wip

hybrid/mpimppi.c

+/*
+* mpimppi.c
+* Neil Gershenfeld  10/8/19
+* MPI+OpenMP pi calculation benchmark
+* pi = 3.14159265358979323846
+*/
+
+#include <stdio.h>
+#include <mpi.h>
+#include <omp.h>
+#include <sys/time.h>
+#include <stdint.h>
+
+#define NPTS 10000000000
+#define NLOOP 10
+
+void main(int argc, char** argv) {
+   uint32_t rank,nproc;
+   uint64_t i,j,istart,iend,start_time,end_time;
+   struct timeval start,end;
+   double sum,pi,mflops,max;
+   MPI_Init(&argc,&argv);
+   MPI_Comm_rank(MPI_COMM_WORLD,&rank);
+   MPI_Comm_size(MPI_COMM_WORLD,&nproc);
+   max = 0;
+   if (rank == 0) {
+      istart = 1+rank*NPTS;
+      iend = (rank+1)*NPTS;
+      for (j = 0; j < NLOOP; ++j) {
+         MPI_Barrier(MPI_COMM_WORLD);
+         gettimeofday(&start,NULL);
+         sum = 0.0;
+         #pragma omp parallel for reduction(+:sum)
+         for (i = istart; i <= iend; ++i)
+            sum += 0.5/((i-0.75)*(i-0.25));
+         MPI_Reduce(&sum,&pi,1,MPI_DOUBLE,MPI_SUM,0,MPI_COMM_WORLD);
+         gettimeofday(&end,NULL);
+         start_time = start.tv_sec*1e6+start.tv_usec;
+         end_time = end.tv_sec*1e6+end.tv_usec;
+         mflops = nproc*NPTS*(5.0/(end_time-start_time));
+         printf("processes = %d, threads = %d, NPTS = %ld, pi = %f\n",nproc,omp_get_max_threads(),NPTS,pi);
+         if (mflops > max) max = mflops;
+         printf("time = %f, estimated MFlops = %f, max = %f\n",(end_time-start_time)/1.0e6,mflops,max);
+         }
+      }
+   else {
+      istart = 1+rank*NPTS;
+      iend = (rank+1)*NPTS;
+      for (j = 0; j < NLOOP; ++j) {
+         MPI_Barrier(MPI_COMM_WORLD);
+         sum = 0.0;
+         #pragma omp parallel for reduction(+:sum)
+         for (i = istart; i <= iend; ++i)
+            sum += 0.5/((i-0.75)*(i-0.25));
+         MPI_Reduce(&sum,&pi,1,MPI_DOUBLE,MPI_SUM,0,MPI_COMM_WORLD);
+         }
+      }
+   MPI_Finalize();
+   }
+