz];
+               y(dx,dy,dz,e) = u[dz];
             }
          }
       }
@@ -1129,8 +1115,8 @@ inline void PAMassApply3D(const int NE,
    });
 }
 
-// Shared memory PA Mass Apply 3D kernel
-template<int T_D1D = 0, int T_Q1D = 0, int TBATCH=1>
+// Shared memory PA Mass Apply 2D kernel
+template<int T_D1D = 0, int T_Q1D = 0>
 inline void SmemPAMassApply3D(const int NE,
                               const Array<real_t> &b_,
                               const Array<real_t> &bt_,
@@ -1140,9 +1126,6 @@ inline void SmemPAMassApply3D(const int NE,
                               const int d1d = 0,
                               const int q1d = 0)
 {
-   static_assert(T_D1D > 0, "T_D1D must be positive");
-   static_assert(T_Q1D > 0, "T_Q1D must be positive");
-   static_assert(TBATCH > 0, "TBATCH must be positive");
    MFEM_CONTRACT_VAR(bt_);
    const int D1D = T_D1D ? T_D1D : d1d;
    const int Q1D = T_Q1D ? T_Q1D : q1d;
@@ -1154,11 +1137,9 @@ inline void SmemPAMassApply3D(const int NE,
    const auto d = d_.Read();
    const auto x = x_.Read();
    auto y = y_.ReadWrite();
-   mfem::forall_2D_batch<T_Q1D * T_Q1D * TBATCH>(NE, Q1D, Q1D, TBATCH,
-                                                 [=] MFEM_HOST_DEVICE(int e)
+   mfem::forall_2D<T_Q1D*T_Q1D>(NE, Q1D, Q1D, [=] MFEM_HOST_DEVICE (int e)
    {
-      internal::SmemPAMassApply3D_Element<T_D1D, T_Q1D, TBATCH>(e, NE, b, d, x,
-                                                                y, d1d, q1d);
+      internal::SmemPAMassApply3D_Element<T_D1D,T_Q1D>(e, NE, b, d, x, y, d1d, q1d);
    });
 }
 
@@ -1413,16 +1394,7 @@ ApplyKernelType MassIntegrator::ApplyPAKernels::Kernel()
 {
    if constexpr (DIM == 1) { return internal::PAMassApply1D; }
    else if constexpr (DIM == 2) { return internal::SmemPAMassApply2D<T_D1D,T_Q1D>; }
-   else if constexpr (DIM == 3)
-   {
-      constexpr int MDQ = T_D1D >= T_Q1D ? T_D1D : T_Q1D;
-      // max 64 threads in z limit in cuda and hip
-      if constexpr (MDQ > 0)
-      {
-         return internal::SmemPAMassApply3D<T_D1D, T_Q1D,
-                internal::mass::NBZ3D(MDQ)>;
-      }
-   }
+   else if constexpr (DIM == 3) { return internal::SmemPAMassApply3D<T_D1D, T_Q1D>; }
    MFEM_ABORT("");
 }
 

fem/intrules.hpp

@@ -43,52 +43,56 @@ public:
       index = i;
    }
 
-   void Set3w(const real_t x1, const real_t x2, const real_t x3, const real_t w)
-   { x = x1; y = x2; z = x3; weight = w; }
-   void Set2w(const real_t x1, const real_t x2, const real_t w)
-   { x = x1; y = x2; weight = w; }
-   void Set1w(const real_t x1, const real_t w)
-   { x = x1; weight = w; }
-
-   void Set3w(const real_t *p) { Set3w(p[0], p[1], p[2], p[3]); }
-   void Set2w(const real_t *p) { Set2w(p[0], p[1], p[2]); }
-   void Set1w(const real_t *p) { Set1w(p[0], p[1]); }
-
-   void Set3(const real_t x1, const real_t x2, const real_t x3)
-   { x = x1; y = x2; z = x3; }
-   void Set2(const real_t x1, const real_t x2)
-   { x = x1; y = x2; }
-   void Set1(const real_t x1)
-   { x = x1; }
-
-   void Set3(const real_t *p) { Set3(p[0], p[1], p[2]); }
-   void Set2(const real_t *p) { Set2(p[0], p[1]); }
-   void Set1(const real_t *p) { Set1(p[0]); }
-
-   void Set(const real_t x1, const real_t x2, const real_t x3, const real_t w)
-   { Set3w(x1, x2, x3, w); }
-
    void Set(const real_t *p, const int dim)
    {
       MFEM_ASSERT(1 <= dim && dim <= 3, "invalid dim: " << dim);
-      switch (dim)
+      x = p[0];
+      if (dim > 1)
       {
-         case 3: Set3(p); break;
-         case 2: Set2(p); break;
-         case 1: Set1(p); break;
+         y = p[1];
+         if (dim > 2)
+         {
+            z = p[2];
+         }
       }
    }
 
    void Get(real_t *p, const int dim) const
    {
       MFEM_ASSERT(1 <= dim && dim <= 3, "invalid dim: " << dim);
-      switch (dim)
+      p[0] = x;
+      if (dim > 1)
       {
-         case 3: p[2] = z;
-         case 2: p[1] = y;
-         case 1: p[0] = x;
+         p[1] = y;
+         if (dim > 2)
+         {
+            p[2] = z;
+         }
       }
    }
+
+   void Set(const real_t x1, const real_t x2, const real_t x3, const real_t w)
+   { x = x1; y = x2; z = x3; weight = w; }
+
+   void Set3w(const real_t *p) { x = p[0]; y = p[1]; z = p[2]; weight = p[3]; }
+
+   void Set3(const real_t x1, const real_t x2, const real_t x3)
+   { x = x1; y = x2; z = x3; }
+
+   void Set3(const real_t *p) { x = p[0]; y = p[1]; z = p[2]; }
+
+   void Set2w(const real_t x1, const real_t x2, const real_t w)
+   { x = x1; y = x2; weight = w; }
+
+   void Set2w(const real_t *p) { x = p[0]; y = p[1]; weight = p[2]; }
+
+   void Set2(const real_t x1, const real_t x2) { x = x1; y = x2; }
+
+   void Set2(const real_t *p) { x = p[0]; y = p[1]; }
+
+   void Set1w(const real_t x1, const real_t w) { x = x1; weight = w; }
+
+   void Set1w(const real_t *p) { x = p[0]; weight = p[1]; }
 };
 
 /// Class for an integration rule - an Array of IntegrationPoint.

fem/lininteg.hpp

@@ -164,8 +164,8 @@ private:
 
 public:
    /// Constructs the domain integrator $ (Q, \nabla v) $
-   DomainLFGradIntegrator(VectorCoefficient &QF, const IntegrationRule *ir = NULL)
-      : DeltaLFIntegrator(QF, ir), Q(QF) { }
+   DomainLFGradIntegrator(VectorCoefficient &QF)
+      : DeltaLFIntegrator(QF), Q(QF) { }
 
    bool SupportsDevice() const override { return true; }
 

fem/nonlinearform.cpp

@@ -53,6 +53,22 @@ void NonlinearForm::SetEssentialBC(const Array<int> &bdr_attr_is_ess,
    }
 }
 
+void NonlinearForm::SetEssentialBC(const Array<int> &bdr_attr_is_ess,
+                                   const Array2D<bool> &bdr_component,
+                                   Vector *rhs)
+{
+   // virtual call, works in parallel too
+   fes->GetEssentialTrueDofs(bdr_attr_is_ess, ess_tdof_list, bdr_component);
+
+   if (rhs)
+   {
+      for (int i = 0; i < ess_tdof_list.Size(); i++)
+      {
+         (*rhs)(ess_tdof_list[i]) = 0.0;
+      }
+   }
+}
+
 void NonlinearForm::SetEssentialVDofs(const Array<int> &ess_vdofs_list)
 {
    if (!P)

fem/nonlinearform.hpp

@@ -165,14 +165,39 @@ public:
    const Array<NonlinearFormIntegrator*> &GetBdrFaceIntegrators() const
    { return bfnfi; }
 
