gmx::Gpu3dFft::Impl Class Reference

#include <gromacs/fft/gpu_3dfft_impl.h>

Inheritance diagram for gmx::Gpu3dFft::Impl:

Description

Impl base class for all FFT backends.

Public Member Functions
	Impl ()
	Default constructor.
	Impl (bool performOutOfPlaceFFT)
	Constructor called from dervied classes.
	Impl (bool allocateRealGrid, MPI_Comm comm, ArrayRef< const int > gridSizesInXForEachRank, ArrayRef< const int > gridSizesInYForEachRank, int nz, bool performOutOfPlaceFFT, const DeviceContext &context, const DeviceStream &pmeStream, ivec realGridSize, ivec realGridSizePadded, ivec complexGridSizePadded, DeviceBuffer< float > *realGrid, DeviceBuffer< float > *complexGrid)
	Constructs GPU FFT plans for performing 3D FFT on a PME grid. More...
virtual	~Impl ()
	Default destructor.
virtual void	perform3dFft (gmx_fft_direction dir, CommandEvent *timingEvent)=0
	Performs the FFT transform in given direction. More...

Protected Member Functions
void	allocateComplexGrid (const ivec complexGridSizePadded, DeviceBuffer< float > *realGrid, DeviceBuffer< float > *complexGrid, const DeviceContext &context)
	Allocate and assign complexGrid.
void	deallocateComplexGrid ()
	free complexGrid

Protected Attributes
bool	performOutOfPlaceFFT_ = false
	A boolean which tells whether the complex and real grids are different or same. Currenty true.
DeviceBuffer< float >	complexGrid_
	FFT complex grid.

Constructor & Destructor Documentation

gmx::Gpu3dFft::Impl::Impl	(	bool	allocateRealGrid,
		MPI_Comm	comm,
		ArrayRef< const int >	gridSizesInXForEachRank,
		ArrayRef< const int >	gridSizesInYForEachRank,
		int	nz,
		bool	performOutOfPlaceFFT,
		const DeviceContext &	context,
		const DeviceStream &	pmeStream,
		ivec	realGridSize,
		ivec	realGridSizePadded,
		ivec	complexGridSizePadded,
		DeviceBuffer< float > *	realGrid,
		DeviceBuffer< float > *	complexGrid
	)

Constructs GPU FFT plans for performing 3D FFT on a PME grid.

Parameters

[in]	allocateRealGrid	True if fft real-grid is to be allocated, false if pre-allocated
[in]	comm	MPI communicator, used with distributed-FFT backends
[in]	gridSizesInXForEachRank	Number of grid points used with each rank in X-dimension
[in]	gridSizesInYForEachRank	Number of grid points used with each rank in Y-dimension
[in]	nz	Grid dimension in Z
[in]	performOutOfPlaceFFT	Whether the FFT will be performed out-of-place
[in]	context	GPU context
[in]	pmeStream	GPU stream for PME
[in,out]	realGridSize	Dimensions of the local real grid, out if allocateRealGrid=true
[in,out]	realGridSizePadded	Dimensions of the local real grid with padding, out if allocateRealGrid=true
[in,out]	complexGridSizePadded	Dimensions of the local complex grid with padding, out if allocateRealGrid=true
[in,out]	realGrid	Device buffer of floats for the local real grid, out if allocateRealGrid=true
[out]	complexGrid	Device buffer of complex floats for the local complex grid

Member Function Documentation

virtual void gmx::Gpu3dFft::Impl::perform3dFft ( gmx_fft_direction  dir, CommandEvent *  timingEvent  )

pure virtual

Performs the FFT transform in given direction.

Parameters

[in]	dir	FFT transform direction specifier
[out]	timingEvent	pointer to the timing event where timing data is recorded

Implemented in gmx::Gpu3dFft::ImplOclVkfft, gmx::Gpu3dFft::ImplSyclVkfft, gmx::Gpu3dFft::ImplCuFftMp, gmx::Gpu3dFft::ImplCuFft, gmx::Gpu3dFft::ImplOcl, gmx::Gpu3dFft::ImplSyclBbfft, gmx::Gpu3dFft::ImplSyclMkl, gmx::Gpu3dFft::ImplSyclRocfft, and gmx::Gpu3dFft::ImplSycl.

---

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

• src/gromacs/fft/gpu_3dfft_impl.h
• src/gromacs/fft/gpu_3dfft_impl.cpp