{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### A test of the performance prediction model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from collections import OrderedDict as odict\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.cm as cm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#(hardware name, number of nodes)\n",
    "files = odict({})\n",
    "files['i5'] = ('i5',1)\n",
    "files['gtx1060'] = ('gtx1060',1)\n",
    "files['skl_mpi1'] = ('skl',1)\n",
    "files['skl_mpi2'] = ('skl',2)\n",
    "files['skl_mpi4'] = ('skl',4)\n",
    "files['knl_mpi1'] = ('knl',1)\n",
    "files['knl_mpi2'] = ('knl',2)\n",
    "files['knl_mpi4'] = ('knl',4)\n",
    "files['p100nv_mpi1'] = ('p100',1)\n",
    "files['p100nv_mpi2'] = ('p100',2)\n",
    "files['p100nv_mpi4'] = ('p100',4)\n",
    "files['v100nv_mpi1'] = ('v100',1)\n",
    "files['v100nv_mpi2'] = ('v100',2)\n",
    "files['v100nv_mpi4'] = ('v100',4)\n",
    "# order by number of nodes to make labeling easier further down\n",
    "files=odict(sorted(files.items(), key= lambda t : t[1][1]))\n",
    "# count number of 1 nodes in dict\n",
    "number=0\n",
    "for k,v in files.items(): \n",
    "    if v[1]==1: number+=1\n",
    "#setup plotting specifications\n",
    "arch = {'knl':(cm.Greens, 450,0.5,0.33),'skl':(cm.Greys,200,0.5,0.75),   'p100':(cm.Blues,  550,0.5,0.43),\n",
    "       'v100':(cm.Purples,850,0.5,0.85), 'i5':(cm.Wistia,30,0.5,0.79),'gtx1060':(cm.Oranges,155,0.5,0.70)}\n",
    "intens={1:0.8, 2:0.6, 4:0.4}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here, we setup the prediction model by giving the number of function calls and memory operations of each of the three types of primitive functions axpby, dot and dxdy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#(axpby,dot,dxdy)\n",
    "latencies = odict()\n",
    "latencies['scal'] = (1,0,0)\n",
    "latencies['axpby'] = (1,0,0)\n",
    "latencies['pointwiseDot'] = (1,0,0)\n",
    "latencies['dot'] = (0,1,0)\n",
    "latencies['dx'] = (0,0,1)\n",
    "latencies['dy'] = (0,0,1)\n",
    "latencies['arakawa'] = (3,0,6) # N = 9\n",
    "latencies['cg'] = (6,2,6) # N = 13\n",
    "latencies['avg']= (9,2,12) # N=23\n",
    "memops = odict()\n",
    "memops['scal']= (2,0,0)\n",
    "memops['axpby']= (3,0,0)\n",
    "memops['pointwiseDot']= (6,0,0)\n",
    "memops['dot']= (0,2,0)\n",
    "memops['dx']= (0,0,3)\n",
    "memops['dy']= (0,0,3)\n",
    "memops['arakawa'] = (16,0,18) # M = 34 -> M/N = 3.78\n",
    "memops['cg'] = (20,4,18) # M = 42 -> M/N = 3.23\n",
    "memops['avg'] = (36,4,36) # M = 76 -> M/N = 3.30"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us read in the previously measured bandwidths and latencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>axpby_bw</th>\n",
       "      <th>axpby_bw_err</th>\n",
       "      <th>dot_bw</th>\n",
       "      <th>dot_bw_err</th>\n",
       "      <th>dxdy2_bw</th>\n",
       "      <th>dxdy2_bw_err</th>\n",
       "      <th>dxdy3_bw</th>\n",
       "      <th>dxdy3_bw_err</th>\n",
       "      <th>dxdy4_bw</th>\n",
       "      <th>dxdy4_bw_err</th>\n",
       "      <th>...</th>\n",
       "      <th>axpby_lat_dist</th>\n",
       "      <th>axpby_lat_dist_err</th>\n",
       "      <th>dot_lat_shared</th>\n",
       "      <th>dot_lat_shared_err</th>\n",
       "      <th>dot_lat_dist</th>\n",
       "      <th>dot_lat_dist_err</th>\n",
