HW8: Basic MPI

• Send the integer 17 from MPI process 1 to process 0 (one send, one receive).
  
• Broadcast the integer 17 from process 0 out to all processes.
• Split up an image just like ~olawlor/481_demo/image_split/, but make sure the final image is computed properly (no black border at the bottom) even if the number of image rows is not an integer multiple of the number of MPI processes.
  
• Open a local GLUT window for each process and repeat the following one thousand times: synchronize all processes and draw a local animating triangle.  You can assume the DISPLAY variable for each rank can be set like this (before glutInit!):
  

  	char myDisplay[100];
  	sprintf(myDisplay,"powerwall%d:0",rank);
  	setenv("DISPLAY",myDisplay,1);

  WARNING: with this setup, since rank 6 is sending its output to DISPLAY=powerwall6:0, you *cannot* use GLSL because rank 6 normally runs on powerwall3 (=6/2, since there are 2 processors per machine), and GLSL doesn't like being sent across the network. Also beware that your reshape routine might *not* get called, so make sure your glViewport and matrices will be correct regardless (for example, you could call main_repaint directly from your main routine). I recommend starting from a very simple OpenGL program like the 481_matrix/compact_noglui.cpp example linked off the main page. (This problem is hard, but this will motivate you to learn the much nicer MPIglut, instead of bare MPI!)