mfem/general/forall.hpp

// 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.

#ifndef MFEM_FORALL_HPP
#define MFEM_FORALL_HPP

#include "../config/config.hpp"
#include "annotation.hpp"
#include "error.hpp"
#include "backends.hpp"
#include "device.hpp"
#include "mem_manager.hpp"
#include "../linalg/dtensor.hpp"
#ifdef MFEM_USE_MPI
#include <_hypre_utilities.h>
#endif

namespace mfem
{

// The following DofQuadLimit_ structs define the maximum values of D1D and Q1D
// often used in the "fallback kernels" for partial assembly. Different limits
// take effect for different architectures. The limits should be queried using
// the public interface in DeviceDofQuadLimits or DofQuadLimits, and generally
// not be directly accessing the structs defined below.
//
// In host code, the limits associated with the currently configured Device can
// be accessed using DeviceDofQuadLimits::Get().
//
// In mfem::forall kernels or MFEM_HOST_DEVICE functions, the limits
// corresponding to the architecture the function is being compiled for can be
// accessed as static constexpr variables using the type alias DofQuadLimits.

namespace internal
{

struct DofQuadLimits_CUDA
{
   static constexpr int MAX_D1D = 14;
   static constexpr int MAX_Q1D = 14;
   static constexpr int MAX_T1D = 32;
   static constexpr int HCURL_MAX_D1D = 5;
   static constexpr int HCURL_MAX_Q1D = 6;
   static constexpr int HDIV_MAX_D1D = 5;
   static constexpr int HDIV_MAX_Q1D = 6;
   static constexpr int MAX_INTERP_1D = 8;
   static constexpr int MAX_DET_1D = 6;
};

struct DofQuadLimits_HIP
{
   static constexpr int MAX_D1D = 10;
   static constexpr int MAX_Q1D = 10;
   static constexpr int MAX_T1D = 32;
   static constexpr int HCURL_MAX_D1D = 5;
   static constexpr int HCURL_MAX_Q1D = 5;
   static constexpr int HDIV_MAX_D1D = 5;
   static constexpr int HDIV_MAX_Q1D = 6;
   static constexpr int MAX_INTERP_1D = 8;
   static constexpr int MAX_DET_1D = 6;
};

struct DofQuadLimits_CPU
{
#ifndef _WIN32
   static constexpr int MAX_D1D = 24;
   static constexpr int MAX_Q1D = 24;
#else
   static constexpr int MAX_D1D = 14;
   static constexpr int MAX_Q1D = 14;
#endif
   static constexpr int MAX_T1D = 32;
   static constexpr int HCURL_MAX_D1D = 10;
   static constexpr int HCURL_MAX_Q1D = 10;
   static constexpr int HDIV_MAX_D1D = 10;
   static constexpr int HDIV_MAX_Q1D = 10;
   static constexpr int MAX_INTERP_1D = MAX_D1D;
   static constexpr int MAX_DET_1D = MAX_D1D;
};

} // namespace internal

/// @brief Maximum number of 1D DOFs or quadrature points for the architecture
/// currently being compiled for (used in fallback kernels).
///
/// DofQuadLimits provides access to the limits as static constexpr member
/// variables for use in mfem::forall kernels or MFEM_HOST_DEVICE functions.
///
/// @sa For accessing the limits according to the runtime configuration of the
/// Device, see DeviceDofQuadLimits.
#if defined(__CUDA_ARCH__)
using DofQuadLimits = internal::DofQuadLimits_CUDA;
#elif defined(__HIP_DEVICE_COMPILE__)
using DofQuadLimits = internal::DofQuadLimits_HIP;
#else
using DofQuadLimits = internal::DofQuadLimits_CPU;
#endif

/// @brief Maximum number of 1D DOFs or quadrature points for the current
/// runtime configuration of the Device (used in fallback kernels).
///
/// DeviceDofQuadLimits can be used in host code to query the limits for the
/// configured device (e.g. when the user has selected GPU execution at
/// runtime).
///
/// @sa For accessing the limits according to the current compiler pass, see
/// DofQuadLimits.
struct DeviceDofQuadLimits
{
   int MAX_D1D; ///< Maximum number of 1D nodal points.
   int MAX_Q1D; ///< Maximum number of 1D quadrature points.
   int HCURL_MAX_D1D; ///< Maximum number of 1D nodal points for H(curl).
   int HCURL_MAX_Q1D; ///< Maximum number of 1D quadrature points for H(curl).
   int HDIV_MAX_D1D; ///< Maximum number of 1D nodal points for H(div).
   int HDIV_MAX_Q1D; ///< Maximum number of 1D quadrature points for H(div).
   int MAX_INTERP_1D; ///< Maximum number of points for use in QuadratureInterpolator.
   int MAX_DET_1D; ///< Maximum number of points for determinant computation in QuadratureInterpolator.

