xref: /libCEED/rust/libceed-sys/c-src/backends/cuda-ref/ceed-cuda-ref-operator.c (revision 59ad764ae9820336980ee591bbbd96481bb41719)

// Copyright (c) 2017-2018, Lawrence Livermore National Security, LLC.
// Produced at the Lawrence Livermore National Laboratory. LLNL-CODE-734707.
// All Rights reserved. See files LICENSE and NOTICE for details.
//
// This file is part of CEED, a collection of benchmarks, miniapps, software
// libraries and APIs for efficient high-order finite element and spectral
// element discretizations for exascale applications. For more information and
// source code availability see http://github.com/ceed.
//
// The CEED research is supported by the Exascale Computing Project 17-SC-20-SC,
// a collaborative effort of two U.S. Department of Energy organizations (Office
// of Science and the National Nuclear Security Administration) responsible for
// the planning and preparation of a capable exascale ecosystem, including
// software, applications, hardware, advanced system engineering and early
// testbed platforms, in support of the nation's exascale computing imperative.

#include <ceed/ceed.h>
#include <ceed/backend.h>
#include <assert.h>
#include <cuda.h>
#include <cuda_runtime.h>
#include <stdbool.h>
#include <string.h>
#include "ceed-cuda-ref.h"
#include "../cuda/ceed-cuda-compile.h"

//------------------------------------------------------------------------------
// Destroy operator
//------------------------------------------------------------------------------
static int CeedOperatorDestroy_Cuda(CeedOperator op) {
  int ierr;
  CeedOperator_Cuda *impl;
  ierr = CeedOperatorGetData(op, &impl); CeedChkBackend(ierr);

  // Apply data
  for (CeedInt i = 0; i < impl->numein + impl->numeout; i++) {
    ierr = CeedVectorDestroy(&impl->evecs[i]); CeedChkBackend(ierr);
  }
  ierr = CeedFree(&impl->evecs); CeedChkBackend(ierr);

  for (CeedInt i = 0; i < impl->numein; i++) {
    ierr = CeedVectorDestroy(&impl->qvecsin[i]); CeedChkBackend(ierr);
  }
  ierr = CeedFree(&impl->qvecsin); CeedChkBackend(ierr);

  for (CeedInt i = 0; i < impl->numeout; i++) {
    ierr = CeedVectorDestroy(&impl->qvecsout[i]); CeedChkBackend(ierr);
  }
  ierr = CeedFree(&impl->qvecsout); CeedChkBackend(ierr);

  // QFunction assembly data
  for (CeedInt i=0; i<impl->qfnumactivein; i++) {
    ierr = CeedVectorDestroy(&impl->qfactivein[i]); CeedChkBackend(ierr);
  }
  ierr = CeedFree(&impl->qfactivein); CeedChkBackend(ierr);

  // Diag data
  if (impl->diag) {
    Ceed ceed;
    ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
    CeedChk_Cu(ceed, cuModuleUnload(impl->diag->module));
    ierr = CeedFree(&impl->diag->h_emodein); CeedChkBackend(ierr);
    ierr = CeedFree(&impl->diag->h_emodeout); CeedChkBackend(ierr);
    ierr = cudaFree(impl->diag->d_emodein); CeedChk_Cu(ceed, ierr);
    ierr = cudaFree(impl->diag->d_emodeout); CeedChk_Cu(ceed, ierr);
    ierr = cudaFree(impl->diag->d_identity); CeedChk_Cu(ceed, ierr);
    ierr = cudaFree(impl->diag->d_interpin); CeedChk_Cu(ceed, ierr);
    ierr = cudaFree(impl->diag->d_interpout); CeedChk_Cu(ceed, ierr);
    ierr = cudaFree(impl->diag->d_gradin); CeedChk_Cu(ceed, ierr);
    ierr = cudaFree(impl->diag->d_gradout); CeedChk_Cu(ceed, ierr);
    ierr = CeedElemRestrictionDestroy(&impl->diag->pbdiagrstr);
    CeedChkBackend(ierr);
    ierr = CeedVectorDestroy(&impl->diag->elemdiag); CeedChkBackend(ierr);
    ierr = CeedVectorDestroy(&impl->diag->pbelemdiag); CeedChkBackend(ierr);
  }
  ierr = CeedFree(&impl->diag); CeedChkBackend(ierr);

  if (impl->asmb) {
    Ceed ceed;
    ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
    CeedChk_Cu(ceed, cuModuleUnload(impl->asmb->module));
    ierr = cudaFree(impl->asmb->d_B_in); CeedChk_Cu(ceed, ierr);
    ierr = cudaFree(impl->asmb->d_B_out); CeedChk_Cu(ceed, ierr);
  }
  ierr = CeedFree(&impl->asmb); CeedChkBackend(ierr);

  ierr = CeedFree(&impl); CeedChkBackend(ierr);
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Setup infields or outfields
//------------------------------------------------------------------------------
static int CeedOperatorSetupFields_Cuda(CeedQFunction qf, CeedOperator op,
                                        bool isinput, CeedVector *evecs,
                                        CeedVector *qvecs, CeedInt starte,
                                        CeedInt numfields, CeedInt Q,
                                        CeedInt numelements) {
  CeedInt dim, ierr, size;
  Ceed ceed;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
  CeedBasis basis;
  CeedElemRestriction Erestrict;
  CeedOperatorField *opfields;
  CeedQFunctionField *qffields;
  CeedVector fieldvec;
  bool strided;
  bool skiprestrict;

  if (isinput) {
    ierr = CeedOperatorGetFields(op, NULL, &opfields, NULL, NULL);
    CeedChkBackend(ierr);
    ierr = CeedQFunctionGetFields(qf, NULL, &qffields, NULL, NULL);
    CeedChkBackend(ierr);
  } else {
    ierr = CeedOperatorGetFields(op, NULL, NULL, NULL, &opfields);
    CeedChkBackend(ierr);
    ierr = CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qffields);
    CeedChkBackend(ierr);
  }

  // Loop over fields
  for (CeedInt i = 0; i < numfields; i++) {
    CeedEvalMode emode;
    ierr = CeedQFunctionFieldGetEvalMode(qffields[i], &emode); CeedChkBackend(ierr);

    strided = false;
    skiprestrict = false;
    if (emode != CEED_EVAL_WEIGHT) {
      ierr = CeedOperatorFieldGetElemRestriction(opfields[i], &Erestrict);
      CeedChkBackend(ierr);

      // Check whether this field can skip the element restriction:
      // must be passive input, with emode NONE, and have a strided restriction with
      // CEED_STRIDES_BACKEND.

      // First, check whether the field is input or output:
      if (isinput) {
        // Check for passive input:
        ierr = CeedOperatorFieldGetVector(opfields[i], &fieldvec); CeedChkBackend(ierr);
        if (fieldvec != CEED_VECTOR_ACTIVE) {
          // Check emode
          if (emode == CEED_EVAL_NONE) {
            // Check for strided restriction
            ierr = CeedElemRestrictionIsStrided(Erestrict, &strided);
            CeedChkBackend(ierr);
            if (strided) {
              // Check if vector is already in preferred backend ordering
              ierr = CeedElemRestrictionHasBackendStrides(Erestrict,
                     &skiprestrict); CeedChkBackend(ierr);
            }
          }
        }
      }
      if (skiprestrict) {
        // We do not need an E-Vector, but will use the input field vector's data
        // directly in the operator application.
        evecs[i + starte] = NULL;
      } else {
        ierr = CeedElemRestrictionCreateVector(Erestrict, NULL,
                                               &evecs[i + starte]);
        CeedChkBackend(ierr);
      }
    }

    switch (emode) {
    case CEED_EVAL_NONE:
      ierr = CeedQFunctionFieldGetSize(qffields[i], &size); CeedChkBackend(ierr);
      ierr = CeedVectorCreate(ceed, numelements * Q * size, &qvecs[i]);
      CeedChkBackend(ierr);
      break;
    case CEED_EVAL_INTERP:
      ierr = CeedQFunctionFieldGetSize(qffields[i], &size); CeedChkBackend(ierr);
      ierr = CeedVectorCreate(ceed, numelements * Q * size, &qvecs[i]);
      CeedChkBackend(ierr);
      break;
    case CEED_EVAL_GRAD:
      ierr = CeedOperatorFieldGetBasis(opfields[i], &basis); CeedChkBackend(ierr);
      ierr = CeedQFunctionFieldGetSize(qffields[i], &size); CeedChkBackend(ierr);
      ierr = CeedBasisGetDimension(basis, &dim); CeedChkBackend(ierr);
      ierr = CeedVectorCreate(ceed, numelements * Q * size, &qvecs[i]);
      CeedChkBackend(ierr);
      break;
    case CEED_EVAL_WEIGHT: // Only on input fields
      ierr = CeedOperatorFieldGetBasis(opfields[i], &basis); CeedChkBackend(ierr);
      ierr = CeedVectorCreate(ceed, numelements * Q, &qvecs[i]); CeedChkBackend(ierr);
      ierr = CeedBasisApply(basis, numelements, CEED_NOTRANSPOSE,
                            CEED_EVAL_WEIGHT, NULL, qvecs[i]); CeedChkBackend(ierr);
      break;
    case CEED_EVAL_DIV:
      break; // TODO: Not implemented
    case CEED_EVAL_CURL:
      break; // TODO: Not implemented
    }
  }
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// CeedOperator needs to connect all the named fields (be they active or passive)
//   to the named inputs and outputs of its CeedQFunction.
//------------------------------------------------------------------------------
static int CeedOperatorSetup_Cuda(CeedOperator op) {
  int ierr;
  bool setupdone;
  ierr = CeedOperatorIsSetupDone(op, &setupdone); CeedChkBackend(ierr);
  if (setupdone)
    return CEED_ERROR_SUCCESS;
  Ceed ceed;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
  CeedOperator_Cuda *impl;
  ierr = CeedOperatorGetData(op, &impl); CeedChkBackend(ierr);
  CeedQFunction qf;
  ierr = CeedOperatorGetQFunction(op, &qf); CeedChkBackend(ierr);
  CeedInt Q, numelements, numinputfields, numoutputfields;
  ierr = CeedOperatorGetNumQuadraturePoints(op, &Q); CeedChkBackend(ierr);
  ierr = CeedOperatorGetNumElements(op, &numelements); CeedChkBackend(ierr);
  CeedChkBackend(ierr);
  CeedOperatorField *opinputfields, *opoutputfields;
  ierr = CeedOperatorGetFields(op, &numinputfields, &opinputfields,
                               &numoutputfields, &opoutputfields);
  CeedChkBackend(ierr);
  CeedQFunctionField *qfinputfields, *qfoutputfields;
  ierr = CeedQFunctionGetFields(qf, NULL, &qfinputfields, NULL, &qfoutputfields);
  CeedChkBackend(ierr);

