carma_obj.h File Reference

#include <carma_context.h>
#include <carma_streams.h>
#include <carma_utils.h>
#include <curand.h>
#include <curand_kernel.h>
#include <iostream>
#include <type_traits>
#include <typeinfo>

Include dependency graph for carma_obj.h:

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

Go to the source code of this file.

Classes
class	CarmaData< T_data >
class	CarmaHostObj< T_data >
	this class provides wrappers to the generic carma host object More...
class	CarmaObj
	this class provides wrappers to the generic carma object More...

Macros
#define	BLOCK_SZ   16

Typedefs
typedef CarmaObj< int >	CarmaObjI
typedef CarmaObj< unsigned int >	CarmaObjUI
typedef CarmaObj< uint16_t >	CarmaObjUSI
typedef CarmaObj< float >	CarmaObjS
typedef CarmaObj< double >	CarmaObjD
typedef CarmaObj< float2 >	CarmaObjS2
typedef CarmaObj< double2 >	CarmaObjD2
typedef CarmaObj< cuFloatComplex >	CarmaObjC
typedef CarmaObj< cuDoubleComplex >	CarmaObjZ

Enumerations
enum	MemType {   MT_DEVICE, MT_DARRAY, MT_HOST, MT_PAGELOCK,   MT_ZEROCPY, MT_PORTABLE, MT_WRICOMB, MT_GENEPIN }

Functions
template<class T_data >
std::ostream &	operator<< (std::ostream &os, CarmaObj< T_data > &obj)
template<class T_data >
void	clip_array (T_data *d_data, T_data min, T_data max, int N, CarmaDevice *device)
template<class T_data >
void	reduce (int size, int threads, int blocks, T_data *d_idata, T_data *d_odata)
template<class T_data >
T_data	reduce (T_data *data, int N)
template<class T_data >
void	init_reduceCubCU (T_data *&cub_data, size_t &cub_data_size, T_data *data, T_data *&o_data, int N)
template<class T_data >
void	reduceCubCU (T_data *cub_data, size_t cub_data_size, T_data *data, T_data *o_data, int N)
template<class T_data >
int	transposeCU (T_data *d_idata, T_data *d_odata, long N1, long N2)
template<class T_data >
int	launch_generic1d (T_data *d_idata, T_data *d_odata, int N, CarmaDevice *device)
template<class T_data >
int	launch_generic2d (T_data *d_odata, T_data *d_idata, int N1, int N2)
int	carma_prng_init (int *seed, const int nb_threads, const int nb_blocks, curandState *state)
template<class T >
int	carma_prng_cu (T *results, const int nb_threads, const int nb_blocks, curandState *state, char gtype, int n, float alpha, float beta)
template<class T >
int	carma_curand_montagn (curandState *state, T *d_odata, int N, CarmaDevice *device)
template<class T_in , class T_out >
cufftType	carma_select_plan ()
template<class T_in , class T_out >
void	carma_initfft (const long *dims_data, cufftHandle *plan, cufftType type_plan)
template<class T_in , class T_out >
int	CarmaFFT (T_in *input, T_out *output, int dir, cufftHandle plan)
template<class T_data >
int	fillindex (T_data *d_odata, T_data *d_idata, int *indx, int N, CarmaDevice *device)
template<class T_data >
int	fillvalues (T_data *d_odata, T_data *val, int N, CarmaDevice *device)
template<class T >
int	getarray2d (T *d_odata, T *d_idata, int x0, int Ncol, int NC, int N, CarmaDevice *device)
template<class T >
int	fillarray2d (T *d_odata, T *d_idata, int x0, int Ncol, int NC, int N, CarmaDevice *device)
template<class T >
int	fillarray2d2 (T *d_odata, T *d_idata, int x0, int Ncol, int NC, int N, CarmaDevice *device)
template<class T >
int	fill_sym_matrix (char src_uplo, T *d_data, int Ncol, int N, CarmaDevice *device)
template<class T >
int	carma_plus (T *d_odata, T elpha, int N, CarmaDevice *device)
template<class T >
int	carma_plusai (T *d_odata, T *i_data, int i, int sgn, int N, CarmaDevice *device)
int	fftconv_unpad (float *d_odata, float *d_idata, int fftW, int dataH, int dataW, int N, int n, int nim)
int	carma_initfftconv (CarmaObjS *data_in, CarmaObjS *kernel_in, CarmaObjS *padded_data, CarmaObjC *padded_spectrum, int kernelY, int kernelX)
int	carma_fftconv (CarmaObjS *data_out, CarmaObjS *padded_data, CarmaObjC *padded_spectrum, int kernelY, int kernelX)
template<class T >
int	extract (T *d_smallimg, const T *d_fullimg, int fullimg_size, int center_pos, int extract_size, bool roll)
	Kernel to extract a part of the image centred on center_pos. More...

