03_Scaling_Python.ipynb

{
 "metadata": {
  "name": "03_Scaling_Python"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "<style>\n",
      "    h1{\n",
      "        text-align:center;\n",
      "    }\n",
      "    h2{\n",
      "        text-align:center;\n",
      "        font-variant:small-caps;\n",
      "    }\n",
      "</style>\n",
      "# Scaling Python\n",
      "# Python in HPC\n",
      "\n",
      "##Supercomputing 2012\n",
      "\n",
      "Presenters:\n",
      "\n",
      "Andy R. Terrel, PhD  \n",
      "Texas Advanced Computing Center  \n",
      "University of Texas at Austin  \n",
      "\n",
      "Travis Oliphant, PhD  \n",
      "Continuum Analytics\n",
      "\n",
      "Aron Ahmadia, PhD  \n",
      "Supercomputing Laboratory  \n",
      "King Abdullah University of Science and Technoglogy\n",
      "\n",
      "![TACC Logo](/files/figures/TACC_logo.png)\n",
      "![Continuum Logo](/files/figures/continuum.png)\n",
      "![KAUST Logo](/files/figures/kaust.png)\n",
      "\n",
      "The contents of this tutorial are distributed under the CC-BY Creative Commons License \n",
      "\n",
      "[![Creative Commons License](http://i.creativecommons.org/l/by/3.0/88x31.png)](http://creativecommons.org/licenses/by/3.0/deed.en_US) \n",
      "  "
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "## About This Tutorial\n",
      "\n",
      "### PyHPC Tutorial on GitHub\n",
      "\n",
      "These presentation materials are part of a continuously updated tutorial on Python for High Performance Computing.  To get the presentation materials as they were delivered today, please use the SC2012 tag on github.  Future versions of this presentation will be found at:\n",
      "\n",
      "https://github.com/pyhpc/pyhpc-tutorial\n",
      "\n",
      "Please set all permanent bookmarks to this URL.\n",
      "\n",
      "### Direct download of this SC2012 presentation\n",
      "You can download a zip or tar ball from the [github SC2012 tag](https://github.com/pyHPC/pyhpc-tutorial/tags):\n",
      "\n",
      "```\n",
      "wget --no-check-certificate https://github.com/pyhpc/pyhpc-tutorial/zipball/SC2012\n",
      "wget --no-check-certificate https://github.com/pyhpc/pyhpc-tutorial/tarball/SC2012\n",
      "```\n",
      "\n",
      "### \n",
      "2) Checkout from git\n",
      "\n",
      "```\n",
      "git clone https://github.com/pyhpc/pyhpc-tutorial.git\n",
      "```\n",
      "\n",
      "### Viewing the read-only version of this presentation on nbviewer: \n",
      "\n",
      "* http://nbviewer.ipython.org/urls/raw.github.com/pyhpc/pyhpc-tutorial/master/03_Scaling_Python.ipynb"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "## Interacting with the Tutorial Slides\n",
      "\n",
      "This tutorial is an interactive worksheet designed to encourage you to try out the lessons during the demonstration.  If you are looking at the PDF version of these slides, we encourage you to download the updated version (see previous slide) and try the interactive version.\n",
      "\n",
      "To run the interactive version of this notebook, you will need a Python 2.7 environment including: \n",
      "\n",
      "* IPython version >= 13.0\n",
      "* numpy version >= 1.6\n",
      "* scipy >= 0.10\n",
      "* matplotlib >= 1.0.0\n",
      "* mpi4py >= 1.2\n",
      "\n",
      "Move to the directory containing the tarball and execute:\n",
      "\n",
      "    ipcluster start --engines=MPI --n 4\n",
      "    ipython notebook --pylab=inline\n",
      "\n",
      "If you are installing the packages yourself, Continuum Analytics (disclaimer, Travis Oliphant is CEO of this company) provides both free community as well as professional versions of the Anaconda installer, which provides all the packages you will need for this portion of the tutorial. The installer is available from the [Anaconda download page at Continuum Analytics](https://store.continuum.io/cshop/anaconda).\n",
      "\n",
      "You are also welcome to use Enthought's Python Distribution (free for Academic users), available from the [EPD download page at Enthought](http://www.enthought.com/products/epd.php).\n",
      "\n",
      "Unfortunately, you will need to upgrade your EPD's IPython to 0.13 if you go this route, and this may be non-trivial, please see the [Enthought discussion thread here](https://mail.enthought.com/pipermail/epd-users/2012-August/000800.html)."
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "--- \n",
      "\n",
      "## Using the Texas Advanced Computing Center IPython Notebooks\n",
      "\n",
      "You will need a modern web browser to interact with the TACC nodes, there are no other requirements.\n",
      "\n",
      "### TACC IPython Notebooks\n",
      "* http://pyhpc01.tacc.utexas.edu/\n",
      "* http://pyhpc02.tacc.utexas.edu/\n",
      "* http://pyhpc03.tacc.utexas.edu/\n",
      "* http://pyhpc04.tacc.utexas.edu/\n",
      "* http://pyhpc05.tacc.utexas.edu/\n",
      "* http://pyhpc06.tacc.utexas.edu/\n",
      "* http://pyhpc07.tacc.utexas.edu/\n",
      "* http://pyhpc08.tacc.utexas.edu/\n",
      "* http://pyhpc09.tacc.utexas.edu/\n",
      "* http://pyhpc10.tacc.utexas.edu/\n",
      "\n",
      "The TACC nodes will not be available after this workshop."
