GPUMATMULTCONF Struct Reference

This structure holds the GPU computation setup for matrix multiplication. More...

#include <cudacomp.h>

Data Fields
int_fast8_t	init
int_fast8_t *	refWFSinit
int_fast8_t	alloc
long	CM_ID
long	CM_cnt
long	timerID
uint_fast32_t	M
uint_fast32_t	N
int_fast8_t	sem
	synchronization More...
int_fast8_t	gpuinit
sem_t **	semptr1
	one semaphore per thread More...
sem_t **	semptr2
sem_t **	semptr3
sem_t **	semptr4
sem_t **	semptr5
float *	cMat
float **	cMat_part
float *	wfsVec
float **	wfsVec_part
float *	wfsRef
float **	wfsRef_part
float *	dmVec
float *	dmVecTMP
float **	dmVec_part
float **	dmRef_part
float **	d_cMat
float **	d_wfsVec
float **	d_dmVec
float **	d_wfsRef
float **	d_dmRef
THDATA *	thdata
int *	iret
pthread_t *	threadarray
int_fast8_t	NBstreams
cudaStream_t *	stream
cublasHandle_t *	handle
uint_fast32_t *	Nsize
uint_fast32_t *	Noffset
int *	GPUdevice
int_fast8_t	orientation
long	IDout

Detailed Description

This structure holds the GPU computation setup for matrix multiplication.

By declaring an array of these structures, several parallel computations can be executed

Field Documentation

int_fast8_t GPUMATMULTCONF::alloc

1 if memory has been allocated

long GPUMATMULTCONF::CM_cnt

long GPUMATMULTCONF::CM_ID

float* GPUMATMULTCONF::cMat

float** GPUMATMULTCONF::cMat_part

float** GPUMATMULTCONF::d_cMat

float** GPUMATMULTCONF::d_dmRef

float** GPUMATMULTCONF::d_dmVec

float** GPUMATMULTCONF::d_wfsRef

float** GPUMATMULTCONF::d_wfsVec

float** GPUMATMULTCONF::dmRef_part

float* GPUMATMULTCONF::dmVec

float** GPUMATMULTCONF::dmVec_part

float* GPUMATMULTCONF::dmVecTMP

int* GPUMATMULTCONF::GPUdevice

int_fast8_t GPUMATMULTCONF::gpuinit

cublasHandle_t* GPUMATMULTCONF::handle

long GPUMATMULTCONF::IDout

int_fast8_t GPUMATMULTCONF::init

1 if initialized

int* GPUMATMULTCONF::iret

uint_fast32_t GPUMATMULTCONF::M

uint_fast32_t GPUMATMULTCONF::N

int_fast8_t GPUMATMULTCONF::NBstreams

uint_fast32_t* GPUMATMULTCONF::Noffset

uint_fast32_t* GPUMATMULTCONF::Nsize

int_fast8_t GPUMATMULTCONF::orientation

int_fast8_t* GPUMATMULTCONF::refWFSinit

reference init

int_fast8_t GPUMATMULTCONF::sem

synchronization

if sem = 1, wait for semaphore to perform computation

sem_t** GPUMATMULTCONF::semptr1

one semaphore per thread

command to start matrix multiplication (input)

sem_t** GPUMATMULTCONF::semptr2

memory transfer to device completed (output)

sem_t** GPUMATMULTCONF::semptr3

computation done (output)

sem_t** GPUMATMULTCONF::semptr4

command to start transfer to host (input)

sem_t** GPUMATMULTCONF::semptr5

output transfer to host completed (output)

cudaStream_t* GPUMATMULTCONF::stream

THDATA* GPUMATMULTCONF::thdata

pthread_t* GPUMATMULTCONF::threadarray

long GPUMATMULTCONF::timerID

float* GPUMATMULTCONF::wfsRef

float** GPUMATMULTCONF::wfsRef_part

float* GPUMATMULTCONF::wfsVec

float** GPUMATMULTCONF::wfsVec_part

---

The documentation for this struct was generated from the following file:

• src/cudacomp/cudacomp.h