33 lines
744 B
Plaintext
33 lines
744 B
Plaintext
#include <stdio.h>
|
|
#include <cuda_runtime.h>
|
|
|
|
__global__ void myKernel() {
|
|
// Kernel code goes here
|
|
}
|
|
|
|
int main() {
|
|
// Set the device (GPU)
|
|
cudaSetDevice(0);
|
|
|
|
// Get device properties
|
|
cudaDeviceProp prop;
|
|
cudaGetDeviceProperties(&prop, 0);
|
|
|
|
// Set kernel parameters (adjust as needed)
|
|
int numThreadsPerBlock = 256;
|
|
int numBlocks = 16;
|
|
|
|
// Compute occupancy
|
|
int activeWarps;
|
|
int maxWarps;
|
|
cudaOccupancyMaxActiveBlocksPerMultiprocessor(&activeWarps, myKernel, numThreadsPerBlock, 0);
|
|
maxWarps = prop.maxThreadsPerMultiProcessor / prop.warpSize;
|
|
|
|
// Compute occupancy ratio
|
|
float occupancy = (float)activeWarps / maxWarps;
|
|
|
|
printf("Occupancy: %f\n", occupancy);
|
|
|
|
return 0;
|
|
}
|