comparatif_BLAS_lib.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparatif entre MKL, OpenBLAS et BLIS"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ce notebook permet d'obtenir les résultats de la figure 2. Nous avons créer plusieurs environnements où numpy est installé avec différentes bibliothèques BLAS-Like. Pour reproduire les résultats il faut donc recréer ces 3 environnements et modifier l'environnement de travail à chaque fois que cela est spécifié dans le notebook. Le processus d'installation est indiqué dans le document ReadME pour BLIS et si vous êtes sur Anaconda votre numpy devrait être adossé à MKL par défaut donc nul besoin de créer cet environnement la. Une autre méthode est d'installer numpy avec différentes libraries en même temps et de changer l'ordre de priorité grâce aux variables d'environnement. Nous avons choisi de séparer les environnement pour des raisons de clareté et pour éviter les erreurs. \n",
    "\n",
    "**Remarque** : pour pouvoir modifier son environnement au sein d'un même notebook : \n",
    "\n",
    "Dans l'environnement, installer ipykernel :\n",
    "``` \n",
    "conda install ipykernel \n",
    "``` \n",
    "Puis installer un nouveau kernel dans l'environnement : \n",
    "``` \n",
    "ipython kernel install --user --name=projectname \n",
    "``` \n",
    "Vous pouvez à présent basculer d'un environnement à l'autre dans la barre du haut à l'onglet Kernel > Change Kernel. La fonction time_count_np vous permet de vérifier que vous travaillez avec le bon numpy à chaque fois que vous lancez les calculs. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color=\"red\"> **Modifier le kernel : BLIS** </font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time \n",
    "import os \n",
    "os.system(\"taskset -p 0xff %d\" % os.getpid())\n",
    "os.environ[\"BLIS_MAIN_FREE\"]='1'\n",
    "os.environ[\"BLIS_JC_NT\"] = \"1\"\n",
    "os.environ[\"BLIS_IC_NT\"] = \"6\"\n",
    "os.environ[\"BLIS_JR_NT\"] = \"1\"\n",
    "os.environ[\"BLIS_IR_NT\"] = \"1\"\n",
    "import numpy as np\n",
    "from fonctions import time_count_np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "blis_info:\n",
      "    libraries = ['blis', 'blis']\n",
      "    library_dirs = ['/home/cesar/Desktop/blis/lib/haswell']\n",
      "    define_macros = [('HAVE_CBLAS', None)]\n",
      "    include_dirs = ['/home/cesar/Desktop/blis/frame/compat/cblas/src']\n",
      "    language = c\n",
      "    runtime_library_dirs = ['/home/cesar/Desktop/blis/lib/haswell']\n",
      "blas_opt_info:\n",
      "    libraries = ['blis', 'blis']\n",
      "    library_dirs = ['/home/cesar/Desktop/blis/lib/haswell']\n",
      "    define_macros = [('HAVE_CBLAS', None)]\n",
      "    include_dirs = ['/home/cesar/Desktop/blis/frame/compat/cblas/src']\n",
      "    language = c\n",
      "    runtime_library_dirs = ['/home/cesar/Desktop/blis/lib/haswell']\n",
      "lapack_mkl_info:\n",
      "  NOT AVAILABLE\n",
      "openblas_lapack_info:\n",
      "  NOT AVAILABLE\n",
      "openblas_clapack_info:\n",
      "  NOT AVAILABLE\n",
      "flame_info:\n",
      "  NOT AVAILABLE\n",
      "atlas_3_10_threads_info:\n",
      "  NOT AVAILABLE\n",
      "atlas_3_10_info:\n",
      "  NOT AVAILABLE\n",