  // Allocate
  ierr = CeedCalloc(numinputfields + numoutputfields, &impl->evecs);
  CeedChkBackend(ierr);

  ierr = CeedCalloc(CEED_FIELD_MAX, &impl->qvecsin); CeedChkBackend(ierr);
  ierr = CeedCalloc(CEED_FIELD_MAX, &impl->qvecsout); CeedChkBackend(ierr);

  impl->numein = numinputfields; impl->numeout = numoutputfields;

  // Set up infield and outfield evecs and qvecs
  // Infields
  ierr = CeedOperatorSetupFields_Cuda(qf, op, true,
                                      impl->evecs, impl->qvecsin, 0,
                                      numinputfields, Q, numelements);
  CeedChkBackend(ierr);

  // Outfields
  ierr = CeedOperatorSetupFields_Cuda(qf, op, false,
                                      impl->evecs, impl->qvecsout,
                                      numinputfields, numoutputfields, Q,
                                      numelements); CeedChkBackend(ierr);

  ierr = CeedOperatorSetSetupDone(op); CeedChkBackend(ierr);
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Setup Operator Inputs
//------------------------------------------------------------------------------
static inline int CeedOperatorSetupInputs_Cuda(CeedInt numinputfields,
    CeedQFunctionField *qfinputfields, CeedOperatorField *opinputfields,
    CeedVector invec, const bool skipactive, CeedScalar *edata[2*CEED_FIELD_MAX],
    CeedOperator_Cuda *impl, CeedRequest *request) {
  CeedInt ierr;
  CeedEvalMode emode;
  CeedVector vec;
  CeedElemRestriction Erestrict;

  for (CeedInt i = 0; i < numinputfields; i++) {
    // Get input vector
    ierr = CeedOperatorFieldGetVector(opinputfields[i], &vec); CeedChkBackend(ierr);
    if (vec == CEED_VECTOR_ACTIVE) {
      if (skipactive)
        continue;
      else
        vec = invec;
    }

    ierr = CeedQFunctionFieldGetEvalMode(qfinputfields[i], &emode);
    CeedChkBackend(ierr);
    if (emode == CEED_EVAL_WEIGHT) { // Skip
    } else {
      // Get input vector
      ierr = CeedOperatorFieldGetVector(opinputfields[i], &vec); CeedChkBackend(ierr);
      // Get input element restriction
      ierr = CeedOperatorFieldGetElemRestriction(opinputfields[i], &Erestrict);
      CeedChkBackend(ierr);
      if (vec == CEED_VECTOR_ACTIVE)
        vec = invec;
      // Restrict, if necessary
      if (!impl->evecs[i]) {
        // No restriction for this field; read data directly from vec.
        ierr = CeedVectorGetArrayRead(vec, CEED_MEM_DEVICE,
                                      (const CeedScalar **) &edata[i]);
        CeedChkBackend(ierr);
      } else {
        ierr = CeedElemRestrictionApply(Erestrict, CEED_NOTRANSPOSE, vec,
                                        impl->evecs[i], request); CeedChkBackend(ierr);
        // Get evec
        ierr = CeedVectorGetArrayRead(impl->evecs[i], CEED_MEM_DEVICE,
                                      (const CeedScalar **) &edata[i]);
        CeedChkBackend(ierr);
      }
    }
  }
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Input Basis Action
//------------------------------------------------------------------------------
static inline int CeedOperatorInputBasis_Cuda(CeedInt numelements,
    CeedQFunctionField *qfinputfields, CeedOperatorField *opinputfields,
    CeedInt numinputfields, const bool skipactive,
    CeedScalar *edata[2*CEED_FIELD_MAX],CeedOperator_Cuda *impl) {
  CeedInt ierr;
  CeedInt elemsize, size;
  CeedElemRestriction Erestrict;
  CeedEvalMode emode;
  CeedBasis basis;

  for (CeedInt i=0; i<numinputfields; i++) {
    // Skip active input
    if (skipactive) {
      CeedVector vec;
      ierr = CeedOperatorFieldGetVector(opinputfields[i], &vec); CeedChkBackend(ierr);
      if (vec == CEED_VECTOR_ACTIVE)
        continue;
    }
    // Get elemsize, emode, size
    ierr = CeedOperatorFieldGetElemRestriction(opinputfields[i], &Erestrict);
    CeedChkBackend(ierr);
    ierr = CeedElemRestrictionGetElementSize(Erestrict, &elemsize);
    CeedChkBackend(ierr);
    ierr = CeedQFunctionFieldGetEvalMode(qfinputfields[i], &emode);
    CeedChkBackend(ierr);
    ierr = CeedQFunctionFieldGetSize(qfinputfields[i], &size); CeedChkBackend(ierr);
    // Basis action
    switch (emode) {
    case CEED_EVAL_NONE:
      ierr = CeedVectorSetArray(impl->qvecsin[i], CEED_MEM_DEVICE,
                                CEED_USE_POINTER, edata[i]); CeedChkBackend(ierr);
      break;
    case CEED_EVAL_INTERP:
      ierr = CeedOperatorFieldGetBasis(opinputfields[i], &basis);
      CeedChkBackend(ierr);
      ierr = CeedBasisApply(basis, numelements, CEED_NOTRANSPOSE,
                            CEED_EVAL_INTERP, impl->evecs[i],
                            impl->qvecsin[i]); CeedChkBackend(ierr);
      break;
    case CEED_EVAL_GRAD:
      ierr = CeedOperatorFieldGetBasis(opinputfields[i], &basis);
      CeedChkBackend(ierr);
      ierr = CeedBasisApply(basis, numelements, CEED_NOTRANSPOSE,
                            CEED_EVAL_GRAD, impl->evecs[i],
                            impl->qvecsin[i]); CeedChkBackend(ierr);
      break;
    case CEED_EVAL_WEIGHT:
      break; // No action
    case CEED_EVAL_DIV:
      break; // TODO: Not implemented
    case CEED_EVAL_CURL:
      break; // TODO: Not implemented
    }
  }
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Restore Input Vectors
//------------------------------------------------------------------------------
static inline int CeedOperatorRestoreInputs_Cuda(CeedInt numinputfields,
    CeedQFunctionField *qfinputfields, CeedOperatorField *opinputfields,
    const bool skipactive, CeedScalar *edata[2*CEED_FIELD_MAX],
    CeedOperator_Cuda *impl) {
  CeedInt ierr;
  CeedEvalMode emode;
  CeedVector vec;

  for (CeedInt i = 0; i < numinputfields; i++) {
    // Skip active input
    if (skipactive) {
      ierr = CeedOperatorFieldGetVector(opinputfields[i], &vec); CeedChkBackend(ierr);
      if (vec == CEED_VECTOR_ACTIVE)
        continue;
    }
    ierr = CeedQFunctionFieldGetEvalMode(qfinputfields[i], &emode);
    CeedChkBackend(ierr);
    if (emode == CEED_EVAL_WEIGHT) { // Skip
    } else {
      if (!impl->evecs[i]) {  // This was a skiprestrict case
        ierr = CeedOperatorFieldGetVector(opinputfields[i], &vec); CeedChkBackend(ierr);
        ierr = CeedVectorRestoreArrayRead(vec,
                                          (const CeedScalar **)&edata[i]);
        CeedChkBackend(ierr);
      } else {
        ierr = CeedVectorRestoreArrayRead(impl->evecs[i],
                                          (const CeedScalar **) &edata[i]);
        CeedChkBackend(ierr);
      }
    }
  }
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Apply and add to output
//------------------------------------------------------------------------------
static int CeedOperatorApplyAdd_Cuda(CeedOperator op, CeedVector invec,
                                     CeedVector outvec, CeedRequest *request) {
  int ierr;
  CeedOperator_Cuda *impl;
  ierr = CeedOperatorGetData(op, &impl); CeedChkBackend(ierr);
  CeedQFunction qf;
  ierr = CeedOperatorGetQFunction(op, &qf); CeedChkBackend(ierr);
  CeedInt Q, numelements, elemsize, numinputfields, numoutputfields, size;
  ierr = CeedOperatorGetNumQuadraturePoints(op, &Q); CeedChkBackend(ierr);
  ierr = CeedOperatorGetNumElements(op, &numelements); CeedChkBackend(ierr);
  CeedChkBackend(ierr);
  CeedOperatorField *opinputfields, *opoutputfields;
  ierr = CeedOperatorGetFields(op, &numinputfields, &opinputfields,
                               &numoutputfields, &opoutputfields);
  CeedChkBackend(ierr);
  CeedQFunctionField *qfinputfields, *qfoutputfields;
  ierr = CeedQFunctionGetFields(qf, NULL, &qfinputfields, NULL, &qfoutputfields);
  CeedChkBackend(ierr);
  CeedEvalMode emode;
  CeedVector vec;
  CeedBasis basis;
  CeedElemRestriction Erestrict;
  CeedScalar *edata[2*CEED_FIELD_MAX] = {0};

  // Setup
  ierr = CeedOperatorSetup_Cuda(op); CeedChkBackend(ierr);

  // Input Evecs and Restriction
  ierr = CeedOperatorSetupInputs_Cuda(numinputfields, qfinputfields,
                                      opinputfields, invec, false, edata,
                                      impl, request); CeedChkBackend(ierr);

  // Input basis apply if needed
  ierr = CeedOperatorInputBasis_Cuda(numelements, qfinputfields, opinputfields,
                                     numinputfields, false, edata, impl);
  CeedChkBackend(ierr);

