tmp_par.cu

#include <stdio.h>
int n = 1; 
int *count; 
cudaMallocManaged(&count, n * sizeof(int)); 
int n = 1; 
int *valor; 
cudaMallocManaged(&valor, n * sizeof(int));
// #include <omp.h>

// #include <stdio.h>

// #include <stdlib.h>





size_t number_of_blocks = 1000;


size_t threads_per_block = 1000;





//Chamada da função GPUFuncion

__global__ void GPUFunction(double nCuda , *valor , double *valor ){


int idx = blockIdx.x * blockDim.x + threadIdx.x;


int totalThreads = gridDim.x * blockDim.x;


for (double idxCuda = idx; idxCuda < nCuda ; idxCuda += totalThreads){


//parte que será paralelizada

valor = valor + 4.0/(1.0 + ((idx + 0.5)*(1.0/(double)2147480000))*((idx + 0.5)*(1.0/(double)2147480000)));

}


}




int main(int argc, char** argv){



double valor = 0;

// double passo;

// passo = 1.0/(double)2147480000;



GPUFunction<<<number_of_blocks, threads_per_block>>>(2147480000, count, valor);


cudaDeviceSynchronize();







valor = valor*(1.0/(double)2147480000);



printf("%f\n", valor);

return 0;

}