      "atlas_threads_info:\n",
      "  NOT AVAILABLE\n",
      "atlas_info:\n",
      "  NOT AVAILABLE\n",
      "accelerate_info:\n",
      "  NOT AVAILABLE\n",
      "lapack_info:\n",
      "    libraries = ['lapack', 'lapack']\n",
      "    library_dirs = ['/usr/lib']\n",
      "    language = f77\n",
      "lapack_opt_info:\n",
      "    libraries = ['lapack', 'lapack', 'blis', 'blis']\n",
      "    library_dirs = ['/usr/lib', '/home/cesar/Desktop/blis/lib/haswell']\n",
      "    language = c\n",
      "    define_macros = [('HAVE_CBLAS', None), ('NO_ATLAS_INFO', 1)]\n",
      "    include_dirs = ['/home/cesar/Desktop/blis/frame/compat/cblas/src']\n",
      "    runtime_library_dirs = ['/home/cesar/Desktop/blis/lib/haswell']\n",
      "### Temps d'execution moyen d'un matrice de taille :\n",
      "\t- (50x50)\t:\t\t1.4781951904296875e-05\ts.\n",
      "\t- (100x100)\t:\t\t4.3487548828125e-05\ts.\n",
      "\t- (150x150)\t:\t\t0.00014190673828125\ts.\n",
      "\t- (200x200)\t:\t\t0.00039930343627929686\ts.\n",
      "\t- (250x250)\t:\t\t0.0021898746490478516\ts.\n",
      "\t- (300x300)\t:\t\t0.00021586418151855468\ts.\n",
      "\t- (350x350)\t:\t\t0.0021803855895996095\ts.\n",
      "\t- (400x400)\t:\t\t0.0004965782165527344\ts.\n",
      "\t- (450x450)\t:\t\t0.0007680892944335937\ts.\n",
      "\t- (500x500)\t:\t\t0.0037577152252197266\ts.\n",
      "\t- (550x550)\t:\t\t0.0037837505340576174\ts.\n",
      "\t- (600x600)\t:\t\t0.001793241500854492\ts.\n",
      "\t- (650x650)\t:\t\t0.0023683547973632813\ts.\n",
      "\t- (700x700)\t:\t\t0.0046978950500488285\ts.\n",
      "\t- (750x750)\t:\t\t0.0058688163757324215\ts.\n",
      "\t- (800x800)\t:\t\t0.008743190765380859\ts.\n",
      "\t- (850x850)\t:\t\t0.005345392227172852\ts.\n",
      "\t- (900x900)\t:\t\t0.006143760681152344\ts.\n",
      "\t- (950x950)\t:\t\t0.007434368133544922\ts.\n",
      "\t- (1000x1000)\t:\t\t0.008436107635498047\ts.\n",
      "\t- (1050x1050)\t:\t\t0.014998245239257812\ts.\n",
      "\t- (1100x1100)\t:\t\t0.011560440063476562\ts.\n",
      "\t- (1150x1150)\t:\t\t0.013124847412109375\ts.\n",
      "\t- (1200x1200)\t:\t\t0.019203710556030273\ts.\n",
      "\t- (1250x1250)\t:\t\t0.016392230987548828\ts.\n",
      "\t- (1300x1300)\t:\t\t0.02571978569030762\ts.\n",
      "\t- (1350x1350)\t:\t\t0.025301599502563478\ts.\n",
      "\t- (1400x1400)\t:\t\t0.028652381896972657\ts.\n",
      "\t- (1450x1450)\t:\t\t0.026220941543579103\ts.\n",
      "\t- (1500x1500)\t:\t\t0.03221597671508789\ts.\n",
      "\t- (1550x1550)\t:\t\t0.03428530693054199\ts.\n",
      "\t- (1600x1600)\t:\t\t0.037882232666015626\ts.\n",
      "\t- (1650x1650)\t:\t\t0.043096494674682614\ts.\n",
      "\t- (1700x1700)\t:\t\t0.04910345077514648\ts.\n",
      "\t- (1750x1750)\t:\t\t0.0493532657623291\ts.\n",
      "\t- (1800x1800)\t:\t\t0.05542488098144531\ts.\n",
      "\t- (1850x1850)\t:\t\t0.06063809394836426\ts.\n",
      "\t- (1900x1900)\t:\t\t0.06275572776794433\ts.\n",
      "\t- (1950x1950)\t:\t\t0.07302165031433105\ts.\n",
      "\t- (2000x2000)\t:\t\t0.07481307983398437\ts.\n",
      "\t- (2050x2050)\t:\t\t0.08610777854919434\ts.\n",
      "\t- (2100x2100)\t:\t\t0.0972074031829834\ts.\n",
      "\t- (2150x2150)\t:\t\t0.09232974052429199\ts.\n",
