forall.hpp File Reference

Go to the source code of this file.

Classes
struct	mfem::DeviceDofQuadLimits
	Maximum number of 1D DOFs or quadrature points for the current runtime configuration of the Device (used in fallback kernels). More...
struct	mfem::RajaCuWrap< 1 >
struct	mfem::RajaCuWrap< 2 >
struct	mfem::RajaCuWrap< 3 >
struct	mfem::RajaHipWrap< 1 >
struct	mfem::RajaHipWrap< 2 >
struct	mfem::RajaHipWrap< 3 >
struct	mfem::CuWrap< 1 >
struct	mfem::CuWrap< 2 >
struct	mfem::CuWrap< 3 >
struct	mfem::HipWrap< 1 >
struct	mfem::HipWrap< 2 >
struct	mfem::HipWrap< 3 >

Namespaces
namespace	mfem

Typedefs
using	mfem::DofQuadLimits = internal::DofQuadLimits_CUDA
	Maximum number of 1D DOFs or quadrature points for the architecture currently being compiled for (used in fallback kernels).
using	mfem::cuda_launch_policy
	RAJA Cuda and Hip backends.
using	mfem::cuda_teams_x
using	mfem::cuda_threads_z
using	mfem::hip_launch_policy
using	mfem::hip_teams_x
using	mfem::hip_threads_z

Functions
template<typename HBODY >
void	mfem::OmpWrap (const int N, HBODY &&h_body)
	OpenMP backend.
template<const int BLOCKS = MFEM_CUDA_BLOCKS, typename DBODY >
void	mfem::RajaCuWrap1D (const int N, DBODY &&d_body)
template<typename DBODY >
void	mfem::RajaCuWrap2D (const int N, DBODY &&d_body, const int X, const int Y, const int BZ)
template<typename DBODY >
void	mfem::RajaCuWrap3D (const int N, DBODY &&d_body, const int X, const int Y, const int Z, const int G)
template<const int BLOCKS = MFEM_HIP_BLOCKS, typename DBODY >
void	mfem::RajaHipWrap1D (const int N, DBODY &&d_body)
template<typename DBODY >
void	mfem::RajaHipWrap2D (const int N, DBODY &&d_body, const int X, const int Y, const int BZ)
template<typename DBODY >
void	mfem::RajaHipWrap3D (const int N, DBODY &&d_body, const int X, const int Y, const int Z, const int G)
template<typename HBODY >
void	mfem::RajaOmpWrap (const int N, HBODY &&h_body)
	RAJA OpenMP backend.
template<typename HBODY >
void	mfem::RajaSeqWrap (const int N, HBODY &&h_body)
	RAJA sequential loop backend.
template<const int BLCK = MFEM_CUDA_BLOCKS, typename DBODY >
void	mfem::CuWrap1D (const int N, DBODY &&d_body)
template<typename DBODY >
void	mfem::CuWrap2D (const int N, DBODY &&d_body, const int X, const int Y, const int BZ)
template<typename DBODY >
void	mfem::CuWrap3D (const int N, DBODY &&d_body, const int X, const int Y, const int Z, const int G)
template<const int BLCK = MFEM_HIP_BLOCKS, typename DBODY >
void	mfem::HipWrap1D (const int N, DBODY &&d_body)
template<typename DBODY >
void	mfem::HipWrap2D (const int N, DBODY &&d_body, const int X, const int Y, const int BZ)
template<typename DBODY >
void	mfem::HipWrap3D (const int N, DBODY &&d_body, const int X, const int Y, const int Z, const int G)
template<const int DIM, typename d_lambda , typename h_lambda >
void	mfem::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)
	The forall kernel body wrapper.
template<const int DIM, typename lambda >
void	mfem::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)
template<typename lambda >
void	mfem::forall (int N, lambda &&body)
template<typename lambda >
void	mfem::forall_switch (bool use_dev, int N, lambda &&body)
template<typename lambda >
void	mfem::forall_2D (int N, int X, int Y, lambda &&body)
template<typename lambda >
void	mfem::forall_2D_batch (int N, int X, int Y, int BZ, lambda &&body)
template<typename lambda >
void	mfem::forall_3D (int N, int X, int Y, int Z, lambda &&body)
template<typename lambda >
void	mfem::forall_3D_grid (int N, int X, int Y, int Z, int G, lambda &&body)
template<typename lambda >
void	mfem::hypre_forall_cpu (int N, lambda &&body)
template<typename lambda >
void	mfem::hypre_forall_gpu (int N, lambda &&body)
template<typename lambda >
void	mfem::hypre_forall (int N, lambda &&body)
MemoryClass	mfem::GetHypreForallMemoryClass ()
template<class T , class B , class R >
void	mfem::reduce (int N, T &res, B &&body, const R &reducer, bool use_dev, Array< T > &workspace)
	Performs a 1D reduction on the range [0,N). res initial value and where the result will be written. body reduction function body. reducer helper for joining two reduced values. use_dev true to perform the reduction on the device, if possible. workspace temporary workspace used for device reductions. May be resized to a larger capacity as needed. Preferably should have MemoryType::MANAGED or MemoryType::HOST_PINNED. TODO: replace with internal temporary workspace vectors once that's added to the memory manager.