   /// Return a const reference to the DeviceDofQuadLimits singleton.
   static const DeviceDofQuadLimits &Get()
   {
      static const DeviceDofQuadLimits dof_quad_limits;
      return dof_quad_limits;
   }

private:
   /// Initialize the limits depending on the configuration of the Device.
   DeviceDofQuadLimits()
   {
      if (Device::Allows(Backend::CUDA_MASK)) { Populate<internal::DofQuadLimits_CUDA>(); }
      else if (Device::Allows(Backend::HIP_MASK)) { Populate<internal::DofQuadLimits_HIP>(); }
      else { Populate<internal::DofQuadLimits_CPU>(); }
   }

   /// @brief Set the limits using the static members of the type @a T.
   ///
   /// @a T should be one of DofQuadLimits_CUDA, DofQuadLimits_HIP, or
   /// DofQuadLimits_CPU.
   template <typename T> void Populate()
   {
      MAX_D1D = T::MAX_D1D;
      MAX_Q1D = T::MAX_Q1D;
      HCURL_MAX_D1D = T::HCURL_MAX_D1D;
      HCURL_MAX_Q1D = T::HCURL_MAX_Q1D;
      HDIV_MAX_D1D = T::HDIV_MAX_D1D;
      HDIV_MAX_Q1D = T::HDIV_MAX_Q1D;
      MAX_INTERP_1D = T::MAX_INTERP_1D;
      MAX_DET_1D = T::MAX_DET_1D;
   }
};

// MFEM pragma macros that can be used inside MFEM_FORALL macros.
#define MFEM_PRAGMA(X) _Pragma(#X)

// MFEM_UNROLL pragma macro that can be used inside MFEM_FORALL macros.
#if defined(MFEM_USE_CUDA) && defined(__CUDA_ARCH__) // Clang cuda or nvcc
#ifdef __NVCC__ // nvcc specifically
#define MFEM_UNROLL(N) MFEM_PRAGMA(unroll(N))
#else // Assuming Clang CUDA
#define MFEM_UNROLL(N) MFEM_PRAGMA(unroll N)
#endif
#else
#define MFEM_UNROLL(N)
#endif

// MFEM_GPU_FORALL: "parallel for" executed with CUDA or HIP based on the MFEM
// build-time configuration (MFEM_USE_CUDA or MFEM_USE_HIP), and if compiling
// with CUDA/HIP language. Otherwise, this macro is a no-op.
#if defined(MFEM_USE_CUDA) && defined(__CUDACC__)
#define MFEM_GPU_FORALL(i, N,...) CuWrap1D(N, [=] MFEM_DEVICE      \
                                       (int i) {__VA_ARGS__})
#elif defined(MFEM_USE_HIP) && defined(__HIP__)
#define MFEM_GPU_FORALL(i, N,...) HipWrap1D(N, [=] MFEM_DEVICE     \
                                        (int i) {__VA_ARGS__})
#else
#define MFEM_GPU_FORALL(i, N,...) do { } while (false)
#endif

// Implementation of MFEM's "parallel for" (forall) device/host kernel
// interfaces supporting RAJA, CUDA, OpenMP, and sequential backends.

// The MFEM_FORALL wrapper
#define MFEM_FORALL(i,N,...) \
   ForallWrap<1>(true,N,[=] MFEM_HOST_DEVICE (int i) {__VA_ARGS__})

// MFEM_FORALL with a 2D CUDA block
#define MFEM_FORALL_2D(i,N,X,Y,BZ,...) \
   ForallWrap<2>(true,N,[=] MFEM_HOST_DEVICE (int i) {__VA_ARGS__},X,Y,BZ)

// MFEM_FORALL with a 3D CUDA block
#define MFEM_FORALL_3D(i,N,X,Y,Z,...) \
   ForallWrap<3>(true,N,[=] MFEM_HOST_DEVICE (int i) {__VA_ARGS__},X,Y,Z)

// MFEM_FORALL with a 3D CUDA block and grid
// With G=0, this is the same as MFEM_FORALL_3D(i,N,X,Y,Z,...)
#define MFEM_FORALL_3D_GRID(i,N,X,Y,Z,G,...) \
   ForallWrap<3>(true,N,[=] MFEM_HOST_DEVICE (int i) {__VA_ARGS__},X,Y,Z,G)

// MFEM_FORALL that uses the basic CPU backend when use_dev is false. See for
// example the functions in vector.cpp, where we don't want to use the mfem
// device for operations on small vectors.
#define MFEM_FORALL_SWITCH(use_dev,i,N,...) \
   ForallWrap<1>(use_dev,N,[=] MFEM_HOST_DEVICE (int i) {__VA_ARGS__})