      "\t- (2200x2200)\t:\t\t0.09516749382019044\ts.\n",
      "\t- (2250x2250)\t:\t\t0.10586991310119628\ts.\n",
      "\t- (2300x2300)\t:\t\t0.10922732353210449\ts.\n",
      "\t- (2350x2350)\t:\t\t0.12443108558654785\ts.\n",
      "\t- (2400x2400)\t:\t\t0.12905464172363282\ts.\n",
      "\t- (2450x2450)\t:\t\t0.12859749794006348\ts.\n",
      "\t- (2500x2500)\t:\t\t0.13225455284118653\ts.\n",
      "\t- (2550x2550)\t:\t\t0.13973550796508788\ts.\n",
      "\t- (2600x2600)\t:\t\t0.1552152156829834\ts.\n",
      "\t- (2650x2650)\t:\t\t0.16859893798828124\ts.\n",
      "\t- (2700x2700)\t:\t\t0.17228922843933106\ts.\n",
      "\t- (2750x2750)\t:\t\t0.1864643096923828\ts.\n",
      "\t- (2800x2800)\t:\t\t0.20004091262817383\ts.\n",
      "\t- (2850x2850)\t:\t\t0.20247392654418944\ts.\n",
      "\t- (2900x2900)\t:\t\t0.21451468467712403\ts.\n",
      "\t- (2950x2950)\t:\t\t0.27777748107910155\ts.\n",
      "\t- (3000x3000)\t:\t\t0.22815184593200682\ts.\n",
      "\t- (3050x3050)\t:\t\t0.23733372688293458\ts.\n",
      "\t- (3100x3100)\t:\t\t0.26398320198059083\ts.\n",
      "\t- (3150x3150)\t:\t\t0.2711483955383301\ts.\n",
      "\t- (3200x3200)\t:\t\t0.3006294727325439\ts.\n",
      "\t- (3250x3250)\t:\t\t0.2895052909851074\ts.\n",
      "\t- (3300x3300)\t:\t\t0.3021736145019531\ts.\n",
      "\t- (3350x3350)\t:\t\t0.33718209266662597\ts.\n",
      "\t- (3400x3400)\t:\t\t0.3450645923614502\ts.\n",
      "\t- (3450x3450)\t:\t\t0.35599555969238283\ts.\n",
      "\t- (3500x3500)\t:\t\t0.3629612922668457\ts.\n",
      "\t- (3550x3550)\t:\t\t0.3741950035095215\ts.\n",
      "\t- (3600x3600)\t:\t\t0.4324484348297119\ts.\n",
      "\t- (3650x3650)\t:\t\t0.40899105072021485\ts.\n",
      "\t- (3700x3700)\t:\t\t0.4318894386291504\ts.\n",
      "\t- (3750x3750)\t:\t\t0.4385229110717773\ts.\n",
      "\t- (3800x3800)\t:\t\t0.45324273109436036\ts.\n",
      "\t- (3850x3850)\t:\t\t0.4856258869171143\ts.\n",
      "\t- (3900x3900)\t:\t\t0.5027763366699218\ts.\n",
      "\t- (3950x3950)\t:\t\t0.5151789665222168\ts.\n",
      "\t- (4000x4000)\t:\t\t0.573549461364746\ts.\n",
      "\t- (4050x4050)\t:\t\t0.5480030536651611\ts.\n",
      "\t- (4100x4100)\t:\t\t0.6175383567810059\ts.\n",
      "\t- (4150x4150)\t:\t\t0.6205345630645752\ts.\n",
      "\t- (4200x4200)\t:\t\t0.632822847366333\ts.\n",
      "\t- (4250x4250)\t:\t\t0.645698070526123\ts.\n",
      "\t- (4300x4300)\t:\t\t0.7225182533264161\ts.\n",
      "\t- (4350x4350)\t:\t\t0.6693305015563965\ts.\n",
      "\t- (4400x4400)\t:\t\t0.7734833240509034\ts.\n",
      "\t- (4450x4450)\t:\t\t0.7387148857116699\ts.\n",
      "\t- (4500x4500)\t:\t\t0.7537085056304932\ts.\n",
      "\t- (4550x4550)\t:\t\t0.7702407836914062\ts.\n",
      "\t- (4600x4600)\t:\t\t0.7874140262603759\ts.\n",
      "\t- (4650x4650)\t:\t\t0.8361196517944336\ts.\n",
      "\t- (4700x4700)\t:\t\t0.8681750774383545\ts.\n",
      "\t- (4750x4750)\t:\t\t0.8966275215148926\ts.\n",
      "\t- (4800x4800)\t:\t\t0.932660722732544\ts.\n",
      "\t- (4850x4850)\t:\t\t0.93081693649292\ts.\n",
      "\t- (4900x4900)\t:\t\t0.9672323703765869\ts.\n",
      "\t- (4950x4950)\t:\t\t1.0029701709747314\ts.\n",
      "\t- (5000x5000)\t:\t\t1.0654780864715576\ts.\n"
     ]
    }
   ],
   "source": [
