Hermite4GPU.cuh File Reference

#include "../Hermite4.hpp"
#include <cassert>
#include <string>

Include dependency graph for Hermite4GPU.cuh:

This graph shows which files directly or indirectly include this file:

Classes
class	Hermite4GPU
	Class which implements on the GPU the structure of the Hermite4 scheme. More...

Macros
#define	HERMITE4GPU_HPP
#define	CSC_NO_SYNC(call)
#define	CSC(call)   CSC_NO_SYNC(call);

Functions
__host__ __device__ double4	operator+ (const double4 &a, const double4 &b)
__host__ __device__ void	operator+= (double4 &a, double4 &b)
__host__ __device__ Forces	operator+ (Forces &a, Forces &b)
__host__ __device__ void	operator+= (Forces &a, Forces &b)
__global__ void	k_init_acc_jrk (Predictor *p, Forces *f, int n, double e2, int dev, int dev_size)
	Initialization kernel, which consider an interaction of the particles. More...
__global__ void	k_prediction (Forces *f, double4 *r, double4 *v, double *t, Predictor *p, int dev_size, double ITIME)
	Predictor kernel, in charge of performing the prediction step of all the particles on each integration step. More...
__device__ void	k_force_calculation (Predictor i_p, Predictor j_p, Forces &f, double e2)
	Force interaction kernel, in charge of performing gravitational interaction computation between two particles. More...
__global__ void	k_update (Predictor *i_p, Predictor *j_p, Forces *fout, int n, int total, double e2)
	Force kernel, in charge of performing distribution of how the particles will be distributed among the GPUs. More...
__global__ void	k_reduce (Forces *in, Forces *out, int shift_id, int shift)
	Force reduction kernel, in charge of summing up all the preliminary results of the forces for the particles. More...
__global__ void	k_energy (double4 *r, double4 *v, double *ekin, double *epot, int n, int dev_size, int dev)
	Energy kernel, in charge of the calculation of the kinetic and potential energy on the GPUs. More...

Macro Definition Documentation

CSC

#define CSC

(

call

)

CSC_NO_SYNC(call);

CSC_NO_SYNC

#define CSC_NO_SYNC

(

call

)

Value:

do {                          \
    cudaError err = call;                                             \
    if( cudaSuccess != err) {                                         \
        fprintf(stderr, "Cuda error in file '%s' in line %i : %s.\n", \
                __FILE__, __LINE__, cudaGetErrorString( err) );       \
        exit(EXIT_FAILURE);                                           \
    } } while (0)

HERMITE4GPU_HPP

#define HERMITE4GPU_HPP

Function Documentation

k_energy()

__global__ void k_energy	(	double4 *	r,
		double4 *	v,
		double *	ekin,
		double *	epot,
		int	n,
		int	dev_size,
		int	dev
	)

Energy kernel, in charge of the calculation of the kinetic and potential energy on the GPUs.

k_force_calculation()

__device__ void k_force_calculation	(	Predictor	i_p,
		Predictor	j_p,
		Forces &	f,
		double	e2
	)

Force interaction kernel, in charge of performing gravitational interaction computation between two particles.

k_init_acc_jrk()

__global__ void k_init_acc_jrk	(	Predictor *	p,
		Forces *	f,
		int	n,
		double	e2,
		int	dev,
		int	dev_size
	)

Initialization kernel, which consider an interaction of the particles.

k_prediction()

__global__ void k_prediction	(	Forces *	f,
		double4 *	r,
		double4 *	v,
		double *	t,
		Predictor *	p,
		int	dev_size,
		double	ITIME
	)

Predictor kernel, in charge of performing the prediction step of all the particles on each integration step.

k_reduce()

__global__ void k_reduce	(	Forces *	in,
		Forces *	out,
		int	shift_id,
		int	shift
	)

Force reduction kernel, in charge of summing up all the preliminary results of the forces for the particles.

k_update()

__global__ void k_update	(	Predictor *	i_p,
		Predictor *	j_p,
		Forces *	fout,
		int	n,
		int	total,
		double	e2
	)

Force kernel, in charge of performing distribution of how the particles will be distributed among the GPUs.

This kernel calls the k_prediction kernel.

operator+() [1/2]

__host__ __device__ double4 operator+ ( const double4 &  a, const double4 &  b  )

inline

operator+() [2/2]

__host__ __device__ Forces operator+ ( Forces &  a, Forces &  b  )

inline

operator+=() [1/2]

__host__ __device__ void operator+= ( double4 &  a, double4 &  b  )

inline

operator+=() [2/2]

__host__ __device__ void operator+= ( Forces &  a, Forces &  b  )

inline