  // Output pointers, as necessary
  for (CeedInt i = 0; i < numoutputfields; i++) {
    ierr = CeedQFunctionFieldGetEvalMode(qfoutputfields[i], &emode);
    CeedChkBackend(ierr);
    if (emode == CEED_EVAL_NONE) {
      // Set the output Q-Vector to use the E-Vector data directly.
      ierr = CeedVectorGetArrayWrite(impl->evecs[i + impl->numein], CEED_MEM_DEVICE,
                                     &edata[i + numinputfields]); CeedChkBackend(ierr);
      ierr = CeedVectorSetArray(impl->qvecsout[i], CEED_MEM_DEVICE,
                                CEED_USE_POINTER, edata[i + numinputfields]);
      CeedChkBackend(ierr);
    }
  }

  // Q function
  ierr = CeedQFunctionApply(qf, numelements * Q, impl->qvecsin, impl->qvecsout);
  CeedChkBackend(ierr);

  // Output basis apply if needed
  for (CeedInt i = 0; i < numoutputfields; i++) {
    // Get elemsize, emode, size
    ierr = CeedOperatorFieldGetElemRestriction(opoutputfields[i], &Erestrict);
    CeedChkBackend(ierr);
    ierr = CeedElemRestrictionGetElementSize(Erestrict, &elemsize);
    CeedChkBackend(ierr);
    ierr = CeedQFunctionFieldGetEvalMode(qfoutputfields[i], &emode);
    CeedChkBackend(ierr);
    ierr = CeedQFunctionFieldGetSize(qfoutputfields[i], &size);
    CeedChkBackend(ierr);
    // Basis action
    switch (emode) {
    case CEED_EVAL_NONE:
      break;
    case CEED_EVAL_INTERP:
      ierr = CeedOperatorFieldGetBasis(opoutputfields[i], &basis);
      CeedChkBackend(ierr);
      ierr = CeedBasisApply(basis, numelements, CEED_TRANSPOSE,
                            CEED_EVAL_INTERP, impl->qvecsout[i],
                            impl->evecs[i + impl->numein]); CeedChkBackend(ierr);
      break;
    case CEED_EVAL_GRAD:
      ierr = CeedOperatorFieldGetBasis(opoutputfields[i], &basis);
      CeedChkBackend(ierr);
      ierr = CeedBasisApply(basis, numelements, CEED_TRANSPOSE,
                            CEED_EVAL_GRAD, impl->qvecsout[i],
                            impl->evecs[i + impl->numein]); CeedChkBackend(ierr);
      break;
    // LCOV_EXCL_START
    case CEED_EVAL_WEIGHT: {
      Ceed ceed;
      ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
      return CeedError(ceed, CEED_ERROR_BACKEND,
                       "CEED_EVAL_WEIGHT cannot be an output evaluation mode");
      break; // Should not occur
    }
    case CEED_EVAL_DIV:
      break; // TODO: Not implemented
    case CEED_EVAL_CURL:
      break; // TODO: Not implemented
      // LCOV_EXCL_STOP
    }
  }

  // Output restriction
  for (CeedInt i = 0; i < numoutputfields; i++) {
    // Restore evec
    ierr = CeedQFunctionFieldGetEvalMode(qfoutputfields[i], &emode);
    CeedChkBackend(ierr);
    if (emode == CEED_EVAL_NONE) {
      ierr = CeedVectorRestoreArray(impl->evecs[i+impl->numein],
                                    &edata[i + numinputfields]);
      CeedChkBackend(ierr);
    }
    // Get output vector
    ierr = CeedOperatorFieldGetVector(opoutputfields[i], &vec);
    CeedChkBackend(ierr);
    // Restrict
    ierr = CeedOperatorFieldGetElemRestriction(opoutputfields[i], &Erestrict);
    CeedChkBackend(ierr);
    // Active
    if (vec == CEED_VECTOR_ACTIVE)
      vec = outvec;

    ierr = CeedElemRestrictionApply(Erestrict, CEED_TRANSPOSE,
                                    impl->evecs[i + impl->numein], vec,
                                    request); CeedChkBackend(ierr);
  }

  // Restore input arrays
  ierr = CeedOperatorRestoreInputs_Cuda(numinputfields, qfinputfields,
                                        opinputfields, false, edata, impl);
  CeedChkBackend(ierr);
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Core code for assembling linear QFunction
//------------------------------------------------------------------------------
static inline int CeedOperatorLinearAssembleQFunctionCore_Cuda(CeedOperator op,
    bool build_objects, CeedVector *assembled, CeedElemRestriction *rstr,
    CeedRequest *request) {
  int ierr;
  CeedOperator_Cuda *impl;
  ierr = CeedOperatorGetData(op, &impl); CeedChkBackend(ierr);
  CeedQFunction qf;
  ierr = CeedOperatorGetQFunction(op, &qf); CeedChkBackend(ierr);
  CeedInt Q, numelements, numinputfields, numoutputfields, size;
  ierr = CeedOperatorGetNumQuadraturePoints(op, &Q); CeedChkBackend(ierr);
  ierr = CeedOperatorGetNumElements(op, &numelements); CeedChkBackend(ierr);
  CeedChkBackend(ierr);
  CeedOperatorField *opinputfields, *opoutputfields;
  ierr = CeedOperatorGetFields(op, &numinputfields, &opinputfields,
                               &numoutputfields, &opoutputfields);
  CeedChkBackend(ierr);
  CeedQFunctionField *qfinputfields, *qfoutputfields;
  ierr = CeedQFunctionGetFields(qf, NULL, &qfinputfields, NULL, &qfoutputfields);
  CeedChkBackend(ierr);
  CeedVector vec;
  CeedInt numactivein = impl->qfnumactivein, numactiveout = impl->qfnumactiveout;
  CeedVector *activein = impl->qfactivein;
  CeedScalar *a, *tmp;
  Ceed ceed, ceedparent;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
  ierr = CeedGetOperatorFallbackParentCeed(ceed, &ceedparent);
  CeedChkBackend(ierr);
  ceedparent = ceedparent ? ceedparent : ceed;
  CeedScalar *edata[2*CEED_FIELD_MAX];

  // Setup
  ierr = CeedOperatorSetup_Cuda(op); CeedChkBackend(ierr);

  // Check for identity
  bool identityqf;
  ierr = CeedQFunctionIsIdentity(qf, &identityqf); CeedChkBackend(ierr);
  if (identityqf)
    // LCOV_EXCL_START
    return CeedError(ceed, CEED_ERROR_BACKEND,
                     "Assembling identity QFunctions not supported");
  // LCOV_EXCL_STOP

  // Input Evecs and Restriction
  ierr = CeedOperatorSetupInputs_Cuda(numinputfields, qfinputfields,
                                      opinputfields, NULL, true, edata,
                                      impl, request); CeedChkBackend(ierr);

  // Count number of active input fields
  if (!numactivein) {
    for (CeedInt i=0; i<numinputfields; i++) {
      // Get input vector
      ierr = CeedOperatorFieldGetVector(opinputfields[i], &vec); CeedChkBackend(ierr);
      // Check if active input
      if (vec == CEED_VECTOR_ACTIVE) {
        ierr = CeedQFunctionFieldGetSize(qfinputfields[i], &size); CeedChkBackend(ierr);
        ierr = CeedVectorSetValue(impl->qvecsin[i], 0.0); CeedChkBackend(ierr);
        ierr = CeedVectorGetArray(impl->qvecsin[i], CEED_MEM_DEVICE, &tmp);
        CeedChkBackend(ierr);
        ierr = CeedRealloc(numactivein + size, &activein); CeedChkBackend(ierr);
        for (CeedInt field = 0; field < size; field++) {
          ierr = CeedVectorCreate(ceed, Q*numelements,
                                  &activein[numactivein+field]); CeedChkBackend(ierr);
          ierr = CeedVectorSetArray(activein[numactivein+field], CEED_MEM_DEVICE,
                                    CEED_USE_POINTER, &tmp[field*Q*numelements]);
          CeedChkBackend(ierr);
        }
        numactivein += size;
        ierr = CeedVectorRestoreArray(impl->qvecsin[i], &tmp); CeedChkBackend(ierr);
      }
    }
    impl->qfnumactivein = numactivein;
    impl->qfactivein = activein;
  }

  // Count number of active output fields
  if (!numactiveout) {
    for (CeedInt i=0; i<numoutputfields; i++) {
      // Get output vector
      ierr = CeedOperatorFieldGetVector(opoutputfields[i], &vec);
      CeedChkBackend(ierr);
      // Check if active output
      if (vec == CEED_VECTOR_ACTIVE) {
        ierr = CeedQFunctionFieldGetSize(qfoutputfields[i], &size);
        CeedChkBackend(ierr);
        numactiveout += size;
      }
    }
    impl->qfnumactiveout = numactiveout;
  }

  // Check sizes
  if (!numactivein || !numactiveout)
    // LCOV_EXCL_START
    return CeedError(ceed, CEED_ERROR_BACKEND,
                     "Cannot assemble QFunction without active inputs "
                     "and outputs");
  // LCOV_EXCL_STOP

  // Build objects if needed
  if (build_objects) {
    // Create output restriction
    CeedInt strides[3] = {1, numelements*Q, Q}; /* *NOPAD* */
    ierr = CeedElemRestrictionCreateStrided(ceedparent, numelements, Q,
                                            numactivein*numactiveout,
                                            numactivein*numactiveout*numelements*Q,
                                            strides, rstr); CeedChkBackend(ierr);
    // Create assembled vector
    ierr = CeedVectorCreate(ceedparent, numelements*Q*numactivein*numactiveout,
                            assembled); CeedChkBackend(ierr);
  }
  ierr = CeedVectorSetValue(*assembled, 0.0); CeedChkBackend(ierr);
  ierr = CeedVectorGetArray(*assembled, CEED_MEM_DEVICE, &a);
  CeedChkBackend(ierr);

  // Input basis apply
  ierr = CeedOperatorInputBasis_Cuda(numelements, qfinputfields, opinputfields,
                                     numinputfields, true, edata, impl);
  CeedChkBackend(ierr);