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "## Presentation mode\n",
      "The slideshow mode is only supported by an IPython development branch version. If you would like to write your own slideshows (the notebooks will work regardless of slideshow capabilities), you will need to install Matthias Bussonnier's slideshow_metadata branch.  Here are some sample command-line instructions.\n",
      "\n",
      "    #!/bin/bash\n",
      "    git clone carreau git://github.com/Carreau/ipython.git # Matthias' IPython fork\n",
      "    git checkout origin/slideshow_metadata                 # Select the right branch\n",
      "    python setup.py develop                                # Install the development version\n",
      "    ipython notebook                                       # Check out the slideshows!"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "## Slideshow Setup Code"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# The below command starts pylab inline mode, which simplifies syntax for many common \n",
      "# scientific computing commands.  All commands starting with % are IPython extensions.\n",
      "%pylab inline\n",
      "\n",
      "# The following two lines of code load the slideshow extension and start it,\n",
      "# you only need to run these if you are writing or running your own slideshow\n",
      "\n",
      "%load_ext slidemode\n",
      "%slideshow_mode"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n",
        "Welcome to pylab, a matplotlib-based Python environment [backend: module://IPython.zmq.pylab.backend_inline].\n",
        "For more information, type 'help(pylab)'.\n"
       ]
      },
      {
       "javascript": [
        "\n",
        "var bind_remote = function(selector){\n",
        "    var cc = $('<input/>')\n",
        "        .attr('id','azerty')\n",
        "        .fadeTo(0.2,0)\n",
        "        .keydown(\n",
        "        function(event){\n",
        "                if(event.which == IPython.utils.keycodes.LEFT_ARROW){\n",
        "                    console.log('go left');\n",
        "                    IPython.slideshow.prev();\n",
        "                }\n",
        "                else if(event.which == IPython.utils.keycodes.RIGHT_ARROW){\n",
        "                    console.log('go right')\n",
        "                    IPython.slideshow.next();   \n",
        "                }\n",
        "            event.preventDefault();\n",
        "            return false;\n",
        "            })\n",
        "            .focusin(function(){$('#qwerty').button('option','label','Slide Control Enabled')})\n",
        "            .focusout(function(){$('#qwerty').button('option','label','Enable Slide Control')})\n",
        "        \n",
        "        var dd = $('<div/>')\n",
        "            .attr('id','qwerty')\n",
        "            .button({label:'Enable Slide Control'})\n",
        "            .attr('style','float:right')\n",
        "            .click(\n",
        "                function(){\n",
        "                    $('#azerty').focus();\n",
        "                    console.log('...');\n",
        "                })\n",
        "            .keydown(\n",
        "                function(event){\n",
        "                    console.log('event append',event);\n",
        "                event.preventDefault();\n",
        "                return false;\n",
        "                })\n",
        "    var hin = function(){$(selector).fadeTo('slow',1);}\n",
        "    var hout= function(){$(selector).fadeTo('slow',0.3);}        \n",
        "    $(selector)\n",
        "        .append(cc)\n",
        "        .append(dd)\n",
        "        .fadeTo('slow',0.3)\n",
        "    .hover(hin,hout)\n",
        "     \n",
        "}\n",
        "    \n",
        "var show_line = function(cell,line_number){\n",
        "    cell.code_mirror.showLine(cell.code_mirror.getLineHandle(line_number-1))    }\n",
        "\n",
        "    IPython = (function(IPython) {\n",
        "    \n",
        "    var Presentation = function(){\n",
        "          this.ccell = 0;\n",
        "      }\n",
        "\n",
        "      var is_marked_cell = function(cell){\n",
        "          return (cell.code_mirror.getValue().match(/(\\n)?(<!--|#)?====(-->)?(\\n)?/) != null)\n",
        "      }\n",
        "      Presentation.prototype.create_toolbar = function(){\n",
        "            var pt = $('<div/>').attr('id','toolbar_present').addClass('toolbar');\n",
        "            var that = this;\n",
        "            this.avc = $('<div/>').button({label:that.eta()});\n",
        "            pt.append(this.avc);\n",
        "\n",
        "            $('#toolbar').after(pt);\n",
        "            var ptoolbar = new IPython.ToolBar('#toolbar_present');\n",
        "            IPython.ptoolbar = ptoolbar;\n",
        "            ptoolbar.addButtonsGroup([{label:'Stop', icon:'ui-icon-stop', callback:function(){that.stop()}}])\n",
        "            ptoolbar.addButtonsGroup([\n",
        "                     {label:'Restart from 0', icon:'ui-icon-arrowthickstop-1-w', callback:function(){that.stop(); that.restart(); that.start()}}\n",
        "                     ])\n",
        "            ptoolbar.addButtonsGroup([\n",
        "                     {label:'Prev Slide', icon:'ui-icon-seek-prev', callback:function(){that.prev_group()}},\n",
        "                     {label:'Next Slide', icon:'ui-icon-seek-next', callback:function(){that.next_group()}},\n",
        "                                    ])\n",
        "            ptoolbar.addButtonsGroup([\n",
        "                     {label:'Step Next', icon:'ui-icon-play', callback:function(){that.next()}}\n",
        "                     ])\n",
        "            bind_remote('#toolbar_present');\n",
        "\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.remove_toolbar = function(){\n",
        "          $('#toolbar_present').remove();\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.ngroups = function(){\n",
        "          var cells = IPython.notebook.get_cells();\n",
        "          var cnt =0\n",
        "          for( var i=0; i< cells.length;i++)\n",
        "              if(is_marked_cell(cells[i])) cnt++;\n",
        "          return cnt\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.cgroups = function(){\n",
        "          var cells = IPython.notebook.get_cells();\n",
        "          var cnt =0\n",
