pme_gpu_program_impl_sycl.cpp File Reference

#include "gmxpre.h"
#include "gromacs/gpu_utils/gmxsycl.h"
#include "gromacs/gpu_utils/syclutils.h"
#include "gromacs/hardware/device_information.h"
#include "pme_gather_sycl.h"
#include "pme_gpu_constants.h"
#include "pme_gpu_internal.h"
#include "pme_gpu_program_impl.h"
#include "pme_gpu_types_host.h"
#include "pme_solve_sycl.h"
#include "pme_spread_sycl.h"

Include dependency graph for pme_gpu_program_impl_sycl.cpp:

Description

Implements PmeGpuProgramImpl, which stores permanent PME GPU context-derived data, such as (compiled) kernel handles.

Author

Aleksei Iupinov a.yup.nosp@m.inov.nosp@m.@gmai.nosp@m.l.co.nosp@m.m

Andrey Alekseenko al42a.nosp@m.nd@g.nosp@m.mail..nosp@m.com

Macros
#define	INSTANTIATE_SPREAD_2(order, computeSplines, spreadCharges, numGrids, writeGlobal, threadsPerAtom, subGroupSize)   extern template class PmeSplineAndSpreadKernel<order, computeSplines, spreadCharges, true, true, numGrids, writeGlobal, threadsPerAtom, subGroupSize>;
#define	INSTANTIATE_SPREAD(order, numGrids, threadsPerAtom, subGroupSize)
#define	INSTANTIATE_GATHER_2(order, numGrids, readGlobal, threadsPerAtom, subGroupSize)   extern template class PmeGatherKernel<order, true, true, numGrids, readGlobal, threadsPerAtom, subGroupSize>;
#define	INSTANTIATE_GATHER(order, numGrids, threadsPerAtom, subGroupSize)
#define	INSTANTIATE_X(x, order, subGroupSize)
#define	INSTANTIATE_SOLVE(subGroupSize)
#define	INSTANTIATE(order, subGroupSize)

Functions
static int	chooseSubGroupSizeForDevice (const DeviceInformation &deviceInfo)
template<int subGroupSize>
static void	setKernelPointers (struct PmeGpuProgramImpl *pmeGpuProgram)
	Helper function to set proper kernel functor pointers.

Variables
constexpr int	c_pmeOrder = 4
constexpr bool	c_wrapX = true
constexpr bool	c_wrapY = true
constexpr int	c_stateA = 0
constexpr int	c_stateB = 1

Macro Definition Documentation

#define INSTANTIATE	(		order,
			subGroupSize
	)

Value:

INSTANTIATE_X(SPREAD, order, subGroupSize); \
    INSTANTIATE_X(GATHER, order, subGroupSize); \
    INSTANTIATE_SOLVE(subGroupSize);

#define INSTANTIATE_GATHER	(		order,
			numGrids,
			threadsPerAtom,
			subGroupSize
	)

Value:

INSTANTIATE_GATHER_2(order, numGrids, true, threadsPerAtom, subGroupSize); \
    INSTANTIATE_GATHER_2(order, numGrids, false, threadsPerAtom, subGroupSize);

#define INSTANTIATE_SOLVE

(

subGroupSize

)

Value:

extern template class PmeSolveKernel<GridOrdering::XYZ, false, c_stateA, subGroupSize>; \
    extern template class PmeSolveKernel<GridOrdering::XYZ, true, c_stateA, subGroupSize>;  \
    extern template class PmeSolveKernel<GridOrdering::YZX, false, c_stateA, subGroupSize>; \
    extern template class PmeSolveKernel<GridOrdering::YZX, true, c_stateA, subGroupSize>;  \
    extern template class PmeSolveKernel<GridOrdering::XYZ, false, c_stateB, subGroupSize>; \
    extern template class PmeSolveKernel<GridOrdering::XYZ, true, c_stateB, subGroupSize>;  \
    extern template class PmeSolveKernel<GridOrdering::YZX, false, c_stateB, subGroupSize>; \
    extern template class PmeSolveKernel<GridOrdering::YZX, true, c_stateB, subGroupSize>;

#define INSTANTIATE_SPREAD	(		order,
			numGrids,
			threadsPerAtom,
			subGroupSize
	)

Value:

INSTANTIATE_SPREAD_2(order, true, true, numGrids, true, threadsPerAtom, subGroupSize);  \
    INSTANTIATE_SPREAD_2(order, true, false, numGrids, true, threadsPerAtom, subGroupSize); \
    INSTANTIATE_SPREAD_2(order, false, true, numGrids, true, threadsPerAtom, subGroupSize); \
    INSTANTIATE_SPREAD_2(order, true, true, numGrids, false, threadsPerAtom, subGroupSize);

#define INSTANTIATE_X	(		x,
			order,
			subGroupSize
	)

Value:

INSTANTIATE_##x(order, 1, ThreadsPerAtom::Order, subGroupSize);        \
    INSTANTIATE_##x(order, 1, ThreadsPerAtom::OrderSquared, subGroupSize); \
    INSTANTIATE_##x(order, 2, ThreadsPerAtom::Order, subGroupSize);        \
    INSTANTIATE_##x(order, 2, ThreadsPerAtom::OrderSquared, subGroupSize);