  // Assemble QFunction
  for (CeedInt in=0; in<numactivein; in++) {
    // Set Inputs
    ierr = CeedVectorSetValue(activein[in], 1.0); CeedChkBackend(ierr);
    if (numactivein > 1) {
      ierr = CeedVectorSetValue(activein[(in+numactivein-1)%numactivein],
                                0.0); CeedChkBackend(ierr);
    }
    // Set Outputs
    for (CeedInt out=0; out<numoutputfields; out++) {
      // Get output vector
      ierr = CeedOperatorFieldGetVector(opoutputfields[out], &vec);
      CeedChkBackend(ierr);
      // Check if active output
      if (vec == CEED_VECTOR_ACTIVE) {
        CeedVectorSetArray(impl->qvecsout[out], CEED_MEM_DEVICE,
                           CEED_USE_POINTER, a); CeedChkBackend(ierr);
        ierr = CeedQFunctionFieldGetSize(qfoutputfields[out], &size);
        CeedChkBackend(ierr);
        a += size*Q*numelements; // Advance the pointer by the size of the output
      }
    }
    // Apply QFunction
    ierr = CeedQFunctionApply(qf, Q*numelements, impl->qvecsin, impl->qvecsout);
    CeedChkBackend(ierr);
  }

  // Un-set output Qvecs to prevent accidental overwrite of Assembled
  for (CeedInt out=0; out<numoutputfields; out++) {
    // Get output vector
    ierr = CeedOperatorFieldGetVector(opoutputfields[out], &vec);
    CeedChkBackend(ierr);
    // Check if active output
    if (vec == CEED_VECTOR_ACTIVE) {
      ierr = CeedVectorTakeArray(impl->qvecsout[out], CEED_MEM_DEVICE, NULL);
      CeedChkBackend(ierr);
    }
  }

  // Restore input arrays
  ierr = CeedOperatorRestoreInputs_Cuda(numinputfields, qfinputfields,
                                        opinputfields, true, edata, impl);
  CeedChkBackend(ierr);

  // Restore output
  ierr = CeedVectorRestoreArray(*assembled, &a); CeedChkBackend(ierr);

  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Assemble Linear QFunction
//------------------------------------------------------------------------------
static int CeedOperatorLinearAssembleQFunction_Cuda(CeedOperator op,
    CeedVector *assembled, CeedElemRestriction *rstr, CeedRequest *request) {
  return CeedOperatorLinearAssembleQFunctionCore_Cuda(op, true, assembled, rstr,
         request);
}

//------------------------------------------------------------------------------
// Update Assembled Linear QFunction
//------------------------------------------------------------------------------
static int CeedOperatorLinearAssembleQFunctionUpdate_Cuda(CeedOperator op,
    CeedVector assembled, CeedElemRestriction rstr, CeedRequest *request) {
  return CeedOperatorLinearAssembleQFunctionCore_Cuda(op, false, &assembled,
         &rstr, request);
}

//------------------------------------------------------------------------------
// Diagonal assembly kernels
//------------------------------------------------------------------------------
// *INDENT-OFF*
static const char *diagonalkernels = QUOTE(

typedef enum {
  /// Perform no evaluation (either because there is no data or it is already at
  /// quadrature points)
  CEED_EVAL_NONE   = 0,
  /// Interpolate from nodes to quadrature points
  CEED_EVAL_INTERP = 1,
  /// Evaluate gradients at quadrature points from input in a nodal basis
  CEED_EVAL_GRAD   = 2,
  /// Evaluate divergence at quadrature points from input in a nodal basis
  CEED_EVAL_DIV    = 4,
  /// Evaluate curl at quadrature points from input in a nodal basis
  CEED_EVAL_CURL   = 8,
  /// Using no input, evaluate quadrature weights on the reference element
  CEED_EVAL_WEIGHT = 16,
} CeedEvalMode;

//------------------------------------------------------------------------------
// Get Basis Emode Pointer
//------------------------------------------------------------------------------
extern "C" __device__ void CeedOperatorGetBasisPointer_Cuda(const CeedScalar **basisptr,
    CeedEvalMode emode, const CeedScalar *identity, const CeedScalar *interp,
    const CeedScalar *grad) {
  switch (emode) {
  case CEED_EVAL_NONE:
    *basisptr = identity;
    break;
  case CEED_EVAL_INTERP:
    *basisptr = interp;
    break;
  case CEED_EVAL_GRAD:
    *basisptr = grad;
    break;
  case CEED_EVAL_WEIGHT:
  case CEED_EVAL_DIV:
  case CEED_EVAL_CURL:
    break; // Caught by QF Assembly
  }
}

//------------------------------------------------------------------------------
// Core code for diagonal assembly
//------------------------------------------------------------------------------
__device__ void diagonalCore(const CeedInt nelem,
    const CeedScalar maxnorm, const bool pointBlock,
    const CeedScalar *identity,
    const CeedScalar *interpin, const CeedScalar *gradin,
    const CeedScalar *interpout, const CeedScalar *gradout,
    const CeedEvalMode *emodein, const CeedEvalMode *emodeout,
    const CeedScalar *__restrict__ assembledqfarray,
    CeedScalar *__restrict__ elemdiagarray) {
  const int tid = threadIdx.x; // running with P threads, tid is evec node
  const CeedScalar qfvaluebound = maxnorm*1e-12;

  // Compute the diagonal of B^T D B
  // Each element
  for (CeedInt e = blockIdx.x*blockDim.z + threadIdx.z; e < nelem;
       e += gridDim.x*blockDim.z) {
    CeedInt dout = -1;
    // Each basis eval mode pair
    for (CeedInt eout = 0; eout < NUMEMODEOUT; eout++) {
      const CeedScalar *bt = NULL;
      if (emodeout[eout] == CEED_EVAL_GRAD)
        dout += 1;
      CeedOperatorGetBasisPointer_Cuda(&bt, emodeout[eout], identity, interpout,
                                      &gradout[dout*NQPTS*NNODES]);
      CeedInt din = -1;
      for (CeedInt ein = 0; ein < NUMEMODEIN; ein++) {
        const CeedScalar *b = NULL;
        if (emodein[ein] == CEED_EVAL_GRAD)
          din += 1;
        CeedOperatorGetBasisPointer_Cuda(&b, emodein[ein], identity, interpin,
                                        &gradin[din*NQPTS*NNODES]);
        // Each component
        for (CeedInt compOut = 0; compOut < NCOMP; compOut++) {
          // Each qpoint/node pair
          if (pointBlock) {
            // Point Block Diagonal
            for (CeedInt compIn = 0; compIn < NCOMP; compIn++) {
              CeedScalar evalue = 0.;
              for (CeedInt q = 0; q < NQPTS; q++) {
                const CeedScalar qfvalue =
                  assembledqfarray[((((ein*NCOMP+compIn)*NUMEMODEOUT+eout)*
                                     NCOMP+compOut)*nelem+e)*NQPTS+q];
                if (abs(qfvalue) > qfvaluebound)
                  evalue += bt[q*NNODES+tid] * qfvalue * b[q*NNODES+tid];
              }
              elemdiagarray[((compOut*NCOMP+compIn)*nelem+e)*NNODES+tid] += evalue;
            }
          } else {
            // Diagonal Only
            CeedScalar evalue = 0.;
            for (CeedInt q = 0; q < NQPTS; q++) {
              const CeedScalar qfvalue =
                assembledqfarray[((((ein*NCOMP+compOut)*NUMEMODEOUT+eout)*
                                   NCOMP+compOut)*nelem+e)*NQPTS+q];
              if (abs(qfvalue) > qfvaluebound)
                evalue += bt[q*NNODES+tid] * qfvalue * b[q*NNODES+tid];
            }
            elemdiagarray[(compOut*nelem+e)*NNODES+tid] += evalue;
          }
        }
      }
    }
  }
}

//------------------------------------------------------------------------------
// Linear diagonal
//------------------------------------------------------------------------------
extern "C" __global__ void linearDiagonal(const CeedInt nelem,
    const CeedScalar maxnorm, const CeedScalar *identity,
    const CeedScalar *interpin, const CeedScalar *gradin,
    const CeedScalar *interpout, const CeedScalar *gradout,
    const CeedEvalMode *emodein, const CeedEvalMode *emodeout,
    const CeedScalar *__restrict__ assembledqfarray,
    CeedScalar *__restrict__ elemdiagarray) {
  diagonalCore(nelem, maxnorm, false, identity, interpin, gradin, interpout,
               gradout, emodein, emodeout, assembledqfarray, elemdiagarray);
}

//------------------------------------------------------------------------------
// Linear point block diagonal
//------------------------------------------------------------------------------
extern "C" __global__ void linearPointBlockDiagonal(const CeedInt nelem,
    const CeedScalar maxnorm, const CeedScalar *identity,
    const CeedScalar *interpin, const CeedScalar *gradin,
    const CeedScalar *interpout, const CeedScalar *gradout,
    const CeedEvalMode *emodein, const CeedEvalMode *emodeout,
    const CeedScalar *__restrict__ assembledqfarray,
    CeedScalar *__restrict__ elemdiagarray) {
  diagonalCore(nelem, maxnorm, true, identity, interpin, gradin, interpout,
               gradout, emodein, emodeout, assembledqfarray, elemdiagarray);
}

);
// *INDENT-ON*

//------------------------------------------------------------------------------
// Create point block restriction
//------------------------------------------------------------------------------
static int CreatePBRestriction(CeedElemRestriction rstr,
                               CeedElemRestriction *pbRstr) {
  int ierr;
  Ceed ceed;
  ierr = CeedElemRestrictionGetCeed(rstr, &ceed); CeedChkBackend(ierr);
  const CeedInt *offsets;
  ierr = CeedElemRestrictionGetOffsets(rstr, CEED_MEM_HOST, &offsets);
  CeedChkBackend(ierr);

  // Expand offsets
  CeedInt nelem, ncomp, elemsize, compstride, max = 1, *pbOffsets;
  ierr = CeedElemRestrictionGetNumElements(rstr, &nelem); CeedChkBackend(ierr);
  ierr = CeedElemRestrictionGetNumComponents(rstr, &ncomp); CeedChkBackend(ierr);
  ierr = CeedElemRestrictionGetElementSize(rstr, &elemsize); CeedChkBackend(ierr);
  ierr = CeedElemRestrictionGetCompStride(rstr, &compstride);
  CeedChkBackend(ierr);
  CeedInt shift = ncomp;
  if (compstride != 1)
    shift *= ncomp;
  ierr = CeedCalloc(nelem*elemsize, &pbOffsets); CeedChkBackend(ierr);
  for (CeedInt i = 0; i < nelem*elemsize; i++) {
    pbOffsets[i] = offsets[i]*shift;
    if (pbOffsets[i] > max)
      max = pbOffsets[i];
  }