-   /// Specify essential boundary conditions.
-   /** This method calls FiniteElementSpace::GetEssentialTrueDofs() and stores
+   /** @brief Specify essential boundary conditions.
+
+       This method calls FiniteElementSpace::GetEssentialTrueDofs() and stores
        the result internally for use by other methods. If the @a rhs pointer is
        not NULL, its essential true dofs will be set to zero. This makes it
        "compatible" with the output vectors from the Mult() method which also
-       have zero entries at the essential true dofs. */
+       have zero entries at the essential true dofs.
+
+       @note The values in the essential vdofs have to come from the initial guess.
+       */
    void SetEssentialBC(const Array<int> &bdr_attr_is_ess, Vector *rhs = NULL);
 
+   /** @brief Specify essential boundary conditions.
+
+       For spaces with 'vdim' > 1, the 'bdr_component' array can be used
+       to restricts the marked tDOFs per boundary to the specified components.
+       If vdim > 1 then one can specify per boundary attribute which components
+       on a boundary are essential by assigning a value of true to its location
+       in the bdr_component array.
+       The bdr_component has dimensions number of boundary attributes x vdim
+
+       This method calls FiniteElementSpace::GetEssentialTrueDofs() and stores
+       the result internally for use by other methods. If the @a rhs pointer is
+       not NULL, its essential true dofs will be set to zero. This makes it
+       "compatible" with the output vectors from the Mult() method which also
+       have zero entries at the essential true dofs.
+
+       @note The values in the essential vdofs have to come from the initial guess.
+       */
+   void SetEssentialBC(const Array<int> &bdr_attr_is_ess,
+                       const Array2D<bool> &bdr_component,
+                       Vector *rhs);
+
    /// Specify essential boundary conditions.
    /** Use either SetEssentialBC() or SetEssentialTrueDofs() if possible. */
    void SetEssentialVDofs(const Array<int> &ess_vdofs_list);

fem/pfespace.cpp

@@ -1122,9 +1122,11 @@ void ParFiniteElementSpace::Synchronize(Array<int> &ldof_marker) const
 
 void ParFiniteElementSpace::GetEssentialVDofs(const Array<int> &bdr_attr_is_ess,
                                               Array<int> &ess_dofs,
-                                              int component) const
+                                              int component,
+                                              bool overwrite) const
 {
-   FiniteElementSpace::GetEssentialVDofs(bdr_attr_is_ess, ess_dofs, component);
+   FiniteElementSpace::GetEssentialVDofs(bdr_attr_is_ess, ess_dofs, component,
+                                         overwrite);
 
    // Make sure that processors without boundary elements mark
    // their boundary dofs (if they have any).
@@ -1184,6 +1186,39 @@ void ParFiniteElementSpace::GetEssentialTrueDofs(const Array<int>
    MarkerToList(true_ess_dofs, ess_tdof_list);
 }
 
+void ParFiniteElementSpace::GetEssentialTrueDofs(const Array<int>
+                                                 &bdr_attr_is_ess,
+                                                 Array<int> &ess_tdof_list,
+                                                 const Array2D<bool> &component)
+{
+   MFEM_VERIFY(!IsVariableOrderH1(),
+               "Variable order H1 spaces are currently not supported with this feature");
+
+   Array<int> ess_dofs, true_ess_dofs;
+
+   GetEssentialVDofsFromComponent(bdr_attr_is_ess, component, ess_dofs);
+   GetRestrictionMatrix()->BooleanMult(ess_dofs, true_ess_dofs);
+
+#ifdef MFEM_DEBUG
+   // Verify that in boolean arithmetic: P^T ess_dofs = R ess_dofs.
+   Array<int> true_ess_dofs2(true_ess_dofs.Size());
+   HypreParMatrix *Pt = Dof_TrueDof_Matrix()->Transpose();
+   const int *ess_dofs_data = ess_dofs.HostRead();
+   Pt->BooleanMult(1, ess_dofs_data, 0, true_ess_dofs2);
+   delete Pt;
+   int counter = 0;
+   const int *ted = true_ess_dofs.HostRead();
+   for (int i = 0; i < true_ess_dofs.Size(); i++)
+   {
+      if (bool(ted[i]) != bool(true_ess_dofs2[i])) { counter++; }
+   }
+   MFEM_VERIFY(counter == 0, "internal MFEM error: counter = " << counter
+               << ", rank = " << MyRank);
+#endif
+
+   MarkerToList(true_ess_dofs, ess_tdof_list);
+}
+
 void ParFiniteElementSpace::GetEssentialTrueDofsVar(const Array<int>
                                                     &bdr_attr_is_ess,
                                                     const Array<int> &ess_dofs,

fem/pfespace.hpp

@@ -420,10 +420,18 @@ public:
        "partially conforming") space. */
    void Synchronize(Array<int> &ldof_marker) const;
 
-   /// Determine the boundary degrees of freedom
-   void GetEssentialVDofs(const Array<int> &bdr_attr_is_ess,
-                          Array<int> &ess_dofs,
-                          int component = -1) const override;
+   /** @brief Mark degrees of freedom associated with boundary elements with
+       the specified boundary attributes (marked in 'bdr_attr_is_ess').
+
+       For spaces with 'vdim' > 1, the 'component' parameter can be used
+       to restricts the marked vDOFs to the specified component.
+       If overwrite is set to false then values in ess_vdofs are preserved
+       and not reset. However, the assumption here is that ess_vdofs is set to
+       the correct size already.*/
+   virtual void GetEssentialVDofs(const Array<int> &bdr_attr_is_ess,
+                                  Array<int> &ess_dofs,
+                                  int component = -1,
+                                  bool overwrite = true) const override;
 
    /** Get a list of essential true dofs, ess_tdof_list, corresponding to the
        boundary attributes marked in the array bdr_attr_is_ess. */
@@ -448,6 +456,19 @@ public:
    void GetExteriorTrueDofs(Array<int> &ext_tdof_list,
                             int component = -1) const override;
 
+   /** @brief Get a list of essential true dofs, ess_tdof_list, corresponding to the
+       boundary attributes marked in the array bdr_attr_is_ess.
+
+       For spaces with 'vdim' > 1, the 'component' array can be used
+       to restricts the marked tDOFs per boundary to the specified components.
+       If vdim > 1 then one can specify per boundary attribute which components
+       on a boundary are essential by assigning a value of true to its location
+       in the component array.
+       The component has dimensions number of boundary attributes x vdim. */
+   virtual void GetEssentialTrueDofs(const Array<int> &bdr_attr_is_ess,
+                                     Array<int> &ess_tdof_list,
+                                     const Array2D<bool> &component) override;
+
    /** If the given ldof is owned by the current processor, return its local
        tdof number, otherwise return -1 */
    int GetLocalTDofNumber(int ldof) const;

fem/pgridfunc.cpp

@@ -717,9 +717,9 @@ void ParGridFunction::ProjectCoefficientElementL2(VectorCoefficient &vcoeff)
 }
 
