pme_gpu_grid_sycl.cpp File Reference

#include "gmxpre.h"
#include "config.h"
#include <cstdlib>
#include "gromacs/fft/parallel_3dfft.h"
#include "gromacs/gpu_utils/gmxsycl.h"
#include "gromacs/gpu_utils/syclutils.h"
#include "gromacs/timing/wallcycle.h"
#include "gromacs/utility/vec.h"
#include "pme_gpu_grid.h"
#include "pme_gpu_types.h"
#include "pme_gpu_types_host.h"
#include "pme_gpu_types_host_impl.h"

Include dependency graph for pme_gpu_grid_sycl.cpp:

Description

Implements PME GPU halo exchange and PME GPU - Host FFT grid conversion functions. These functions are used for PME decomposition in mixed-mode.

Author

Mark Abraham mark..nosp@m.j.ab.nosp@m.raham.nosp@m.@gma.nosp@m.il.co.nosp@m.m

Classes
class	PackHaloExternal
	Handles a kernel which packs non-contiguous overlap data in all 8 neighboring directions. More...
class	UnpackHaloExternal
	Handles a kernel which gathers data from halo region in all 8 neighboring directions. More...
class	UnpackAndAddHaloInternal
	Handles a kernel which adds grid overlap data received from neighboring ranks. More...
class	PackHaloInternal
	Handles a kernel which packs non-contiguous overlap data in all 8 neighboring directions. More...
class	GridConverter< pmeToFft >
	Builds a kernel to convert between PME and FFT grids. More...

Typedefs
using	mode = sycl::access_mode

Functions
void	pmeGpuGridHaloExchange (const PmeGpu *pmeGpu, gmx_wallcycle *wcycle)
	Grid Halo exchange after PME spread ToDo: Current implementation transfers halo region from/to only immediate neighbours And, expects that overlapSize <= local grid width. Implement exchange with multiple neighbors to remove this limitation ToDo: Current implementation synchronizes pmeStream to make sure data is ready on GPU after spread. Consider using events for this synchnozation. More...
void	pmeGpuGridHaloExchangeReverse (const PmeGpu *pmeGpu, gmx_wallcycle *wcycle)
	Grid reverse Halo exchange before PME gather ToDo: Current implementation transfers halo region from/to only immediate neighbours And, expects that overlapSize <= local grid width. Implement exchange with multiple neighbors to remove this limitation ToDo: Current implementation synchronizes pmeStream to make sure data is ready on GPU after FFT to PME grid conversion. Consider using events for this synchnozation. More...
template<bool pmeToFft>
void	convertPmeGridToFftGrid (const PmeGpu *pmeGpu, float *h_fftRealGrid, gmx_parallel_3dfft *fftSetup, const int gridIndex)
	Copy PME Grid with overlap region to host FFT grid and vice-versa. Used in mixed mode PME decomposition. More...
template<bool pmeToFft>
void	convertPmeGridToFftGrid (const PmeGpu *pmeGpu, DeviceBuffer< float > *d_fftRealGrid, const int gridIndex)
	Copy PME Grid with overlap region to device FFT grid and vice-versa. Used in full GPU PME decomposition. More...
template void	convertPmeGridToFftGrid< true > (const PmeGpu *pmeGpu, float *h_fftRealGrid, gmx_parallel_3dfft *fftSetup, const int gridIndex)
template void	convertPmeGridToFftGrid< false > (const PmeGpu *pmeGpu, float *h_fftRealGrid, gmx_parallel_3dfft *fftSetup, const int gridIndex)
template void	convertPmeGridToFftGrid< true > (const PmeGpu *pmeGpu, DeviceBuffer< float > *d_fftRealGrid, const int gridIndex)
template void	convertPmeGridToFftGrid< false > (const PmeGpu *pmeGpu, DeviceBuffer< float > *d_fftRealGrid, const int gridIndex)

Variables
static constexpr int	sc_subGroupSize = 32
	Sub-group size for conversion kernels. More...

Function Documentation

template<bool pmeToFft>

void convertPmeGridToFftGrid	(	const PmeGpu *	pmeGpu,
		float *	h_fftRealGrid,
		gmx_parallel_3dfft *	fftSetup,
		int	gridIndex
	)

Copy PME Grid with overlap region to host FFT grid and vice-versa. Used in mixed mode PME decomposition.

Parameters

[in]	pmeGpu	The PME GPU structure.
[in]	h_fftRealGrid	FFT grid on host
[in]	fftSetup	Host FFT setup structure
[in]	gridIndex	Grid index which is to be converted

Template Parameters

pmeToFft

A boolean which tells if this is conversion from PME grid to FFT grid or reverse

template<bool pmeToFft>

void convertPmeGridToFftGrid	(	const PmeGpu *	pmeGpu,
		DeviceBuffer< float > *	d_fftRealGrid,
		int	gridIndex
	)

Copy PME Grid with overlap region to device FFT grid and vice-versa. Used in full GPU PME decomposition.

Parameters

[in]	pmeGpu	The PME GPU structure.
[in]	d_fftRealGrid	FFT grid on device
[in]	gridIndex	Grid index which is to be converted

Template Parameters

pmeToFft

A boolean which tells if this is conversion from PME grid to FFT grid or reverse

void pmeGpuGridHaloExchange	(	const PmeGpu *	pmeGpu,
		gmx_wallcycle *	wcycle
	)

Grid Halo exchange after PME spread ToDo: Current implementation transfers halo region from/to only immediate neighbours And, expects that overlapSize <= local grid width. Implement exchange with multiple neighbors to remove this limitation ToDo: Current implementation synchronizes pmeStream to make sure data is ready on GPU after spread. Consider using events for this synchnozation.

Parameters

[in]	pmeGpu	The PME GPU structure.
[in]	wcycle	The wallclock counter.

void pmeGpuGridHaloExchangeReverse	(	const PmeGpu *	pmeGpu,
		gmx_wallcycle *	wcycle
	)

Grid reverse Halo exchange before PME gather ToDo: Current implementation transfers halo region from/to only immediate neighbours And, expects that overlapSize <= local grid width. Implement exchange with multiple neighbors to remove this limitation ToDo: Current implementation synchronizes pmeStream to make sure data is ready on GPU after FFT to PME grid conversion. Consider using events for this synchnozation.

Parameters

[in]	pmeGpu	The PME GPU structure.
[in]	wcycle	The wallclock counter.

Variable Documentation

constexpr int sc_subGroupSize = 32

static

Sub-group size for conversion kernels.

Chosen to match relevant hardware widths on supported hardware. In particular, PonteVecchio is 32-wide.