  // Create new restriction
  ierr = CeedElemRestrictionCreate(ceed, nelem, elemsize, ncomp*ncomp, 1,
                                   max + ncomp*ncomp, CEED_MEM_HOST,
                                   CEED_OWN_POINTER, pbOffsets, pbRstr);
  CeedChkBackend(ierr);

  // Cleanup
  ierr = CeedElemRestrictionRestoreOffsets(rstr, &offsets); CeedChkBackend(ierr);

  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Assemble diagonal setup
//------------------------------------------------------------------------------
static inline int CeedOperatorAssembleDiagonalSetup_Cuda(CeedOperator op,
    const bool pointBlock) {
  int ierr;
  Ceed ceed;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
  CeedQFunction qf;
  ierr = CeedOperatorGetQFunction(op, &qf); CeedChkBackend(ierr);
  CeedInt numinputfields, numoutputfields;
  ierr = CeedQFunctionGetNumArgs(qf, &numinputfields, &numoutputfields);
  CeedChkBackend(ierr);

  // Determine active input basis
  CeedOperatorField *opfields;
  CeedQFunctionField *qffields;
  ierr = CeedOperatorGetFields(op, NULL, &opfields, NULL, NULL);
  CeedChkBackend(ierr);
  ierr = CeedQFunctionGetFields(qf, NULL, &qffields, NULL, NULL);
  CeedChkBackend(ierr);
  CeedInt numemodein = 0, ncomp = 0, dim = 1;
  CeedEvalMode *emodein = NULL;
  CeedBasis basisin = NULL;
  CeedElemRestriction rstrin = NULL;
  for (CeedInt i = 0; i < numinputfields; i++) {
    CeedVector vec;
    ierr = CeedOperatorFieldGetVector(opfields[i], &vec); CeedChkBackend(ierr);
    if (vec == CEED_VECTOR_ACTIVE) {
      CeedElemRestriction rstr;
      ierr = CeedOperatorFieldGetBasis(opfields[i], &basisin); CeedChkBackend(ierr);
      ierr = CeedBasisGetNumComponents(basisin, &ncomp); CeedChkBackend(ierr);
      ierr = CeedBasisGetDimension(basisin, &dim); CeedChkBackend(ierr);
      ierr = CeedOperatorFieldGetElemRestriction(opfields[i], &rstr);
      CeedChkBackend(ierr);
      if (rstrin && rstrin != rstr)
        // LCOV_EXCL_START
        return CeedError(ceed, CEED_ERROR_BACKEND,
                         "Multi-field non-composite operator diagonal assembly not supported");
      // LCOV_EXCL_STOP
      rstrin = rstr;
      CeedEvalMode emode;
      ierr = CeedQFunctionFieldGetEvalMode(qffields[i], &emode);
      CeedChkBackend(ierr);
      switch (emode) {
      case CEED_EVAL_NONE:
      case CEED_EVAL_INTERP:
        ierr = CeedRealloc(numemodein + 1, &emodein); CeedChkBackend(ierr);
        emodein[numemodein] = emode;
        numemodein += 1;
        break;
      case CEED_EVAL_GRAD:
        ierr = CeedRealloc(numemodein + dim, &emodein); CeedChkBackend(ierr);
        for (CeedInt d = 0; d < dim; d++)
          emodein[numemodein+d] = emode;
        numemodein += dim;
        break;
      case CEED_EVAL_WEIGHT:
      case CEED_EVAL_DIV:
      case CEED_EVAL_CURL:
        break; // Caught by QF Assembly
      }
    }
  }

  // Determine active output basis
  ierr = CeedOperatorGetFields(op, NULL, NULL, NULL, &opfields);
  CeedChkBackend(ierr);
  ierr = CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qffields);
  CeedChkBackend(ierr);
  CeedInt numemodeout = 0;
  CeedEvalMode *emodeout = NULL;
  CeedBasis basisout = NULL;
  CeedElemRestriction rstrout = NULL;
  for (CeedInt i = 0; i < numoutputfields; i++) {
    CeedVector vec;
    ierr = CeedOperatorFieldGetVector(opfields[i], &vec); CeedChkBackend(ierr);
    if (vec == CEED_VECTOR_ACTIVE) {
      CeedElemRestriction rstr;
      ierr = CeedOperatorFieldGetBasis(opfields[i], &basisout); CeedChkBackend(ierr);
      ierr = CeedOperatorFieldGetElemRestriction(opfields[i], &rstr);
      CeedChkBackend(ierr);
      if (rstrout && rstrout != rstr)
        // LCOV_EXCL_START
        return CeedError(ceed, CEED_ERROR_BACKEND,
                         "Multi-field non-composite operator diagonal assembly not supported");
      // LCOV_EXCL_STOP
      rstrout = rstr;
      CeedEvalMode emode;
      ierr = CeedQFunctionFieldGetEvalMode(qffields[i], &emode); CeedChkBackend(ierr);
      switch (emode) {
      case CEED_EVAL_NONE:
      case CEED_EVAL_INTERP:
        ierr = CeedRealloc(numemodeout + 1, &emodeout); CeedChkBackend(ierr);
        emodeout[numemodeout] = emode;
        numemodeout += 1;
        break;
      case CEED_EVAL_GRAD:
        ierr = CeedRealloc(numemodeout + dim, &emodeout); CeedChkBackend(ierr);
        for (CeedInt d = 0; d < dim; d++)
          emodeout[numemodeout+d] = emode;
        numemodeout += dim;
        break;
      case CEED_EVAL_WEIGHT:
      case CEED_EVAL_DIV:
      case CEED_EVAL_CURL:
        break; // Caught by QF Assembly
      }
    }
  }

  // Operator data struct
  CeedOperator_Cuda *impl;
  ierr = CeedOperatorGetData(op, &impl); CeedChkBackend(ierr);
  ierr = CeedCalloc(1, &impl->diag); CeedChkBackend(ierr);
  CeedOperatorDiag_Cuda *diag = impl->diag;
  diag->basisin = basisin;
  diag->basisout = basisout;
  diag->h_emodein = emodein;
  diag->h_emodeout = emodeout;
  diag->numemodein = numemodein;
  diag->numemodeout = numemodeout;

  // Assemble kernel
  CeedInt nnodes, nqpts;
  ierr = CeedBasisGetNumNodes(basisin, &nnodes); CeedChkBackend(ierr);
  ierr = CeedBasisGetNumQuadraturePoints(basisin, &nqpts); CeedChkBackend(ierr);
  diag->nnodes = nnodes;
  ierr = CeedCompileCuda(ceed, diagonalkernels, &diag->module, 5,
                         "NUMEMODEIN", numemodein,
                         "NUMEMODEOUT", numemodeout,
                         "NNODES", nnodes,
                         "NQPTS", nqpts,
                         "NCOMP", ncomp
                        ); CeedChk_Cu(ceed, ierr);
  ierr = CeedGetKernelCuda(ceed, diag->module, "linearDiagonal",
                           &diag->linearDiagonal); CeedChk_Cu(ceed, ierr);
  ierr = CeedGetKernelCuda(ceed, diag->module, "linearPointBlockDiagonal",
                           &diag->linearPointBlock);
  CeedChk_Cu(ceed, ierr);

  // Basis matrices
  const CeedInt qBytes = nqpts * sizeof(CeedScalar);
  const CeedInt iBytes = qBytes * nnodes;
  const CeedInt gBytes = qBytes * nnodes * dim;
  const CeedInt eBytes = sizeof(CeedEvalMode);
  const CeedScalar *interpin, *interpout, *gradin, *gradout;

  // CEED_EVAL_NONE
  CeedScalar *identity = NULL;
  bool evalNone = false;
  for (CeedInt i=0; i<numemodein; i++)
    evalNone = evalNone || (emodein[i] == CEED_EVAL_NONE);
  for (CeedInt i=0; i<numemodeout; i++)
    evalNone = evalNone || (emodeout[i] == CEED_EVAL_NONE);
  if (evalNone) {
    ierr = CeedCalloc(nqpts*nnodes, &identity); CeedChkBackend(ierr);
    for (CeedInt i=0; i<(nnodes<nqpts?nnodes:nqpts); i++)
      identity[i*nnodes+i] = 1.0;
    ierr = cudaMalloc((void **)&diag->d_identity, iBytes); CeedChk_Cu(ceed, ierr);
    ierr = cudaMemcpy(diag->d_identity, identity, iBytes,
                      cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);
  }

  // CEED_EVAL_INTERP
  ierr = CeedBasisGetInterp(basisin, &interpin); CeedChkBackend(ierr);
  ierr = cudaMalloc((void **)&diag->d_interpin, iBytes); CeedChk_Cu(ceed, ierr);
  ierr = cudaMemcpy(diag->d_interpin, interpin, iBytes,
                    cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);
  ierr = CeedBasisGetInterp(basisout, &interpout); CeedChkBackend(ierr);
  ierr = cudaMalloc((void **)&diag->d_interpout, iBytes); CeedChk_Cu(ceed, ierr);
  ierr = cudaMemcpy(diag->d_interpout, interpout, iBytes,
                    cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);

  // CEED_EVAL_GRAD
  ierr = CeedBasisGetGrad(basisin, &gradin); CeedChkBackend(ierr);
  ierr = cudaMalloc((void **)&diag->d_gradin, gBytes); CeedChk_Cu(ceed, ierr);
  ierr = cudaMemcpy(diag->d_gradin, gradin, gBytes,
                    cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);
  ierr = CeedBasisGetGrad(basisout, &gradout); CeedChkBackend(ierr);
  ierr = cudaMalloc((void **)&diag->d_gradout, gBytes); CeedChk_Cu(ceed, ierr);
  ierr = cudaMemcpy(diag->d_gradout, gradout, gBytes,
                    cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);

  // Arrays of emodes
  ierr = cudaMalloc((void **)&diag->d_emodein, numemodein * eBytes);
  CeedChk_Cu(ceed, ierr);
  ierr = cudaMemcpy(diag->d_emodein, emodein, numemodein * eBytes,
                    cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);
  ierr = cudaMalloc((void **)&diag->d_emodeout, numemodeout * eBytes);
  CeedChk_Cu(ceed, ierr);
  ierr = cudaMemcpy(diag->d_emodeout, emodeout, numemodeout * eBytes,
                    cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);

