Is there a better way to get number of threads per block is my way is correct and bug free?

Question

I was looking for on the internet for a official way to get the maximum number of threads per block for the current NVIDIA GPU and I found nothing.

I was read that the number must be 2^N so I did some code that find it but I am not sure that it is 100% correct and maybe there is a better way.

In my case it give me the output 1024 (for my NVIDIA GTX 660M).

This is my code:

#include "cuda_runtime.h"
#include "device_launch_parameters.h"

#include <stdio.h>
#include <iostream>
#include <Windows.h>




// Easy print macros:
#define pp std::cout <<
#define ss << " " <<
#define ee << std::endl;
#define ww while(1){Sleep(100);} return 0;



namespace nameMyCudaSystem {

    int threadsPerBlock = 0;


    __global__ void ThreadPerBlockCounter(){/*Do nothing*/}


    int InitializeCuda()
    {
        cudaError_t cudaStatus;


        // Choose which GPU to run on, change this on a multi-GPU system.
        cudaStatus = cudaSetDevice(0);
        if (cudaStatus != cudaSuccess) return -1;


        threadsPerBlock = 1;
        do
        {
            if (threadsPerBlock > 1)
            {
                cudaStatus = cudaDeviceSynchronize();
                if (cudaStatus != cudaSuccess)
                {
                    return -2;
                    break;
                }
            }
            threadsPerBlock *= 2;
            ThreadPerBlockCounter << <1, threadsPerBlock >> >();
        } while (cudaGetLastError() == cudaSuccess);

        threadsPerBlock /= 2;

        if (threadsPerBlock == 0) return -3;

        return 0;

    }



}






int main()
{

    if (nameMyCudaSystem::InitializeCuda() != 0) return -1; // error (some error inside InitializeCuda)

    pp nameMyCudaSystem::threadsPerBlock ee; // print the maximum number of thread per block

    ww; // just a loop to keep the console window open...

    return 0;
}