       "      <th>dxdy_lat_shared</th>\n",
       "      <th>dxdy_lat_shared_err</th>\n",
       "      <th>dxdy_lat_dist</th>\n",
       "      <th>dxdy_lat_dist_err</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>i5</th>\n",
       "      <td>29.99</td>\n",
       "      <td>0.19</td>\n",
       "      <td>9.31</td>\n",
       "      <td>0.04</td>\n",
       "      <td>27.79</td>\n",
       "      <td>2.97</td>\n",
       "      <td>29.12</td>\n",
       "      <td>2.84</td>\n",
       "      <td>25.58</td>\n",
       "      <td>1.49</td>\n",
       "      <td>...</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "      <td>4.76</td>\n",
       "      <td>0.23</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "      <td>0.00</td>\n",
       "      <td>1.44</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gtx1060</th>\n",
       "      <td>157.05</td>\n",
       "      <td>0.06</td>\n",
       "      <td>26.50</td>\n",
       "      <td>0.10</td>\n",
       "      <td>130.63</td>\n",
       "      <td>0.40</td>\n",
       "      <td>111.23</td>\n",
       "      <td>1.11</td>\n",
       "      <td>83.82</td>\n",
       "      <td>13.83</td>\n",
       "      <td>...</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "      <td>92.06</td>\n",
       "      <td>8.70</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.82</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>skl</th>\n",
       "      <td>206.71</td>\n",
       "      <td>5.87</td>\n",
       "      <td>192.05</td>\n",
       "      <td>18.31</td>\n",
       "      <td>181.56</td>\n",
       "      <td>35.38</td>\n",
       "      <td>161.75</td>\n",
       "      <td>13.00</td>\n",
       "      <td>118.06</td>\n",
       "      <td>18.39</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.26</td>\n",
       "      <td>17.28</td>\n",
       "      <td>2.32</td>\n",
       "      <td>37.93</td>\n",
       "      <td>4.14</td>\n",
       "      <td>22.70</td>\n",
       "      <td>2.11</td>\n",
       "      <td>28.52</td>\n",
       "      <td>2.10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>knl</th>\n",
       "      <td>393.15</td>\n",
       "      <td>22.19</td>\n",
       "      <td>141.36</td>\n",
       "      <td>6.63</td>\n",
       "      <td>239.04</td>\n",
       "      <td>17.02</td>\n",
       "      <td>172.69</td>\n",
       "      <td>26.80</td>\n",
       "      <td>126.04</td>\n",
       "      <td>18.59</td>\n",
       "      <td>...</td>\n",
       "      <td>9.16</td>\n",
       "      <td>0.09</td>\n",
       "      <td>54.83</td>\n",
       "      <td>1.79</td>\n",
       "      <td>119.59</td>\n",
       "      <td>5.14</td>\n",
       "      <td>9.93</td>\n",
       "      <td>0.70</td>\n",
       "      <td>52.67</td>\n",
       "      <td>3.72</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>p100</th>\n",
       "      <td>550.51</td>\n",
       "      <td>1.23</td>\n",
       "      <td>375.61</td>\n",
       "      <td>1.94</td>\n",
       "      <td>293.25</td>\n",
       "      <td>7.11</td>\n",
       "      <td>238.99</td>\n",
       "      <td>12.63</td>\n",
       "      <td>208.44</td>\n",
       "      <td>7.05</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.27</td>\n",
       "      <td>50.89</td>\n",
       "      <td>7.06</td>\n",
       "      <td>51.67</td>\n",