        "          for( var i=0; i<= this.ccell ;i++)\n",
        "              if(is_marked_cell(cells[i])) cnt++;\n",
        "          return cnt\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.eta = function(){\n",
        "          return this.cgroups()+'/'+this.ngroups()\n",
        "          }\n",
        "\n",
        "      Presentation.prototype.next_marked_cell_n = function()\n",
        "      {\n",
        "          for(var i=this.ccell+1; i< $('.cell').length; i++)\n",
        "          {\n",
        "              if(is_marked_cell(IPython.notebook.get_cell(i)))\n",
        "                { return i; }\n",
        "          }\n",
        "          return null;\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.prev_marked_cell_n = function()\n",
        "      {\n",
        "          for(var i=this.ccell-1; i> 0; i--)\n",
        "          {\n",
        "              if(is_marked_cell(IPython.notebook.get_cell(i))){\n",
        "                 return i\n",
        "              }\n",
        "          }\n",
        "          return 0;\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.start = function(){\n",
        "          this.restart();\n",
        "          this.resume();\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.restart = function(){\n",
        "          this.ccell = 0;\n",
        "          $(this.avc).remove();\n",
        "          delete this.avc;\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.resume = function(){\n",
        "          this.create_toolbar();\n",
        "          $('#menubar, #pager_splitter, #pager, #header,#toolbar').addClass('pmode');\n",
        "          $('.cell').fadeOut();\n",
        "          if(this.current_is_marked()){\n",
        "              $('.cell:nth('+this.ccell+')').fadeIn();\n",
        "          } else {\n",
        "              for( var i=this.prev_marked_cell_n() ; i<= this.ccell; i++){\n",
        "                  $('.cell:nth('+i+')').fadeIn();\n",
        "              }\n",
        "          }\n",
        "          var that=this;\n",
        "          if(this.avc != undefined)\n",
        "            $(this.avc).button('option','label',that.eta())\n",
        "          return this;\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.stop = function(){\n",
        "          $('.cell').fadeIn();\n",
        "          $('.pmode').removeClass('pmode');\n",
        "          $('div#notebook').removeClass('pove');\n",
        "          this.remove_toolbar();\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.next = function(){\n",
        "          this.ccell = this.ccell+1;\n",
        "          var that = this;\n",
        "          if(this.ccell >= $('.cell').length ){\n",
        "              this.restart();\n",
        "              this.stop();\n",
        "              return;\n",
        "          }\n",
        "          var nnext = this.ccell;\n",
        "          var ncell = IPython.notebook.get_cell(nnext)\n",
        "\n",
        "          if(is_marked_cell(ncell)){\n",
        "              $('.cell').fadeOut(500);\n",
        "              setTimeout(function(){$('.cell:nth('+nnext+')').fadeIn(500)},600);\n",
        "          } else {\n",
        "              setTimeout(function(){$('.cell:nth('+nnext+')').fadeIn(500)},0);\n",
        "          }\n",
        "          $(this.avc).button('option','label',that.eta())\n",
        "          return this;\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.next_group = function(){\n",
        "          this.ccell = this.next_marked_cell_n();\n",
        "          var that = this;\n",
        "          $('.cell').fadeOut(500);\n",
        "          setTimeout(function(){\n",
        "              $('.cell:nth('+that.ccell+')').fadeIn(500)\n",
        "                  },600);\n",
        "          $(this.avc).button('option','label',that.eta())\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.prev_group = function(){\n",
        "          this.ccell = this.prev_marked_cell_n();\n",
        "          var that = this\n",
        "          $('.cell').fadeOut(500);\n",
        "          setTimeout(function(){$('.cell:nth('+that.ccell+')').fadeIn(500)},600);\n",
        "          $(this.avc).button('option','label',that.eta())\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.is_n_marked = function(n){\n",
        "          return is_marked_cell(IPython.notebook.get_cell(n))\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.current_is_marked = function(n){\n",
        "          return is_marked_cell(IPython.notebook.get_cell(this.ccell));\n",
        "      }\n",
        "\n",
        "      Presentation.prototype.prev = function(){\n",
        "          if(is_marked_cell(IPython.notebook.get_cell(this.ccell))){\n",
        "              var pmcell = this.prev_marked_cell_n();\n",
        "              $('.cell').fadeOut(500);\n",
        "              for( var i=pmcell; i< this.ccell ; i++ ){\n",
        "                  (function(val){\n",
        "                      return function(){\n",
        "                                  setTimeout( function(){\n",
        "                                      $('.cell:nth('+val+')').fadeIn(500)\n",
        "                                      },600);\n",
        "                      }\n",
        "                   })(i)();\n",
        "              }\n",
        "          } else {\n",
        "              $('.cell:nth('+this.ccell+')').fadeOut(500);\n",
        "          }\n",
        "          this.ccell = this.ccell -1;\n",
        "          return this;\n",
        "      }\n",
        "\n",
        "      IPython.Presentation = Presentation;\n",
        "      return IPython;\n",
        "\n",
        "    })(IPython);\n",
        "\n",
        "    $('body').append($('<style/>').text('.pmode{ display: none !important }'));\n",
        "    IPython.slideshow = new IPython.Presentation();\n",
        "\n",
        "    var sid = 'start_pmode'\n",
        "    if(($('#'+sid)).length == 0) {\n",
        "          IPython.toolbar.addButtonsGroup([\n",
        "                  {'label'  :'Start Slideshow',\n",
        "                    'icon'  :'ui-icon-image',\n",
        "                    'callback':function(){IPython.slideshow.resume()},'id':sid},\n",
        "              ])\n",
        "         }\n"
       ],
       "output_type": "display_data",
       "text": [
        "<IPython.core.display.Javascript at 0x36d3e50>"
