cuBLAS speed test

cuBLAS.ipynb

{
 "metadata": {
  "name": "cuBLAS speed test"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "import pycuda.gpuarray as gpuarray\n",
      "import pycuda.autoinit\n",
      "import numpy as np\n",
      "from scikits.cuda import linalg, misc\n",
      "linalg.init()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 1
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "a = np.asarray(np.random.rand(10000, 10000), np.float32)\n",
      "b = np.asarray(np.random.rand(20000, 10000), np.float32)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 2
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Next, test the time to transfer data from main memory into GPU memory"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%timeit -n 10 a_gpu = gpuarray.to_gpu(a)\n",
      "%timeit -n 10 b_gpu = gpuarray.to_gpu(b.T)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "10 loops, best of 3: 62.5 ms per loop\n",
        "10 loops, best of 3: 124 ms per loop"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      }
     ],
     "prompt_number": 3
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "a_gpu = gpuarray.to_gpu(a)\n",
      "b_gpu = gpuarray.to_gpu(b.T)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 4
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Then, I test the time to compute dot product between 10k points and 20k points with 10k dims"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%timeit -n 10 c_gpu = linalg.dot(a_gpu, b_gpu)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "10 loops, best of 3: 1.33 s per loop\n"
       ]
      }
     ],
     "prompt_number": 5
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Finally, I test the time to get the distance back into main memory"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "c_gpu = linalg.dot(a_gpu, b_gpu)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 6
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%timeit -n 10 c = c_gpu.get()"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "10 loops, best of 3: 388 ms per loop\n"
       ]
      }
     ],
     "prompt_number": 7
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "print c_gpu.nbytes"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "800000000\n"
       ]
      }
     ],
     "prompt_number": 10
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": []
    }
   ],
   "metadata": {}
  }
 ]
}