enlarge the matrix size

main.cu

 #include "test.h"
 #include "device_launch_parameters.h"
 #include "cuda_runtime.h"
+#include <ctime>
+#include "chrono"
+
+using namespace std::chrono;
 
 template <typename T>
 void printMat(const T &mat, 
@@ -29,9 +33,9 @@ int main(void) {
     std::cout << "Max Threads Batch Per MultiProcessor: | " << devProp.maxThreadsPerMultiProcessor / 32 << std::endl;
     std::cout << "---------------------------------------------------------" << std::endl;
 
-    int m = 4;
-    int k = 4;
-    int n = 4;
+    int m = 1000;
+    int k = 600;
+    int n = 1000;
     int nBytes_a = m * k * sizeof(float);
     int nBytes_b = k * n * sizeof(float);
     int nBytes_c = m * n * sizeof(float);
@@ -46,14 +50,14 @@ int main(void) {
     mat_c = (float *)malloc(nBytes_c);
     
     for (int i = 0; i < m * k; i++) {
-        mat_a[i] = i;
+        mat_a[i] = 10;
     }
-    printMat(mat_a, m, k);
+    // printMat(mat_a, m, k);
 
     for (int i = 0; i < k * n; i++) {
-        mat_b[i] = i;
+        mat_b[i] = 10;
     }
-    printMat(mat_b, k, n);
+    // printMat(mat_b, k, n);
 
     for (int i = 0; i < m * n; i++) {
         mat_c[i] = 0.0;
@@ -64,6 +68,8 @@ int main(void) {
     cudaMalloc((void **)&d_b, nBytes_b);
     cudaMalloc((void **)&d_c, nBytes_c);
 
+    auto start = system_clock::now();
+
     cudaMemcpy((void *)d_a, (void *)mat_a, nBytes_a, cudaMemcpyHostToDevice);
     cudaMemcpy((void *)d_b, (void *)mat_b, nBytes_b, cudaMemcpyHostToDevice);
     cudaMemcpy((void *)d_c, (void *)mat_c, nBytes_c, cudaMemcpyHostToDevice);
@@ -78,15 +84,24 @@ int main(void) {
     if (cudaStatus != cudaSuccess) {
         std::cout << "cudaError: " << cudaGetErrorString(cudaStatus) << std::endl;
     }
+
     cudaMemcpy((void *)mat_c, (void *)d_c, nBytes_c, cudaMemcpyDeviceToHost);
     cudaDeviceSynchronize();
-    printMat(mat_c, m, n);
+
+    auto end = system_clock::now();
+    auto duration = duration_cast<microseconds>(end - start);
     
+    std::cout << "GPU time: "
+         << double(duration.count()) * microseconds::period::num / microseconds::period::den << " seconds" << std::endl;
+
+    // printMat(mat_c, m, n);
+    /* 
     float maxError = 0.0;
     for (int i = 0; i < m * n; i++) {
         maxError = fmax(maxError, fabs(*(mat_c + i) - 100.0));
     }
     std::cout << "Max Error: " << maxError << std::endl;
+    */
 
     cudaFree(d_a);
     cudaFree(d_b);
@@ -96,7 +111,7 @@ int main(void) {
     free(mat_b);
     free(mat_c);
 
-    test<<<1, 1>>>();
-    std::cout << "Hello CUDA!" << std::endl;
+    // test<<<1, 1>>>();
+    // std::cout << "Hello CUDA!" << std::endl;
     return 0;
 }