Macro Definition Documentation

BLOCK_SZ

#define BLOCK_SZ   16

Definition at line 79 of file carma_obj.h.

Typedef Documentation

CarmaObjC

typedef CarmaObj<cuFloatComplex> CarmaObjC

Definition at line 381 of file carma_obj.h.

CarmaObjD

typedef CarmaObj<double> CarmaObjD

Definition at line 378 of file carma_obj.h.

CarmaObjD2

typedef CarmaObj<double2> CarmaObjD2

Definition at line 380 of file carma_obj.h.

CarmaObjI

typedef CarmaObj<int> CarmaObjI

Definition at line 374 of file carma_obj.h.

CarmaObjS

typedef CarmaObj<float> CarmaObjS

Definition at line 377 of file carma_obj.h.

CarmaObjS2

typedef CarmaObj<float2> CarmaObjS2

Definition at line 379 of file carma_obj.h.

CarmaObjUI

typedef CarmaObj<unsigned int> CarmaObjUI

Definition at line 375 of file carma_obj.h.

CarmaObjUSI

typedef CarmaObj<uint16_t> CarmaObjUSI

Definition at line 376 of file carma_obj.h.

CarmaObjZ

typedef CarmaObj<cuDoubleComplex> CarmaObjZ

Definition at line 382 of file carma_obj.h.

Enumeration Type Documentation

MemType

enum MemType

Enumerator
MT_DEVICE
MT_DARRAY
MT_HOST
MT_PAGELOCK
MT_ZEROCPY
MT_PORTABLE
MT_WRICOMB
MT_GENEPIN

Definition at line 80 of file carma_obj.h.

             {
  MT_DEVICE,
  MT_DARRAY,
  MT_HOST,
  MT_PAGELOCK,
  MT_ZEROCPY,
  MT_PORTABLE,
  MT_WRICOMB,
  MT_GENEPIN
};

Function Documentation

carma_curand_montagn()

template<class T >

int carma_curand_montagn	(	curandState *	state,
		T *	d_odata,
		int	N,
		CarmaDevice *	device
	)

carma_fftconv()

int carma_fftconv	(	CarmaObjS *	data_out,
		CarmaObjS *	padded_data,
		CarmaObjC *	padded_spectrum,
		int	kernelY,
		int	kernelX
	)

carma_initfft()

template<class T_in , class T_out >

void carma_initfft	(	const long *	dims_data,
		cufftHandle *	plan,
		cufftType	type_plan
	)

carma_initfftconv()

int carma_initfftconv	(	CarmaObjS *	data_in,
		CarmaObjS *	kernel_in,
		CarmaObjS *	padded_data,
		CarmaObjC *	padded_spectrum,
		int	kernelY,
		int	kernelX
	)

carma_plus()

template<class T >

int carma_plus	(	T *	d_odata,
		T	elpha,
		int	N,
		CarmaDevice *	device
	)

carma_plusai()

template<class T >

int carma_plusai	(	T *	d_odata,
		T *	i_data,
		int	i,
		int	sgn,
		int	N,
		CarmaDevice *	device
	)

carma_prng_cu()

template<class T >

int carma_prng_cu	(	T *	results,
		const int	nb_threads,
		const int	nb_blocks,
		curandState *	state,
		char	gtype,
		int	n,
		float	alpha,
		float	beta
	)

carma_prng_init()

int carma_prng_init	(	int *	seed,
		const int	nb_threads,
		const int	nb_blocks,
		curandState *	state
	)

carma_select_plan()

template<class T_in , class T_out >

cufftType carma_select_plan

(

)

CarmaFFT()

template<class T_in , class T_out >

int CarmaFFT	(	T_in *	input,
		T_out *	output,
		int	dir,
		cufftHandle	plan
	)

clip_array()

template<class T_data >

void clip_array	(	T_data *	d_data,
		T_data	min,
		T_data	max,
		int	N,
		CarmaDevice *	device
	)

extract()

template<class T >

int extract	(	T *	d_smallimg,
		const T *	d_fullimg,
		int	fullimg_size,
		int	center_pos,
		int	extract_size,
		bool	roll
	)

Kernel to extract a part of the image centred on center_pos.

