Simplest Jupyter notebook that demonstrates GPU vs CPU performance possibilities - Sorting an array - Created while learning in NVidia AI Workbench

numpy_cupy_sort.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "610ee839-1c46-4ce3-bec0-935eeb9524b8",
   "metadata": {},
   "source": [
    "# Sorting an Array\n",
    "Code from [this medium article](https://medium.com/@geminae.stellae/introduction-to-gpu-programming-with-python-cuda-577bfdaa47f3)\n",
    "\n",
    "Tested with the PyTorch NVIDIA image because it already has cupy installed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "06460746-3590-4da7-bec7-57622e636003",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "# create the array - 67,108,864 elements\n",
    "size = 8192 * 8192\n",
    "array = np.random.random(size).astype(np.float32)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e6fc99e8-2ca9-40d1-bbb8-f282683456c4",
   "metadata": {},
   "source": [
    "# CPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "44f38e11-d4ec-4edb-b11b-93972207232b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6.24 s ± 0 ns per loop (mean ± std. dev. of 1 run, 1 loop each)\n"
     ]
    }
   ],
   "source": [
    "# Time sorting the array using the CPU with \n",
    "%timeit -n 1 -r 1 result = np.sort(array)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eacc0ae5-4a95-4b12-a275-e48fb86b9899",
   "metadata": {},
   "source": [
    "# GPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b4e036f1-383a-4e4d-a140-014de167519c",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import cupy as cp\n",
    "# cp.show_config()\n",
    "array_gpu = cp.asarray(array)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "3646ba15-e068-44a8-8b2d-8a1d1b65d2d3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "34.8 ms ± 0 ns per loop (mean ± std. dev. of 1 run, 7 loops each)\n"
     ]
    }
   ],
   "source": [
    "# Time sorting the gpu based array\n",
    "%timeit -n 7 -r 1 result_gpu = cp.sort(array_gpu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "279cbad0-7e59-4832-a6ec-8a5b126d0955",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}