       ]
      }
     ],
     "prompt_number": 1
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "Acknowledgements\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "* Much of this tutorial adapts slide material from [William Gropp](http://www.cs.uiuc.edu/~wgropp/), University of Illinois and [Lisandro Dalcin](http://plus.google.com/107621373684536061961/about), CONICET\n",
      "* [mpi4py](http://code.google.com/p/mpi4py/) is a [Cythonized](http://www.cython.org/) wrapper around [MPI](http://www.mpi-forum.org/) originally developed by [Lisandro Dalcin](http://plus.google.com/107621373684536061961/about), CONICET\n"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "Connecting to MPI via IPython\n",
      "-------------------------------------\n",
      "\n",
      "To run the examples in parallel, you must run the command:\n",
      "\n",
      "`ipcluster start --engines=MPI --n 4` \n",
      "\n",
      "Or use notebook Appendix_03_Launch_MPI_Engines \n",
      "(if you are using the VMs or have this configured for your environment).\n",
      "\n",
      "Then execute the next cell:"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from IPython.parallel import Client\n",
      "c = Client()\n",
      "view = c[:]\n",
      "\n",
      "%load_ext parallelmagic\n",
      "view.activate()\n",
      "view.block = True"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "outputs": [],
     "prompt_number": 2
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "# A Quick Review of Concepts of Scalability"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "# The Multiple Forms of Parallelism\n",
      "\n",
      "* **instruction** - multiple program instructions are simultaneously dispatched in a pipeline or to multiple execution units (superscalar)\n",
      "* **data** - the same program instructions are carried out simultaneously on multiple data items (SIMD)\n",
      "* **task** - different program instructions on different data (MIMD)\n",
      "* **collective** - single program, multiple data, not necessarily synchronized at individual operation level (SPMD)"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "source": [
      "> *This part of the tutorial focuses on data and collective parallelism*"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "# Parallel Architectures\n",
      "<!--====-->\n",
      "\n",
      "* All modern supercomputing architectures are composed of distributed memory compute nodes connected over toroidal networks\n",
      "  \n",
      "  - close mapping to physically n-dimensional problems\n",
      "  - efficient algorithms exist for local point-to-point and collective communications across toroids\n",
      "\n",
      "* We expect to see higher dimensional interconnects and power-efficient networks that consume less power when not sending traffic\n"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "# Parallel Programming Paradigms\n",
      "<!--====-->\n",
      "\n",
      "* a parallel programming paradigm is a specific approach to exploiting parallelism in hardware\n",
      "* many programming paradigms are very tightly coupled to the hardware beneath!\n",
      "  \n",
      " - CUDA assumes large register files, Same Instruction Multiple Thread parallelism, and a mostly flat, structured memory model, matching the underlying GPU hardware\n",
      "\n",
      " - OpenMP exposes loop level parallelism with a fork/join model, assumes the presence of shared memory and atomics\n",
      "\n",
      " - OpenCL tries to generalize CUDA, but still assumes a 'coprocessor' approach, where kernels are shipped from a master core to worker cores\n"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "The Message Passing Model\n",
      "-------------------------------------\n",
      "\n",
      "* a process is (traditionally) a program counter for instructions and an address space for data\n",
      "* processes may have multiple threads (program counters and associated stacks) sharing a single address space\n",
      "* message passing is for communication among processes, which have separate address spaces\n",
      "* interprocess communication consists of\n",
      "  \n",
      "  - synchronization\n",
      "  - movement of data from one process's address space to another's\n"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "Why MPI?\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "* **communicators** encapsulate communication spaces for library safety\n",
      "* **datatypes** reduce copying costs and permit heterogeneity\n",
      "* multiple **communication modes** allow more control of memory buffer management\n",
      "* extensive **collective operations** for scalable global communication\n",
      "* **process topologies** permit efficient process placement, user views of process layout\n",
      "* **profiling interface** encourages portable tools\n",
      "\n",
      "It Scales!"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "Python + MPI Does It Scale?\n",
      "-------------------------------------\n",
      "\n",
      "![concurrency](/files/figures/does_it_scale.png)\n"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "The Scalability of Python + MPI\n",
      "-------------------------------------\n",
      "\n",
      "![concurrency](/files/figures/euler_weak_scaling.png)\n"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "MPI - Quick Review\n",
      "-------------------------------------\n",
      "\n",
      "* processes can be collected into **groups**\n",
      "* each message is sent in a **context**, and must be received in the same context\n",
      "* a  **communicator** encapsulates a context for a specific group\n",
      "* a given program may have many communicators with any level of overlap\n",
      "* two initial communicators\n",
      " \n",
      "  - ``MPI_COMM_WORLD`` (all processes)\n",
      "  - ``MPI_COMM_SELF`` (current process)"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "First Example: Hello World\n",
      "-------------------------------------\n",
      "\n",
      "For interactive convenience, we load the parallel magic extensions and make this view\n",
      "the active one for the automatic parallelism magics.\n",
      "\n",
      "This is not necessary and in production codes likely won't be used, as the engines will \n",
      "load their own MPI codes separately.  But it makes it easy to illustrate everything from\n",
      "within a single notebook here."