       "      <td>0.59</td>\n",
       "      <td>26.23</td>\n",
       "      <td>0.05</td>\n",
       "      <td>54.40</td>\n",
       "      <td>0.35</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>titanXp</th>\n",
       "      <td>431.24</td>\n",
       "      <td>3.45</td>\n",
       "      <td>61.37</td>\n",
       "      <td>0.12</td>\n",
       "      <td>372.85</td>\n",
       "      <td>4.16</td>\n",
       "      <td>308.92</td>\n",
       "      <td>9.47</td>\n",
       "      <td>246.73</td>\n",
       "      <td>7.92</td>\n",
       "      <td>...</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "      <td>44.37</td>\n",
       "      <td>5.15</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "      <td>2.38</td>\n",
       "      <td>0.57</td>\n",
       "      <td>nan</td>\n",
       "      <td>nan</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v100</th>\n",
       "      <td>846.42</td>\n",
       "      <td>0.95</td>\n",
       "      <td>610.15</td>\n",
       "      <td>5.99</td>\n",
       "      <td>794.43</td>\n",
       "      <td>20.52</td>\n",
       "      <td>735.42</td>\n",
       "      <td>33.02</td>\n",
       "      <td>696.49</td>\n",
       "      <td>15.14</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.31</td>\n",
       "      <td>88.49</td>\n",
       "      <td>4.68</td>\n",
       "      <td>97.58</td>\n",
       "      <td>0.79</td>\n",
       "      <td>4.20</td>\n",
       "      <td>0.02</td>\n",
       "      <td>37.19</td>\n",
       "      <td>0.42</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>7 rows × 24 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         axpby_bw  axpby_bw_err  dot_bw  dot_bw_err  dxdy2_bw  dxdy2_bw_err  \\\n",
       "i5          29.99          0.19    9.31        0.04     27.79          2.97   \n",
       "gtx1060    157.05          0.06   26.50        0.10    130.63          0.40   \n",
       "skl        206.71          5.87  192.05       18.31    181.56         35.38   \n",
       "knl        393.15         22.19  141.36        6.63    239.04         17.02   \n",
       "p100       550.51          1.23  375.61        1.94    293.25          7.11   \n",
       "titanXp    431.24          3.45   61.37        0.12    372.85          4.16   \n",
       "v100       846.42          0.95  610.15        5.99    794.43         20.52   \n",
       "\n",
       "         dxdy3_bw  dxdy3_bw_err  dxdy4_bw  dxdy4_bw_err  ...  axpby_lat_dist  \\\n",
       "i5          29.12          2.84     25.58          1.49  ...             nan   \n",
       "gtx1060    111.23          1.11     83.82         13.83  ...             nan   \n",
       "skl        161.75         13.00    118.06         18.39  ...            0.00   \n",
       "knl        172.69         26.80    126.04         18.59  ...            9.16   \n",
       "p100       238.99         12.63    208.44          7.05  ...            0.00   \n",
       "titanXp    308.92          9.47    246.73          7.92  ...             nan   \n",
       "v100       735.42         33.02    696.49         15.14  ...            0.00   \n",
       "\n",
       "         axpby_lat_dist_err  dot_lat_shared  dot_lat_shared_err  dot_lat_dist  \\\n",
       "i5                      nan            4.76                0.23           nan   \n",
       "gtx1060                 nan           92.06                8.70           nan   \n",
