CPU vs GPU performance.ipynb

{
 "cells": [
  {
   "cell_type": "code",

   "execution_count": 1,


   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [

      "Todos devices: {'/gpu:0': [], '/cpu:0': []}\n",

      "####### Calculating on the /cpu:0 #######\n",


      "{'/cpu:0': [0.7611141204833984, 0.9495096206665039, 1.2992866039276123, 1.407411813735962, 2.224921464920044, 2.779550313949585, 3.4322919845581055]}\n"

     ]
    },
    {
     "data": {

      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAcEAAAEWCAYAAAAegCx/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTIr2AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1bc0329b518>"

     ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from __future__ import print_function\n",
    "import matplotlib\n",
    "import tensorflow as tf\n",
    "import time\n",
    "from tensorflow.python.client import device_lib\n",
    "import numpy as np \n",
    "import matplotlib.pyplot as pplot\n",
    "\n",
    "def get_times(devs, matrix_sizes):\n",
    "    device_times = dict()\n",
    "    \n",
    "    for d in devs:\n",
    "        device_times[d] = []\n",
    "\n",
    "    print(\"Todos devices: %s\" % device_times)\n",
    "    \n",
    "    for device_name in device_times.keys():\n",
    "        print(\"####### Calculating on the \" + device_name + \" #######\")\n",
    "        for size in matrix_sizes:\n",
    "            shape = (size,size)\n",
    "            data_type = tf.float16\n",
    "            with tf.device(device_name):\n",
    "                r1 = tf.random_uniform(shape=shape, minval=0, maxval=1, dtype=data_type)\n",
    "                r2 = tf.random_uniform(shape=shape, minval=0, maxval=1, dtype=data_type)\n",
    "                dot_operation = tf.matmul(r2, r1)\n",
    "                \n",
    "            with tf.Session(config=tf.ConfigProto(log_device_placement=True)) as session:\n",
    "                    start_time = time.time()\n",
    "                    result = session.run(dot_operation)\n",
    "                    time_taken = time.time() - start_time\n",
    "                    device_times[device_name].append(time_taken)\n",
    "#                     print(\"device: %s Size: %s Time: %s\" % (device_name,size, time_taken))\n",
    "    \n",
    "    return device_times\n",
    "\n",
    "# Lista devices disponíveis:\n",
    "# \"/cpu:0\": The CPU of your machine.\n",
    "# \"/gpu:0\": The GPU of your machine, if you have one.\n",
    "# \"/gpu:1\": The second GPU of your machine, etc.\n",
    "    \n",
    "local_devices = device_lib.list_local_devices()\n",
    "devices = [x.name for x in local_devices]\n",
    "matrix_sizes = range(500,850,50)\n",
    "device_times = get_times(devices, matrix_sizes)\n",
    "print(device_times)\n",
    "\n",
    "# Plota resultados\n",
    "pplot.figure()\n",
    "pplot.title(\"Mulitplication Time\")\n",
    "pplot.xlabel('Matrix size')\n",
    "pplot.ylabel('Time')\n",
    "pplot.grid()\n",
    "for d in device_times:\n",
    "    pplot.plot(matrix_sizes[:len(device_times[d])], device_times[d], 'o-', label=d)\n",
    "pplot.legend(loc='center left', bbox_to_anchor=(1, 0.5))\n",
    "pplot.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "####### Calculating #######\n",
      "Size: 500 Time: 1.3599655628204346\n",
      "Size: 550 Time: 1.8012802600860596\n",
      "Size: 600 Time: 2.4217376708984375\n",
      "Size: 650 Time: 3.029149293899536\n",
      "Size: 700 Time: 3.670607089996338\n",
      "Size: 750 Time: 4.531220436096191\n",
      "Size: 800 Time: 5.456874132156372\n"
     ]
    }
   ],
   "source": [
    "# Testa no device default, sem espeficiar qual device será usado na Session\n",
    "from __future__ import print_function\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "import time\n",
    "from tensorflow.python.client import device_lib\n",
    "import numpy as np \n",
    "\n",
    "def get_times(matrix_sizes):\n",
    "    device_times = dict()\n",
    "    device_times[\"teste\"] = []\n",
    "\n",
    "    print(\"####### Calculating #######\")\n",
    "    for size in matrix_sizes:\n",
    "        shape = (size,size)\n",
    "        data_type = tf.float16\n",
    "        r1 = tf.random_uniform(shape=shape, minval=0, maxval=1, dtype=data_type)\n",
    "        r2 = tf.random_uniform(shape=shape, minval=0, maxval=1, dtype=data_type)\n",
    "        dot_operation = tf.matmul(r2, r1)\n",
    "\n",
    "        with tf.Session() as session:\n",
    "            start_time = time.time()\n",
    "            result = session.run(dot_operation)\n",
    "            time_taken = time.time() - start_time\n",
    "            device_times[\"teste\"].append(time_taken)\n",
    "            print(\"Size: %s Time: %s\" % (size, time_taken))\n",
    "\n",
    "    return device_times\n",
    "\n",
    "matrix_sizes = range(500,850,50)\n",
    "device_times = get_times(matrix_sizes)\n"
   ]
  }
 ],
 "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": "3.6.1"
   }
   },
 "nbformat": 4,
 "nbformat_minor": 2
}