benchmarks_sampling_only.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import rmm\n",
    "\n",
    "rmm.reinitialize(pool_allocator=True,initial_pool_size=5e+9, maximum_pool_size=20e+9)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cudf\n",
    "import cugraph\n",
    "from cugraph.experimental import PropertyGraph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cugraph\n",
    "import cudf\n",
    "from ogb.nodeproppred import NodePropPredDataset\n",
    "\n",
    "dataset = NodePropPredDataset(name = 'ogbn-mag', root='/work/dataset/') \n",
    "\n",
    "data = dataset[0]\n",
    "\n",
    "edgelist = cudf.DataFrame()\n",
    "for elt in list(data[0]['edge_index_dict'].values()):\n",
    "    elt_df = cudf.DataFrame({\n",
    "        'src': cudf.from_dlpack(elt[0].__dlpack__()),\n",
    "        'dst': cudf.from_dlpack(elt[1].__dlpack__())\n",
    "    })\n",
    "    edgelist = cudf.concat([edgelist, elt_df])\n",
    "edgelist['weight']=1.\n",
    "\n",
    "sg = cugraph.Graph()\n",
    "sg.from_cudf_edgelist(edgelist, source='src', destination='dst', edge_attr='weight', legacy_renum_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def exec_gpu_cugraph(batch, fanout):\n",
    "    cugraph.uniform_neighbor_sample(sg, batch, [fanout,fanout], True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from datetime import datetime\n",
    "import numpy as np\n",
    "import cupy\n",
    "from time import sleep\n",
    "\n",
    "batch_size = 1000\n",
    "tries = 10\n",
    "\n",
    "res_cugraph = np.empty(99)\n",
    "\n",
    "for i, fanout in enumerate(range(1,100)):\n",
    "    batch = cudf.Series(cupy.random.randint(low=0, high=sg.number_of_vertices(), size=batch_size))\n",
    "    start_time = datetime.now()\n",
    "    for _ in range(tries):\n",
    "        exec_gpu_cugraph(batch, fanout)\n",
    "    end_time = datetime.now()\n",
    "    sleep(0.01)\n",
    "    total_time = (end_time - start_time).total_seconds() / tries\n",
    "    res_cugraph[i] = total_time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch_geometric.data import Data\n",
    "\n",
    "data = Data(\n",
    "    edge_index=torch.stack([\n",
    "        torch.from_dlpack(sg.edgelist.edgelist_df.dst.to_dlpack()),\n",
    "        torch.from_dlpack(sg.edgelist.edgelist_df.src.to_dlpack())\n",
    "    ]),\n",
    "    num_nodes=sg.number_of_vertices()\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch_geometric.loader.utils import to_csc\n",
    "sample_fn = torch.ops.torch_sparse.neighbor_sample\n",
    "\n",
    "out = to_csc(data, device='cpu', share_memory=False,\n",
    "                is_sorted=False)\n",