/// OpenMP backend
template <typename HBODY>
void OmpWrap(const int N, HBODY &&h_body)
{
#ifdef MFEM_USE_OPENMP
   #pragma omp parallel for
   for (int k = 0; k < N; k++)
   {
      h_body(k);
   }
#else
   MFEM_CONTRACT_VAR(N);
   MFEM_CONTRACT_VAR(h_body);
   MFEM_ABORT("OpenMP requested for MFEM but OpenMP is not enabled!");
#endif
}

template <typename HBODY>
void OmpWrap2D(const int Nx, const int Ny, HBODY &&h_body)
{
#ifdef MFEM_USE_OPENMP
   // requires OpenMP 3.1
   #pragma omp parallel for collapse(2)
   for (int j = 0; j < Ny; j++)
   {
      for (int i = 0; i < Nx; i++)
      {
         h_body(i, j);
      }
   }
#else
   MFEM_CONTRACT_VAR(Nx);
   MFEM_CONTRACT_VAR(Ny);
   MFEM_CONTRACT_VAR(h_body);
   MFEM_ABORT("OpenMP requested for MFEM but OpenMP is not enabled!");
#endif
}

template <typename HBODY>
void OmpWrap3D(const int Nx, const int Ny, const int Nz, HBODY &&h_body)
{
#ifdef MFEM_USE_OPENMP
   // requires OpenMP 3.1
   #pragma omp parallel for collapse(3)
   for (int k = 0; k < Nz; k++)
   {
      for (int j = 0; j < Ny; j++)
      {
         for (int i = 0; i < Nx; i++)
         {
            h_body(i, j, k);
         }
      }
   }
#else
   MFEM_CONTRACT_VAR(Nx);
   MFEM_CONTRACT_VAR(Ny);
   MFEM_CONTRACT_VAR(Nz);
   MFEM_CONTRACT_VAR(h_body);
   MFEM_ABORT("OpenMP requested for MFEM but OpenMP is not enabled!");
#endif
}


/// RAJA Cuda and Hip backends
#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_CUDA) && defined(__CUDACC__)
using cuda_launch_policy =
   RAJA::LaunchPolicy<RAJA::cuda_launch_t<true>>;
using cuda_teams_x =
   RAJA::LoopPolicy<RAJA::cuda_block_x_direct>;
using cuda_threads_z =
   RAJA::LoopPolicy<RAJA::cuda_thread_z_direct>;
#endif

#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_HIP) && defined(__HIP__)
using hip_launch_policy =
   RAJA::LaunchPolicy<RAJA::hip_launch_t<true>>;
using hip_teams_x =
   RAJA::LoopPolicy<RAJA::hip_block_x_direct>;
using hip_threads_z =
   RAJA::LoopPolicy<RAJA::hip_thread_z_direct>;
#endif

#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_CUDA) && defined(__CUDACC__)
template <typename DBODY>
void RajaCuWrap1D(const int N, DBODY &&d_body)
{
   //true denotes asynchronous kernel
   RAJA::forall<RAJA::cuda_exec<MFEM_CUDA_BLOCKS,true>>(RAJA::RangeSegment(0,N),
                                                        d_body);
}

template <typename DBODY>
void RajaCuWrap2D(const int N, DBODY &&d_body,
                  const int X, const int Y, const int BZ)
{
   MFEM_VERIFY(BZ>0, "");
   const int G = (N+BZ-1)/BZ;

   using namespace RAJA;
   using RAJA::RangeSegment;

   launch<cuda_launch_policy>
   (LaunchParams(Teams(G), Threads(X, Y, BZ)),
    [=] RAJA_DEVICE (LaunchContext ctx)
   {

      loop<cuda_teams_x>(ctx, RangeSegment(0, G), [&] (const int n)
      {

         loop<cuda_threads_z>(ctx, RangeSegment(0, BZ), [&] (const int tz)
         {

            const int k = n*BZ + tz;
            if (k >= N) { return; }
            d_body(k);

         });

      });

   });

   MFEM_GPU_CHECK(cudaGetLastError());
}

template <typename DBODY>
void RajaCuWrap3D(const int N, DBODY &&d_body,
                  const int X, const int Y, const int Z, const int G)
{
   const int GRID = G == 0 ? N : G;
   using namespace RAJA;
   using RAJA::RangeSegment;

   launch<cuda_launch_policy>
   (LaunchParams(Teams(GRID), Threads(X, Y, Z)),
    [=] RAJA_DEVICE (LaunchContext ctx)
   {

      loop<cuda_teams_x>(ctx, RangeSegment(0, N), d_body);

   });

   MFEM_GPU_CHECK(cudaGetLastError());
}

template <int Dim>
struct RajaCuWrap;

template <>
struct RajaCuWrap<1>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      RajaCuWrap1D(N, d_body);
   }
};

template <>
struct RajaCuWrap<2>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      RajaCuWrap2D(N, d_body, X, Y, Z);
   }
};

template <>
struct RajaCuWrap<3>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      RajaCuWrap3D(N, d_body, X, Y, Z, G);
   }
};

#endif

#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_HIP) && defined(__HIP__)
template <typename DBODY>
void RajaHipWrap1D(const int N, DBODY &&d_body)
{
   //true denotes asynchronous kernel
   RAJA::forall<RAJA::hip_exec<MFEM_HIP_BLOCKS,true>>(RAJA::RangeSegment(0,N),
                                                      d_body);
}

template <typename DBODY>
void RajaHipWrap2D(const int N, DBODY &&d_body,
                   const int X, const int Y, const int BZ)
{
   MFEM_VERIFY(BZ>0, "");
   const int G = (N+BZ-1)/BZ;

   using namespace RAJA;
   using RAJA::RangeSegment;

   launch<hip_launch_policy>
   (LaunchParams(Teams(G), Threads(X, Y, BZ)),
    [=] RAJA_DEVICE (LaunchContext ctx)
   {

      loop<hip_teams_x>(ctx, RangeSegment(0, G), [&] (const int n)
      {

         loop<hip_threads_z>(ctx, RangeSegment(0, BZ), [&] (const int tz)
         {

            const int k = n*BZ + tz;
            if (k >= N) { return; }
            d_body(k);