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from IPython.parallel import Client\n",
      "c = Client()\n",
      "view = c[:]\n",
      "\n",
      "%load_ext parallelmagic\n",
      "view.activate()\n",
      "view.block = True"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "outputs": [],
     "prompt_number": 3
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "source": [
      "Use the autopx magic to make the rest of this cell execute on the engines instead\n",
      "of locally"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%autopx"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "%autopx enabled\n"
       ]
      }
     ],
     "prompt_number": 4
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "source": [
      "With autopx enabled, the next cell will actually execute *entirely on each engine*:"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "\n",
      "size = MPI.COMM_WORLD.Get_size()\n",
      "rank = MPI.COMM_WORLD.Get_rank()\n",
      "name = MPI.Get_processor_name()\n",
      "\n",
      "print(\"Hello World! I am process %d of %d on %s.\\n\" % (rank, size, name))"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:0] \n",
        "Hello World! I am process 2 of 4 on pyhpc01.\n",
        "\n",
        "[stdout:1] \n",
        "Hello World! I am process 3 of 4 on pyhpc01.\n",
        "\n",
        "[stdout:2] \n",
        "Hello World! I am process 0 of 4 on pyhpc01.\n",
        "\n",
        "[stdout:3] \n",
        "Hello World! I am process 1 of 4 on pyhpc01.\n",
        "\n"
       ]
      }
     ],
     "prompt_number": 5
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "Communicators\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "* processes can be collected into **groups**\n",
      "* each message is sent in a **context**, and must be received in the same context\n",
      "* a  **communicator** encapsulates a context for a specific group\n",
      "* a given program may have many communicators with any level of overlap\n",
      "* two initial communicators\n",
      " \n",
      "  - ``MPI_COMM_WORLD`` (all processes)\n",
      "  - ``MPI_COMM_SELF`` (current process)"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "Datatypes\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "* the data in a message to send or receive is described by address, count and datatype\n",
      "* a datatype is recursively defined as:\n",
      "\n",
      "  + predefined, corresponding to a data type from the language (e.g., ``MPI_INT``, ``MPI_DOUBLE``)\n",
      "  + a contiguous, strided block, or indexed array of blocks of MPI datatypes\n",
      "  + an arbitrary structure of datatypes\n",
      "\n",
      "* there are MPI functions to construct custom datatypes"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "source": [
      "---\n",
      "\n",
      "Tags\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "* messages are sent with an accompanying user-defined integer tag to assist the receiving process in identifying the message\n",
      "* messages can be screened at the receiving end by specifying the expected tag, or not screened by using ``MPI_ANY_TAG`` "
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "source": [
      "---\n",
      "\n",
      "Functionality\n",
      "-------------------------------------\n",
      "\n",
      "* There are hundreds of functions in the MPI standard, not all of them are necessarily available in MPI4Py, most commonly used are\n",
      "* No need to call MPI_Init() or MPI_Finalize()\n",
      " - MPI_Init() is called when you import the module\n",
      " - MPI_Finalize() is called before the Python process ends\n",
      "* To launch:\n",
      "    mpirun --np < number of process > -machinefile < hostlist > python < my MPI4Py python script >\n",
      "* IPython automatically handles calling mpirun for you with the `ipcluster` command"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true,
       "slide_step": false
      }
     },
     "source": [
      "---\n",
      "\n",
      "MPI Basic (Blocking) Send\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "```\n",
      "   int MPI_Send(void* buf, int count, MPI_Datatype type, \n",
      "   int dest, int tag, MPI_Comm comm)\n",
      "```\n",
      "\n",
      "Python (mpi4py)\n",
      "\n",
      "```\n",
      "  Comm.Send(self, buf, int dest=0, int tag=0)\n",
      "  Comm.send(self, obj=None, int dest=0, int tag=0)\n",
      "```"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "MPI Basic (Blocking) Recv\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "```\n",
      "   int MPI_Recv(void* buf, int count, MPI_Datatype type, \n",
      "   int source, int tag, MPI_Comm comm, MPI_Status status)\n",
      "```\n",
      "\n",
      "Python (mpi4py)\n",
      "\n",
      "```\n",
      "  comm.Recv(self, buf, int source=0, int tag=0, \n",
      "  Status status=None)\n",
      "  comm.recv(self, obj=None, int source=0, \n",
      "  int tag=0, Status status=None)\n",