       "skl                    0.26           17.28                2.32         37.93   \n",
       "knl                    0.09           54.83                1.79        119.59   \n",
       "p100                   0.27           50.89                7.06         51.67   \n",
       "titanXp                 nan           44.37                5.15           nan   \n",
       "v100                   0.31           88.49                4.68         97.58   \n",
       "\n",
       "         dot_lat_dist_err  dxdy_lat_shared  dxdy_lat_shared_err  \\\n",
       "i5                    nan             0.00                 1.44   \n",
       "gtx1060               nan             0.00                 0.82   \n",
       "skl                  4.14            22.70                 2.11   \n",
       "knl                  5.14             9.93                 0.70   \n",
       "p100                 0.59            26.23                 0.05   \n",
       "titanXp               nan             2.38                 0.57   \n",
       "v100                 0.79             4.20                 0.02   \n",
       "\n",
       "         dxdy_lat_dist  dxdy_lat_dist_err  \n",
       "i5                 nan                nan  \n",
       "gtx1060            nan                nan  \n",
       "skl              28.52               2.10  \n",
       "knl              52.67               3.72  \n",
       "p100             54.40               0.35  \n",
       "titanXp            nan                nan  \n",
       "v100             37.19               0.42  \n",
       "\n",
       "[7 rows x 24 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "theo = pd.read_csv('performance.csv',delimiter=' ')\n",
    "theo.set_index('arch',inplace=True)\n",
    "theo.index.name = None\n",
    "pd.set_option('display.float_format', lambda x: '%.2f' % x)\n",
    "theo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#define conversion function \n",
    "def toString(x): \n",
    "    if pd.isnull(x) : return 'n/a'\n",
    "    #string = '%.1f'% x\n",
    "    string = '%d' %np.ceil(x)\n",
    "    #if np.ceil(x)<100 : string = '0'+string\n",
    "    if np.ceil(x)<10 : string = '0'+string\n",
    "    return string"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In the followin cell we construct a table that shows the average bandwiths and latencies among a typical selection of primitive algorithms. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>B(P=2) [GB/s]</th>\n",
       "      <th>B(P=3) [GB/s]</th>\n",
       "      <th>B(P=4) [GB/s]</th>\n",
       "      <th>B(P=5) [GB/s]</th>\n",
       "      <th>$T_{lat}(1)$ [$\\mu$s]</th>\n",
       "      <th>$T_{lat}(4)$ [$\\mu$s]</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>i5</th>\n",
       "      <td>26 $\\pm$ 02</td>\n",
       "      <td>27 $\\pm$ 02</td>\n",
       "      <td>26 $\\pm$ 01</td>\n",
       "      <td>23 $\\pm$ 02</td>\n",
       "      <td>01 $\\pm$ 01</td>\n",
       "      <td>n/a</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gtx1060</th>\n",
       "      <td>116 $\\pm$ 01</td>\n",
       "      <td>108 $\\pm$ 01</td>\n",
       "      <td>94 $\\pm$ 09</td>\n",
       "      <td>85 $\\pm$ 12</td>\n",
       "      <td>09 $\\pm$ 01</td>\n",
       "      <td>n/a</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>skl</th>\n",
       "      <td>194 $\\pm$ 20</td>\n",