  // Restriction
  diag->diagrstr = rstrout;

  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Assemble diagonal common code
//------------------------------------------------------------------------------
static inline int CeedOperatorAssembleDiagonalCore_Cuda(CeedOperator op,
    CeedVector assembled, CeedRequest *request, const bool pointBlock) {
  int ierr;
  Ceed ceed;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
  CeedOperator_Cuda *impl;
  ierr = CeedOperatorGetData(op, &impl); CeedChkBackend(ierr);

  // Assemble QFunction
  CeedVector assembledqf;
  CeedElemRestriction rstr;
  ierr = CeedOperatorLinearAssembleQFunctionBuildOrUpdate(op, &assembledqf,
         &rstr, request); CeedChkBackend(ierr);
  ierr = CeedElemRestrictionDestroy(&rstr); CeedChkBackend(ierr);
  CeedScalar maxnorm = 0;
  ierr = CeedVectorNorm(assembledqf, CEED_NORM_MAX, &maxnorm);
  CeedChkBackend(ierr);

  // Setup
  if (!impl->diag) {
    ierr = CeedOperatorAssembleDiagonalSetup_Cuda(op, pointBlock);
    CeedChkBackend(ierr);
  }
  CeedOperatorDiag_Cuda *diag = impl->diag;
  assert(diag != NULL);

  // Restriction
  if (pointBlock && !diag->pbdiagrstr) {
    CeedElemRestriction pbdiagrstr;
    ierr = CreatePBRestriction(diag->diagrstr, &pbdiagrstr); CeedChkBackend(ierr);
    diag->pbdiagrstr = pbdiagrstr;
  }
  CeedElemRestriction diagrstr = pointBlock ? diag->pbdiagrstr : diag->diagrstr;

  // Create diagonal vector
  CeedVector elemdiag = pointBlock ? diag->pbelemdiag : diag->elemdiag;
  if (!elemdiag) {
    ierr = CeedElemRestrictionCreateVector(diagrstr, NULL, &elemdiag);
    CeedChkBackend(ierr);
    if (pointBlock)
      diag->pbelemdiag = elemdiag;
    else
      diag->elemdiag = elemdiag;
  }
  ierr = CeedVectorSetValue(elemdiag, 0.0); CeedChkBackend(ierr);

  // Assemble element operator diagonals
  CeedScalar *elemdiagarray;
  const CeedScalar *assembledqfarray;
  ierr = CeedVectorGetArray(elemdiag, CEED_MEM_DEVICE, &elemdiagarray);
  CeedChkBackend(ierr);
  ierr = CeedVectorGetArrayRead(assembledqf, CEED_MEM_DEVICE, &assembledqfarray);
  CeedChkBackend(ierr);
  CeedInt nelem;
  ierr = CeedElemRestrictionGetNumElements(diagrstr, &nelem);
  CeedChkBackend(ierr);

  // Compute the diagonal of B^T D B
  int elemsPerBlock = 1;
  int grid = nelem/elemsPerBlock+((nelem/elemsPerBlock*elemsPerBlock<nelem)?1:0);
  void *args[] = {(void *) &nelem, (void *) &maxnorm, &diag->d_identity,
                  &diag->d_interpin, &diag->d_gradin, &diag->d_interpout,
                  &diag->d_gradout, &diag->d_emodein, &diag->d_emodeout,
                  &assembledqfarray, &elemdiagarray
                 };
  if (pointBlock) {
    ierr = CeedRunKernelDimCuda(ceed, diag->linearPointBlock, grid,
                                diag->nnodes, 1, elemsPerBlock, args);
    CeedChkBackend(ierr);
  } else {
    ierr = CeedRunKernelDimCuda(ceed, diag->linearDiagonal, grid,
                                diag->nnodes, 1, elemsPerBlock, args);
    CeedChkBackend(ierr);
  }

  // Restore arrays
  ierr = CeedVectorRestoreArray(elemdiag, &elemdiagarray); CeedChkBackend(ierr);
  ierr = CeedVectorRestoreArrayRead(assembledqf, &assembledqfarray);
  CeedChkBackend(ierr);

  // Assemble local operator diagonal
  ierr = CeedElemRestrictionApply(diagrstr, CEED_TRANSPOSE, elemdiag,
                                  assembled, request); CeedChkBackend(ierr);

  // Cleanup
  ierr = CeedVectorDestroy(&assembledqf); CeedChkBackend(ierr);

  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Assemble composite diagonal common code
//------------------------------------------------------------------------------
static inline int CeedOperatorLinearAssembleAddDiagonalCompositeCore_Cuda(
  CeedOperator op, CeedVector assembled, CeedRequest *request,
  const bool pointBlock) {
  int ierr;
  CeedInt numSub;
  CeedOperator *subOperators;
  ierr = CeedOperatorGetNumSub(op, &numSub); CeedChkBackend(ierr);
  ierr = CeedOperatorGetSubList(op, &subOperators); CeedChkBackend(ierr);
  for (CeedInt i = 0; i < numSub; i++) {
    ierr = CeedOperatorAssembleDiagonalCore_Cuda(subOperators[i], assembled,
           request, pointBlock); CeedChkBackend(ierr);
  }
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Assemble Linear Diagonal
//------------------------------------------------------------------------------
static int CeedOperatorLinearAssembleAddDiagonal_Cuda(CeedOperator op,
    CeedVector assembled, CeedRequest *request) {
  int ierr;
  bool isComposite;
  ierr = CeedOperatorIsComposite(op, &isComposite); CeedChkBackend(ierr);
  if (isComposite) {
    return CeedOperatorLinearAssembleAddDiagonalCompositeCore_Cuda(op, assembled,
           request, false);
  } else {
    return CeedOperatorAssembleDiagonalCore_Cuda(op, assembled, request, false);
  }
}

//------------------------------------------------------------------------------
// Assemble Linear Point Block Diagonal
//------------------------------------------------------------------------------
static int CeedOperatorLinearAssembleAddPointBlockDiagonal_Cuda(CeedOperator op,
    CeedVector assembled, CeedRequest *request) {
  int ierr;
  bool isComposite;
  ierr = CeedOperatorIsComposite(op, &isComposite); CeedChkBackend(ierr);
  if (isComposite) {
    return CeedOperatorLinearAssembleAddDiagonalCompositeCore_Cuda(op, assembled,
           request, true);
  } else {
    return CeedOperatorAssembleDiagonalCore_Cuda(op, assembled, request, true);
  }
}

//------------------------------------------------------------------------------
// Matrix assembly kernel for low-order elements (2D thread block)
//------------------------------------------------------------------------------
// *INDENT-OFF*
static const char *assemblykernel = QUOTE(
extern "C" __launch_bounds__(BLOCK_SIZE)
           __global__ void linearAssemble(const CeedScalar *B_in, const CeedScalar *B_out,
                   const CeedScalar *__restrict__ qf_array,
                   CeedScalar *__restrict__ values_array) {

  // This kernel assumes B_in and B_out have the same number of quadrature points and
  // basis points.
  // TODO: expand to more general cases
  const int i = threadIdx.x; // The output row index of each B^TDB operation
  const int l = threadIdx.y; // The output column index of each B^TDB operation
			     // such that we have (Bout^T)_ij D_jk Bin_kl = C_il

  // Strides for final output ordering, determined by the reference (interface) implementation of
  // the symbolic assembly, slowest --> fastest: element, comp_in, comp_out, node_row, node_col
  const CeedInt comp_out_stride = NNODES * NNODES;
  const CeedInt comp_in_stride = comp_out_stride * NCOMP;
  const CeedInt e_stride = comp_in_stride * NCOMP;
  // Strides for QF array, slowest --> fastest:  emode_in, comp_in, emode_out, comp_out, elem, qpt
  const CeedInt qe_stride = NQPTS;
  const CeedInt qcomp_out_stride = NELEM * qe_stride;
  const CeedInt qemode_out_stride = qcomp_out_stride * NCOMP;
  const CeedInt qcomp_in_stride = qemode_out_stride * NUMEMODEOUT;
  const CeedInt qemode_in_stride = qcomp_in_stride * NCOMP;

  // Loop over each element (if necessary)
  for (CeedInt e = blockIdx.x*blockDim.z + threadIdx.z; e < NELEM;
         e += gridDim.x*blockDim.z) {
    for (CeedInt comp_in = 0; comp_in < NCOMP; comp_in++) {
      for (CeedInt comp_out = 0; comp_out < NCOMP; comp_out++) {
        CeedScalar result = 0.0;
        CeedInt qf_index_comp = qcomp_in_stride * comp_in + qcomp_out_stride * comp_out + qe_stride * e;
        for (CeedInt emode_in = 0; emode_in < NUMEMODEIN; emode_in++) {
          CeedInt b_in_index = emode_in * NQPTS * NNODES;
      	  for (CeedInt emode_out = 0; emode_out < NUMEMODEOUT; emode_out++) {
             CeedInt b_out_index = emode_out * NQPTS * NNODES;
             CeedInt qf_index = qf_index_comp + qemode_out_stride * emode_out + qemode_in_stride * emode_in;
 	     // Perform the B^T D B operation for this 'chunk' of D (the qf_array)
            for (CeedInt j = 0; j < NQPTS; j++) {
     	      result += B_out[b_out_index + j * NNODES  + i] * qf_array[qf_index + j] * B_in[b_in_index + j * NNODES + l];
	    }

          }// end of emode_out
        } // end of emode_in
        CeedInt val_index = comp_in_stride * comp_in + comp_out_stride * comp_out + e_stride * e + NNODES * i + l;
   	values_array[val_index] = result;
      } // end of out component
    } // end of in component
  } // end of element loop
}
);

//------------------------------------------------------------------------------
// Fallback kernel for larger orders (1D thread block)
//------------------------------------------------------------------------------
static const char *assemblykernelbigelem = QUOTE(
extern "C" __launch_bounds__(BLOCK_SIZE)
           __global__ void linearAssemble(const CeedScalar *B_in, const CeedScalar *B_out,
                   const CeedScalar *__restrict__ qf_array,
                   CeedScalar *__restrict__ values_array) {