      "```"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Send/Receive Example (lowercase convenience methods)\n",
      "-------------------------------------\n",
      "<!--====-->\n"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "\n",
      "comm = MPI.COMM_WORLD\n",
      "rank = comm.Get_rank()\n",
      "\n",
      "if rank == 0:\n",
      "   data = {'a': 7, 'b': 3.14}\n",
      "   comm.send(data, dest=1, tag=11)\n",
      "elif rank == 1:\n",
      "   data = comm.recv(source=0, tag=11)\n",
      "   print data"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:3] {'a': 7, 'b': 3.14}\n"
       ]
      }
     ],
     "prompt_number": 6
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Send/Receive Example (MPI API on numpy)\n",
      "-------------------"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "import numpy\n",
      "\n",
      "comm = MPI.COMM_WORLD\n",
      "rank = comm.Get_rank()\n",
      "\n",
      "# pass explicit MPI datatypes\n",
      "if rank == 0:\n",
      "   data = numpy.arange(1000, dtype='i')\n",
      "   comm.Send([data, MPI.INT], dest=1, tag=77)\n",
      "elif rank == 1:\n",
      "   data = numpy.empty(1000, dtype='i')\n",
      "   comm.Recv([data, MPI.INT], source=0, tag=77)\n",
      "\n",
      "# or take advantage of automatic MPI datatype discovery\n",
      "if rank == 0:\n",
      "   data = numpy.arange(100, dtype=numpy.float64)\n",
      "   comm.Send(data, dest=1, tag=13)\n",
      "elif rank == 1:\n",
      "   data = numpy.empty(100, dtype=numpy.float64)\n",
      "   comm.Recv(data, source=0, tag=13)\n",
      "   print data"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:3] \n",
        "[  0.   1.   2.   3.   4.   5.   6.   7.   8.   9.  10.  11.  12.  13.  14.\n",
        "  15.  16.  17.  18.  19.  20.  21.  22.  23.  24.  25.  26.  27.  28.  29.\n",
        "  30.  31.  32.  33.  34.  35.  36.  37.  38.  39.  40.  41.  42.  43.  44.\n",
        "  45.  46.  47.  48.  49.  50.  51.  52.  53.  54.  55.  56.  57.  58.  59.\n",
        "  60.  61.  62.  63.  64.  65.  66.  67.  68.  69.  70.  71.  72.  73.  74.\n",
        "  75.  76.  77.  78.  79.  80.  81.  82.  83.  84.  85.  86.  87.  88.  89.\n",
        "  90.  91.  92.  93.  94.  95.  96.  97.  98.  99.]\n"
       ]
      }
     ],
     "prompt_number": 7
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Synchronization\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "```\n",
      "   int MPI_Barrier(MPI_Comm comm)\n",
      "```\n",
      "\n",
      "Python (mpi4py)\n",
      "\n",
      "```\n",
      "   comm.Barrier(self)\n",
      "   comm.barrier(self)\n",
      "```"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "\n",
      "comm = MPI.COMM_WORLD\n",
      "rank = comm.Get_rank()\n",
      "\n",
      "for r_id in range(comm.Get_size()):\n",
      "    if rank == r_id:\n",
      "        print \"Hello from proc:\", rank\n",
      "    comm.Barrier()\n"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:0] Hello from proc: 2\n",
        "[stdout:1] Hello from proc: 3\n",
        "[stdout:2] Hello from proc: 0\n",
        "[stdout:3] Hello from proc: 1\n"
       ]
      }
     ],
     "prompt_number": 8
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "\n",
      "Timing and Profiling\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "the elapsed (wall-clock) time between two points in an MPI program can be computed using MPI_Wtime:"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "t1 = MPI.Wtime()\n",
      "t2 = MPI.Wtime()\n",
      "print(\"time elapsed is: %e\\n\" % (t2-t1))"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:0] \n",
        "time elapsed is: 6.008148e-05\n",
        "\n",
        "[stdout:1] \n",
        "time elapsed is: 8.392334e-05\n",
        "\n",
        "[stdout:2] \n",
        "time elapsed is: 8.511543e-05\n",
        "\n",
        "[stdout:3] \n",
        "time elapsed is: 6.198883e-05\n",
        "\n"
       ]
      }
     ],
     "prompt_number": 9
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Send/Receive Example (lowercase convenience methods)\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "![broadcast_scatter_gather](/files/figures/broadcast_scatter_gather.png)\n"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Broadcast Example\n",
      "-------------------------------------\n",
      "<!--====-->\n",
      "\n",
      "```\n",
      "   int MPI_Bcast(void *buf, int count, MPI_Datatype type, \n",
      "   int root, MPI_Comm comm)\n",
      "```\n",
      "\n",
      "Python (mpi4py)\n",
      "\n",
      "``` \n",
      "  comm.Bcast(self, buf, int root=0)\n",
      "  comm.bcast(self, obj=None, int root=0)\n",
      "```"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "\n",
      "comm = MPI.COMM_WORLD\n",
      "rank = comm.Get_rank()\n",
      "\n",
      "if rank == 0:\n",
      "   data = {'key1' : [7, 2.72, 2+3j],\n",
      "           'key2' : ( 'abc', 'xyz')}\n",
      "else:\n",
      "   data = None\n",