       "      <td>183 $\\pm$ 09</td>\n",
       "      <td>153 $\\pm$ 15</td>\n",
       "      <td>147 $\\pm$ 07</td>\n",
       "      <td>14 $\\pm$ 02</td>\n",
       "      <td>19 $\\pm$ 02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>knl</th>\n",
       "      <td>281 $\\pm$ 13</td>\n",
       "      <td>232 $\\pm$ 24</td>\n",
       "      <td>188 $\\pm$ 20</td>\n",
       "      <td>160 $\\pm$ 18</td>\n",
       "      <td>13 $\\pm$ 01</td>\n",
       "      <td>42 $\\pm$ 02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>titanXp</th>\n",
       "      <td>310 $\\pm$ 02</td>\n",
       "      <td>287 $\\pm$ 04</td>\n",
       "      <td>259 $\\pm$ 05</td>\n",
       "      <td>230 $\\pm$ 21</td>\n",
       "      <td>06 $\\pm$ 01</td>\n",
       "      <td>n/a</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>p100</th>\n",
       "      <td>383 $\\pm$ 06</td>\n",
       "      <td>336 $\\pm$ 12</td>\n",
       "      <td>306 $\\pm$ 08</td>\n",
       "      <td>267 $\\pm$ 21</td>\n",
       "      <td>22 $\\pm$ 01</td>\n",
       "      <td>33 $\\pm$ 01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>v100</th>\n",
       "      <td>806 $\\pm$ 11</td>\n",
       "      <td>776 $\\pm$ 18</td>\n",
       "      <td>755 $\\pm$ 09</td>\n",
       "      <td>691 $\\pm$ 43</td>\n",
       "      <td>11 $\\pm$ 01</td>\n",
       "      <td>28 $\\pm$ 01</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        B(P=2) [GB/s] B(P=3) [GB/s] B(P=4) [GB/s] B(P=5) [GB/s]  \\\n",
       "i5        26 $\\pm$ 02   27 $\\pm$ 02   26 $\\pm$ 01   23 $\\pm$ 02   \n",
       "gtx1060  116 $\\pm$ 01  108 $\\pm$ 01   94 $\\pm$ 09   85 $\\pm$ 12   \n",
       "skl      194 $\\pm$ 20  183 $\\pm$ 09  153 $\\pm$ 15  147 $\\pm$ 07   \n",
       "knl      281 $\\pm$ 13  232 $\\pm$ 24  188 $\\pm$ 20  160 $\\pm$ 18   \n",
       "titanXp  310 $\\pm$ 02  287 $\\pm$ 04  259 $\\pm$ 05  230 $\\pm$ 21   \n",
       "p100     383 $\\pm$ 06  336 $\\pm$ 12  306 $\\pm$ 08  267 $\\pm$ 21   \n",
       "v100     806 $\\pm$ 11  776 $\\pm$ 18  755 $\\pm$ 09  691 $\\pm$ 43   \n",
       "\n",
       "        $T_{lat}(1)$ [$\\mu$s] $T_{lat}(4)$ [$\\mu$s]  \n",
       "i5                01 $\\pm$ 01                   n/a  \n",
       "gtx1060           09 $\\pm$ 01                   n/a  \n",
       "skl               14 $\\pm$ 02           19 $\\pm$ 02  \n",
       "knl               13 $\\pm$ 01           42 $\\pm$ 02  \n",
       "titanXp           06 $\\pm$ 01                   n/a  \n",
       "p100              22 $\\pm$ 01           33 $\\pm$ 01  \n",
       "v100              11 $\\pm$ 01           28 $\\pm$ 01  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lines=[]\n",
    "#now compute and plot the prediction \n",
    "archs = ['i5','gtx1060','skl','knl','titanXp','p100','v100']\n",
    "for k in archs : \n",
    "    line =[]\n",
    "    for q,l in latencies.items():\n",
    "        m = memops[q]\n",
    "        M = m[0]+m[1]+m[2]\n",
    "        for n in [2,3,4,5]:\n",
    "            bw = [theo.loc[k,'axpby_bw'],theo.loc[k,'dot_bw'],theo.loc[k,'dxdy'+str(n)+'_bw']]\n",
    "            err_bw = [theo.loc[k,'axpby_bw_err'],theo.loc[k,'dot_bw_err'],theo.loc[k,'dxdy'+str(n)+'_bw_err']]\n",
    "            bandwidth = M/(m[0]/bw[0] + m[1]/bw[1] + m[2]/bw[2])\n",
    "            err_bandwidth = bandwidth/(m[0]/bw[0] + m[1]/bw[1] + m[2]/bw[2])*np.sqrt(\n",
