Add OMP/MPI run

-run shell to execute with 16 thread each process pinned to l3cache

Author: sbstndbs

hybride.c

@@ -7,12 +7,15 @@
 
 int main(int argc, char** argv){
 
+	unsigned long long N = 1000000 ; 
+	unsigned long long M = 100000;
+
 	int world_rank;
 	int world_size;
 
 	int namelen ; 
-	int iam = 0 ; 
-	int np = 1 ; 
+	int iam  ; 
+	int np ; 
 	MPI_Init(&argc,&argv);
 	MPI_Comm_rank(MPI_COMM_WORLD, &world_rank);
 	MPI_Comm_size(MPI_COMM_WORLD, &world_size);
@@ -26,7 +29,10 @@ int main(int argc, char** argv){
 	printf("Hello from thread %d out of %d from process %d out of %d on %s\n",
 	iam, np, world_rank, world_size, processor_name);
 
-	
+	float * array = malloc(N * sizeof(float));
+	for (unsigned long long i  = 0 ; i < N ; i++){
+		array[i]=0.0;
+	}
 
 	float value = 0.0 ; 
 	float recieved = 0.0 ; 
@@ -44,6 +50,9 @@ int main(int argc, char** argv){
 		}
 		// recieve values from every other processes
 		// in a sync way
+		
+		
+		
 		for (int i = 1 ; i < world_size ; i++){
 			MPI_Recv(&recieved, 1 , MPI_FLOAT , i ,i , MPI_COMM_WORLD, MPI_STATUS_IGNORE) ; 
 			if (recieved == (float)i){
@@ -75,28 +84,47 @@ int main(int argc, char** argv){
 			MPI_Isend(&value, 1 , MPI_FLOAT , i , i , MPI_COMM_WORLD, &request) ;  
 		}
 
+		
+		
+		
+		
 		// recieve values from every other processes
 		for (int i = 1 ; i < world_size ; i++){
 			MPI_Irecv(&recieved, 1 , MPI_FLOAT , i ,i , MPI_COMM_WORLD, &request) ; 
 			MPI_Wait(&request , &status) ; 
-			if (recieved == (float)i){
-				printf("process %d correctly send back\n", i);
+			if (recieved == (float)i*M){
+				printf("process %d correctly send back %f\n", i, recieved);
 			} 
 			else{
-				printf("process %d badly send back %f (%f excepted) \n", i, recieved , (float)i);				
+				printf("process %d badly send back %f (%f excepted) \n", i, recieved , (float)i*M);				
 			}
-		}			
+		}
+		
+		for (unsigned long long j = 0 ; j < M ; j++){
+			#pragma omp parallel for
+			for (unsigned long long i = 0 ; i < N ; i++){
+				array[i] += world_rank*1.0 ;  
+			}
+		}					
 	}
 	else if (world_rank < world_size){
 
 		MPI_Irecv(&value , 1 , MPI_FLOAT , 0 , world_rank , MPI_COMM_WORLD , &request ) ; 
+		MPI_Wait(&request , &status) ; 		
 		int inc = 0 ;
-		#pragma omp parallel for
-		for (int i = 0 ; i < 1000 ; i++){
-			inc += 1 ; 
-		}
-		MPI_Wait(&request , &status) ; 
-		MPI_Isend(&value , 1 , MPI_FLOAT , 0 , world_rank , MPI_COMM_WORLD, &request ) ;	
+
+
+
+		value = array[N-1];		
+		MPI_Isend(&value , 1 , MPI_FLOAT , 0 , world_rank , MPI_COMM_WORLD, &request ) ;
+		
+		for (unsigned long long j = 0 ; j < M ; j++){
+			#pragma omp parallel for
+			for (unsigned long long i = 0 ; i < N ; i++){
+				array[i] += world_rank*1.0 ;  
+			}
+		}		
+			
 	}		
 
 

run_with_omp.sh

@@ -0,0 +1,2 @@
+export OMP_NUM_THREADS=16
+mpirun -np 2 --map-by l3cache:PE=8 --bind-to core ./hybride