  // This kernel assumes B_in and B_out have the same number of quadrature points and
  // basis points.
  // TODO: expand to more general cases
  const int l = threadIdx.x; // The output column index of each B^TDB operation
			     // such that we have (Bout^T)_ij D_jk Bin_kl = C_il

  // Strides for final output ordering, determined by the reference (interface) implementation of
  // the symbolic assembly, slowest --> fastest: element, comp_in, comp_out, node_row, node_col
  const CeedInt comp_out_stride = NNODES * NNODES;
  const CeedInt comp_in_stride = comp_out_stride * NCOMP;
  const CeedInt e_stride = comp_in_stride * NCOMP;
  // Strides for QF array, slowest --> fastest:  emode_in, comp_in, emode_out, comp_out, elem, qpt
  const CeedInt qe_stride = NQPTS;
  const CeedInt qcomp_out_stride = NELEM * qe_stride;
  const CeedInt qemode_out_stride = qcomp_out_stride * NCOMP;
  const CeedInt qcomp_in_stride = qemode_out_stride * NUMEMODEOUT;
  const CeedInt qemode_in_stride = qcomp_in_stride * NCOMP;

    // Loop over each element (if necessary)
  for (CeedInt e = blockIdx.x*blockDim.z + threadIdx.z; e < NELEM;
         e += gridDim.x*blockDim.z) {
    for (CeedInt comp_in = 0; comp_in < NCOMP; comp_in++) {
      for (CeedInt comp_out = 0; comp_out < NCOMP; comp_out++) {
        for (CeedInt i = 0; i < NNODES; i++) {
          CeedScalar result = 0.0;
          CeedInt qf_index_comp = qcomp_in_stride * comp_in + qcomp_out_stride * comp_out + qe_stride * e;
          for (CeedInt emode_in = 0; emode_in < NUMEMODEIN; emode_in++) {
            CeedInt b_in_index = emode_in * NQPTS * NNODES;
        	  for (CeedInt emode_out = 0; emode_out < NUMEMODEOUT; emode_out++) {
               CeedInt b_out_index = emode_out * NQPTS * NNODES;
               CeedInt qf_index = qf_index_comp + qemode_out_stride * emode_out + qemode_in_stride * emode_in;
   	     // Perform the B^T D B operation for this 'chunk' of D (the qf_array)
              for (CeedInt j = 0; j < NQPTS; j++) {
       	      result += B_out[b_out_index + j * NNODES  + i] * qf_array[qf_index + j] * B_in[b_in_index + j * NNODES + l];
  	    }

            }// end of emode_out
          } // end of emode_in
          CeedInt val_index = comp_in_stride * comp_in + comp_out_stride * comp_out + e_stride * e + NNODES * i + l;
     	  values_array[val_index] = result;
        } // end of loop over element node index, i
      } // end of out component
    } // end of in component
  } // end of element loop
}
);
// *INDENT-ON*

//------------------------------------------------------------------------------
// Single operator assembly setup
//------------------------------------------------------------------------------
static int CeedSingleOperatorAssembleSetup_Cuda(CeedOperator op) {
  int ierr;
  Ceed ceed;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
  CeedOperator_Cuda *impl;
  ierr = CeedOperatorGetData(op, &impl); CeedChkBackend(ierr);

  // Get intput and output fields
  CeedInt num_input_fields, num_output_fields;
  CeedOperatorField *input_fields;
  CeedOperatorField *output_fields;
  ierr = CeedOperatorGetFields(op, &num_input_fields, &input_fields,
                               &num_output_fields, &output_fields); CeedChk(ierr);

  // Determine active input basis eval mode
  CeedQFunction qf;
  ierr = CeedOperatorGetQFunction(op, &qf); CeedChk(ierr);
  CeedQFunctionField *qf_fields;
  ierr = CeedQFunctionGetFields(qf, NULL, &qf_fields, NULL, NULL); CeedChk(ierr);
  // Note that the kernel will treat each dimension of a gradient action separately;
  // i.e., when an active input has a CEED_EVAL_GRAD mode, num_emode_in will increment
  // by dim.  However, for the purposes of loading the B matrices, it will be treated
  // as one mode, and we will load/copy the entire gradient matrix at once, so
  // num_B_in_mats_to_load will be incremented by 1.
  CeedInt num_emode_in = 0, dim = 1, num_B_in_mats_to_load = 0, size_B_in = 0;
  CeedEvalMode *eval_mode_in = NULL; //will be of size num_B_in_mats_load
  CeedBasis basis_in = NULL;
  CeedInt nqpts, esize;
  CeedElemRestriction rstr_in = NULL;
  for (CeedInt i=0; i<num_input_fields; i++) {
    CeedVector vec;
    ierr = CeedOperatorFieldGetVector(input_fields[i], &vec); CeedChk(ierr);
    if (vec == CEED_VECTOR_ACTIVE) {
      ierr = CeedOperatorFieldGetBasis(input_fields[i], &basis_in);
      CeedChk(ierr);
      ierr = CeedBasisGetDimension(basis_in, &dim); CeedChk(ierr);
      ierr = CeedBasisGetNumQuadraturePoints(basis_in, &nqpts); CeedChk(ierr);
      ierr = CeedOperatorFieldGetElemRestriction(input_fields[i], &rstr_in);
      ierr = CeedElemRestrictionGetElementSize(rstr_in, &esize); CeedChk(ierr);
      CeedChk(ierr);
      CeedEvalMode eval_mode;
      ierr = CeedQFunctionFieldGetEvalMode(qf_fields[i], &eval_mode);
      CeedChk(ierr);
      if (eval_mode != CEED_EVAL_NONE) {
        ierr = CeedRealloc(num_B_in_mats_to_load + 1, &eval_mode_in); CeedChk(ierr);
        eval_mode_in[num_B_in_mats_to_load] = eval_mode;
        num_B_in_mats_to_load += 1;
        if (eval_mode == CEED_EVAL_GRAD) {
          num_emode_in += dim;
          size_B_in += dim * esize * nqpts;
        } else {
          num_emode_in +=1;
          size_B_in += esize * nqpts;
        }
      }
    }
  }

  // Determine active output basis; basis_out and rstr_out only used if same as input, TODO
  ierr = CeedQFunctionGetFields(qf, NULL, NULL, NULL, &qf_fields); CeedChk(ierr);
  CeedInt num_emode_out = 0, num_B_out_mats_to_load = 0, size_B_out = 0;
  CeedEvalMode *eval_mode_out = NULL;
  CeedBasis basis_out = NULL;
  CeedElemRestriction rstr_out = NULL;
  for (CeedInt i=0; i<num_output_fields; i++) {
    CeedVector vec;
    ierr = CeedOperatorFieldGetVector(output_fields[i], &vec); CeedChk(ierr);
    if (vec == CEED_VECTOR_ACTIVE) {
      ierr = CeedOperatorFieldGetBasis(output_fields[i], &basis_out);
      CeedChk(ierr);
      ierr = CeedOperatorFieldGetElemRestriction(output_fields[i], &rstr_out);
      CeedChk(ierr);
      if (rstr_out && rstr_out != rstr_in)
        // LCOV_EXCL_START
        return CeedError(ceed, CEED_ERROR_BACKEND,
                         "Multi-field non-composite operator assembly not supported");
      // LCOV_EXCL_STOP
      CeedEvalMode eval_mode;
      ierr = CeedQFunctionFieldGetEvalMode(qf_fields[i], &eval_mode);
      CeedChk(ierr);
      if (eval_mode != CEED_EVAL_NONE) {
        ierr = CeedRealloc(num_B_out_mats_to_load + 1, &eval_mode_out); CeedChk(ierr);
        eval_mode_out[num_B_out_mats_to_load] = eval_mode;
        num_B_out_mats_to_load += 1;
        if (eval_mode == CEED_EVAL_GRAD) {
          num_emode_out += dim;
          size_B_out += dim * esize * nqpts;
        } else {
          num_emode_out +=1;
          size_B_out += esize * nqpts;
        }
      }
    }
  }

  if (num_emode_in == 0 || num_emode_out == 0)
    // LCOV_EXCL_START
    return CeedError(ceed, CEED_ERROR_UNSUPPORTED,
                     "Cannot assemble operator without inputs/outputs");
  // LCOV_EXCL_STOP

  CeedInt nelem, ncomp;
  ierr = CeedElemRestrictionGetNumElements(rstr_in, &nelem); CeedChk(ierr);
  ierr = CeedElemRestrictionGetNumComponents(rstr_in, &ncomp); CeedChk(ierr);

  ierr = CeedCalloc(1, &impl->asmb); CeedChkBackend(ierr);
  CeedOperatorAssemble_Cuda *asmb = impl->asmb;
  asmb->nelem = nelem;

  // Compile kernels
  int elemsPerBlock = 1;
  asmb->elemsPerBlock = elemsPerBlock;
  CeedInt block_size = esize * esize * elemsPerBlock;
  Ceed_Cuda *cuda_data;
  ierr = CeedGetData(ceed, &cuda_data); CeedChkBackend(ierr);
  if (block_size > cuda_data->device_prop.maxThreadsPerBlock) {
    // Use fallback kernel with 1D threadblock
    block_size = esize * elemsPerBlock;
    asmb->block_size_x = esize;
    asmb->block_size_y = 1;
    ierr = CeedCompileCuda(ceed, assemblykernelbigelem, &asmb->module, 7,
                           "NELEM", nelem,
                           "NUMEMODEIN", num_emode_in,
                           "NUMEMODEOUT", num_emode_out,
                           "NQPTS", nqpts,
                           "NNODES", esize,
                           "BLOCK_SIZE", block_size,
                           "NCOMP", ncomp
                          ); CeedChk_Cu(ceed, ierr);
  } else {  // Use kernel with 2D threadblock
    asmb->block_size_x = esize;
    asmb->block_size_y = esize;
    ierr = CeedCompileCuda(ceed, assemblykernel, &asmb->module, 7,
                           "NELEM", nelem,
                           "NUMEMODEIN", num_emode_in,
                           "NUMEMODEOUT", num_emode_out,
                           "NQPTS", nqpts,
                           "NNODES", esize,
                           "BLOCK_SIZE", block_size,
                           "NCOMP", ncomp
                          ); CeedChk_Cu(ceed, ierr);
  }
  ierr = CeedGetKernelCuda(ceed, asmb->module, "linearAssemble",
                           &asmb->linearAssemble); CeedChk_Cu(ceed, ierr);