 
-void ParGridFunction::ProjectDiscCoefficient(
-   std::variant<Coefficient*, VectorCoefficient*> coeff)
+void ParGridFunction::ProjectDiscCoefficient(VectorCoefficient &coeff)
 {
+   MFEM_VERIFY(VectorDim() == coeff.GetVDim(), "coeff vdim != VectorDim()");
    // local maximal element attribute for each dof
    Array<int> ldof_attr;
 

fem/pgridfunc.hpp

@@ -63,12 +63,6 @@ protected:
    void ProjectBdrCoefficient(Coefficient *coeff[], VectorCoefficient *vcoeff,
                               const Array<int> &attr);
 
-   /** @brief Project a discontinuous (vector) coefficient as a grid function on
-       a continuous finite element space. The values in shared dofs are
-       determined from the element with maximal attribute. */
-   virtual void ProjectDiscCoefficient(
-      std::variant<Coefficient*, VectorCoefficient*> coeff) override;
-
 public:
    ParGridFunction() { pfes = NULL; }
 
@@ -274,6 +268,11 @@ public:
                            ProjectType type = ProjectType::DEFAULT) override;
 
    using GridFunction::ProjectDiscCoefficient;
+   /** @brief Project a discontinuous vector coefficient as a grid function on
+       a continuous finite element space. The values in shared dofs are
+       determined from the element with maximal attribute. */
+   void ProjectDiscCoefficient(VectorCoefficient &coeff) override;
+
    void ProjectDiscCoefficient(Coefficient &coeff, AvgType type) override;
 
    void ProjectDiscCoefficient(VectorCoefficient &vcoeff, AvgType type) override;

general/forall.hpp

@@ -1090,12 +1090,6 @@ inline void forall_2D_batch(int N, int X, int Y, int BZ, lambda &&body)
    ForallWrap<2>(true, N, body, X, Y, BZ);
 }
 
-template<int MAX_THREADS_PER_BLOCK, typename lambda>
-inline void forall_2D_batch(int N, int X, int Y, int BZ, lambda &&body)
-{
-   ForallWrap<2, MAX_THREADS_PER_BLOCK>(true, N, body, X, Y, BZ);
-}
-
 template<typename lambda>
 inline void forall_3D(int N, int X, int Y, int Z, lambda &&body)
 {

makefile

@@ -123,20 +123,15 @@ EXAMPLE_SUBDIRS = amgx caliper ginkgo hiop petsc pumi sundials superlu moonolith
 EXAMPLE_DIRS := examples $(addprefix examples/,$(EXAMPLE_SUBDIRS))
 EXAMPLE_TEST_DIRS := examples
 
-MINIAPP_ALL_SUBDIRS = common electromagnetics meshing performance tools \
+MINIAPP_SUBDIRS = common electromagnetics meshing performance tools \
  toys nurbs gslib adjoint solvers shifted mtop parelag tribol autodiff dfem \
  hooke multidomain dpg hdiv-linear-solver spde diag-smoothers contact \
  fluids/navier fluids/schrodinger-flow plasma plasma/pic
-MINIAPP_RECURSIVE_SUBDIRS = plasma/pic
-MINIAPP_SUBDIRS := $(filter-out \
-   $(MINIAPP_RECURSIVE_SUBDIRS),$(MINIAPP_ALL_SUBDIRS))
-MINIAPP_ALL_DIRS := $(addprefix miniapps/,$(MINIAPP_ALL_SUBDIRS))
 MINIAPP_DIRS := $(addprefix miniapps/,$(MINIAPP_SUBDIRS))
 MINIAPP_TEST_DIRS := $(filter-out %/common,$(MINIAPP_DIRS))
 MINIAPP_USE_COMMON := $(addprefix miniapps/,electromagnetics meshing tools \
- toys gslib shifted dpg diag-smoothers fluids/navier plasma plasma/pic)
+ toys shifted dpg diag-smoothers fluids/navier plasma plasma/pic)
 
-EM_ALL_DIRS = $(EXAMPLE_DIRS) $(MINIAPP_ALL_DIRS)
 EM_DIRS = $(EXAMPLE_DIRS) $(MINIAPP_DIRS)
 
 TEST_SUBDIRS = unit
@@ -151,7 +146,7 @@ MFEM_BUILD_DIR ?= .
 BUILD_DIR := $(MFEM_BUILD_DIR)
 BUILD_REAL_DIR := $(abspath $(BUILD_DIR))
 ifneq ($(BUILD_REAL_DIR),$(MFEM_REAL_DIR))
-   BUILD_SUBDIRS = $(DIRS) config $(EM_ALL_DIRS) doc $(TEST_DIRS)
+   BUILD_SUBDIRS = $(DIRS) config $(EM_DIRS) doc $(TEST_DIRS)
    CONFIG_FILE_DEF = -DMFEM_CONFIG_FILE='"$(BUILD_REAL_DIR)/config/_config.hpp"'
    BLD := $(if $(BUILD_REAL_DIR:$(CURDIR)=),$(BUILD_DIR)/,)
    $(if $(word 2,$(BLD)),$(error Spaces in BLD = "$(BLD)" are not supported))
@@ -488,10 +483,10 @@ $(OBJECT_FILES): $(BLD)%.o: $(SRC)%.cpp $(CONFIG_MK)
 
 all: examples miniapps $(TEST_DIRS)
 
-.PHONY: miniapps $(EM_ALL_DIRS) $(TEST_DIRS)
+.PHONY: miniapps $(EM_DIRS) $(TEST_DIRS)
 miniapps: $(MINIAPP_DIRS)
 $(MINIAPP_USE_COMMON): miniapps/common
-$(EM_ALL_DIRS) $(TEST_DIRS): lib
+$(EM_DIRS) $(TEST_DIRS): lib
 	$(MAKE) -C $(BLD)$(@)
 
 .PHONY: doc
@@ -699,7 +694,7 @@ local-config:
 .PHONY: build-config
 build-config:
 	for d in $(BUILD_SUBDIRS); do mkdir -p $(BLD)$${d}; done
-	for dir in "" $(addsuffix /,config $(EM_ALL_DIRS) doc $(TEST_DIRS)); do\
+	for dir in "" $(addsuffix /,config $(EM_DIRS) doc $(TEST_DIRS)); do \
 	   printf "# Auto-generated file.\n%s\n%s\n" \
 	      "MFEM_DIR = $(MFEM_REAL_DIR)" \
 	      "include \$$(MFEM_DIR)/$${dir}makefile" \
@@ -801,15 +796,13 @@ status info:
 
 ASTYLE = $(ASTYLE_BIN) --options=$(SRC)config/mfem.astylerc
 ASTYLE_VER = "Artistic Style Version 3.1"
-FORMAT_FILES = $(foreach dir,$(DIRS) $(EM_ALL_DIRS) config,$(dir)/*.?pp)
+FORMAT_FILES = $(foreach dir,$(DIRS) $(EM_DIRS) config,$(dir)/*.?pp)
 TESTS_SUBDIRS = unit benchmarks convergence mem_manager par-mesh-format
-UNIT_TESTS_SUBDIRS = general linalg mesh fem miniapps ceed enzyme dfem
-MINIAPPS_SUBDIRS = dpg/util hooke/operators hooke/preconditioners \
-   hooke/materials hooke/kernels
+UNIT_TESTS_SUBDIRS = general linalg mesh fem miniapps ceed enzyme
+MINIAPPS_SUBDIRS = dpg/util hooke/operators hooke/preconditioners hooke/materials hooke/kernels
 FORMAT_FILES += $(foreach dir,$(TESTS_SUBDIRS),tests/$(dir)/*.?pp)
 FORMAT_FILES += $(foreach dir,$(UNIT_TESTS_SUBDIRS),tests/unit/$(dir)/*.?pp)
 FORMAT_FILES += $(foreach dir,$(MINIAPPS_SUBDIRS),miniapps/$(dir)/*.?pp)
-FORMAT_FILES += config/cmake/config.hpp.in config/config.hpp.in mfem*.hpp
 FORMAT_EXCLUDE = general/tinyxml2.cpp tests/unit/catch.hpp
 FORMAT_LIST = $(filter-out $(FORMAT_EXCLUDE),$(wildcard $(FORMAT_FILES)))
 