         });

      });

   });

   MFEM_GPU_CHECK(hipGetLastError());
}

template <typename DBODY>
void RajaHipWrap3D(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
{
   const int GRID = G == 0 ? N : G;
   using namespace RAJA;
   using RAJA::RangeSegment;

   launch<hip_launch_policy>
   (LaunchParams(Teams(GRID), Threads(X, Y, Z)),
    [=] RAJA_DEVICE (LaunchContext ctx)
   {

      loop<hip_teams_x>(ctx, RangeSegment(0, N), d_body);

   });

   MFEM_GPU_CHECK(hipGetLastError());
}

template <int Dim>
struct RajaHipWrap;

template <>
struct RajaHipWrap<1>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      RajaHipWrap1D(N, d_body);
   }
};

template <>
struct RajaHipWrap<2>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      RajaHipWrap2D(N, d_body, X, Y, Z);
   }
};

template <>
struct RajaHipWrap<3>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      RajaHipWrap3D(N, d_body, X, Y, Z, G);
   }
};

#endif

/// RAJA OpenMP backend
#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_OPENMP)

template <typename HBODY>
void RajaOmpWrap(const int N, HBODY &&h_body)
{
   RAJA::forall<RAJA::omp_parallel_for_exec>(RAJA::RangeSegment(0,N), h_body);
}

template <typename HBODY>
void RajaOmpWrap2D(const int Nx, const int Ny, HBODY &&h_body)
{
   using omp_launch_policy = RAJA::LaunchPolicy<RAJA::omp_launch_t>;
   using global_thread_xy = RAJA::LoopPolicy<RAJA::omp_for_exec>;
   RAJA::RangeSegment xrange(0, Nx);
   RAJA::RangeSegment yrange(0, Ny);
   RAJA::launch<omp_launch_policy>(RAJA::ExecPlace::HOST, RAJA::LaunchParams(),
                                   [=](RAJA::LaunchContext ctx)
   {
      // contiguous in x
      RAJA::expt::loop<global_thread_xy>(ctx, xrange, yrange, [&](int i, int j)
      {
         h_body(i, j);
      });
   });
}

template <typename HBODY>
void RajaOmpWrap3D(const int Nx, const int Ny, const int Nz, HBODY &&h_body)
{
   using omp_launch_policy = RAJA::LaunchPolicy<RAJA::omp_launch_t>;
   using global_thread_xyz = RAJA::LoopPolicy<RAJA::omp_for_exec>;
   RAJA::RangeSegment xrange(0, Nx);
   RAJA::RangeSegment yrange(0, Ny);
   RAJA::RangeSegment zrange(0, Nz);
   RAJA::launch<omp_launch_policy>(RAJA::ExecPlace::HOST, RAJA::LaunchParams(),
                                   [=](RAJA::LaunchContext ctx)
   {
      // contiguous in x
      RAJA::expt::loop<global_thread_xyz>(ctx, xrange, yrange, zrange,
                                          [&](int i, int j, int k)
      { h_body(i, j, k); });
   });
}

#endif


/// RAJA sequential loop backend
template <typename HBODY>
void RajaSeqWrap(const int N, HBODY &&h_body)
{
#ifdef MFEM_USE_RAJA

#if (RAJA_VERSION_MAJOR >= 2023)
   //loop_exec was marked deprecated in RAJA version 2023.06.0
   //and will be removed. We now use seq_exec.
   using raja_forall_pol = RAJA::seq_exec;
#else
   using raja_forall_pol = RAJA::loop_exec;
#endif

   RAJA::forall<raja_forall_pol>(RAJA::RangeSegment(0,N), h_body);
#else
   MFEM_CONTRACT_VAR(N);
   MFEM_CONTRACT_VAR(h_body);
   MFEM_ABORT("RAJA requested but RAJA is not enabled!");
#endif
}


/// CUDA backend
#if defined(MFEM_USE_CUDA) && defined(__CUDACC__)

template <typename BODY> __global__ static
void CuKernel1D(const int N, BODY body)
{
   const int k = blockDim.x*blockIdx.x + threadIdx.x;
   if (k >= N) { return; }
   body(k);
}

template <typename BODY> __global__ static
void CuKernel2D(const int N, BODY body)
{
   const int k = blockIdx.x*blockDim.z + threadIdx.z;
   if (k >= N) { return; }
   body(k);
}

