Opm::gpuistl::GPUAwareMPISender< field_type, block_size, OwnerOverlapCopyCommunicationType > Class Template Reference

Derived class of GPUSender that handles MPI made with CUDA aware MPI The copOwnerToAll function uses MPI calls refering to data that resides on the GPU in order to send it directly to other GPUs, skipping the staging step on the CPU. More...

#include <GpuAwareMPISender.hpp>

Inheritance diagram for Opm::gpuistl::GPUAwareMPISender< field_type, block_size, OwnerOverlapCopyCommunicationType >:

Public Types
using	X = GpuVector<field_type>
Public Types inherited from Opm::gpuistl::GPUSender< field_type, OwnerOverlapCopyCommunicationType >
using	X = GpuVector<field_type>

Public Member Functions
	GPUAwareMPISender (const OwnerOverlapCopyCommunicationType &cpuOwnerOverlapCopy)
void	copyOwnerToAll (const X &source, X &dest) const override
	copyOwnerToAll will copy the data in source to all processes.
Public Member Functions inherited from Opm::gpuistl::GPUSender< field_type, OwnerOverlapCopyCommunicationType >
	GPUSender (const OwnerOverlapCopyCommunicationType &cpuOwnerOverlapCopy)
void	project (X &x) const
	project will project x to the owned subspace
void	dot (const X &x, const X &y, field_type &output) const
	dot will carry out the dot product between x and y on the owned indices, then sum up the result across MPI processes.
field_type	norm (const X &x) const
	norm computes the l^2-norm of x across processes.
const ::Dune::Communication< MPI_Comm > &	communicator () const
	communicator returns the MPI communicator used by this GPUSender

Additional Inherited Members
Protected Attributes inherited from Opm::gpuistl::GPUSender< field_type, OwnerOverlapCopyCommunicationType >
std::once_flag	m_initializedIndices
std::unique_ptr< GpuVector< int > >	m_indicesOwner
std::unique_ptr< GpuVector< int > >	m_indicesCopy
const OwnerOverlapCopyCommunicationType &	m_cpuOwnerOverlapCopy

Detailed Description

template<class field_type, int block_size, class OwnerOverlapCopyCommunicationType>
class Opm::gpuistl::GPUAwareMPISender< field_type, block_size, OwnerOverlapCopyCommunicationType >

Derived class of GPUSender that handles MPI made with CUDA aware MPI The copOwnerToAll function uses MPI calls refering to data that resides on the GPU in order to send it directly to other GPUs, skipping the staging step on the CPU.

Template Parameters

field_type	is float or double
block_size	is the blocksize of the blockelements in the matrix
OwnerOverlapCopyCommunicationType	is typically a Dune::LinearOperator::communication_type

Member Function Documentation

copyOwnerToAll()

template<class field_type, int block_size, class OwnerOverlapCopyCommunicationType>

void Opm::gpuistl::GPUAwareMPISender< field_type, block_size, OwnerOverlapCopyCommunicationType >::copyOwnerToAll ( const X & source, X & dest ) const

inlineoverridevirtual

copyOwnerToAll will copy the data in source to all processes.

Note

Depending on the implementation, this may or may not use GPU aware MPI. If it does not use GPU aware MPI, the data will be copied to the CPU before the communication.

Parameters

[in]	source
[out]	dest

Implements Opm::gpuistl::GPUSender< field_type, OwnerOverlapCopyCommunicationType >.

---

The documentation for this class was generated from the following file:

• opm/simulators/linalg/gpuistl/GpuAwareMPISender.hpp