@@ -840,29 +833,14 @@ mfem_check_command = \
 # Verify the C++ code styling in MFEM and check that std::cout and std::cerr are
 # not used in the library (use mfem::out and mfem::err instead).
 style:
+	@echo "Applying C++ code style..."
 	@astyle_version="$$($(ASTYLE_BIN) --version)";\
 	 if [ "$$astyle_version" != $(ASTYLE_VER) ]; then\
 	    printf "%s\n" "Invalid astyle version: '$$astyle_version'"\
 	           "Please use: '"$(ASTYLE_VER)"'";\
 	    exit 1;\
 	 fi
-	@err_code=0; \
-	if command -v git 2>&1 > /dev/null && [ -d $(MFEM_DIR)/.git ]; then \
-	   echo "Checking if all git files are selected for formatting ..."; \
-	   ls -1 $(FORMAT_FILES) | sort > format-files-make.txt; \
-	   git -C $(MFEM_DIR) ls-files '*.[ch]pp*' | sort \
-	      > format-files-git.txt; \
-	   cat format-files-make.txt format-files-git.txt | sort | uniq \
-	      > format-files-make-plus-git.txt; \
-	   rm -f format-files-git.txt; \
-	   $(call mfem_check_command,\
-	      diff format-files-make.txt format-files-make-plus-git.txt | \
-	      grep "^> ",\
-	      "All git files are selected for formatting",\
-	      "The above git files are NOT selected for formatting"); \
-	   rm -f format-files-make.txt format-files-make-plus-git.txt; \
-	fi; \
-	echo "Applying C++ code style...";\
+	@err_code=0;\
 	$(call mfem_check_command,\
 	    $(ASTYLE) $(FORMAT_LIST) | grep Formatted,\
 	    "No source files were changed",\

mesh/mesh.cpp

@@ -1616,9 +1616,7 @@ Element::Type Mesh::GetFaceElementType(int Face) const
 
 Array<int> Mesh::GetFaceToBdrElMap() const
 {
-   Array<int> face_to_be(Dim == 1 ? NumOfVertices :
-                         Dim == 2 ? NumOfEdges :
-                         Dim == 3 ? NumOfFaces : 0);
+   Array<int> face_to_be(Dim == 2 ? NumOfEdges : NumOfFaces);
    face_to_be = -1;
    for (int i = 0; i < NumOfBdrElements; i++)
    {

mesh/mesh_operators.cpp

@@ -63,6 +63,7 @@ ThresholdRefiner::ThresholdRefiner(ErrorEstimator &est)
 
    threshold = 0.0;
    num_marked_elements = 0LL;
+   current_sequence = -1;
 
    non_conforming = -1;
    nc_limit = 0;
@@ -86,6 +87,7 @@ int ThresholdRefiner::MarkWithoutRefining(Mesh & mesh,
    threshold = 0.0;
    num_marked_elements = 0LL;
    refinements.SetSize(0);
+   current_sequence = mesh.GetSequence();
 
    const long long num_elements = mesh.GetGlobalNE();
    if (num_elements >= max_elements) { return STOP; }
@@ -147,6 +149,7 @@ int ThresholdRefiner::ApplyImpl(Mesh &mesh)
 void ThresholdRefiner::Reset()
 {
    estimator.Reset();
+   current_sequence = -1;
    num_marked_elements = 0LL;
    // marked_elements.SetSize(0); // not necessary
 }

mesh/mesh_operators.hpp

@@ -188,6 +188,7 @@ protected:
    long long num_marked_elements;
 
    Array<Refinement> marked_elements;
+   long current_sequence;
 
    int non_conforming;
    int nc_limit;

mesh/submesh/psubmesh.hpp

@@ -227,29 +227,15 @@ public:
                                            const ParGridFunction &dst);
 
    /**
-   * @brief Check if Mesh @a m is a ParSubMesh.
+   * @brief Check if ParMesh @a m is a ParSubMesh.
    *
-   * @param m The input Mesh
+   * @param m The input ParMesh
    */
-   static bool IsParSubMesh(const Mesh *m)
+   static bool IsParSubMesh(const ParMesh *m)
    {
       return dynamic_cast<const ParSubMesh *>(m) != nullptr;
    }
 
-   /**
-   * @brief Check if Mesh @a sub is a ParSubMesh of Mesh @a parent.
-   *
-   * @param sub The potential submesh Mesh
-   * @param parent The potential parent Mesh
-   */
-   static bool IsParSubMesh(const Mesh* sub, const Mesh* parent)
-   {
-      while (IsParSubMesh(sub) &&
-             (sub = static_cast<const ParSubMesh *>(sub)->GetParent()) &&
-             sub != parent);
-      return sub == parent;
-   }
-
 private:
    ParSubMesh(const ParMesh &parent, SubMesh::From from,
               const Array<int> &attributes);

mesh/submesh/submesh.hpp

@@ -225,20 +225,6 @@ public:
       return dynamic_cast<const SubMesh *>(m) != nullptr;
    }
 
-   /**
-   * @brief Check if Mesh @a sub is a SubMesh of Mesh @a parent.
-   *
-   * @param sub The potential submesh Mesh
-   * @param parent The potential parent Mesh
-   */
-   static bool IsSubMesh(const Mesh* sub, const Mesh* parent)
-   {
-      while (IsSubMesh(sub) &&
-             (sub = static_cast<const SubMesh *>(sub)->GetParent()) &&
-             sub != parent);
-      return sub == parent;
-   }
-
 private:
    /// Private constructor
    SubMesh(const Mesh &parent, From from, const Array<int> &attributes);

miniapps/electromagnetics/CMakeLists.txt

@@ -43,39 +43,19 @@ endif()
   
   # Add the corresponding tests to the "test" target
   if (MFEM_ENABLE_TESTING)
-    add_test(NAME tesla_1_np=${MFEM_MPI_NP}
+    add_test(NAME tesla_np=4
       COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
       ${MPIEXEC_PREFLAGS}
       $<TARGET_FILE:tesla> -no-vis -maxit 2 -cr "0 0 -0.2 0 0 0.2 0.2 0.4 1"
       ${MPIEXEC_POSTFLAGS})
-    add_test(NAME tesla_2_np=${MFEM_MPI_NP}
+
+    add_test(NAME volta_np=4
       COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
       ${MPIEXEC_PREFLAGS}
-      $<TARGET_FILE:tesla>
-      -no-vis -maxit 2 -m ../../data/inline-hex.mesh -ubbc "0 0 1"
+      $<TARGET_FILE:volta> -no-vis -maxit 2 -dbcs 1 -dbcg -ds "0.0 0.0 0.0 0.2 8.0"
       ${MPIEXEC_POSTFLAGS})
 