    "                (m[0]/bw[0]**2*err_bw[0])**2 + (m[1]/bw[1]**2*err_bw[1])**2 + (m[2]/bw[2]**2*err_bw[2])**2  )\n",
    "            line.append( toString( bandwidth)+\" $\\pm$ \"+toString(err_bandwidth))\n",
    "        L = l[0]+l[1]+l[2]\n",
    "        for dist in ['shared','dist']:\n",
    "            lat = [theo.loc[k,'axpby_lat_'+dist], theo.loc[k,'dot_lat_'+dist], theo.loc[k,'dxdy_lat_'+dist]]\n",
    "            err_lat = [theo.loc[k,'axpby_lat_'+dist+'_err'], theo.loc[k,'dot_lat_'+dist+'_err'], theo.loc[k,'dxdy_lat_'+dist+'_err']]\n",
    "            latency = ( l[0]*lat[0]+ l[1]* lat[1] + l[2]*lat[2])/L #in us\n",
    "            err_latency = np.sqrt( (l[0]*err_lat[0])**2 + (l[1]*err_lat[1])**2 + (l[2]*err_lat[2])**2 )/L\n",
    "            if (dist == 'dist') and ((k == 'i5') or (k=='gtx1060') or (k=='titanXp')):\n",
    "                line.append(toString( None))\n",
    "            else: line.append(toString( latency)+\" $\\pm$ \"+toString(err_latency))\n",
    "        #print(q,latency)\n",
    "\n",
    "    lines.append(line)\n",
    "index = archs\n",
    "tuples=[]  \n",
    "for p in latencies.keys():\n",
    "    for q in ['B(P=2) [GB/s]','B(P=3) [GB/s]','B(P=4) [GB/s]','B(P=5) [GB/s]',\n",
    "              '$T_{lat}(1)$ [$\\mu$s]','$T_{lat}(4)$ [$\\mu$s]']:\n",
    "        tuples.append((p,q))\n",
    "    \n",
    "\n",
    "cols=pd.MultiIndex.from_tuples(tuples)\n",
    "\n",
    "avg = pd.DataFrame(lines, index=index, columns=cols)\n",
    "#avg.sort_values(by=('avg','B(P=2) [GB/s]'), inplace=True)\n",
    "#avg.loc[:,('cg','size')]=(avg['cg']['P=3']/1e3*9/34\n",
    "#                              )*avg['cg']['lat 4 [$\\mu$s]']\n",
    "pd.set_option('display.float_format', lambda x: '%.0f' % x)\n",
    "#avg.loc['i5',('avg','$T_{lat}(4)$ [$\\mu$s]')]='n/a'\n",
    "#avg.loc['gtx1060',('avg','$T_{lat}(4)$ [$\\mu$s]')]='n/a'\n",
    "filename='avg.tex'\n",
    "with open(filename, 'wb') as f:\n",
    "    f.write(bytes(avg['avg'].to_latex(\n",
    "        escape=False,column_format='lp{1.5cm}p{1.5cm}p{1.5cm}p{1.5cm}p{1.2cm}p{1.2cm}',\n",
    "                               bold_rows=True),'UTF-8'))\n",
    "avg['avg']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In the following we compare the predicted with the measured values for the Arakawa and CG algorithm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 480x296 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 480x296 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#now compute and plot the prediction \n",
    "del latencies['avg']\n",
    "del memops['avg']\n",
    "for n in ['arakawa','cg'] :#latencies.keys():\n",
    "    fig,ax=plt.subplots(1,1,figsize=(6,3.7),dpi= 80, facecolor='w', edgecolor='k')\n",
    "    xs = np.array([0.1,1000])\n",
    "    ys = np.array([1.0,1.0])\n",
    "    for frac in [1.0,4/3,8/4]:\n",
    "        plt.plot(xs,frac*ys,ls='--',color=cm.Greys(0.8/frac))\n",
    "        plt.plot(xs,1/frac*ys,ls='--',color=cm.Greys(0.8/frac))\n",
    "        \n",
    "    for f, v in files.items() :#{'knl_mpi2':('knl',2)}.items():\n",
    "        df=pd.read_csv('benchmark_'+f+'.csv', delimiter=' ')\n",
    "        #add size and get rid of non-relevant columns\n",
    "        df.insert(0,'size', 8*df['n']*df['n']*df['Nx']*df['Ny']/1e6/v[1])\n",
    "        dfr = df[['n','Nx','Ny','size']+list(memops.keys())]\n",