// __launch_bounds__ second argument is omitted to get the default behavior
template <int MAX_THREADS_PER_BLOCK, typename BODY>
__global__
MFEM_LAUNCH_BOUNDS(MAX_THREADS_PER_BLOCK)
static void CuKernel2DLaunchBounds(const int N, BODY body)
{
   const int k = blockIdx.x*blockDim.z + threadIdx.z;
   if (k >= N) { return; }
   body(k);
}

template <typename BODY> __global__ static
void CuKernel3D(const int N, BODY body)
{
   for (int k = blockIdx.x; k < N; k += gridDim.x) { body(k); }
}

template <int MAX_THREADS_PER_BLOCK, typename BODY>
__global__
MFEM_LAUNCH_BOUNDS(MAX_THREADS_PER_BLOCK)
static void CuKernel3DLaunchBounds(const int N, BODY body)
{
   for (int k = blockIdx.x; k < N; k += gridDim.x) { body(k); }
}

template <const int BLCK = MFEM_CUDA_BLOCKS, typename DBODY>
void CuWrap1D(const int N, DBODY &&d_body)
{
   if (N==0) { return; }
   const int GRID = (N+BLCK-1)/BLCK;
   CuKernel1D<<<GRID,BLCK>>>(N, d_body);
   MFEM_GPU_CHECK(cudaGetLastError());
}

template <typename DBODY>
void CuWrap2D(const int N, DBODY &&d_body,
              const int X, const int Y, const int BZ)
{
   if (N==0) { return; }
   // required for optimized GCC/NVCC builds to prevent runtime
   // ODR/linkage violations of inlined templated kernel helpers
   MFEM_VERIFY(BZ>0, "");
   const int GRID = (N+BZ-1)/BZ;
   const dim3 BLCK(X,Y,BZ);
   CuKernel2D<<<GRID,BLCK>>>(N,d_body);
   MFEM_GPU_CHECK(cudaGetLastError());
}

template <int MAX_THREADS_PER_BLOCK, typename DBODY>
void CuWrap2DLaunchBounds(const int N, DBODY &&d_body,
                          const int X, const int Y, const int BZ)
{
   if (N==0) { return; }
   MFEM_VERIFY(BZ>0, "");
   const int GRID = (N+BZ-1)/BZ;
   const dim3 BLCK(X,Y,BZ);
   static_assert(MAX_THREADS_PER_BLOCK > 0);
   CuKernel2DLaunchBounds<MAX_THREADS_PER_BLOCK><<<GRID,BLCK>>>(N, d_body);
   MFEM_GPU_CHECK(cudaGetLastError());
}

template <typename DBODY>
void CuWrap3D(const int N, DBODY &&d_body,
              const int X, const int Y, const int Z, const int G)
{
   if (N==0) { return; }
   const int GRID = G == 0 ? N : G;
   const dim3 BLCK(X,Y,Z);
   CuKernel3D<<<GRID,BLCK>>>(N,d_body);
   MFEM_GPU_CHECK(cudaGetLastError());
}

template <int MAX_THREADS_PER_BLOCK, typename DBODY>
void CuWrap3DLaunchBounds(const int N, DBODY &&d_body,
                          const int X, const int Y, const int Z, const int G)
{
   if (N==0) { return; }
   const int GRID = G == 0 ? N : G;
   const dim3 BLCK(X,Y,Z);
   static_assert(MAX_THREADS_PER_BLOCK > 0);
   CuKernel3DLaunchBounds<MAX_THREADS_PER_BLOCK><<<GRID, BLCK>>>(N, d_body);
   MFEM_GPU_CHECK(cudaGetLastError());
}

template <int Dim, int MAX_THREADS_PER_BLOCK> struct CuWrap;

template <int MAX_THREADS_PER_BLOCK>
struct CuWrap<1, MAX_THREADS_PER_BLOCK>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      CuWrap1D<MFEM_CUDA_BLOCKS>(N, d_body);
   }
};

template <>
struct CuWrap<2, 0>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      CuWrap2D(N, d_body, X, Y, Z);
   }
};

template <int MAX_THREADS_PER_BLOCK>
struct CuWrap<2, MAX_THREADS_PER_BLOCK>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      static_assert(MAX_THREADS_PER_BLOCK > 0);
      CuWrap2DLaunchBounds<MAX_THREADS_PER_BLOCK>(N, d_body, X, Y, Z);
   }
};

template <>
struct CuWrap<3, 0>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      CuWrap3D(N, d_body, X, Y, Z, G);
   }
};

template <int MAX_THREADS_PER_BLOCK>
struct CuWrap<3, MAX_THREADS_PER_BLOCK>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      CuWrap3DLaunchBounds<MAX_THREADS_PER_BLOCK>(N, d_body, X, Y, Z, G);
   }
};

#endif // defined(MFEM_USE_CUDA) && defined(__CUDACC__)