-    add_test(NAME volta_1_np=${MFEM_MPI_NP}
-      COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
-      ${MPIEXEC_PREFLAGS}
-      $<TARGET_FILE:volta>
-      -no-vis -maxit 2 -dbcs 1 -dbcg -ds "0.0 0.0 0.0 0.2 8.0"
-      ${MPIEXEC_POSTFLAGS})
-    add_test(NAME volta_2_np=${MFEM_MPI_NP}
-      COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
-      ${MPIEXEC_PREFLAGS}
-      $<TARGET_FILE:volta>
-      -no-vis -maxit 2 -m ../../data/square-disc.mesh -dbcs "1 2 3 4 5 6 7 8"
-      -dbcv "0 0 0 0 1 1 1 1"
-      ${MPIEXEC_POSTFLAGS})
-    add_test(NAME volta_3_np=${MFEM_MPI_NP}
-      COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
-      ${MPIEXEC_PREFLAGS}
-      $<TARGET_FILE:volta>
-      -no-vis -maxit 2 -m ../../data/inline-hex.mesh -dbcs "1 6" -dbcv "0 1"
-      ${MPIEXEC_POSTFLAGS})
-
-    add_test(NAME joule_np=${MFEM_MPI_NP}
+    add_test(NAME joule_np=4
       COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
       ${MPIEXEC_PREFLAGS}
       $<TARGET_FILE:joule>
@@ -83,41 +63,12 @@ endif()
       ${MPIEXEC_POSTFLAGS})
 
     if (MFEM_USE_DOUBLE) # otherwise returns MFEM_SKIP_RETURN_VALUE
-      add_test(NAME maxwell_np=${MFEM_MPI_NP}
+      add_test(NAME maxwell_np=4
         COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
         ${MPIEXEC_PREFLAGS}
         $<TARGET_FILE:maxwell>
         -no-vis -abcs "-1" -dp "-0.3 0.0 0.0 0.3 0.0 0.0 0.1 1 .5 .5"
         ${MPIEXEC_POSTFLAGS})
     endif()
-
-    if (MFEM_USE_GSLIB)
-      add_test(NAME lorentz_1_np=${MFEM_MPI_NP}
-        COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
-        ${MPIEXEC_PREFLAGS}
-        $<TARGET_FILE:lorentz>
-        -no-vis -er Volta-AMR-Parallel -ec 2 -npt 100 -xmin "0.0 0.0 0.0"
-        -xmax "1.0 1.0 1.0" -pmin "1 0 0" -pmax "1 0 0" -rdf 0 -vt 0 -nt 100
-        ${MPIEXEC_POSTFLAGS})
-      # Setup dependency on volta_3_np=<np>
-      set_tests_properties(volta_3_np=${MFEM_MPI_NP}
-        PROPERTIES FIXTURES_SETUP Volta3)
-      set_tests_properties(lorentz_1_np=${MFEM_MPI_NP}
-        PROPERTIES FIXTURES_REQUIRED Volta3)
-
-      add_test(NAME lorentz_2_np=${MFEM_MPI_NP}
-        COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} ${MFEM_MPI_NP}
-        ${MPIEXEC_PREFLAGS}
-        $<TARGET_FILE:lorentz>
-        -no-vis -br Tesla-AMR-Parallel -bc 2 -npt 10 -xmin "0.0 0.0 0.0"
-        -xmax "1.0 1.0 1.0" -pmin "0 0.1 0.05" -pmax "0 0.4 0.1" -nt 1000 -rdf 0
-        -vt 0
-        ${MPIEXEC_POSTFLAGS})
-      # Setup dependency on tesla_2_np=<np>
-      set_tests_properties(tesla_2_np=${MFEM_MPI_NP}
-        PROPERTIES FIXTURES_SETUP Tesla2)
-      set_tests_properties(lorentz_2_np=${MFEM_MPI_NP}
-        PROPERTIES FIXTURES_REQUIRED Tesla2)
-    endif()
   endif()
 endif()

miniapps/electromagnetics/makefile

@@ -117,10 +117,10 @@ joule-test-par: joule
 lorentz-test-par: lorentz-test-1 lorentz-test-2
 lorentz-test-1: lorentz volta-test-3
 	@$(call mfem-test,$<, $(RUN_MPI), Electromagnetic miniapp,\
-	-er Volta-AMR-Parallel -ec 2 -npt 100 -xmin '0.0 0.0 0.0' -xmax '1.0 1.0 1.0' -pmin '1 0 0' -pmax '1 0 0' -rdf 0 -vt 0 -nt 100)
+	-er Volta-AMR-Parallel -ec 2 -npt 100 -xmin '0.0 0.0 0.0' -xmax '1.0 1.0 1.0' -pmin '1 0 0' -pmax '1 0 0' -rdf 0 -vt 0 -nt 100')
 lorentz-test-2: lorentz tesla-test-2
 	@$(call mfem-test,$<, $(RUN_MPI), Electromagnetic miniapp,\
-	-br Tesla-AMR-Parallel -bc 2 -npt 10 -xmin '0.0 0.0 0.0' -xmax '1.0 1.0 1.0' -pmin '0 0.1 0.05' -pmax '0 0.4 0.1' -nt 1000 -rdf 0 -vt 0)
+	-br Tesla-AMR-Parallel -bc 2 -br Tesla-AMR-Parallel -npt 10 -xmin '0.0 0.0 0.0' -xmax '1.0 1.0 1.0' -pmin '0 0.1 0.05' -pmax '0 0.4 0.1' -nt 1000 -rdf 0 -vt 0)
 
 # Testing: "test" target and mfem-test* variables are defined in config/test.mk
 

miniapps/gslib/CMakeLists.txt

@@ -46,16 +46,8 @@ if (MFEM_USE_GSLIB)
 
   if (MFEM_ENABLE_TESTING)
     foreach (test "schwarz_ex1" "field-diff" "findpts" "field-interp")
-      if (MFEM_USE_MPI)
-        add_test(NAME ${test}
-          COMMAND ${MPIEXEC} ${MPIEXEC_NUMPROC_FLAG} 1
-          ${MPIEXEC_PREFLAGS}
-          $<TARGET_FILE:${test}> -no-vis
-          ${MPIEXEC_POSTFLAGS})
-      else()
-        add_test(NAME ${test}
-          COMMAND $<TARGET_FILE:${test}> -no-vis)
-      endif()
+      add_test(NAME ${test}
+        COMMAND $<TARGET_FILE:${test}> -no-vis)
     endforeach()
   endif()
 

miniapps/gslib/makefile

@@ -83,16 +83,11 @@ include $(MFEM_TEST_MK)
 
 # Testing: Parallel vs. serial runs
 RUN_MPI = $(MFEM_MPIEXEC) $(MFEM_MPIEXEC_NP) $(MFEM_MPI_NP)
-ifeq ($(MFEM_USE_MPI),YES)
-   RUN_MPI_SER = $(MFEM_MPIEXEC) $(MFEM_MPIEXEC_NP) 1
-else
-   RUN_MPI_SER =
-endif
 TEST_NAME := GSLIB miniapp
 %-test-par: %
 	@$(call mfem-test,$<, $(RUN_MPI), $(TEST_NAME))
 %-test-seq: %
-	@$(call mfem-test,$<, $(RUN_MPI_SER), $(TEST_NAME))
+	@$(call mfem-test,$<,, $(TEST_NAME))
 
 # Testing: "test" target and mfem-test* variables are defined in config/test.mk
 

miniapps/plasma/makefile

@@ -28,12 +28,6 @@ endif
 
 PLASMA_SUBDIRS = pic
 
-SUBDIRS_ALL = $(addsuffix /all,$(PLASMA_SUBDIRS))
-SUBDIRS_TEST = $(addsuffix /test,$(PLASMA_SUBDIRS))
-SUBDIRS_TEST_NOCLEAN = $(addsuffix /test-noclean,$(PLASMA_SUBDIRS))
-SUBDIRS_CLEAN = $(addsuffix /clean,$(PLASMA_SUBDIRS))
-SUBDIRS_TPRINT = $(addsuffix /test-print,$(PLASMA_SUBDIRS))
-
 .SUFFIXES:
 .SUFFIXES: .o .cpp .mk
 .PHONY: all lib-common clean clean-build clean-exec
@@ -43,24 +37,31 @@ COMMON_LIB = -L$(MFEM_BUILD_DIR)/miniapps/common -lmfem-common
 
