Section_3_Memory_Management.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numba import vectorize, cuda\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The following attriutes will be used in this Section!\n",
    "- cuda.device_array\n",
    "- cuda.copy_to_host()\n",
    "- cuda.to_device()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Managing GPU Memory"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "During the benchmarking in the previous notebook, we used NumPy arrays on the CPU as inputs and outputs. If you want to reduce the impact of host-to-device/device-to-host bandwidth, <b>it is best to copy data to the GPU explicitly</b> and leave it there to amortize the cost over multiple function calls. In addition, allocating device memory can be relatively slow, so <b> allocating GPU arrays once and refilling them with data from the host can also be a performance improvement. </b>\n",
    "\n",
    "Let's create our example addition ufunc again:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Simple function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "@vectorize(['float32(float32, float32)'], target='cuda') # ???\n",
    "def add_ufunc(x, y):\n",
    "    return x + y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "n = 100000\n",
    "x = np.arange(n).astype(np.float32)\n",
    "y = 2 * x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.58 ms ± 595 µs per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "%timeit add_ufunc(x, y)  # Baseline performance with host arrays"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(cuda.gpus)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "b'GeForce GTX 1050 Ti'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cuda.gpus[0].name"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<b>The numba.cuda module includes a function that will copy host data to the GPU and return a CUDA device array:</b>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<numba.cuda.cudadrv.devicearray.DeviceNDArray object at 0x000001550875FC88>\n",
      "(100000,)\n",
      "float32\n"
     ]
    }
   ],
   "source": [
    "x_device = cuda.to_device(x)\n",
    "y_device = cuda.to_device(y)\n",
    "\n",
    "print(x_device)\n",
    "print(x_device.shape)\n",
    "print(x_device.dtype)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Device arrays can be passed to CUDA functions just like NumPy arrays, but without the copy overhead:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "829 µs ± 61.7 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
     ]
    }
   ],
   "source": [
    "%timeit add_ufunc(x_device, y_device) # it is faster corresponding to the first case"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "That's a big performance improvement already, but we are <b>still allocating a device array for the output of the ufunc and copying it back to the host</b>. We can create the output buffer with the numba.cuda.device_array() function:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "out_device = cuda.device_array(shape=(n,), dtype=np.float32)  # does not initialize the contents, like np.empty()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "688 µs ± 59.4 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
     ]
    }
   ],
   "source": [
    "%timeit add_ufunc(x_device, y_device, out=out_device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now that we <b>have removed the device allocation and copy steps</b>, the computation runs much faster than before. When we want to bring the device array back to the host memory, we can use the <b>copy_to_host()</b> method"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.  3.  6.  9. 12. 15. 18. 21. 24. 27.]\n",
      "(100000,)\n"
     ]
    }
   ],
   "source": [
    "out_host = out_device.copy_to_host()\n",
    "print(out_host[:10])\n",
    "print(out_host.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<numba.cuda.cudadrv.devicearray.DeviceNDArray object at 0x0000015508734240>\n",
      "(100000,)\n"
     ]
    }
   ],
   "source": [
    "print(out_device[:10]) # How to watch it?\n",
    "print(out_device.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Exercise"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "\n",
    "@vectorize(['float32(float32, float32, float32)'], target='cuda')\n",
    "def make_pulses(i, period, amplitude):\n",
    "    return max(math.sin(i / period) - 0.3, 0.0) * amplitude\n",
    "\n",
    "n = 100000\n",
    "noise = (np.random.normal(size=n) * 3).astype(np.float32)\n",
    "t = np.arange(n, dtype=np.float32)\n",
    "period = n / 23"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Convert this code to use device allocations so that there are only host<->device copies at the beginning and end and benchmark performance change:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "pulses = make_pulses(t, period, 100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "waveform = add_ufunc(pulses, noise) # It uses GPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x15508780240>]"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "from matplotlib import pyplot as plt\n",
    "plt.plot(waveform)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "GPU allocation, not complited?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "pulse_device = cuda.to_device(pulses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "out_device = cuda.device_array(shape=(n,), dtype=np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<numba.cuda.cudadrv.devicearray.DeviceNDArray at 0x15508761908>"
      ]
     },
     "execution_count": 87,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "add_ufunc(x_device, y_device, out=out_device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "out_host = out_device.copy_to_host()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0.,  3.,  6.,  9., 12., 15., 18., 21., 24., 27.], dtype=float32)"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "out_host[0:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Full GPU allocation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "No matching version.  GPU ufunc requires array arguments to have the exact types.  This behaves like regular ufunc with casting='no'.",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-94-8caa585863af>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[0mperiod_device\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcuda\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mto_device\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mperiod\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[0mamp_device\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcuda\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mto_device\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m100.0\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 5\u001b[1;33m \u001b[0mmake_pulses\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mt_device\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mperiod_device\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mamp_device\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mout\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mpulse_device\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# Change types of arguments according to function\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\numba\\cuda\\dispatcher.py\u001b[0m in \u001b[0;36m__call__\u001b[1;34m(self, *args, **kws)\u001b[0m\n\u001b[0;32m     86\u001b[0m                       \u001b[0mthe\u001b[0m \u001b[0minput\u001b[0m \u001b[0marguments\u001b[0m\u001b[1;33m.\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     87\u001b[0m         \"\"\"\n\u001b[1;32m---> 88\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mCUDAUFuncMechanism\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcall\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfunctions\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0margs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mkws\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     89\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     90\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mreduce\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0marg\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mstream\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\numba\\npyufunc\\deviceufunc.py\u001b[0m in \u001b[0;36mcall\u001b[1;34m(cls, typemap, args, kws)\u001b[0m\n\u001b[0;32m    254\u001b[0m         \u001b[1;31m# Begin call resolution\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    255\u001b[0m         \u001b[0mcr\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcls\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtypemap\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 256\u001b[1;33m         \u001b[0margs\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcr\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_arguments\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    257\u001b[0m         \u001b[0mresty\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfunc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcr\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_function\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    258\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\numba\\npyufunc\\deviceufunc.py\u001b[0m in \u001b[0;36mget_arguments\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    204\u001b[0m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fill_arrays\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    205\u001b[0m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_fill_argtypes\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 206\u001b[1;33m         \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_resolve_signature\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    207\u001b[0m         \u001b[0marys\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_actual_args\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    208\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_broadcast\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0marys\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\lib\\site-packages\\numba\\npyufunc\\deviceufunc.py\u001b[0m in \u001b[0;36m_resolve_signature\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    144\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    145\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mmatches\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 146\u001b[1;33m             raise TypeError(\"No matching version.  GPU ufunc requires array \"\n\u001b[0m\u001b[0;32m    147\u001b[0m                             \u001b[1;34m\"arguments to have the exact types.  This behaves \"\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    148\u001b[0m                             \"like regular ufunc with casting='no'.\")\n",
      "\u001b[1;31mTypeError\u001b[0m: No matching version.  GPU ufunc requires array arguments to have the exact types.  This behaves like regular ufunc with casting='no'."
     ]
    }
   ],
   "source": [
    "pulse_device = cuda.device_array(shape=(n,), dtype=np.float32)\n",
    "t_device = cuda.to_device(t)\n",
    "period_device = cuda.to_device(period)\n",
    "amp_device = cuda.to_device(100.0)\n",
    "make_pulses(t_device, period_device, amp_device, out=pulse_device) # Change types of arguments according to function"
   ]
  }
 ],
 "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": "3.5.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}