ex7: start

exercise7/wave_2d_parallel.cu

@@ -1,3 +1,4 @@
+#include <cuda_runtime_api.h>
 #include <errno.h>
 #include <inttypes.h>
 #include <math.h>
@@ -11,7 +12,8 @@
 // TASK: T1
 // Include the cooperative groups library
 // BEGIN: T1
-;
+#include <cooperative_groups.h>
+namespace cg = cooperative_groups; // TODO
 // END: T1
 
 // Convert 'struct timeval' into seconds in double prec. floating point
@@ -172,6 +174,28 @@ void occupancy(void) {
 // Make sure at least one CUDA-capable device exists
 static bool init_cuda() {
   // BEGIN: T2
+  int count;
+  if (cudaGetDeviceCount(&count) != cudaSuccess)
+    return false;
+
+  printf("CUDA device count: %d\n", count);
+
+  if (count > 0) {
+    cudaDeviceProp p;
+    if (cudaSetDevice(0) != cudaSuccess)
+      return false;
+    if (cudaGetDeviceProperties(&p, 0) != cudaSuccess)
+      return false;
+
+    printf("CUDA device #0:\n");
+    printf("  Name: %s\n", p.name);
+    printf("  Compute capability: %d.%d\n", p.major, p.minor);
+    printf("  Multiprocessors: %d\n", p.multiProcessorCount);
+    printf("  Warp size: %d\n", p.warpSize);
+    printf("  Global memory: %.1fGiB bytes\n", p.totalGlobalMem / (1024.0 * 1024.0 * 1024.0));
+    printf("  Per-block shared memory: %.1fKiB\n", p.sharedMemPerBlock / 1024.0);
+    printf("  Per-block registers: %d\n", p.regsPerBlock);
+  }
   return true;
   // END: T2
 }