    "results_blis = time_count_np(\"BLIS\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color=\"red\"> **Modifier le kernel : OpenBLAS** </font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time \n",
    "import os \n",
    "os.system(\"taskset -p 0xff %d\" % os.getpid())\n",
    "os.environ[\"OPENBLAS_MAIN_FREE\"]='1'\n",
    "os.environ[\"OMP_NUM_THREADS\"] = \"6\"\n",
    "import numpy as np \n",
    "from fonctions import time_count_np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "openblas_info:\n",
      "    libraries = ['openblas', 'openblas']\n",
      "    library_dirs = ['/opt/OpenBLAS/lib']\n",
      "    language = c\n",
      "    define_macros = [('HAVE_CBLAS', None)]\n",
      "    runtime_library_dirs = ['/opt/OpenBLAS/lib']\n",
      "blas_opt_info:\n",
      "    libraries = ['openblas', 'openblas']\n",
      "    library_dirs = ['/opt/OpenBLAS/lib']\n",
      "    language = c\n",
      "    define_macros = [('HAVE_CBLAS', None)]\n",
      "    runtime_library_dirs = ['/opt/OpenBLAS/lib']\n",
      "lapack_mkl_info:\n",
      "  NOT AVAILABLE\n",
      "openblas_lapack_info:\n",
      "    libraries = ['openblas', 'openblas']\n",
      "    library_dirs = ['/opt/OpenBLAS/lib']\n",
      "    language = c\n",
      "    define_macros = [('HAVE_CBLAS', None)]\n",
      "    runtime_library_dirs = ['/opt/OpenBLAS/lib']\n",
      "lapack_opt_info:\n",
      "    libraries = ['openblas', 'openblas']\n",
      "    library_dirs = ['/opt/OpenBLAS/lib']\n",
      "    language = c\n",
      "    define_macros = [('HAVE_CBLAS', None)]\n",
      "    runtime_library_dirs = ['/opt/OpenBLAS/lib']\n",
      "### Temps d'execution moyen d'un matrice de taille :\n",
      "\t- (50x50)\t:\t\t1.5974044799804688e-05\ts.\n",
      "\t- (100x100)\t:\t\t4.3964385986328126e-05\ts.\n",
      "\t- (150x150)\t:\t\t0.00014333724975585939\ts.\n",
      "\t- (200x200)\t:\t\t0.000403594970703125\ts.\n",
      "\t- (250x250)\t:\t\t0.0024315357208251954\ts.\n",
      "\t- (300x300)\t:\t\t0.001009225845336914\ts.\n",
      "\t- (350x350)\t:\t\t0.0006368637084960937\ts.\n",
      "\t- (400x400)\t:\t\t0.0004897117614746094\ts.\n",
      "\t- (450x450)\t:\t\t0.0007611751556396484\ts.\n",
      "\t- (500x500)\t:\t\t0.0037759780883789063\ts.\n",
      "\t- (550x550)\t:\t\t0.0017389774322509766\ts.\n",
      "\t- (600x600)\t:\t\t0.0029342174530029297\ts.\n",
      "\t- (650x650)\t:\t\t0.0022451400756835936\ts.\n",
      "\t- (700x700)\t:\t\t0.0027587413787841797\ts.\n",
      "\t- (750x750)\t:\t\t0.004973411560058594\ts.\n",
      "\t- (800x800)\t:\t\t0.007848644256591797\ts.\n",
      "\t- (850x850)\t:\t\t0.006254911422729492\ts.\n",
      "\t- (900x900)\t:\t\t0.005950164794921875\ts.\n",
      "\t- (950x950)\t:\t\t0.006958723068237305\ts.\n",
      "\t- (1000x1000)\t:\t\t0.008203554153442382\ts.\n",
      "\t- (1050x1050)\t:\t\t0.013584709167480469\ts.\n",
      "\t- (1100x1100)\t:\t\t0.011244678497314453\ts.\n",
      "\t- (1150x1150)\t:\t\t0.012560081481933594\ts.\n",
      "\t- (1200x1200)\t:\t\t0.017682981491088868\ts.\n",
      "\t- (1250x1250)\t:\t\t0.015900897979736327\ts.\n",
      "\t- (1300x1300)\t:\t\t0.025356292724609375\ts.\n",
      "\t- (1350x1350)\t:\t\t0.025716400146484374\ts.\n",
      "\t- (1400x1400)\t:\t\t0.027046680450439453\ts.\n",
