All dependancies are included under libs with the exception of MPI. The output of the eight body run is located in the file eightoutput.txt. That includes 8 bodies on one processor with 100 ticks Pictures are in the screenshots/ folder with varius confiurations. To run without a display and without MPI the first edit the header file nbody.h and set DISPLAY_ON to false and MPI_ON to false. Compile with : $ make clean; make then run with : $./bin/nbody You can adjust the number of bodies and the amount of time in the header file also. To run with MPI set MPI_ON in nbody.h to true then: $ make clean; make; and run $job.sh or $ mpiexec -n 16 -hostfile hostfile ./bin/nbody ----------------====== Benchmark results: ======----------------- My results weren't as successful as the cluster was in use at the time. 8 bodies, 1 process 100 ticks -- real 0m0.131s user 0m0.009s sys 0m0.010s 8 bodies, 8 processes, 100 ticks -- real 0m1.191s user 0m0.140s sys 0m0.133s 800 bodies, 1 process, 100 ticks -- real 0m4.558s user 0m4.431s sys 0m0.009s 800 bodies, 4 processes, 100 ticks -- real 0m6.991s user 0m14.813s sys 0m0.669s 800 bodies, 16 processes, 100 ticks -- real 0m7.385s user 0m14.286s sys 0m8.320s 800 bodies, 1 process, 1000 ticks -- real 0m44.676s user 0m44.476s sys 0m0.024s 800 bodies, 16 process, 1000 ticks -- real 0m59.973s user 2m22.241s sys 1m16.705s Someone else is using the cluster at this time. Sunday afternoon / night. The cluster goes faster however only at times. Seeing as the host computer doesn't actually do any processing for the cluster, it appears to get better results. I can only conclude from my results that the parallel computing on negitivily reflects my results.