{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Performance of Randperm on CPU/GPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Node Random perm on host: host duration 6384.1ms, device duration 6384.4ms\n",
      "Node Random perm on device: host duration 2525.0ms, device duration 2525.0ms\n"
     ]
    }
   ],
   "source": [
    "from common import *\n",
    "from utils.profile import debug_profile\n",
    "from utils.mem_profiler import MemProfiler\n",
    "\n",
    "with debug_profile(\"Random perm on host\"):\n",
    "    torch.randperm(1024 * 1024 * 100)\n",
    "\n",
    "with debug_profile(\"Random perm on device\"),\\\n",
    "        MemProfiler(\"Random perm on host\", device=\"cuda:3\"):\n",
    "    torch.randperm(1024 * 1024 * 100, device=\"cuda:3\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# \\_\\_getattribute\\_\\_ Method"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "a.a: 1\n",
      "a.b: 2\n"
     ]
    },
    {
     "ename": "AttributeError",
     "evalue": "'A' object has no attribute 'c'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mAttributeError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-3-bdb259af4410>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     24\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"a.a:\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0ma\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     25\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"a.b:\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 26\u001b[0;31m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"a.c:\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mc\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m<ipython-input-3-bdb259af4410>\u001b[0m in \u001b[0;36m__getattribute__\u001b[0;34m(self, _A__name)\u001b[0m\n\u001b[1;32m     16\u001b[0m                 \u001b[0;32mpass\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     17\u001b[0m             \u001b[0merr\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 18\u001b[0;31m         \u001b[0;32mraise\u001b[0m \u001b[0merr\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     19\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     20\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-3-bdb259af4410>\u001b[0m in \u001b[0;36m__getattribute__\u001b[0;34m(self, _A__name)\u001b[0m\n\u001b[1;32m      9\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__getattribute__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0m__name\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mstr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     10\u001b[0m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 11\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__getattribute__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m__name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     12\u001b[0m         \u001b[0;32mexcept\u001b[0m \u001b[0mAttributeError\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     13\u001b[0m             \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mAttributeError\u001b[0m: 'A' object has no attribute 'c'"
     ]
    }
   ],
   "source": [
    "class A(object):\n",
    "\n",
    "\n",
    "    def __init__(self, a, **extra) -> None:\n",
    "        super().__init__()\n",
    "        self.a = a\n",
    "        self.extra = extra\n",
    "\n",
    "    def __getattribute__(self, __name: str):\n",
    "        try:\n",
    "            return super().__getattribute__(__name)\n",
    "        except AttributeError as e:\n",
    "            try:\n",
    "                return self.extra[__name]\n",
    "            except KeyError:\n",
    "                pass\n",
    "            err = e\n",
    "        raise err\n",
    "\n",
    "\n",
    "a = A(a=1, b=2)\n",
    "\n",
    "\n",
    "print(\"a.a:\", a.a)\n",
    "print(\"a.b:\", a.b)\n",
    "print(\"a.c:\", a.c)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Performance of Various Select/Scatter Methods"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.010159730911254883 tensor(100., device='cuda:0')\n",
      "0.011237859725952148 tensor(1400., device='cuda:0')\n",
      "0.032263755798339844 tensor(2700., device='cuda:0')\n",
      "0.02148723602294922 tensor(4.1723e-07, device='cuda:0')\n",
      "0.009927511215209961 tensor(4.1723e-07, device='cuda:0')\n",
      "Mask set 0.02173590660095215\n",
      "Inplace mask scatter 0.0041882991790771484\n",
      "Mask scatter 0.00580906867980957\n",
      "Index set 0.03358888626098633\n",
      "Index put 0.01044917106628418\n"
     ]
    }
   ],
   "source": [
    "from common import *\n",
    "from time import time\n",
    "\n",
    "a = torch.zeros(200000, device=\"cuda\")\n",
    "i = torch.randint(0, a.shape[0], [500000], device=\"cuda\")\n",
    "start = time()\n",
    "for _ in range(100):\n",
    "    a[i] += 1\n",
    "end = time()\n",
    "print(end - start, a.max())\n",
    "start = time()\n",
    "for _ in range(100):\n",
    "    a.index_add_(0, i, torch.ones_like(i, dtype=torch.float))\n",
    "end = time()\n",
    "print(end - start, a.max())\n",
    "\n",
    "start = time()\n",
    "for _ in range(100):\n",
    "    ui, n = i.unique(return_counts=True)\n",
    "    a[ui] += n\n",
    "end = time()\n",
    "print(end - start, a.max())\n",
    "\n",
    "\n",
    "a = torch.rand(2000, 2000, device=\"cuda\") - .5\n",
    "m = a > 0\n",
    "\n",
    "start = time()\n",
    "for _ in range(100):\n",
    "    b = a[m]\n",
    "end = time()\n",
    "print(end - start, b.min())\n",
    "\n",
    "start = time()\n",
    "for _1 in range(20):\n",
    "    m1 = m.nonzero(as_tuple=True)\n",
    "    for _ in range(5):\n",
    "        b = a[m1]\n",
    "end = time()\n",
    "print(end - start, b.min())\n",
    "\n",
    "\n",
    "c = torch.rand_like(b)\n",
    "\n",
    "start = time()\n",
    "for _ in range(100):\n",
    "    a[m] = c\n",
    "end = time()\n",
    "print(\"Mask set\", end - start)\n",
    "\n",
    "start = time()\n",
    "for _ in range(100):\n",
    "    a.masked_scatter_(m, c)\n",
    "end = time()\n",
    "print(\"Inplace mask scatter\", end - start)\n",
    "\n",
    "\n",
    "start = time()\n",
    "for _ in range(100):\n",
    "    a = a.masked_scatter(m, c)\n",
    "end = time()\n",
    "print(\"Mask scatter\", end - start)\n",
    "\n",
    "start = time()\n",
    "for _1 in range(20):\n",
    "    m1 = m.nonzero(as_tuple=True)\n",
    "    for _ in range(5):\n",
    "        a[m1] = b\n",
    "end = time()\n",
    "print(\"Index set\", end - start)\n",
    "\n",
    "\n",
    "start = time()\n",
    "for _1 in range(20):\n",
    "    m1 = m.nonzero(as_tuple=True)\n",
    "    for _ in range(5):\n",
    "        a.index_put_(m1, b)\n",
    "end = time()\n",
    "print(\"Index put\", end - start)"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "65406b00395a48e1d89cf658ae895e7869e05878f5469716b06a752a3915211c"
  },
  "kernelspec": {
   "display_name": "Python 3.8.12 ('base')",
   "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": "3.8.12"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}