Blockwise info of data how to get this info: Block0 operating on xyz, Blok1 operating on abc


The following program simply generates n numbers on host , copy it to the device and then again copy back to the host and prints the numbers. Basically I need it somewhere in a code that I am writing. My requirement is that I must be able to generate and pass more than 1024 elements. In the code below since I am loading two elements per thread so keeping in mind maximum threads per block limitation of 512, I can access not more than 1024 elements, as one thread is loading two elements. So I introduced multiple blocks by using the following statements

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

The program is working fine. But for debugging purpose I want to know which block of thread is actually working (that is copying data) on which data set. For example suppose we generate 8 numbers

g_idata = [1 2 3 4 5 6 7 8 ]

And two blocks, B0 and B1.

Now, once I run this program I want the output something like this:


I am Block B0. I am operating on the data set [1 2 3 4]

I am Block B1. I am operating on the data set [5 6 7 8 ]


In short I want to know data corresponding to each block. I would be grateful to you guys for helping me here.

Thanks for your precious time.

__global__ void modified(float *out, float *g_idata, float *g_blockSums, unsigned int n)


	extern __shared__ float temp[];// allocated on invocation

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

	int ai = thid;

	int bi = thid + n/2;

	temp[ai]  = g_idata[ai];

	temp[bi] = g_idata[bi];

	out[ai] = temp[ai];

	out[bi] = temp[bi];


int  main()


unsigned int num_Elements, blockSize; //Number of data elements 


printf("Please enter the number of elements");

scanf("%d", &num_Elements);

printf("Please enter the Block Size");

scanf("%d", &blockSize);

	size_t size=sizeof(float)*num_Elements;//Size Of Memory

	size_t sizeBlockSums=sizeof(float)*1;

	float *out, *g_idata, *m_odata, *m_idata, *g_blockSums, *m_blockSums;//Declaration of Host and Device Pointer

	m_idata=(float*)malloc(size);//MemoryAllocation On Host

	m_odata=(float*)malloc(size);//Same above


	cudaMalloc((void**)&g_idata,size);//Memory Allocation On Device

	cudaMalloc((void**)&out,size);//Same Above

	// g-->global memory, m--> main memory 

//Intialization of Elements

	for (int i = 0; i < num_Elements; i++)

{		m_idata[i] = i+1;


	cudaMemcpy(g_idata,m_idata,size,cudaMemcpyHostToDevice);//copy the data m_idata from main memory into g_idata in device using cudaMemcpy 

int nBlocks=num_Elements/blockSize; //Number of Blocks

unsigned int nThreads=blockSize/2; //blockDimx.x, number of threads, /2 as one thread is loading two elements

printf("\nNumber of blocks is %d\n and number of threads per block(Block Dimesion) is %d\n",  nBlocks, nThreads); 

modified<<<nBlocks, nThreads ,size >>>(out, g_idata, g_blockSums, num_Elements);//Call the kernel 

cudaMemcpy(m_odata,out ,size,cudaMemcpyDeviceToHost);//copy the data g_odata from Device memory into m_odata in Global Memory

for(unsigned int i=0;i<num_Elements;i++)


		printf("m_odata[%d] = %f\n",i+1, m_odata[i]);


//Free Memory of Host and Device