      "data = comm.bcast(data, root=0)\n",
      "print \"bcast finished and data \\\n",
      " on rank %d is: \"%comm.rank, data\n"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:0] bcast finished and data  on rank 2 is:  {'key2': ('abc', 'xyz'), 'key1': [7, 2.72, (2+3j)]}\n",
        "[stdout:1] bcast finished and data  on rank 3 is:  {'key2': ('abc', 'xyz'), 'key1': [7, 2.72, (2+3j)]}\n",
        "[stdout:2] bcast finished and data  on rank 0 is:  {'key2': ('abc', 'xyz'), 'key1': [7, 2.72, (2+3j)]}\n",
        "[stdout:3] bcast finished and data  on rank 1 is:  {'key2': ('abc', 'xyz'), 'key1': [7, 2.72, (2+3j)]}\n"
       ]
      }
     ],
     "prompt_number": 10
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "\n",
      "---\n",
      "Scatter Example:\n",
      "-------------------------------------\n",
      "<!--====-->\n"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "\n",
      "comm = MPI.COMM_WORLD\n",
      "size = comm.Get_size()\n",
      "rank = comm.Get_rank()\n",
      "\n",
      "if rank == 0:\n",
      "   data = [(i+1)**2 for i in range(size)]\n",
      "else:\n",
      "   data = None\n",
      "data = comm.scatter(data, root=0)\n",
      "assert data == (rank+1)**2\n",
      "print \"data on rank %d is: \"%comm.rank, data"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:0] data on rank 2 is:  9\n",
        "[stdout:1] data on rank 3 is:  16\n",
        "[stdout:2] data on rank 0 is:  1\n",
        "[stdout:3] data on rank 1 is:  4\n"
       ]
      }
     ],
     "prompt_number": 11
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "\n",
      "---\n",
      "Gather (and Barrier) Example:\n",
      "-------------------------------------\n",
      "<!--====-->\n"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "\n",
      "comm = MPI.COMM_WORLD\n",
      "size = comm.Get_size()\n",
      "rank = comm.Get_rank()\n",
      "\n",
      "data = (rank+1)**2\n",
      "print \"before gather, data on \\\n",
      "  rank %d is: \"%rank, data\n",
      "\n",
      "comm.Barrier()\n",
      "data = comm.gather(data, root=0)\n",
      "if rank == 0:\n",
      "   for i in range(size):\n",
      "       assert data[i] == (i+1)**2\n",
      "else:\n",
      "   assert data is None\n",
      "print \"data on rank: %d is: \"%rank, data"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:0] \n",
        "before gather, data on   rank 2 is:  9\n",
        "data on rank: 2 is:  None\n",
        "[stdout:1] \n",
        "before gather, data on   rank 3 is:  16\n",
        "data on rank: 3 is:  None\n",
        "[stdout:2] \n",
        "before gather, data on   rank 0 is:  1\n",
        "data on rank: 0 is:  [1, 4, 9, 16]\n",
        "[stdout:3] \n",
        "before gather, data on   rank 1 is:  4\n",
        "data on rank: 1 is:  None\n"
       ]
      }
     ],
     "prompt_number": 12
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Collective Example\n",
      "-------------------\n",
      "<!--====-->\n",
      "\n",
      "![reduce_scan](/files/figures/reduce_scan.png)\n"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "\n",
      "---\n",
      "Reduce Example:\n",
      "-------------------------------------\n",
      "<!--====-->\n"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "comm = MPI.COMM_WORLD\n",
      "\n",
      "sendmsg = comm.rank\n",
      "\n",
      "recvmsg1 = comm.reduce(sendmsg, op=MPI.SUM, root=0)\n",
      "\n",
      "recvmsg2 = comm.allreduce(sendmsg)\n",
      "print recvmsg2"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:0] 6\n",
        "[stdout:1] 6\n",
        "[stdout:2] 6\n",
        "[stdout:3] 6\n"
       ]
      }
     ],
     "prompt_number": 13
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Compute Pi Example\n",
      "-------------------\n",
      "<!--====-->\n"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from mpi4py import MPI\n",
      "import math\n",
      "\n",
      "def compute_pi(n, start=0, step=1):\n",
      "    h = 1.0 / n\n",
      "    s = 0.0\n",
      "    for i in range(start, n, step):\n",
      "        x = h * (i + 0.5)\n",
      "        s += 4.0 / (1.0 + x**2)\n",
      "    return s * h\n",
      "\n",
      "comm = MPI.COMM_WORLD\n",
      "nprocs = comm.Get_size()\n",
      "myrank = comm.Get_rank()\n",
      "if myrank == 0:\n",
      "    n = 10\n",
      "else:\n",
      "    n = None\n",
      "\n",
      "n = comm.bcast(n, root=0)\n",
      "\n",
      "mypi = compute_pi(n, myrank, nprocs)\n",
      "\n",
      "pi = comm.reduce(mypi, op=MPI.SUM, root=0)\n",
      "\n",
      "if myrank == 0:\n",
      "    error = abs(pi - math.pi)\n",
      "    print (\"pi is approximately %.16f, error is %.16f\" % (pi, error))"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[stdout:2] pi is approximately 3.1424259850010983, error is 0.0008333314113051\n"
       ]
      }
     ],
     "prompt_number": 14
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Mandelbrot Set Example\n",