/// HIP backend
#if defined(MFEM_USE_HIP) && defined(__HIP__)

template <typename BODY> __global__ static
void HipKernel1D(const int N, BODY body)
{
   const int k = hipBlockDim_x*hipBlockIdx_x + hipThreadIdx_x;
   if (k >= N) { return; }
   body(k);
}

template <typename BODY> __global__ static
void HipKernel2D(const int N, BODY body)
{
   const int k = hipBlockIdx_x*hipBlockDim_z + hipThreadIdx_z;
   if (k >= N) { return; }
   body(k);
}

template <int MAX_THREADS_PER_BLOCK, typename BODY>
__global__
MFEM_LAUNCH_BOUNDS(MAX_THREADS_PER_BLOCK)
static void HipKernel2DLaunchBounds(const int N, BODY body)
{
   const int k = hipBlockIdx_x*hipBlockDim_z + hipThreadIdx_z;
   if (k >= N) { return; }
   body(k);
}

template <typename BODY> __global__ static
void HipKernel3D(const int N, BODY body)
{
   for (int k = hipBlockIdx_x; k < N; k += hipGridDim_x) { body(k); }
}

template <int MAX_THREADS_PER_BLOCK, typename BODY>
__global__
MFEM_LAUNCH_BOUNDS(MAX_THREADS_PER_BLOCK)
static void HipKernel3DLaunchBounds(const int N, BODY body)
{
   for (int k = hipBlockIdx_x; k < N; k += hipGridDim_x) { body(k); }
}

template <int BLCK = MFEM_HIP_BLOCKS, typename DBODY>
void HipWrap1D(const int N, DBODY &&d_body)
{
   if (N==0) { return; }
   const int GRID = (N+BLCK-1)/BLCK;
   hipLaunchKernelGGL(HipKernel1D,GRID,BLCK,0,nullptr,N,d_body);
   MFEM_GPU_CHECK(hipGetLastError());
}

template <typename DBODY>
void HipWrap2D(const int N, DBODY &&d_body,
               const int X, const int Y, const int BZ)
{
   if (N==0) { return; }
   MFEM_VERIFY(BZ>0, "");
   const int GRID = (N+BZ-1)/BZ;
   const dim3 BLCK(X,Y,BZ);
   hipLaunchKernelGGL(HipKernel2D,GRID,BLCK,0,nullptr,N,d_body);
   MFEM_GPU_CHECK(hipGetLastError());
}

template <int MAX_THREADS_PER_BLOCK, typename DBODY>
void HipWrap2DLaunchBounds(const int N, DBODY &&d_body,
                           const int X, const int Y, const int BZ)
{
   if (N==0) { return; }
   MFEM_VERIFY(BZ>0, "");
   const int GRID = (N+BZ-1)/BZ;
   const dim3 BLCK(X,Y,BZ);
   static_assert(MAX_THREADS_PER_BLOCK > 0);
   HipKernel2DLaunchBounds<MAX_THREADS_PER_BLOCK><<<dim3(GRID), dim3(BLCK), 0, 0>>>
   (N, d_body);
   MFEM_GPU_CHECK(hipGetLastError());
}

template <typename DBODY>
void HipWrap3D(const int N, DBODY &&d_body,
               const int X, const int Y, const int Z, const int G)
{
   if (N==0) { return; }
   const int GRID = G == 0 ? N : G;
   const dim3 BLCK(X,Y,Z);
   hipLaunchKernelGGL(HipKernel3D,GRID,BLCK,0,nullptr,N,d_body);
   MFEM_GPU_CHECK(hipGetLastError());
}

template <int MAX_THREADS_PER_BLOCK, typename DBODY>
void HipWrap3DLaunchBounds(const int N, DBODY &&d_body,
                           const int X, const int Y, const int Z, const int G)
{
   if (N==0) { return; }
   const int GRID = G == 0 ? N : G;
   const dim3 BLCK(X,Y,Z);
   static_assert(MAX_THREADS_PER_BLOCK > 0);
   HipKernel3DLaunchBounds<MAX_THREADS_PER_BLOCK><<<dim3(GRID), dim3(BLCK), 0, 0>>>
   (N, d_body);
   MFEM_GPU_CHECK(hipGetLastError());
}

template <int Dim, int MAX_THREADS_PER_BLOCK> struct HipWrap;

template <int MAX_THREADS_PER_BLOCK>
struct HipWrap<1, MAX_THREADS_PER_BLOCK>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      HipWrap1D<MFEM_HIP_BLOCKS>(N, d_body);
   }
};

template <>
struct HipWrap<2, 0>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      HipWrap2D(N, d_body, X, Y, Z);
   }
};

template <int MAX_THREADS_PER_BLOCK>
struct HipWrap<2, MAX_THREADS_PER_BLOCK>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      HipWrap2DLaunchBounds<MAX_THREADS_PER_BLOCK>(N, d_body, X, Y, Z);
   }
};

template <>
struct HipWrap<3, 0>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      HipWrap3D(N, d_body, X, Y, Z, G);
   }
};

template <int MAX_THREADS_PER_BLOCK>
struct HipWrap<3, MAX_THREADS_PER_BLOCK>
{
   template <typename DBODY>
   static void run(const int N, DBODY &&d_body,
                   const int X, const int Y, const int Z, const int G)
   {
      HipWrap3DLaunchBounds<MAX_THREADS_PER_BLOCK>(N, d_body, X, Y, Z, G);
   }
};