  // Build 'full' B matrices (not 1D arrays used for tensor-product matrices)
  const CeedScalar *interp_in, *grad_in;
  ierr = CeedBasisGetInterp(basis_in, &interp_in); CeedChkBackend(ierr);
  ierr = CeedBasisGetGrad(basis_in, &grad_in); CeedChkBackend(ierr);

  // Load into B_in, in order that they will be used in eval_mode
  const CeedInt inBytes = size_B_in * sizeof(CeedScalar);
  CeedInt mat_start = 0;
  ierr = cudaMalloc((void **) &asmb->d_B_in, inBytes); CeedChk_Cu(ceed, ierr);
  for (int i = 0; i < num_B_in_mats_to_load; i++) {
    CeedEvalMode eval_mode = eval_mode_in[i];
    if (eval_mode == CEED_EVAL_INTERP) {
      ierr = cudaMemcpy(&asmb->d_B_in[mat_start], interp_in,
                        esize * nqpts * sizeof(CeedScalar),
                        cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);
      mat_start += esize * nqpts;
    } else if (eval_mode == CEED_EVAL_GRAD) {
      ierr = cudaMemcpy(asmb->d_B_in, grad_in,
                        dim * esize * nqpts * sizeof(CeedScalar),
                        cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);
      mat_start += dim * esize * nqpts;
    }
  }

  const CeedScalar *interp_out, *grad_out;
  // Note that this function currently assumes 1 basis, so this should always be true
  // for now
  if (basis_out == basis_in) {
    interp_out = interp_in;
    grad_out = grad_in;
  } else {
    ierr = CeedBasisGetInterp(basis_out, &interp_out); CeedChkBackend(ierr);
    ierr = CeedBasisGetGrad(basis_out, &grad_out); CeedChkBackend(ierr);
  }

  // Load into B_out, in order that they will be used in eval_mode
  const CeedInt outBytes = size_B_out * sizeof(CeedScalar);
  mat_start = 0;
  ierr = cudaMalloc((void **) &asmb->d_B_out, outBytes); CeedChk_Cu(ceed, ierr);
  for (int i = 0; i < num_B_out_mats_to_load; i++) {
    CeedEvalMode eval_mode = eval_mode_out[i];
    if (eval_mode == CEED_EVAL_INTERP) {
      ierr = cudaMemcpy(&asmb->d_B_out[mat_start], interp_out,
                        esize * nqpts * sizeof(CeedScalar),
                        cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);
      mat_start += esize * nqpts;
    } else if (eval_mode == CEED_EVAL_GRAD) {
      ierr = cudaMemcpy(&asmb->d_B_out[mat_start], grad_out,
                        dim * esize * nqpts * sizeof(CeedScalar),
                        cudaMemcpyHostToDevice); CeedChk_Cu(ceed, ierr);
      mat_start += dim * esize * nqpts;
    }
  }
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Single operator assembly
//------------------------------------------------------------------------------
static int CeedSingleOperatorAssemble_Cuda(CeedOperator op, CeedInt offset,
    CeedVector values) {

  int ierr;
  Ceed ceed;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
  CeedOperator_Cuda *impl;
  ierr = CeedOperatorGetData(op, &impl); CeedChkBackend(ierr);

  // Setup
  if (!impl->asmb) {
    ierr = CeedSingleOperatorAssembleSetup_Cuda(op);
    CeedChkBackend(ierr);
  }

  // Assemble QFunction
  CeedVector assembled_qf;
  CeedElemRestriction rstr_q;
  ierr = CeedOperatorLinearAssembleQFunctionBuildOrUpdate(
           op, &assembled_qf, &rstr_q, CEED_REQUEST_IMMEDIATE); CeedChk(ierr);
  ierr = CeedElemRestrictionDestroy(&rstr_q); CeedChkBackend(ierr);
  CeedScalar *values_array;
  ierr = CeedVectorGetArrayWrite(values, CEED_MEM_DEVICE, &values_array);
  CeedChkBackend(ierr);
  values_array += offset;
  const CeedScalar *qf_array;
  ierr = CeedVectorGetArrayRead(assembled_qf, CEED_MEM_DEVICE, &qf_array);
  CeedChkBackend(ierr);

  // Compute B^T D B
  const CeedInt nelem = impl->asmb->nelem;
  const CeedInt elemsPerBlock = impl->asmb->elemsPerBlock;
  const CeedInt grid = nelem/elemsPerBlock+((
                         nelem/elemsPerBlock*elemsPerBlock<nelem)?1:0);
  void *args[] = {&impl->asmb->d_B_in, &impl->asmb->d_B_out,
                  &qf_array, &values_array
                 };
  ierr = CeedRunKernelDimCuda(ceed, impl->asmb->linearAssemble, grid,
                              impl->asmb->block_size_x, impl->asmb->block_size_y,
                              elemsPerBlock, args);
  CeedChkBackend(ierr);


  // Restore arrays
  ierr = CeedVectorRestoreArray(values, &values_array); CeedChkBackend(ierr);
  ierr = CeedVectorRestoreArrayRead(assembled_qf, &qf_array);
  CeedChkBackend(ierr);

  // Cleanup
  ierr = CeedVectorDestroy(&assembled_qf); CeedChkBackend(ierr);

  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Assemble matrix data for COO matrix of assembled operator.
// The sparsity pattern is set by CeedOperatorLinearAssembleSymbolic.
//
// Note that this (and other assembly routines) currently assume only one
// active input restriction/basis per operator (could have multiple basis eval
// modes).
// TODO: allow multiple active input restrictions/basis objects
//------------------------------------------------------------------------------
int CeedOperatorLinearAssemble_Cuda(CeedOperator op, CeedVector values) {

  // As done in the default implementation, loop through suboperators
  // for composite operators, or call single operator assembly otherwise
  bool is_composite;
  CeedInt ierr;
  ierr = CeedOperatorIsComposite(op, &is_composite); CeedChk(ierr);

  CeedElemRestriction rstr;
  CeedInt num_elem, elem_size, num_comp;

  CeedInt offset = 0;
  if (is_composite) {
    CeedInt num_suboperators;
    ierr = CeedOperatorGetNumSub(op, &num_suboperators); CeedChk(ierr);
    CeedOperator *sub_operators;
    ierr = CeedOperatorGetSubList(op, &sub_operators); CeedChk(ierr);
    for (int k = 0; k < num_suboperators; ++k) {
      ierr = CeedSingleOperatorAssemble_Cuda(sub_operators[k], offset, values);
      CeedChk(ierr);
      ierr = CeedOperatorGetActiveElemRestriction(sub_operators[k], &rstr);
      CeedChk(ierr);
      ierr = CeedElemRestrictionGetNumElements(rstr, &num_elem); CeedChk(ierr);
      ierr = CeedElemRestrictionGetElementSize(rstr, &elem_size); CeedChk(ierr);
      ierr = CeedElemRestrictionGetNumComponents(rstr, &num_comp); CeedChk(ierr);
      offset += elem_size*num_comp * elem_size*num_comp * num_elem;
    }
  } else {
    ierr = CeedSingleOperatorAssemble_Cuda(op, offset, values); CeedChk(ierr);
  }

  return CEED_ERROR_SUCCESS;
}
//------------------------------------------------------------------------------
// Create operator
//------------------------------------------------------------------------------
int CeedOperatorCreate_Cuda(CeedOperator op) {
  int ierr;
  Ceed ceed;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);
  CeedOperator_Cuda *impl;

  ierr = CeedCalloc(1, &impl); CeedChkBackend(ierr);
  ierr = CeedOperatorSetData(op, impl); CeedChkBackend(ierr);

  ierr = CeedSetBackendFunction(ceed, "Operator", op, "LinearAssembleQFunction",
                                CeedOperatorLinearAssembleQFunction_Cuda);
  CeedChkBackend(ierr);
  ierr = CeedSetBackendFunction(ceed, "Operator", op,
                                "LinearAssembleQFunctionUpdate",
                                CeedOperatorLinearAssembleQFunctionUpdate_Cuda);
  CeedChkBackend(ierr);
  ierr = CeedSetBackendFunction(ceed, "Operator", op, "LinearAssembleAddDiagonal",
                                CeedOperatorLinearAssembleAddDiagonal_Cuda);
  CeedChkBackend(ierr);
  ierr = CeedSetBackendFunction(ceed, "Operator", op,
                                "LinearAssembleAddPointBlockDiagonal",
                                CeedOperatorLinearAssembleAddPointBlockDiagonal_Cuda);
  CeedChkBackend(ierr);
  ierr = CeedSetBackendFunction(ceed, "Operator", op,
                                "LinearAssemble", CeedOperatorLinearAssemble_Cuda);
  CeedChkBackend(ierr);
  ierr = CeedSetBackendFunction(ceed, "Operator", op, "ApplyAdd",
                                CeedOperatorApplyAdd_Cuda); CeedChkBackend(ierr);
  ierr = CeedSetBackendFunction(ceed, "Operator", op, "Destroy",
                                CeedOperatorDestroy_Cuda); CeedChkBackend(ierr);
  return CEED_ERROR_SUCCESS;
}

//------------------------------------------------------------------------------
// Composite Operator Create
//------------------------------------------------------------------------------
int CeedCompositeOperatorCreate_Cuda(CeedOperator op) {
  int ierr;
  Ceed ceed;
  ierr = CeedOperatorGetCeed(op, &ceed); CeedChkBackend(ierr);

  ierr = CeedSetBackendFunction(ceed, "Operator", op, "LinearAssembleAddDiagonal",
                                CeedOperatorLinearAssembleAddDiagonal_Cuda);
  CeedChkBackend(ierr);
  ierr = CeedSetBackendFunction(ceed, "Operator", op,
                                "LinearAssembleAddPointBlockDiagonal",
                                CeedOperatorLinearAssembleAddPointBlockDiagonal_Cuda);
  CeedChkBackend(ierr);
  ierr = CeedSetBackendFunction(ceed, "Operator", op,
                                "LinearAssemble", CeedOperatorLinearAssemble_Cuda);
  CeedChkBackend(ierr);
  return CEED_ERROR_SUCCESS;
}
//------------------------------------------------------------------------------