      "-------------------\n",
      "<!--====-->\n"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "def mandelbrot (x, y, maxit):\n",
      "    c = x + y*1j\n",
      "    z = 0 + 0j\n",
      "    it = 0\n",
      "    while abs(z) < 2 and it < maxit:\n",
      "        z = z**2 + c\n",
      "        it += 1\n",
      "    return it\n",
      "\n",
      "x1, x2 = -2.0, 1.0\n",
      "y1, y2 = -1.0, 1.0\n",
      "w, h = 150, 100\n",
      "maxit = 127\n",
      "\n",
      "from mpi4py import MPI\n",
      "import numpy\n",
      "  \n",
      "comm = MPI.COMM_WORLD\n",
      "size = comm.Get_size()\n",
      "rank = comm.Get_rank()\n",
      "\n",
      "# number of rows to compute here\n",
      "N = h // size + (h % size > rank)\n",
      "\n",
      "# first row to compute here\n",
      "start = comm.scan(N)-N\n",
      "\n",
      "# array to store local result\n",
      "Cl = numpy.zeros([N, w], dtype='i')\n",
      "\n",
      "# compute owned rows\n",
      "\n",
      "dx = (x2 - x1) / w\n",
      "dy = (y2 - y1) / h\n",
      "\n",
      "for i in range(N):\n",
      "    y = y1 + (i + start) * dy\n",
      "    for j in range(w):\n",
      "        x = x1 + j * dx\n",
      "        Cl[i, j] = mandelbrot(x, y, maxit)\n",
      "\n",
      "# gather results at root (process 0)\n",
      "counts = comm.gather(N, root=0)\n",
      "C = None\n",
      "if rank == 0:\n",
      "    C = numpy.zeros([h, w], dtype='i')\n",
      "\n",
      "rowtype = MPI.INT.Create_contiguous(w)\n",
      "rowtype.Commit()\n",
      "\n",
      "comm.Gatherv(sendbuf=[Cl, MPI.INT], recvbuf=[C, (counts, None), rowtype],root=0)\n",
      "\n",
      "rowtype.Free()"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false,
       "slide_step": true
      }
     },
     "outputs": [],
     "prompt_number": 15
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "source": [
      "We now need to \"pull\" the C array for plotting so we disable autopx. Make sure to re-enable it later on"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%autopx"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "%autopx disabled\n"
       ]
      }
     ],
     "prompt_number": 16
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "# CC is an array of C from all ranks, so we use CC[0]\n",
      "CC = view['C']\n",
      "ranks = view['rank']\n",
      "\n",
      "# Do the plotting\n",
      "from matplotlib import pyplot\n",
      "# Some magic to get to MPI4PY rank 0, not necessarily engine id 0\n",
      "pyplot.imshow(CC[ranks.index(0)], aspect='equal')\n",
      "pyplot.spectral()\n",
      "pyplot.show()"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "display_data",
       "png": "iVBORw0KGgoAAAANSUhEUgAAAWwAAAD8CAYAAABTjp5OAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n"
      }
     ],
     "prompt_number": 17
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "source": [
      "Toggle autopx back"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%autopx"
     ],
     "language": "python",
     "metadata": {
      "slideshow": {
       "slide_start": false
      }
     },
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "%autopx enabled\n"
       ]
      }
     ],
     "prompt_number": 18
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "Advanced Capabilities\n",
      "----------------------------\n",
      "\n",
      "Other features supported by mpi4py\n",
      "\n",
      "* dynamic process spawning: Spawn(), Connect() and Disconnect()\n",
      "* one sided communication: Put(), Get(), Accumulate()\n",
      "* MPI-IO: Open(), Close(), Get_view() and Set_view()"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {
      "slideshow": {
       "slide_start": true
      }
     },
     "source": [
      "---\n",
      "More Comprehensive mpi4py Tutorials\n",
      "----------------------------\n",
      "<!--====-->\n",
      "\n",
      "* basics - http://mpi4py.scipy.org/docs/usrman/tutorial.html\n",
      "* advanced - http://www.bu.edu/pasi/files/2011/01/Lisandro-Dalcin-mpi4py.pdf\n",
      "\n",
      "Interesting Scalable Applications and Tools\n",
      "---------------------------------------------\n",
      "* PyTrilinos - http://trilinos.sandia.gov/packages/pytrilinos/\n",
      "* petsc4py - http://code.google.com/p/petsc4py/\n",
      "* PyClaw - http://numerics.kaust.edu.sa/pyclaw/\n",
      "* GPAW - https://wiki.fysik.dtu.dk/gpaw/\n",
      "\n",
      "Addressing the Import Problem\n",
      "-----------------------------\n",
      "* Static compilation from Pat Marion - https://github.com/patmarion/NumpyBuiltinExample\n",
      "* Path caching from Asher Langton - https://github.com/langton/MPI_Import\n",
      "* ld.so interception plus zipimport-enabled path caching from Aron Ahmadia, Jed Brown, and William Scullin - https://github.com/ahmadia/collfs\n",
      "\n",
      "--> See Appendix 01 References for more resources\n",
      "----------------------------------------------"
     ]
    }
   ],
   "metadata": {}
  }
 ]
}