    "        #compute mean and standard derivation of 'same' groups \n",
    "        dfr=dfr.groupby(['n', 'Nx','Ny','size']).mean()\n",
    "        dfr=dfr.reset_index(level=['n','Nx','Ny','size'])\n",
    "\n",
    "        dfr['FirstLevel']='measured'\n",
    "        dfr.columns=pd.MultiIndex.from_product([dfr.columns,['measured']])\n",
    "        del dfr['FirstLevel']\n",
    "        \n",
    "        dfr['dxdy_bw'] = dfr.apply( \n",
    "            lambda row: theo.loc[v[0],'dxdy'+str(row['n','measured'].astype(int))+'_bw'], axis=1)\n",
    "\n",
    "        dxdystring = 'dxdy_lat_shared'\n",
    "        dotstring = 'dot_lat_shared'\n",
    "        if v[1] > 1 : \n",
    "            dxdystring = 'dxdy_lat_dist'\n",
    "            dotstrint = 'dot_lat_dist'\n",
    "        for q,l in latencies.items():\n",
    "            m = memops[q]\n",
    "            dfr.loc[:,(q,'predicted')] = (\n",
    "                                    ( l[0]*theo.loc[v[0],'axpby_lat_shared']+\n",
    "                                      l[1]*theo.loc[v[0],dotstring] +\n",
    "                                      l[2]*theo.loc[v[0],dxdystring]\n",
    "                                    )*1e-6+\n",
    "                                    (m[0]/theo.loc[v[0],'axpby_bw'] + m[1]/theo.loc[v[0],'dot_bw'] + m[2]/dfr['dxdy_bw',''])\n",
    "                                       *dfr[('size','measured')]/1000)\n",
    "            dfr.loc[:,(q,'meas/pred')]=dfr[(q,'measured')]/dfr[(q,'predicted')]\n",
    "\n",
    "        toPlot = dfr[n].join(dfr[('size')],rsuffix='_size')\n",
    "        toPlot.plot(kind='scatter',ax=ax,color=arch[v[0]][0](intens[v[1]]),edgecolors='k',\n",
    "                     x='measured_size', y='meas/pred',label=v[0],s=64)\n",
    "    handles, labels = plt.gca().get_legend_handles_labels()\n",
    "    handles = handles[0:number]; labels = labels[0:number]\n",
    "\n",
    "    #plt.plot(xs,ys)\n",
    "\n",
    "    \n",
    "    plt.legend(handles, labels, loc='upper right',ncol=2,\n",
    "               scatterpoints=1,fontsize='medium',framealpha=0.5)\n",
    "    plt.xscale('log')\n",
    "    plt.xlim(xs[0],xs[1])\n",
    "    #plt.xlabel('measured time in s')\n",
    "    plt.xlabel('array size [MB] / # of nodes')\n",
    "    plt.ylabel('measured / predicted time')\n",
    "    plt.yscale('log', subsy=[0])#log scale, turn minor ticks off\n",
    "    plt.ylim(1/3,3)\n",
    "    plt.yticks([1/3,0.5,0.75,1,1.33,2,3],['1/3','1/2','3/4',1,'4/3',2,3])\n",
    "    #plt.title(n)\n",
    "    plt.savefig(n+'.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Observations\n",
    "- plots show deviations from the predicted timing\n",
    "- the goal of the discussion is to prove that the time formula is correct because then performance can be discussed analytically and makes \"scaling\" plots somewhat obsolete\n",
    "- there seems to be a systematic overestimation of the knl MPI scaling (is this the fault of the implemenation? What is wrong there?)\n",
    "- skl and i5 for small sizes are mostly faster than predicted because cache effects are not included in the model\n",
    "- there seems to a drop in efficiency in GPUs when the problem size nears the full size of the GPU memory (last single node points in GTX, P100 and V100)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}