Adding material of OpenACC/OpenMP talk

presentations/2024_11_25_openacc_openmp/mandelbrot.c

@@ -0,0 +1,70 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <omp.h>
+#include <openacc.h>
+
+#define MAX_ITER 10000
+#define MAX_PALETTE 100
+#define f 1
+
+static const int Nx = 1920 * f;
+static const int Ny = 1080 * f;
+
+int mandelbrot(int i, int j, int Nx, int Ny);
+void write_to_file(int* M);
+
+int main(int argc, char* argv[]){
+
+    int* M = calloc (Nx * Ny, sizeof(int));
+
+    //#pragma acc parallel loop                                                 // Use OpenACC to parallelize the for loops
+    //#pragma omp target parallel for map(to: Nx, Ny) map(from: M[0: Nx*Ny])    // Use OpenMP  to parallelize the for loops 
+    for(int j = 0; j < Ny; j++){
+        for(int i = 0; i < Nx; i++){
+            M[j * Nx + i] = mandelbrot(i, j, Nx, Ny) % MAX_PALETTE;
+        }
+    }
+
+    //write_to_file(M);  // Comment out to just check the performances
+
+    return 0;
+
+}
+
+int mandelbrot(int i, int j, int Nx, int Ny){
+    static const double x_L = -1.7687784;
+    static const double x_R = -1.7687792;
+    static const double y_L = -0.0017386;
+    static const double y_R = -0.0017392;
+
+    double px = x_L + i * (x_R - x_L) / Nx;
+    double py = y_L + j * (y_R - y_L) / Ny;
+
+    double xc = px;
+    double temp_xc = xc;
+    double yc = py;
+    short int iter = 0;
+    while(xc*xc + yc*yc < 4 && iter < MAX_ITER){
+        temp_xc = xc*xc - yc*yc + px;
+        yc = 2.*xc*yc + py;
+        xc = temp_xc;
+
+        iter++;
+    }
+    return iter;
+}
+
+
+void write_to_file(int* M){
+    FILE *file = fopen("matrix.txt", "w");
+    for (int j = 0; j < Ny; j++) {
+        for (int i = 0; i < Nx; i++) {
+
+            fprintf(file, "%d ", M[j * Nx + i]);
+        }
+        fprintf(file, "\n");  // Newline after each row
+    }
+    fclose(file);
+}
+
+

presentations/2024_11_25_openacc_openmp/test_openacc.c

@@ -0,0 +1,25 @@
+#include <stdio.h>
+#include <unistd.h>
+#include <openacc.h>
+
+/*************************************************/
+/**  HOW TO COMPILE: nvc -acc test_openacc.c    **/
+/*************************************************/
+
+int main(){
+
+  #pragma acc parallel   // Open a parallel region with OpenACC. Whatever is inside the brackets will be run on the GPU
+  {
+    if(acc_on_device(acc_device_default)){   // Function provided by OpenACC to check whether a code is being run on the GPU (the device)
+      printf("Running on GPU\n");
+    }else{
+      printf("Running on CPU\n");
+    }
+  } // acc parallel
+
+  sleep(10);  // Wait for 10 seconds, to check that nvidia-smi shows that a code is running on the GPU
+
+  return 0;
+
+
+}

presentations/2024_11_25_openacc_openmp/test_openmp.c

@@ -0,0 +1,23 @@
+#include <stdio.h>
+#include <unistd.h>
+#include <omp.h>
+
+/***************************************************/
+/**  HOW TO COMPILE: nvc -mp=gpu test_openmp.c    **/
+/***************************************************/
+
+int main(){
+
+  #pragma omp target // Open a parallel region with OpenMP. Whatever is inside the brackets will be run on the GPU
+  {
+    if(!omp_is_initial_device()){ // Function provided by OpenMP to check whether a code is being run on the GPU or on the CPU (the "initial_device")
+      printf("Running on GPU\n");
+    }else{
+      printf("Running on CPU\n");
+    }
+  } // omp target
+
+  sleep(10);  // Wait for 10 seconds, to check that nvidia-smi shows that a code is running on the GPU
+
+  return 0;
+}