 # If MFEM_SHARED is set, add the ../common rpath
 COMMON_LIB += $(if $(MFEM_SHARED:YES=),,\
-   $(MFEM_XLINKER)-rpath,$(abspath $(MFEM_BUILD_DIR)/miniapps/common))
+   $(if $(MFEM_USE_CUDA:YES=),$(CXX_XLINKER),$(CUDA_XLINKER))-rpath,$(abspath\
+   $(MFEM_BUILD_DIR)/miniapps/common))
+
+COMMON_O=
 
 # Remove built-in rules
 %: %.cpp
 %.o: %.cpp
 
-all: $(MINIAPPS) $(SUBDIRS_ALL) 
+all: $(MINIAPPS) subdirs
 
-.PHONY: $(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_TEST_NOCLEAN) \
-   $(SUBDIRS_CLEAN) $(SUBDIRS_TPRINT)
-$(SUBDIRS_ALL) $(SUBDIRS_TEST) $(SUBDIRS_TEST_NOCLEAN) $(SUBDIRS_CLEAN):
-	$(MAKE) -C $(@D) $(@F)
-$(SUBDIRS_TPRINT):
-	@$(MAKE) -C $(@D) $(@F)
+.PHONY: subdirs $(PLASMA_SUBDIRS)
+subdirs: $(PLASMA_SUBDIRS)
+$(PLASMA_SUBDIRS): lib-common
+	$(MAKE) -C $(BLD)$(@)
 
 # Rules for building the miniapps
-%: $(SRC)%.cpp $(MFEM_LIB_FILE) $(CONFIG_MK) | lib-common
-	$(MFEM_CXX) $(MFEM_LINK_FLAGS) $< -o $@ $(COMMON_LIB) $(MFEM_LIBS)
+%: $(SRC)%.cpp $(COMMON_O) $(MFEM_LIB_FILE) $(CONFIG_MK) | lib-common
+	$(MFEM_CXX) $(MFEM_LINK_FLAGS) $< -o $@ $(COMMON_O) $(COMMON_LIB) \
+        $(MFEM_LIBS)
+
+# Rules for compiling miniapp dependencies
+$(COMMON_O) $(addsuffix _solver.o,$(MINIAPPS)): \
+%.o: $(SRC)%.cpp $(SRC)%.hpp $(CONFIG_MK)
+	$(MFEM_CXX) $(MFEM_FLAGS) -c $(<) -o $(@)
 
 # Rule for building lib-common
 lib-common:
@@ -68,9 +69,6 @@ lib-common:
 
 MFEM_TESTS = MINIAPPS
 include $(MFEM_TEST_MK)
-test: $(SUBDIRS_TEST)
-test-noclean: $(SUBDIRS_TEST_NOCLEAN)
-test-print: $(SUBDIRS_TPRINT)
 
 # Testing: Specific execution options
 RUN_MPI = $(MFEM_MPIEXEC) $(MFEM_MPIEXEC_NP) $(MFEM_MPI_NP)
@@ -81,9 +79,14 @@ RUN_MPI = $(MFEM_MPIEXEC) $(MFEM_MPIEXEC_NP) $(MFEM_MPI_NP)
 $(MFEM_LIB_FILE):
 	$(error The MFEM library is not built)
 
-clean: clean-build clean-exec $(SUBDIRS_CLEAN)
+ALL_CLEAN_SUBDIRS = $(addsuffix /clean,$(PLASMA_SUBDIRS))
+.PHONY: $(ALL_CLEAN_SUBDIRS)
+$(ALL_CLEAN_SUBDIRS):
+	$(MAKE) -C $(BLD)$(@D) $(@F)
 
-clean-build:
+clean: clean-build clean-exec
+
+clean-build: $(addsuffix /clean,$(PLASMA_SUBDIRS))
 	rm -f *.o *~ $(SEQ_MINIAPPS) $(PAR_MINIAPPS)
 	rm -rf *.dSYM *.TVD.*breakpoints
 

miniapps/plasma/pic/makefile

@@ -20,7 +20,6 @@ CONFIG_MK = $(or $(wildcard $(MFEM_BUILD_DIR)/config/config.mk),\
 MFEM_LIB_FILE = mfem_is_not_built
 -include $(CONFIG_MK)
 
-SEQ_MINIAPPS =
 PAR_MINIAPPS =
 
 ifeq ($(MFEM_USE_GSLIB),YES)
@@ -28,9 +27,9 @@ ifeq ($(MFEM_USE_GSLIB),YES)
 endif
 
 ifeq ($(MFEM_USE_MPI),NO)
-   MINIAPPS = $(SEQ_MINIAPPS)
+   MINIAPPS =
 else
-   MINIAPPS = $(PAR_MINIAPPS) $(SEQ_MINIAPPS)
+   MINIAPPS = $(PAR_MINIAPPS)
 endif
 
 .SUFFIXES:
@@ -51,7 +50,7 @@ COMMON_LIB += $(if $(MFEM_SHARED:YES=),,\
 all: $(MINIAPPS)
 
 # Rules for building the miniapps
-%: $(SRC)%.cpp $(MFEM_LIB_FILE) $(CONFIG_MK) | lib-common
+electrostatic-pic: electrostatic-pic.cpp $(MFEM_LIB_FILE) $(CONFIG_MK) | lib-common
 	$(MFEM_CXX) $(MFEM_FLAGS) -c $<
 	$(MFEM_CXX) $(MFEM_LINK_FLAGS) -o $@ $@.o $(COMMON_LIB) $(MFEM_LIBS)
 

miniapps/tools/CMakeLists.txt

@@ -61,10 +61,6 @@ if (MFEM_USE_MPI)
       LIBRARIES mfem-common)
       add_dependencies(gridfunction-bounds copy_miniapps_tools_data)
 
-    add_mfem_miniapp(random-gridfunction-bounds
-      MAIN random-gridfunction-bounds.cpp
-      LIBRARIES mfem)
-
     add_mfem_miniapp(plor-transfer
       MAIN plor-transfer.cpp LIBRARIES mfem)
 

miniapps/tools/makefile

@@ -23,8 +23,7 @@ MFEM_LIB_FILE = mfem_is_not_built
 SEQ_MINIAPPS = display-basis load-dc convert-dc get-values lor-transfer \
 	tmop-check-metric tmop-metric-magnitude compare-dc
 
