parallize outer loop

jacobi.c

@@ -67,16 +67,15 @@ int Jacobi_Iterate(int neighbor_dim0_left, int neighbor_dim0_right, int neighbor
 	double delta_a;
 	double **swap;
 	double maxdiff = 0;
-	double inner_maxdiff;
+	int i,j;
+	#pragma omp parallel for private(delta_a, j, i) reduction(max: maxdiff) schedule(dynamic, 1) default(shared)
 	for(
-		int i = (neighbor_dim0_left == MPI_PROC_NULL) ? 2 : 1; // catch edges
+		i = (neighbor_dim0_left == MPI_PROC_NULL) ? 2 : 1; // catch edges
 		i < pi.end_m - pi.start_m + ((neighbor_dim0_right == MPI_PROC_NULL) ? 1 : 2); 
 		i++
 	) {
-		inner_maxdiff = 0;
-		#pragma omp parallel for private(delta_a) reduction(max: inner_maxdiff) schedule(dynamic, 512) default(shared) num_threads(48)
 		for(
-			int j = (neighbor_dim1_left == MPI_PROC_NULL) ? 2 : 1; // catch edges
+			j = (neighbor_dim1_left == MPI_PROC_NULL) ? 2 : 1; // catch edges
 			j < pi.end_n - pi.start_n + ((neighbor_dim1_right == MPI_PROC_NULL) ? 1 : 2); 
 			j++
 		) {
@@ -87,11 +86,9 @@ int Jacobi_Iterate(int neighbor_dim0_left, int neighbor_dim0_right, int neighbor
 			delta_a = delta_a * delta_t;
 			(*partial_field_clipboard)[i][j] = (*partial_field)[i][j] + delta_a;
 			
-			if(delta_a > inner_maxdiff)
-				inner_maxdiff = delta_a;
+			if(delta_a > maxdiff)
+				maxdiff = delta_a;
 		}
-		if(inner_maxdiff > maxdiff)
-			maxdiff = inner_maxdiff;
 	}
 	swap = *partial_field_clipboard;
 	*partial_field_clipboard = *partial_field;

makefile

 all: main
 
 main: matrix.c main.c args.c pid0.c pi.c cart.c mpi_util.c
-	mpicc matrix.c main.c args.c pid0.c pi.c cart.c mpi_util.c jacobi.c -std=c99 -lm -o main
+	mpicc matrix.c main.c args.c pid0.c pi.c cart.c mpi_util.c jacobi.c -fopenmp -std=c99 -lm -o main
 
 ViewMatrix.class: ViewMatrix.java
 	javac ViewMatrix.java
@@ -9,4 +9,4 @@ ViewMatrix.class: ViewMatrix.java
 clean:
 	rm -f *.o *.c~ heat Matrix.txt
 	rm -f ViewMatrix.class
- 
+ 
\ No newline at end of file