InnovArul

#include <stdio.h>

/**
* A dummy cuda asynchronous function 
*/
__global__ void fillContents(int N, int* output)
{
   int correctIndex = threadIdx.x * N;
   
   for(int i = correctIndex; i < N; i++)

InnovArul

/**
 * API to report the memory usage of the GPU
 */
static void reportMemStatus() {

	// show memory usage of GPU
	size_t free_byte;
	size_t total_byte;
	size_t malloc_byte;

InnovArul

#define MALLOC_LIMIT 1024*1024*1024 //1 GB

//eliminate calling cudaDeviceSetLimit multiple times
bool IsMallocSet = false;

/**
 * API to set the malloc limit of GPU
 */
static void setMallocLimit() {

InnovArul

#include "stdio.h"

//enable error check
#define CUDA_ERROR_CHECK

//check the synchronous function call errorcode 'err' if it is a cudaSuccess
#define CudaSafeCall( err ) __cudaSafeCall( err, __FILE__, __LINE__ )

//check if any error happened during asynchronous execution of Cuda kernel __global__ function
#define CudaCheckError()    __cudaCheckError( __FILE__, __LINE__ )

InnovArul

require 'torch';
require 'math'
require 'io'
require 'cutorch';
require 'cunn';


opt = 'DOUBLE'  -- FLOAT / DOUBLE / CUDA

-----------------------------------------------------------