-PAR_MINIAPPS = nodal-transfer plor-transfer gridfunction-bounds \
-	random-gridfunction-bounds
+PAR_MINIAPPS = nodal-transfer plor-transfer gridfunction-bounds
 
 ifeq ($(MFEM_USE_MPI),NO)
    MINIAPPS = $(SEQ_MINIAPPS)
@@ -80,7 +79,7 @@ RUN_MPI = $(MFEM_MPIEXEC) $(MFEM_MPIEXEC_NP) $(MFEM_MPI_NP)
 # Do not test: display-basis, load-dc, convert-dc, get-values, lor-transfer, plor-transfer
 NO_TEST_APPS = display-basis load-dc convert-dc get-values lor-transfer \
 	plor-transfer tmop-check-metric tmop-metric-magnitude gridfunction-bounds \
-	random-gridfunction-bounds compare-dc
+	compare-dc
 $(foreach app,$(NO_TEST_APPS),$(app)-test-seq $(app)-test-par):
 	@true
 

miniapps/tools/random-gridfunction-bounds.cpp

@@ -1,263 +0,0 @@
-// Copyright (c) 2010-2025, 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.
-//
-//    ---------------------------------------------------------------------
-//     Compute bounds of a random grid function on a generated tensor mesh
-//    ---------------------------------------------------------------------
-//
-// This miniapp generates a 1D segment mesh or 2D quad mesh, builds a random
-// discontinuous grid function, computes element-wise piecewise linear bounds,
-// and visualizes the input field together with the lower and upper bounds.
-//
-// Compile with: make random-gridfunction-bounds
-//
-// Sample runs:
-//   mpirun -np 4 random-gridfunction-bounds
-//   mpirun -np 4 random-gridfunction-bounds -nx 64 -o 6 -ref 3 -d hip
-
-#include "mfem.hpp"
-
-#include <algorithm>
-#include <type_traits>
-
-using namespace mfem;
-using namespace std;
-
-void VisualizeField(ParMesh &pmesh, ParGridFunction &input,
-                    char *title, int pos_x, int pos_y);
-
-int main(int argc, char *argv[])
-{
-   Mpi::Init(argc, argv);
-   Hypre::Init();
-
-   int dim = 2;
-   int nx = 16;
-   int order = 4;
-   int num_comp = 2;
-   int ref = 2;
-   int niter = 1000;
-   int seed = 12345;
-   bool kernel_only = true;
-   bool visualization = false;
-   const char *device_config = "cpu";
-
-   OptionsParser args(argc, argv);
-   args.AddOption(&dim, "-dim", "--dimension",
-                  "Dimension of the generated tensor-product mesh (1 or 2).");
-   args.AddOption(&nx, "-nx", "--num-elements",
-                  "Number of elements in each mesh direction.");
-   args.AddOption(&order, "-o", "--order",
-                  "Polynomial degree of the random discontinuous field.");
-   args.AddOption(&num_comp, "-nc", "--num-components",
-                  "Number of vector components in the ParFiniteElementSpace.");
-   args.AddOption(&ref, "-ref", "--piecewise-linear-ref-factor",
-                  "Scaling factor for the resolution of the piecewise linear "
-                  "bounds. If less than 2, the resolution is picked "
-                  "automatically.");
-   args.AddOption(&niter, "-ni", "--num-iters",
-                  "Number of times to evaluate the bounds.");
-   args.AddOption(&seed, "-rs", "--random-seed",
-                  "Random seed used to initialize the field.");
-   args.AddOption(&device_config, "-d", "--device",
-                  "Device configuration string, see Device::Configure().");
-   args.AddOption(&kernel_only, "-ko", "--kernel-only",
-                  "-no-ko", "--no-kernel-only",
-                  "Run only PLBound::GetElementBoundsKernel on a prebuilt "
-                  "element E-vector.");
-   args.AddOption(&visualization, "-vis", "--visualization", "-no-vis",
-                  "--no-visualization",
-                  "Enable or disable GLVis visualization.");
-   args.ParseCheck();
-
-   MFEM_VERIFY(dim == 1 || dim == 2, "dim must be 1 or 2.");
-   MFEM_VERIFY(nx > 0, "nx must be positive.");
-   MFEM_VERIFY(order >= 0, "order must be non-negative.");
-   MFEM_VERIFY(num_comp > 0, "num_comp must be positive.");
-   MFEM_VERIFY(niter > 0, "niter must be positive.");
-
-   Device device(device_config);
-   if (Mpi::Root()) { device.Print(); }
-
-   Mesh mesh = (dim == 1) ?
-               Mesh::MakeCartesian1D(nx, 1.0) :
-               Mesh::MakeCartesian2D(nx, nx, Element::QUADRILATERAL, true,
-                                     1.0, 1.0);
-   ParMesh pmesh(MPI_COMM_WORLD, mesh);
-
-   const int mesh_dim = pmesh.Dimension();
-   L2_FECollection fec(order, mesh_dim, BasisType::GaussLobatto);
-   ParFiniteElementSpace fes(&pmesh, &fec, num_comp, Ordering::byNODES);
-   ParGridFunction input(&fes);
-   input.Randomize(seed + Mpi::WorldRank());
-   input.UseDevice(true);
-
-   L2_FECollection fec_pc(0, mesh_dim);
-   ParFiniteElementSpace fes_pc(&pmesh, &fec_pc, num_comp, Ordering::byNODES);
-   ParGridFunction lowerb(&fes_pc), upperb(&fes_pc);
-   Vector lower_vec, upper_vec;
-
-   PLBound plb(&fes, ref*(fes.GetMaxElementOrder() + 1));
-   if (kernel_only)
-   {
-      const FiniteElement &fe = *fes.GetTypicalFE();
-      const int rdim = fe.GetDim();
-      const int nd = fe.GetDof();
-      const int fes_dim = fes.GetVDim();
-      Vector e_vec(nd*fes_dim*fes.GetNE(), Device::GetDeviceMemoryType());
-      e_vec.UseDevice(true);
-
-      const ElementRestrictionOperator *elem_restr =
-         fes.GetElementRestriction(ElementDofOrdering::LEXICOGRAPHIC);
-      MFEM_VERIFY(elem_restr != nullptr,
-                  "Element restriction is required for kernel-only mode.");
-      elem_restr->Mult(input, e_vec);
-
-      for (int i = 0; i < niter; i++)
-      {
-         plb.GetElementBoundsKernel(rdim, fes_dim, e_vec, lower_vec, upper_vec);
-      }
-   }
-   else
-   {
-      for (int i = 0; i < niter; i++)
-      {
-         input.GetElementBounds(plb, lower_vec, upper_vec);
-      }
-   }
-
-   const real_t *lower_data = lower_vec.HostRead();
-   const real_t *upper_data = upper_vec.HostRead();
-
-   // Build a host reference from the lexicographic E-vector and the scalar
-   // PLBound::GetNDBounds path to avoid re-entering the device dispatch.
-   const bool use_dev = input.UseDevice();
-   PLBound plb_host(&fes, ref*(fes.GetMaxElementOrder() + 1));
-   Vector lower_ref, upper_ref;
-   const FiniteElement &fe = *fes.GetTypicalFE();
-   const int rdim = fe.GetDim();
-   const int nd = fe.GetDof();
-   const int nel = fes.GetNE();
-   const int fes_dim = fes.GetVDim();
-   Vector e_vec_ref(nd*fes_dim*nel);
-   lower_ref.SetSize(nel*fes_dim);
-   upper_ref.SetSize(nel*fes_dim);
-   const ElementRestrictionOperator *elem_restr =
-      fes.GetElementRestriction(ElementDofOrdering::LEXICOGRAPHIC);
-   MFEM_VERIFY(elem_restr != nullptr,
-               "Element restriction is required for host reference.");
-   input.UseDevice(false);
-   input.HostRead();
-   elem_restr->Mult(input, e_vec_ref);
-   input.UseDevice(use_dev);
-   const real_t *e_ref_data = e_vec_ref.HostRead();
-
-   for (int d = 0; d < fes_dim; d++)
-   {
-      for (int e = 0; e < nel; e++)
-      {
-         Vector coeff(nd);
-         for (int i = 0; i < nd; i++)
-         {
-            coeff(i) = e_ref_data[i + nd*(d + fes_dim*e)];
-         }
-         Vector lower_c, upper_c;
-         plb_host.GetNDBounds(rdim, coeff, lower_c, upper_c);
-         lower_ref(e + d*nel) = lower_c.Min();
-         upper_ref(e + d*nel) = upper_c.Max();
-      }
-   }
-   const real_t *lower_ref_data = lower_ref.HostRead();
-   const real_t *upper_ref_data = upper_ref.HostRead();
-
-   MFEM_VERIFY(lower_vec.Size() == lower_ref.Size() &&
-               upper_vec.Size() == upper_ref.Size(),
-               "Reference element-bound vectors have inconsistent sizes.");
-
-   real_t lower_diff = 0.0;
-   real_t upper_diff = 0.0;
-   for (int i = 0; i < lower_vec.Size(); i++)
-   {
-      lower_diff = std::max(lower_diff,
-                            std::abs(lower_data[i] - lower_ref_data[i]));
-   }
-   for (int i = 0; i < upper_vec.Size(); i++)
-   {
-      upper_diff = std::max(upper_diff,
-                            std::abs(upper_data[i] - upper_ref_data[i]));
-   }
-   MPI_Allreduce(MPI_IN_PLACE, &lower_diff, 1, MPITypeMap<real_t>::mpi_type,
-                 MPI_MAX, pmesh.GetComm());
-   MPI_Allreduce(MPI_IN_PLACE, &upper_diff, 1, MPITypeMap<real_t>::mpi_type,
-                 MPI_MAX, pmesh.GetComm());
-
-   const real_t verify_tol = std::is_same<real_t, float>::value ?
-                             real_t(1.0e-5) : real_t(1.0e-12);
-   MFEM_VERIFY(lower_diff <= verify_tol && upper_diff <= verify_tol,
-               "Device element bounds do not match host reference.");
-
-   lowerb = lower_vec;
-   upperb = upper_vec;
-
-   real_t lower_min = lowerb.Min();
-   real_t upper_max = upperb.Max();
-   MPI_Allreduce(MPI_IN_PLACE, &lower_min, 1, MPITypeMap<real_t>::mpi_type,
-                 MPI_MIN, pmesh.GetComm());
-   MPI_Allreduce(MPI_IN_PLACE, &upper_max, 1, MPITypeMap<real_t>::mpi_type,
-                 MPI_MAX, pmesh.GetComm());
-
-   if (Mpi::Root())
-   {
-      cout << "dim: " << mesh_dim << '\n'
-           << "nx: " << nx << '\n'
-           << "order: " << order << '\n'
-           << "num components: " << num_comp << '\n'
-           << "PL bound control-point factor: " << ref << '\n'
-           << "iterations: " << niter << '\n'
-           << "kernel-only mode: " << (kernel_only ? "yes" : "no") << '\n'
-           << "host/device lower max diff: " << lower_diff << '\n'
-           << "host/device upper max diff: " << upper_diff << '\n'
-           << "global lower bound minimum: " << lower_min << '\n'
-           << "global upper bound maximum: " << upper_max << endl;
-   }
-
-   if (visualization)
-   {
-      char title1[] = "Random input gridfunction";
-      char title2[] = "Element-wise lower bound";
-      char title3[] = "Element-wise upper bound";
-      VisualizeField(pmesh, input, title1, 0, 0);
-      VisualizeField(pmesh, lowerb, title2, 450, 0);
-      VisualizeField(pmesh, upperb, title3, 900, 0);
-   }
-
-   return 0;
-}
-
-void VisualizeField(ParMesh &pmesh, ParGridFunction &input,
-                    char *title, int pos_x, int pos_y)
-{
-   socketstream sock;
-   if (pmesh.GetMyRank() == 0)
-   {
-      sock.open("localhost", 19916);
-      sock << "solution\n";
-   }
-   pmesh.PrintAsOne(sock);
-   input.SaveAsOne(sock);
-   if (pmesh.GetMyRank() == 0)
-   {
-      sock << "window_title '" << title << "'\n"
-           << "window_geometry "
-           << pos_x << " " << pos_y << " " << 400 << " " << 400 << "\n"
-           << "keys jRmclApppppppppppp//]]]]]]]]" << endl;
-   }
-}