      "\t- (1450x1450)\t:\t\t0.02521939277648926\ts.\n",
      "\t- (1500x1500)\t:\t\t0.030060815811157226\ts.\n",
      "\t- (1550x1550)\t:\t\t0.03285770416259766\ts.\n",
      "\t- (1600x1600)\t:\t\t0.03563985824584961\ts.\n",
      "\t- (1650x1650)\t:\t\t0.04075665473937988\ts.\n",
      "\t- (1700x1700)\t:\t\t0.042883777618408205\ts.\n",
      "\t- (1750x1750)\t:\t\t0.06492013931274414\ts.\n",
      "\t- (1800x1800)\t:\t\t0.08980374336242676\ts.\n",
      "\t- (1850x1850)\t:\t\t0.05865764617919922\ts.\n",
      "\t- (1900x1900)\t:\t\t0.061463212966918944\ts.\n",
      "\t- (1950x1950)\t:\t\t0.06977553367614746\ts.\n",
      "\t- (2000x2000)\t:\t\t0.07092037200927734\ts.\n",
      "\t- (2050x2050)\t:\t\t0.0908961296081543\ts.\n",
      "\t- (2100x2100)\t:\t\t0.0962367057800293\ts.\n",
      "\t- (2150x2150)\t:\t\t0.0961421012878418\ts.\n",
      "\t- (2200x2200)\t:\t\t0.10139679908752441\ts.\n",
      "\t- (2250x2250)\t:\t\t0.11239871978759766\ts.\n",
      "\t- (2300x2300)\t:\t\t0.10980477333068847\ts.\n",
      "\t- (2350x2350)\t:\t\t0.1266796112060547\ts.\n",
      "\t- (2400x2400)\t:\t\t0.12764511108398438\ts.\n",
      "\t- (2450x2450)\t:\t\t0.1339871883392334\ts.\n",
      "\t- (2500x2500)\t:\t\t0.1370861530303955\ts.\n",
      "\t- (2550x2550)\t:\t\t0.1529756546020508\ts.\n",
      "\t- (2600x2600)\t:\t\t0.1611790657043457\ts.\n",
      "\t- (2650x2650)\t:\t\t0.1799314022064209\ts.\n",
      "\t- (2700x2700)\t:\t\t0.19195008277893066\ts.\n",
      "\t- (2750x2750)\t:\t\t0.19640884399414063\ts.\n",
      "\t- (2800x2800)\t:\t\t0.21481637954711913\ts.\n",
      "\t- (2850x2850)\t:\t\t0.2269810199737549\ts.\n",
      "\t- (2900x2900)\t:\t\t0.22592592239379883\ts.\n",
      "\t- (2950x2950)\t:\t\t0.24353342056274413\ts.\n",
      "\t- (3000x3000)\t:\t\t0.2369283676147461\ts.\n",
      "\t- (3050x3050)\t:\t\t0.2570371627807617\ts.\n",
      "\t- (3100x3100)\t:\t\t0.40555539131164553\ts.\n",
      "\t- (3150x3150)\t:\t\t0.2962620258331299\ts.\n",
      "\t- (3200x3200)\t:\t\t0.31392807960510255\ts.\n",
      "\t- (3250x3250)\t:\t\t0.3072353839874268\ts.\n",
      "\t- (3300x3300)\t:\t\t0.3361405849456787\ts.\n",
      "\t- (3350x3350)\t:\t\t0.37067198753356934\ts.\n",
      "\t- (3400x3400)\t:\t\t0.3832277774810791\ts.\n",
      "\t- (3450x3450)\t:\t\t0.40422840118408204\ts.\n",
      "\t- (3500x3500)\t:\t\t0.38924589157104494\ts.\n",
      "\t- (3550x3550)\t:\t\t0.4180763244628906\ts.\n",
      "\t- (3600x3600)\t:\t\t0.44397869110107424\ts.\n",
      "\t- (3650x3650)\t:\t\t0.4523275852203369\ts.\n",
      "\t- (3700x3700)\t:\t\t0.4770036220550537\ts.\n",
      "\t- (3750x3750)\t:\t\t0.4802483081817627\ts.\n",
      "\t- (3800x3800)\t:\t\t0.48940334320068357\ts.\n",
      "\t- (3850x3850)\t:\t\t0.5426775455474854\ts.\n",
      "\t- (3900x3900)\t:\t\t0.554539966583252\ts.\n",
      "\t- (3950x3950)\t:\t\t0.5691837787628173\ts.\n",
      "\t- (4000x4000)\t:\t\t0.6655501365661621\ts.\n",
      "\t- (4050x4050)\t:\t\t0.6254015922546386\ts.\n",
      "\t- (4100x4100)\t:\t\t0.6856752395629883\ts.\n",
      "\t- (4150x4150)\t:\t\t0.7106784820556641\ts.\n",
      "\t- (4200x4200)\t:\t\t0.7133233070373535\ts.\n",
      "\t- (4250x4250)\t:\t\t0.712743091583252\ts.\n",
