docs/performance/speed-test-prime.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "subjective-appreciation",
   "metadata": {},
   "source": [
    "# GF(p) speed tests"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "hidden-costa",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import galois"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "uniform-accuracy",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GF(8009):\n",
      "  characteristic: 8009\n",
      "  degree: 1\n",
      "  order: 8009\n",
      "  irreducible_poly: Poly(x + 8006, GF(8009))\n",
      "  is_primitive_poly: True\n",
      "  primitive_element: GF(3, order=8009)\n",
      "  dtypes: ['uint16', 'uint32', 'int16', 'int32', 'int64']\n",
      "  ufunc_mode: 'jit-lookup'\n",
      "  ufunc_target: 'cpu'\n"
     ]
    }
   ],
   "source": [
    "prime = galois.next_prime(8000)\n",
    "\n",
    "GF = galois.GF(prime)\n",
    "print(GF.properties)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "later-convention",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['jit-lookup', 'jit-calculate']\n",
      "['cpu', 'parallel']\n"
     ]
    }
   ],
   "source": [
    "modes = GF.ufunc_modes\n",
    "targets = GF.ufunc_targets\n",
    "targets.remove(\"cuda\")  # Can't test with a GPU on my machine\n",
    "print(modes)\n",
    "print(targets)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "described-placement",
   "metadata": {},
   "outputs": [],
   "source": [
    "def speed_test(GF, N):\n",
    "    a = GF.Random(N)\n",
    "    b = GF.Random(N, low=1)\n",
    "\n",
    "    for operation in [np.add, np.multiply]:\n",
    "        print(f\"Operation: {operation.__name__}\")\n",
    "        for target in targets:\n",
    "            for mode in modes:\n",
    "                GF.compile(mode, target)\n",
    "                print(f\"Target: {target}, Mode: {mode}\", end=\"\\n    \")\n",
    "                %timeit operation(a, b)\n",
    "        print()\n",
    "\n",
    "    for operation in [np.reciprocal, np.log]:\n",
    "        print(f\"Operation: {operation.__name__}\")\n",
    "        for target in targets:\n",
    "            for mode in modes:\n",
    "                GF.compile(mode, target)\n",
    "                print(f\"Target: {target}, Mode: {mode}\", end=\"\\n    \")\n",
    "                %timeit operation(b)\n",
    "        print()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "saving-housing",
   "metadata": {},
   "source": [
    "## N = 10k"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "superb-disposal",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Operation: add\n",
      "Target: cpu, Mode: jit-lookup\n",
      "    104 µs ± 1.57 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: cpu, Mode: jit-calculate\n",
      "    71.3 µs ± 2.95 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: parallel, Mode: jit-lookup\n",
      "    The slowest run took 436.22 times longer than the fastest. This could mean that an intermediate result is being cached.\n",
      "10.2 ms ± 18.6 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)\n",
      "Target: parallel, Mode: jit-calculate\n",
      "    The slowest run took 24.46 times longer than the fastest. This could mean that an intermediate result is being cached.\n",
      "3.41 ms ± 2.38 ms per loop (mean ± std. dev. of 7 runs, 1000 loops each)\n",
      "\n",
      "Operation: multiply\n",
      "Target: cpu, Mode: jit-lookup\n",
      "    93.1 µs ± 1.33 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: cpu, Mode: jit-calculate\n",
      "    72.1 µs ± 1.8 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: parallel, Mode: jit-lookup\n",
      "    163 µs ± 19.9 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: parallel, Mode: jit-calculate\n",
      "    2.5 ms ± 680 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "\n",
      "Operation: reciprocal\n",
      "Target: cpu, Mode: jit-lookup\n",
      "    67.3 µs ± 929 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: cpu, Mode: jit-calculate\n",
      "    6.01 ms ± 61.4 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n",
      "Target: parallel, Mode: jit-lookup\n",
      "    152 µs ± 15.9 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: parallel, Mode: jit-calculate\n",
      "    11.1 ms ± 1.05 ms per loop (mean ± std. dev. of 7 runs, 100 loops each)\n",
      "\n",
      "Operation: log\n",
      "Target: cpu, Mode: jit-lookup\n",
      "    75.3 µs ± 242 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: cpu, Mode: jit-calculate\n",
      "    149 ms ± 846 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n",
      "Target: parallel, Mode: jit-lookup\n",
      "    175 µs ± 16.4 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n",
      "Target: parallel, Mode: jit-calculate\n",
      "    56.2 ms ± 20.1 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "speed_test(GF, 10_000)"
   ]
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}