tests/unit/CMakeLists.txt

@@ -122,6 +122,7 @@ set(UNIT_TESTS_SRCS
   fem/test_fe_pos.cpp
   fem/test_fe_symmetry.cpp
   fem/test_fe.cpp
+  fem/test_fespace_get_ess_true_dofs.cpp
   fem/test_get_value.cpp
   fem/test_getderivative.cpp
   fem/test_getgradient.cpp

tests/unit/fem/test_fespace_get_ess_true_dofs.cpp

@@ -0,0 +1,74 @@
+// Copyright (c) 2010-2025, 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.
+
+#include "mfem.hpp"
+#include "unit_tests.hpp"
+
+using namespace mfem;
+
+TEST_CASE("FESpace Get Essential True DOFs",
+          "[FESpace Get Essential True DOFs]")
+{
+   std::cout << "Testing get essential true dofs" << std::endl;
+   int order_h1 = 3, n = 2, dim = 3;
+
+   Mesh mesh = Mesh::MakeCartesian3D(
+                  n, n, n, Element::HEXAHEDRON, 1.0, 1.0, 1.0);
+   mesh.SetCurvature(order_h1);
+
+   H1_FECollection fec(order_h1, dim);
+   FiniteElementSpace fe_space(&mesh, &fec, dim);
+
+
+   const int num_bdr_attr = fe_space.GetMesh()->bdr_attributes.Max();
+
+
+   Array<int> ess_tdofs_2d, ess_tdofs_1d, ess_tdofs_tmp, ess_bdrs;
+   Array2D<bool> comps(num_bdr_attr, dim);
+   ess_bdrs.SetSize(num_bdr_attr);
+   comps = false;
+   ess_bdrs = 0;
+
+   // simple xy boundary condition on all surfaces
+   // could do something more complex but don't really want to...
+   for (int i = 0; i < num_bdr_attr; i++)
+   {
+      ess_bdrs[i] = 1;
+      comps(i, 0) = true;
+      comps(i, 2) = true;
+   }
+
+   fe_space.GetEssentialTrueDofs(ess_bdrs, ess_tdofs_2d, comps);
+
+   // Now for the old way
+   fe_space.GetEssentialTrueDofs(ess_bdrs, ess_tdofs_tmp, 0);
+   ess_tdofs_1d.Append(ess_tdofs_tmp);
+   ess_tdofs_tmp.DeleteAll();
+   fe_space.GetEssentialTrueDofs(ess_bdrs, ess_tdofs_tmp, 2);
+   ess_tdofs_1d.Append(ess_tdofs_tmp);
+   // Sort the 2 arrays in order to compare them
+   ess_tdofs_2d.Sort();
+   ess_tdofs_1d.Sort();
+
+   int diff = 0;
+
+   for (int i = 0; i < ess_tdofs_2d.Size(); i++)
+   {
+      diff += std::abs(ess_tdofs_2d[i] - ess_tdofs_1d[i]);
+   }
+
+   std::cout << "Difference in essential tdofs approaches is: " << diff <<
+             std::endl;
+
+   REQUIRE(diff == 0);
+
+}
+

tests/unit/fem/test_particleset.cpp

@@ -296,7 +296,7 @@ void TestRedistribute(Ordering::Type ordering)
       int wrong_proc_count = 0;
       for (int i = 0; i < procs.Size(); i++)
       {
-         if (static_cast<unsigned>(rank) != procs[i])
+         if (rank != procs[i])
          {
             wrong_proc_count++;
          }