Template Parameters

T	type of the image items

Parameters

d_smallimg	extracted small image of size extract_size*extract_size
d_fullimg	full image of size fullimg_size*fullimg_size
fullimg_size	size of the d_fullimg leading dimension
center_pos	position of the center of d_smallimg in d_fullimg
extract_size	size of the d_smallimg leading dimension
roll	get pixels as if d_fullimg need to be roll

fftconv_unpad()

int fftconv_unpad	(	float *	d_odata,
		float *	d_idata,
		int	fftW,
		int	dataH,
		int	dataW,
		int	N,
		int	n,
		int	nim
	)

fill_sym_matrix()

template<class T >

int fill_sym_matrix	(	char	src_uplo,
		T *	d_data,
		int	Ncol,
		int	N,
		CarmaDevice *	device
	)

fillarray2d()

template<class T >

int fillarray2d	(	T *	d_odata,
		T *	d_idata,
		int	x0,
		int	Ncol,
		int	NC,
		int	N,
		CarmaDevice *	device
	)

fillarray2d2()

template<class T >

int fillarray2d2	(	T *	d_odata,
		T *	d_idata,
		int	x0,
		int	Ncol,
		int	NC,
		int	N,
		CarmaDevice *	device
	)

fillindex()

template<class T_data >

int fillindex	(	T_data *	d_odata,
		T_data *	d_idata,
		int *	indx,
		int	N,
		CarmaDevice *	device
	)

fillvalues()

template<class T_data >

int fillvalues	(	T_data *	d_odata,
		T_data *	val,
		int	N,
		CarmaDevice *	device
	)

getarray2d()

template<class T >

int getarray2d	(	T *	d_odata,
		T *	d_idata,
		int	x0,
		int	Ncol,
		int	NC,
		int	N,
		CarmaDevice *	device
	)

init_reduceCubCU()

template<class T_data >

void init_reduceCubCU	(	T_data *&	cub_data,
		size_t &	cub_data_size,
		T_data *	data,
		T_data *&	o_data,
		int	N
	)

launch_generic1d()

template<class T_data >

int launch_generic1d	(	T_data *	d_idata,
		T_data *	d_odata,
		int	N,
		CarmaDevice *	device
	)

launch_generic2d()

template<class T_data >

int launch_generic2d	(	T_data *	d_odata,
		T_data *	d_idata,
		int	N1,
		int	N2
	)

operator<<()

template<class T_data >

std::ostream& operator<<	(	std::ostream &	os,
		CarmaObj< T_data > &	obj
	)

Definition at line 390 of file carma_obj.h.

                                                              {
  os << "-----------------------" << std::endl;
  os << "CarmaObj<" << typeid(T_data).name() << "> object on GPU"
     << obj.get_device() << std::endl;
  long ndims = obj.get_dims(0);
  os << "ndims = " << ndims << std::endl;
  for (long dim = 0; dim < ndims; dim++) {
    os << "dim[" << dim << "] = " << obj.get_dims(dim + 1) << std::endl;
  }
  os << "nbElem = " << obj.get_nb_elements() << std::endl;
  os << "sizeof(" << typeid(T_data).name() << ") = " << sizeof(T_data)
     << std::endl;
  os << "-----------------------" << std::endl;
  return os;
}

Here is the call graph for this function:

reduce() [1/2]

template<class T_data >

void reduce	(	int	size,
		int	threads,
		int	blocks,
		T_data *	d_idata,
		T_data *	d_odata
	)

reduce() [2/2]

template<class T_data >

T_data reduce	(	T_data *	data,
		int	N
	)

reduceCubCU()

template<class T_data >

void reduceCubCU	(	T_data *	cub_data,
		size_t	cub_data_size,
		T_data *	data,
		T_data *	o_data,
		int	N
	)

transposeCU()

template<class T_data >

int transposeCU	(	T_data *	d_idata,
		T_data *	d_odata,
		long	N1,
		long	N2
	)