      "\t- (4300x4300)\t:\t\t0.7346175670623779\ts.\n",
      "\t- (4350x4350)\t:\t\t0.7496411800384521\ts.\n",
      "\t- (4400x4400)\t:\t\t0.8030953884124756\ts.\n",
      "\t- (4450x4450)\t:\t\t0.8026645660400391\ts.\n",
      "\t- (4500x4500)\t:\t\t0.8242684841156006\ts.\n",
      "\t- (4550x4550)\t:\t\t0.855997896194458\ts.\n",
      "\t- (4600x4600)\t:\t\t0.8681862831115723\ts.\n",
      "\t- (4650x4650)\t:\t\t0.9324178218841552\ts.\n",
      "\t- (4700x4700)\t:\t\t0.9538925170898438\ts.\n",
      "\t- (4750x4750)\t:\t\t1.0525447845458984\ts.\n",
      "\t- (4800x4800)\t:\t\t1.0373545169830323\ts.\n",
      "\t- (4850x4850)\t:\t\t1.051058292388916\ts.\n",
      "\t- (4900x4900)\t:\t\t1.0670233726501466\ts.\n",
      "\t- (4950x4950)\t:\t\t1.0806942939758302\ts.\n",
      "\t- (5000x5000)\t:\t\t1.0815137386322022\ts.\n"
     ]
    }
   ],
   "source": [
    "results_blas = time_count_np('OpenBLAS')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<font color=\"red\"> **Modifier le kernel : MKL** </font>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "import os\n",
    "os.system(\"taskset -p 0xff %d\" % os.getpid())\n",
    "os.environ[\"MKL_MAIN_FREE\"]='1'\n",
    "os.environ[\"MKL_NUM_THREADS\"] = \"6\"\n",
    "os.environ[\"OMP_NUM_THREADS\"] = \"6\"\n",
    "import numpy as np \n",
    "from fonctions import time_count_np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mkl_info:\n",
      "    libraries = ['mkl_rt', 'pthread']\n",
      "    library_dirs = ['/home/cesar/.anaconda3/lib']\n",
      "    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]\n",
      "    include_dirs = ['/home/cesar/.anaconda3/include']\n",
      "blas_mkl_info:\n",
      "    libraries = ['mkl_rt', 'pthread']\n",
      "    library_dirs = ['/home/cesar/.anaconda3/lib']\n",
      "    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]\n",
      "    include_dirs = ['/home/cesar/.anaconda3/include']\n",
      "blas_opt_info:\n",
      "    libraries = ['mkl_rt', 'pthread']\n",
      "    library_dirs = ['/home/cesar/.anaconda3/lib']\n",
      "    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]\n",
      "    include_dirs = ['/home/cesar/.anaconda3/include']\n",
      "lapack_mkl_info:\n",
      "    libraries = ['mkl_rt', 'pthread']\n",
      "    library_dirs = ['/home/cesar/.anaconda3/lib']\n",
      "    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]\n",
      "    include_dirs = ['/home/cesar/.anaconda3/include']\n",
      "lapack_opt_info:\n",
      "    libraries = ['mkl_rt', 'pthread']\n",
      "    library_dirs = ['/home/cesar/.anaconda3/lib']\n",
      "    define_macros = [('SCIPY_MKL_H', None), ('HAVE_CBLAS', None)]\n",
      "    include_dirs = ['/home/cesar/.anaconda3/include']\n",
      "### Temps d'execution moyen d'un matrice de taille :\n",
      "\t- (50x50)\t:\t\t0.004104471206665039\ts.\n",
      "\t- (100x100)\t:\t\t6.327629089355468e-05\ts.\n",
      "\t- (150x150)\t:\t\t0.0010676383972167969\ts.\n",
      "\t- (200x200)\t:\t\t0.00032167434692382814\ts.\n",
      "\t- (250x250)\t:\t\t0.0005551815032958984\ts.\n",
      "\t- (300x300)\t:\t\t0.00016994476318359374\ts.\n",
      "\t- (350x350)\t:\t\t0.0007976055145263672\ts.\n",
      "\t- (400x400)\t:\t\t0.0013753414154052735\ts.\n",
