a very simple problem

Accelerated Computing CUDA CUDA Programming and Performance
1

I am just starting (like today) to program in CUDA. So i copied the makefile off one of the SDK examples and renamed some variables to make it make my own file.

The problem is that i cant get the compiler to recognise the built in variable threadIdx. I have a very simple program and i am using the command make emu=1 so that i can have printf statements in there.

Here is my code

#include <stdlib.h>

#include <stdio.h>

#include <string.h>

#include <math.h>

// includes, project

#include <cufft.h>

#include <cutil.h>

void some(){

        int test = threadIdx.x;

        printf("%i\n",test);

}

int main(int argc, char** argv)

{

        some();

   CUT_EXIT(argc, argv);

}

and the compiler error message i get is:

-bash-3.2$ make emu=1

dave1.cu: In function 'void some()':

dave1.cu:13: error: 'threadIdx' was not declared in this scope

make: *** [obj/emurelease/dave1.cu_o] Error 255

-bash-3.2$

What am i doing wrong?

2

You need to declare the function a CUDA function or “kernel”, like so:

global void some(){
int test = threadIdx.x;
printf(“%i\n”,test);
}

Also, you’ll need to pass in a grid and block thread to call that function as a CUDA kernel:

dim3 gridConfig(xBlocksInGrid, yBlocksInGrid);
dim3 blockConfig(xThreadsPerBlock, yThreadsPerBlock);
some<<<gridConfig, blockConfig>>>();