walberla::gpu::communication::MemcpyPackInfo< GPUFieldType > Class Template Reference

Detailed Description

template<typename GPUFieldType>
class walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >

Data packing/unpacking for ghost layer based communication of a GPUField.

Encapsulate information on how to extract data from blocks that should be communicated to neighboring blocks (see pack()) and how to inject this data in a receiving block (see unpack()). This involves a device memory buffer and two device-to-device memory copy operations.

A special method exists for communication between two blocks which are allocated on the same process (see communicateLocal()). In this case the data does not have be communicated via a device buffer, but can be sent directly. This involves a single device-to-device memory copy operation.

Data that is packed in direction "dir" at one block is unpacked in direction "stencil::inverseDir[dir]" at the neighboring block. This behavior must be implemented in communicateLocal()!

Template Parameters

GPUFieldType

A fully qualified GPUField.

#include <MemcpyPackInfo.h>

Inheritance diagram for walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >:

Public Member Functions
	MemcpyPackInfo (BlockDataID fieldId, const uint_t numGhostLayers=0)
	~MemcpyPackInfo () override=default
void	pack (stencil::Direction dir, unsigned char *buffer, IBlock *block, gpuStream_t stream) override
	Pack data from a block into a send buffer.
void	communicateLocal (stencil::Direction dir, const IBlock *sender, IBlock *receiver, gpuStream_t stream) override
	Copy data from one local block to another local block.
void	unpack (stencil::Direction dir, unsigned char *buffer, IBlock *block, gpuStream_t stream) override
	Unpack data from a receive buffer into a block.
uint_t	size (stencil::Direction dir, IBlock *block) override
Public Member Functions inherited from walberla::gpu::GeneratedGPUPackInfo
	GeneratedGPUPackInfo ()=default
virtual	~GeneratedGPUPackInfo ()=default

Private Member Functions
uint_t	numberOfGhostLayersToCommunicate (const GPUFieldType *const field) const

Private Attributes
BlockDataID	fieldId_
uint_t	numberOfGhostLayers_ { 0 }
bool	communicateAllGhostLayers_ { true }

Constructor & Destructor Documentation

MemcpyPackInfo()

template<typename GPUFieldType >

walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::MemcpyPackInfo ( BlockDataID fieldId, const uint_t numGhostLayers = 0 )

inline

Parameters

fieldId_	BlockDataID of the field to be communicated
numGhostLayers	Number of ghost layers to be communicated. If left at zero, the number will be inferred from the field.

~MemcpyPackInfo()

template<typename GPUFieldType >

walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::~MemcpyPackInfo ( )

overridedefault

Member Function Documentation

communicateLocal()

template<typename GPUFieldType >

void walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::communicateLocal ( stencil::Direction dir, const IBlock * sender, IBlock * receiver, gpuStream_t stream )

overridevirtual

Copy data from one local block to another local block.

Both blocks are allocated on the same MPI rank.

Parameters

dir	the direction of the communication (from sender to receiver)
sender	id of block where the data should be copied from
receiver	id of block where the data should be copied to
stream	GPU stream

Implements walberla::gpu::GeneratedGPUPackInfo.

numberOfGhostLayersToCommunicate()

template<typename GPUFieldType >

uint_t walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::numberOfGhostLayersToCommunicate ( const GPUFieldType *const field ) const

private

pack()

template<typename GPUFieldType >

void walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::pack ( stencil::Direction dir, unsigned char * buffer, IBlock * block, gpuStream_t stream )

overridevirtual

Pack data from a block into a send buffer.

Parameters

dir	pack data for neighbor in this direction
buffer	buffer for writing the data into
block	the block whose data should be packed into a buffer
stream	GPU stream

Implements walberla::gpu::GeneratedGPUPackInfo.

size()

template<typename GPUFieldType >

uint_t walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::size ( stencil::Direction dir, IBlock * block )

overridevirtual

Implements walberla::gpu::GeneratedGPUPackInfo.

unpack()

template<typename GPUFieldType >

void walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::unpack ( stencil::Direction dir, unsigned char * buffer, IBlock * block, gpuStream_t stream )

overridevirtual

Unpack data from a receive buffer into a block.

Parameters

dir	receive data from neighbor in this direction
buffer	buffer for reading the data from
block	the block where the unpacked data should be stored into
stream	GPU stream

Implements walberla::gpu::GeneratedGPUPackInfo.

Member Data Documentation

communicateAllGhostLayers_

template<typename GPUFieldType >

bool walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::communicateAllGhostLayers_ { true }

private

fieldId_

template<typename GPUFieldType >

BlockDataID walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::fieldId_

private

numberOfGhostLayers_

template<typename GPUFieldType >

uint_t walberla::gpu::communication::MemcpyPackInfo< GPUFieldType >::numberOfGhostLayers_ { 0 }

private

---

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

• /builds/administration/walberla-website/walberla/src/gpu/communication/MemcpyPackInfo.h
• /builds/administration/walberla-website/walberla/src/gpu/communication/MemcpyPackInfo.impl.h