      "\t- (450x450)\t:\t\t0.0006307125091552735\ts.\n",
      "\t- (500x500)\t:\t\t0.002046489715576172\ts.\n",
      "\t- (550x550)\t:\t\t0.0018691062927246095\ts.\n",
      "\t- (600x600)\t:\t\t0.0025100231170654295\ts.\n",
      "\t- (650x650)\t:\t\t0.0020216941833496095\ts.\n",
      "\t- (700x700)\t:\t\t0.0025557518005371095\ts.\n",
      "\t- (750x750)\t:\t\t0.004774618148803711\ts.\n",
      "\t- (800x800)\t:\t\t0.005783939361572265\ts.\n",
      "\t- (850x850)\t:\t\t0.00844879150390625\ts.\n",
      "\t- (900x900)\t:\t\t0.0056134223937988285\ts.\n",
      "\t- (950x950)\t:\t\t0.006815576553344726\ts.\n",
      "\t- (1000x1000)\t:\t\t0.008598709106445312\ts.\n",
      "\t- (1050x1050)\t:\t\t0.013880538940429687\ts.\n",
      "\t- (1100x1100)\t:\t\t0.01628537178039551\ts.\n",
      "\t- (1150x1150)\t:\t\t0.012958908081054687\ts.\n",
      "\t- (1200x1200)\t:\t\t0.026468896865844728\ts.\n",
      "\t- (1250x1250)\t:\t\t0.028520584106445312\ts.\n",
      "\t- (1300x1300)\t:\t\t0.03327999114990234\ts.\n",
      "\t- (1350x1350)\t:\t\t0.03427119255065918\ts.\n",
      "\t- (1400x1400)\t:\t\t0.036669015884399414\ts.\n",
      "\t- (1450x1450)\t:\t\t0.0420072078704834\ts.\n",
      "\t- (1500x1500)\t:\t\t0.04722785949707031\ts.\n",
      "\t- (1550x1550)\t:\t\t0.052477121353149414\ts.\n",
      "\t- (1600x1600)\t:\t\t0.05690450668334961\ts.\n",
      "\t- (1650x1650)\t:\t\t0.059403276443481444\ts.\n",
      "\t- (1700x1700)\t:\t\t0.06837606430053711\ts.\n",
      "\t- (1750x1750)\t:\t\t0.07584972381591797\ts.\n",
      "\t- (1800x1800)\t:\t\t0.07884669303894043\ts.\n",
      "\t- (1850x1850)\t:\t\t0.0837447166442871\ts.\n",
      "\t- (1900x1900)\t:\t\t0.09448156356811524\ts.\n",
      "\t- (1950x1950)\t:\t\t0.10016307830810547\ts.\n",
      "\t- (2000x2000)\t:\t\t0.11074843406677246\ts.\n",
      "\t- (2050x2050)\t:\t\t0.08475017547607422\ts.\n",
      "\t- (2100x2100)\t:\t\t0.08867425918579101\ts.\n",
      "\t- (2150x2150)\t:\t\t0.09929776191711426\ts.\n",
      "\t- (2200x2200)\t:\t\t0.10419297218322754\ts.\n",
      "\t- (2250x2250)\t:\t\t0.10832133293151855\ts.\n",
      "\t- (2300x2300)\t:\t\t0.11440234184265137\ts.\n",
      "\t- (2350x2350)\t:\t\t0.1291450023651123\ts.\n",
      "\t- (2400x2400)\t:\t\t0.13884801864624025\ts.\n",
      "\t- (2450x2450)\t:\t\t0.1396559715270996\ts.\n",
      "\t- (2500x2500)\t:\t\t0.15611791610717773\ts.\n",
      "\t- (2550x2550)\t:\t\t0.15758914947509767\ts.\n",
      "\t- (2600x2600)\t:\t\t0.16736989021301268\ts.\n",
      "\t- (2650x2650)\t:\t\t0.1740485668182373\ts.\n",
      "\t- (2700x2700)\t:\t\t0.19033727645874024\ts.\n",
      "\t- (2750x2750)\t:\t\t0.19807929992675782\ts.\n",
      "\t- (2800x2800)\t:\t\t0.2066197872161865\ts.\n",
      "\t- (2850x2850)\t:\t\t0.21204004287719727\ts.\n",
      "\t- (2900x2900)\t:\t\t0.23372859954833985\ts.\n",
      "\t- (2950x2950)\t:\t\t0.24254312515258789\ts.\n",
      "\t- (3000x3000)\t:\t\t0.25392837524414064\ts.\n",
      "\t- (3050x3050)\t:\t\t0.26154794692993166\ts.\n",
      "\t- (3100x3100)\t:\t\t0.28391156196594236\ts.\n",
      "\t- (3150x3150)\t:\t\t0.29736766815185545\ts.\n",
      "\t- (3200x3200)\t:\t\t0.296784496307373\ts.\n",
      "\t- (3250x3250)\t:\t\t0.31046347618103026\ts.\n",