#endif // defined(MFEM_USE_HIP) && defined(__HIP__)


///////////////////////////////////////////////////////////////////////////////
/// Forall host & device kernel dispatch
template <int DIM, int MAX_THREADS_PER_BLOCK = 0,
          typename d_lambda, typename h_lambda>
inline void ForallWrap(const bool use_dev, const int N,
                       d_lambda &&d_body, h_lambda &&h_body,
                       const int X=0, const int Y=0, const int Z=0,
                       const int G=0)
{
   MFEM_CONTRACT_VAR(X);
   MFEM_CONTRACT_VAR(Y);
   MFEM_CONTRACT_VAR(Z);
   MFEM_CONTRACT_VAR(G);
   MFEM_CONTRACT_VAR(d_body);
   if (!use_dev) { goto backend_cpu; }

#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_CUDA) && defined(__CUDACC__)
   // If Backend::RAJA_CUDA is allowed, use it
   if (Device::Allows(Backend::RAJA_CUDA))
   {
      return RajaCuWrap<DIM>::run(N, d_body, X, Y, Z, G);
   }
#endif

#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_HIP) && defined(__HIP__)
   // If Backend::RAJA_HIP is allowed, use it
   if (Device::Allows(Backend::RAJA_HIP))
   {
      return RajaHipWrap<DIM>::run(N, d_body, X, Y, Z, G);
   }
#endif

#if defined(MFEM_USE_CUDA) && defined(__CUDACC__)
   // If Backend::CUDA is allowed, use it
   if (Device::Allows(Backend::CUDA))
   {
      return CuWrap<DIM, MAX_THREADS_PER_BLOCK>::run(N, d_body, X, Y, Z, G);
   }
#endif

#if defined(MFEM_USE_HIP) && defined(__HIP__)
   // If Backend::HIP is allowed, use it
   if (Device::Allows(Backend::HIP))
   {
      return HipWrap<DIM, MAX_THREADS_PER_BLOCK>::run(N, d_body, X, Y, Z, G);
   }
#endif

   // If Backend::DEBUG_DEVICE is allowed, use it
   if (Device::Allows(Backend::DEBUG_DEVICE)) { goto backend_cpu; }

#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_OPENMP)
   // If Backend::RAJA_OMP is allowed, use it
   if (Device::Allows(Backend::RAJA_OMP)) { return RajaOmpWrap(N, h_body); }
#endif

#ifdef MFEM_USE_OPENMP
   // If Backend::OMP is allowed, use it
   if (Device::Allows(Backend::OMP)) { return OmpWrap(N, h_body); }
#endif

#ifdef MFEM_USE_RAJA
   // If Backend::RAJA_CPU is allowed, use it
   if (Device::Allows(Backend::RAJA_CPU)) { return RajaSeqWrap(N, h_body); }
#endif

backend_cpu:
   // Handle Backend::CPU. This is also a fallback for any allowed backends not
   // handled above, e.g. OCCA_CPU with configuration 'occa-cpu,cpu', or
   // OCCA_OMP with configuration 'occa-omp,cpu'.
   for (int k = 0; k < N; k++) { h_body(k); }
}

///////////////////////////////////////////////////////////////////////////////
/// Forall host & device kernel wrappers
template <int DIM, typename lambda>
inline void ForallWrap(const bool use_dev, const int N, lambda &&body,
                       const int X=0, const int Y=0, const int Z=0,
                       const int G=0)
{
   ForallWrap<DIM>(use_dev, N, body, body, X, Y, Z, G);
}

template <int DIM, int MAX_THREADS_PER_BLOCK, typename lambda>
inline void ForallWrap(const bool use_dev, const int N, lambda &&body,
                       const int X=0, const int Y=0, const int Z=0,
                       const int G=0)
{
   ForallWrap<DIM, MAX_THREADS_PER_BLOCK>(use_dev, N, body, body, X, Y, Z, G);
}

///////////////////////////////////////////////////////////////////////////////
// forall interfaces
template<typename lambda>
inline void forall(int N, lambda &&body) { ForallWrap<1>(true, N, body); }

template<typename lambda>
inline void forall(int Nx, int Ny, lambda &&body)
{
   if (Device::Allows(Backend::DEVICE_MASK))
   {
      mfem::forall(Nx * Ny, [=] MFEM_HOST_DEVICE(int idx)
      {
         int j = idx / Nx;
         int i = idx % Nx;
         body(i, j);
      });
   }
#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_OPENMP)
   else if (Device::Allows(Backend::RAJA_OMP))
   {
      return RajaOmpWrap2D(Nx, Ny, body);
   }
#endif
#ifdef MFEM_USE_OPENMP
   else if (Device::Allows(Backend::OMP))
   {
      return OmpWrap2D(Nx, Ny, body);
   }
#endif
   else
   {
      for (int j = 0; j < Ny; ++j)
      {
         for (int i = 0; i < Nx; ++i)
         {
            body(i, j);
         }
      }
   }
}

