cudaBlas-no-PCA

gistfile1.txt

{
 "metadata": {
  "name": "cuBLAS-no-PCA"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "import pycuda.gpuarray as gpuarray\nimport pycuda.autoinit\nimport numpy as np\nfrom scikits.cuda import linalg, misc\nlinalg.init()",
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 1
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": "a_dims = [(20000, 100000), (2000, 800000), (20000, 32768)]\nb_dims = [(100000, 1), (800000, 1), (32768, 5000)]\n\nfor a_dim, b_dim in zip(a_dims, b_dims):\n    print a_dim, b_dim\n    a = np.asarray(np.random.rand(a_dim[0], a_dim[1]), np.float32)\n    b = np.asarray(np.random.rand(b_dim[0], b_dim[1]), np.float32)\n    print 'time to transfer A into GPU memory'\n    %timeit  a_gpu = gpuarray.to_gpu(a)\n    print 'time to transfer B into GPU memory'\n    %timeit b_gpu = gpuarray.to_gpu(b)\n    a_gpu = gpuarray.to_gpu(a)\n    b_gpu = gpuarray.to_gpu(b)\n    print 'time for C = A*B'\n    %timeit c_gpu = linalg.dot(a_gpu, b_gpu)\n    c_gpu = linalg.dot(a_gpu, b_gpu)\n    print 'time for getting C back into memory'\n    %timeit c = c_gpu.get()\n    del a_gpu, b_gpu, c_gpu",
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "(20000, 100000) (100000, 1)\ntime to transfer A into GPU memory"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n1 loops, best of 3: 1.24 s per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime to transfer B into GPU memory\n10000 loops, best of 3: 136 us per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime for C = A*B"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n10 loops, best of 3: 149 ms per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime for getting C back into memory\n10000 loops, best of 3: 37 us per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n(2000, 800000) (800000, 1)\ntime to transfer A into GPU memory"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n1 loops, best of 3: 994 ms per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime to transfer B into GPU memory\n1000 loops, best of 3: 734 us per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime for C = A*B"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n10 loops, best of 3: 1.06 s per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime for getting C back into memory\n1 loops, best of 3: 20 us per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n(20000, 32768) (32768, 5000)\ntime to transfer A into GPU memory"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n1 loops, best of 3: 403 ms per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime to transfer B into GPU memory\n10 loops, best of 3: 102 ms per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime for C = A*B"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n10 loops, best of 3: 2.14 s per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\ntime for getting C back into memory\n1 loops, best of 3: 183 ms per loop"
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": "\n"
      }
     ],
     "prompt_number": 14
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": ""
    }
   ],
   "metadata": {}
  }
 ]
}