      "\t- (3300x3300)\t:\t\t0.33387112617492676\ts.\n",
      "\t- (3350x3350)\t:\t\t0.34405155181884767\ts.\n",
      "\t- (3400x3400)\t:\t\t0.35758562088012696\ts.\n",
      "\t- (3450x3450)\t:\t\t0.36993632316589353\ts.\n",
      "\t- (3500x3500)\t:\t\t0.4019219398498535\ts.\n",
      "\t- (3550x3550)\t:\t\t0.4272460460662842\ts.\n",
      "\t- (3600x3600)\t:\t\t0.4252453327178955\ts.\n",
      "\t- (3650x3650)\t:\t\t0.4549867630004883\ts.\n",
      "\t- (3700x3700)\t:\t\t0.47396340370178225\ts.\n",
      "\t- (3750x3750)\t:\t\t0.4787610530853271\ts.\n",
      "\t- (3800x3800)\t:\t\t0.48857502937316893\ts.\n",
      "\t- (3850x3850)\t:\t\t0.5262506008148193\ts.\n",
      "\t- (3900x3900)\t:\t\t0.5404400825500488\ts.\n",
      "\t- (3950x3950)\t:\t\t0.5474565029144287\ts.\n",
      "\t- (4000x4000)\t:\t\t0.562882661819458\ts.\n",
      "\t- (4050x4050)\t:\t\t0.6240564346313476\ts.\n",
      "\t- (4100x4100)\t:\t\t0.602271556854248\ts.\n",
      "\t- (4150x4150)\t:\t\t0.6501219272613525\ts.\n",
      "\t- (4200x4200)\t:\t\t0.6553904056549072\ts.\n",
      "\t- (4250x4250)\t:\t\t0.6962255477905274\ts.\n",
      "\t- (4300x4300)\t:\t\t0.7364336490631104\ts.\n",
      "\t- (4350x4350)\t:\t\t0.7024241924285889\ts.\n",
      "\t- (4400x4400)\t:\t\t0.7453198432922363\ts.\n",
      "\t- (4450x4450)\t:\t\t0.7949317932128906\ts.\n",
      "\t- (4500x4500)\t:\t\t0.8423180103302002\ts.\n",
      "\t- (4550x4550)\t:\t\t0.8335808277130127\ts.\n",
      "\t- (4600x4600)\t:\t\t0.8343969821929932\ts.\n",
      "\t- (4650x4650)\t:\t\t0.9126011371612549\ts.\n",
      "\t- (4700x4700)\t:\t\t0.9434250831604004\ts.\n",
      "\t- (4750x4750)\t:\t\t0.964992618560791\ts.\n",
      "\t- (4800x4800)\t:\t\t0.9596626281738281\ts.\n",
      "\t- (4850x4850)\t:\t\t1.04030499458313\ts.\n",
      "\t- (4900x4900)\t:\t\t1.049760913848877\ts.\n",
      "\t- (4950x4950)\t:\t\t1.0362091541290284\ts.\n",
      "\t- (5000x5000)\t:\t\t1.0932655334472656\ts.\n"
     ]
    }
   ],
   "source": [
    "results_mkl = time_count_np('MKL')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Graphique "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "import os \n",
    "PATH = os.path.abspath('')\n",
    "results_blis = pickle.load(open(PATH+\"/data/\"+'BLIS'+\".pkl\",\"rb\"))\n",
    "x_blis, y_blis = zip(*sorted(results_blis.items()))\n",
    "results_blas = pickle.load(open(PATH+\"/data/\"+'OpenBLAS'+\".pkl\",\"rb\"))\n",
    "x_blas, y_blas = zip(*sorted(results_blas.items()))\n",
    "results_mkl = pickle.load(open(PATH+\"/data/\"+'MKL'+\".pkl\",\"rb\"))\n",
    "x_mkl, y_mkl = zip(*sorted(results_mkl.items()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt \n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(5,5))\n",
    "sns.set()\n",
    "plt.plot(x_blas,y_blas,label='numpy_blas',color='black',linestyle='--')\n",
    "plt.plot(x_mkl,y_mkl,label='numpy_mkl',color='black')\n",
    "plt.plot(x_blis,y_blis,label='numpy_blis',linewidth=2,color='red')\n",
    "plt.xlabel('Taille de la matrice (m=n)')\n",
    "plt.ylabel('Temps moyen (s)')\n",
    "#plt.title(\"Comparatif temps d'execution sur le produit matriciel (m=n), NUM_THREADS=2\")\n",
    "plt.legend()\n",
    "plt.savefig(PATH+\"/fig/comparaison.pdf\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}