Some PyCUDA examples

pycuda_article.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 1.17558968  0.26406515  0.80357075 -1.79517698]\n",
      " [-2.23722863  1.30630934 -0.92374867  0.92402256]\n",
      " [ 1.09881485  0.3819513  -0.02121204  0.18705048]\n",
      " [-1.19093406  1.9892801   0.13481987 -2.36596513]]\n",
      "[[ 2.  2.  2.  2.]\n",
      " [ 2.  2.  2.  2.]\n",
      " [ 2.  2.  2.  2.]\n",
      " [ 2.  2.  2.  2.]]\n",
      "[[ 0.58779484  0.13203257  0.40178537 -0.89758849]\n",
      " [-1.11861432  0.65315467 -0.46187434  0.46201128]\n",
      " [ 0.54940742  0.19097565 -0.01060602  0.09352524]\n",
      " [-0.59546703  0.99464005  0.06740993 -1.18298256]]\n"
     ]
    }
   ],
   "source": [
    "import pycuda.driver as cuda\n",
    "import pycuda.autoinit\n",
    "import numpy as np\n",
    "import pycuda.compiler\n",
    "\n",
    "a = np.random.randn(4, 4).astype(np.float32)\n",
    "a_gpu = cuda.mem_alloc(a.nbytes)\n",
    "cuda.memcpy_htod(a_gpu, a) #host to device\n",
    "\n",
    "\n",
    "mod = pycuda.compiler.SourceModule(\"\"\"\n",
    "    __global__ void multiply_by_two(float *a)\n",
    "    {\n",
    "        int idx = threadIdx.x + threadIdx.y * 4;\n",
    "        a[idx] *= 2;\n",
    "    }\n",
    "    \"\"\")\n",
    "\n",
    "func = mod.get_function(\"multiply_by_two\")\n",
    "func(a_gpu, block=(4,4,1))\n",
    "a_doubled = np.empty_like(a)\n",
    "cuda.memcpy_dtoh(a_doubled, a_gpu) #device to host\n",
    "print(a_doubled, a_doubled/a, a, sep='\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 0.74762076 -1.6985631  -1.50394118  0.92405546]\n",
      " [-4.40119934 -0.40460199  3.72085404 -0.54394537]\n",
      " [ 1.6078819  -1.56536436 -1.75098097 -2.33173347]\n",
      " [ 2.76836824 -0.34919369  0.37828121 -0.55695456]]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pycuda.autoinit\n",
    "import pycuda.gpuarray as gpuarray\n",
    "\n",
    "a_gpu = gpuarray.to_gpu(\n",
    "            np.random.randn(4,4).astype(np.float32)\n",
    "            )\n",
    "a_doubled = (2*a_gpu).get()\n",
    "print(a_doubled)"
   ]
  }
 ],
 "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": "3.5.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}