template<typename lambda>
inline void forall(int Nx, int Ny, int Nz, lambda &&body)
{
   if (Device::Allows(Backend::DEVICE_MASK))
   {
      mfem::forall(Nx * Ny * Nz, [=] MFEM_HOST_DEVICE(int idx)
      {
         int i = idx % Nx;
         int j = idx / Nx;
         int k = j / Ny;
         j = j % Ny;
         body(i, j, k);
      });
   }
#if defined(MFEM_USE_RAJA) && defined(RAJA_ENABLE_OPENMP)
   else if (Device::Allows(Backend::RAJA_OMP))
   {
      return RajaOmpWrap3D(Nx, Ny, Nz, body);
   }
#endif
#ifdef MFEM_USE_OPENMP
   else if (Device::Allows(Backend::OMP))
   {
      return OmpWrap3D(Nx, Ny, Nz, body);
   }
#endif
   else
   {
      for (int k = 0; k < Nz; ++k)
      {
         for (int j = 0; j < Ny; ++j)
         {
            for (int i = 0; i < Nx; ++i)
            {
               body(i, j, k);
            }
         }
      }
   }
}

template<typename lambda>
inline void forall_switch(bool use_dev, int N, lambda &&body)
{
   ForallWrap<1>(use_dev, N, body);
}

template<typename lambda>
inline void forall_2D(int N, int X, int Y, lambda &&body)
{
   ForallWrap<2>(true, N, body, X, Y, 1);
}

template<int MAX_THREADS_PER_BLOCK, typename lambda>
inline void forall_2D(int N, int X, int Y, lambda &&body)
{
   ForallWrap<2, MAX_THREADS_PER_BLOCK>(true, N, body, X, Y, 1);
}

template<typename lambda>
inline void forall_2D_batch(int N, int X, int Y, int BZ, lambda &&body)
{
   ForallWrap<2>(true, N, body, X, Y, BZ);
}

template<typename lambda>
inline void forall_3D(int N, int X, int Y, int Z, lambda &&body)
{
   ForallWrap<3>(true, N, body, X, Y, Z, 0);
}

template<int MAX_THREADS_PER_BLOCK, typename lambda>
inline void forall_3D(int N, int X, int Y, int Z, lambda &&body)
{
   ForallWrap<3, MAX_THREADS_PER_BLOCK>(true, N, body, X, Y, Z, 0);
}

template<typename lambda>
inline void forall_3D_grid(int N, int X, int Y, int Z, int G, lambda &&body)
{
   ForallWrap<3>(true, N, body, X, Y, Z, G);
}

#ifdef MFEM_USE_MPI

// Function mfem::hypre_forall_cpu() similar to mfem::forall, but it always
// executes on the CPU using sequential or OpenMP-parallel execution based on
// the hypre build time configuration.
template<typename lambda>
inline void hypre_forall_cpu(int N, lambda &&body)
{
#ifdef HYPRE_USING_OPENMP
   #pragma omp parallel for HYPRE_SMP_SCHEDULE
#endif
   for (int i = 0; i < N; i++) { body(i); }
}

// Function mfem::hypre_forall_gpu() similar to mfem::forall, but it always
// executes on the GPU device that hypre was configured with at build time.
#if defined(HYPRE_USING_GPU)
template<typename lambda>
inline void hypre_forall_gpu(int N, lambda &&body)
{
#if defined(HYPRE_USING_CUDA)
   CuWrap1D(N, body);
#elif defined(HYPRE_USING_HIP)
   HipWrap1D(N, body);
#else
#error Unknown HYPRE GPU backend!
#endif
}
#endif

// Function mfem::hypre_forall() similar to mfem::forall, but it executes on the
// device, CPU or GPU, that hypre was configured with at build time (when the
// HYPRE version is < 2.31.0) or at runtime (when HYPRE was configured with GPU
// support at build time and HYPRE's version is >= 2.31.0). This selection is
// generally independent of what device was selected in MFEM's runtime
// configuration.
template<typename lambda>
inline void hypre_forall(int N, lambda &&body)
{
#if !defined(HYPRE_USING_GPU)
   hypre_forall_cpu(N, body);
#elif MFEM_HYPRE_VERSION < 23100
   hypre_forall_gpu(N, body);
#else // HYPRE_USING_GPU is defined and MFEM_HYPRE_VERSION >= 23100
   if (!HypreUsingGPU())
   {
      hypre_forall_cpu(N, body);
   }
   else
   {
      hypre_forall_gpu(N, body);
   }
#endif
}

// Return the most general MemoryClass that can be used with mfem::hypre_forall
// kernels. The returned MemoryClass is the same as the one returned by
// GerHypreMemoryClass() except when hypre is configured to use UVM, in which
// case this function returns MemoryClass::HOST or MemoryClass::DEVICE depending
// on the result of HypreUsingGPU().
inline MemoryClass GetHypreForallMemoryClass()
{
   return HypreUsingGPU() ? MemoryClass::DEVICE : MemoryClass::HOST;
}

#endif // MFEM_USE_MPI

} // namespace mfem

#endif // MFEM_FORALL_HPP