    "colptr, row, perm = out\n",
    "\n",
    "def exec_cpu_pyg(batch, fanout):\n",
    "\n",
    "\n",
    "    sample_fn(\n",
    "        colptr,\n",
    "        row,\n",
    "        batch,\n",
    "        [fanout, fanout],\n",
    "        True,\n",
    "        True,\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "from datetime import datetime\n",
    "import numpy as np\n",
    "from time import sleep\n",
    "\n",
    "import torch\n",
    "\n",
    "batch_size = 1000\n",
    "tries = 10\n",
    "\n",
    "res_pyg_cpu = np.empty(99)\n",
    "\n",
    "for i, fanout in enumerate(range(1,100)):\n",
    "    batch = torch.Tensor(np.random.randint(low=0, high=sg.number_of_vertices(), size=batch_size)).to(torch.long)\n",
    "    start_time = datetime.now()\n",
    "    for _ in range(tries):\n",
    "        exec_cpu_pyg(batch, fanout)\n",
    "    end_time = datetime.now()\n",
    "    sleep(0.01)\n",
    "    total_time = (end_time - start_time).total_seconds() / tries\n",
    "    res_pyg_cpu[i] = total_time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8qNh9FAAAACXBIWXMAAAsTAAALEwEAmpwYAABMI0lEQVR4nO2dd3wVVfbAvychITTpBJQSwIIUCUUQBQRZFRVUbNjB7trbFssq6y6r/tZ117UulhUFBLvYdZGoIEgTkSpIUaT3hJbyzu+POy+ZhJdkgnl5Sd75fj7zeTO3zbkz8+659VxRVQzDMIz4JSHWAhiGYRixxRSBYRhGnGOKwDAMI84xRWAYhhHnmCIwDMOIc0wRGIZhxDmmCOIEEVkkIgMChl0tIr8pxm+AiKwtT9kqIyKSJiIqIjW8649EZESU7vVbEdkoIlki0jga9zCMkjBFAIhITRF5QUTWiEimiHwrIqeVEueAwlJERorItCjIFy6UPijiPk5ERgVJQ1U7qWpGectW3ohIAxF5UUQ2eO/iBxH5Q6zlUtXTVHVseacrIknAY8ApqlpXVbeWQ5qrRSRbRJoUcZ/vfUdpRdxHee69IqTVU0TeF5HtIrJDRBaLyGgRaVjMvUeJSI737sLv70kRaVEG+TNE5Oqg4SsCT6btIlKziPtAEZkqIjtFZHWEeGme/x4RWRqhzLjYK3d2i8g7ItIoylmJiCkCRw3gZ+BEoD7wJ+C1on+YSsBxInJCrIUoL8K17SL8E6gLHI17F2cCP1akXBVMKpACLCprRHEU9x9eBVzkC9sFqBUpDeAyYBswoojf8UAGMB3ooKoNgMFALtC1BNEmqWo9oBEwDGgOzC2LMqhMeOVAP0Bx36Of3cCLwO+Kif4q8C3QGLgXeENEmnrpdgL+g3v+qcAe4OlyFj8YqmpHhANYAJxbgv9q4DdF3EYC03zXR+P+SDtwf/QzfX4vAc8CnwGZwBdAm2LulYb7CP8ATPW5jwNG+a6HAPO9+30NHBNJXlyBMBbYDiwBfg+sLRL2Lu8Z7AQmASme3wBgLXAPsMULe4kvbn3gZWAzsAa4D0jwPZ/puMJ+G/DXCHldCJxdwnN/HKe0dwFzgX4+v1HA695zyQS+B44E7gY2efFO8YXPAB4CZnn5fBdoVOSZ1/CFvdr/noFHvWe4CjjNl25b4EtPhv8BTwHjIuTlSFxBokAW8Lnnfjww25NpNnB8EZlHe89xL3B4Md/mfcBsn9ujuIJIgTSfe38vnUuBrUCyz28a8EQZ/zejiuYVSAS+Ax71rhsC73vfyHbvvKXnNxrIA/Z5z+TJ0t57BBlK+waLfXfFpHe/97wfA94vJsxvgNUR3u9+oJ7P7Svgeu/8b8AEn197INsfvqIOaxFEQERScS+xzLU0XxpJwHvAp0Az4GZgvIgc5Qt2CfAXoAmuAB9fSrJPAUdG6r8Xke64msl1uNrHf4DJRZuyHg/gCrp2wMm4QqAoF+Bqf22BY3B/oDDNPZkPw9Uix/jy9QTuj9gO18K6HLjCF7c3sBL3TEZHuO9MYLSIXCEiR0Twnw2k42qbE4DXRSTF5z8UeAVX2HwLfIJr+R4GPIh7Ln4uB64EDsXVdP8d4Z6R6A0swz2H/wNe8GrXeHLNwr2HUbga3wGo6g9AJ++ygaqe5HUNfODJ0RhX+HxQZOzgMuBaoB6uoIvETOAQETlaRBKB4TgFWZQRuO90knc9BEBE6gB9gDeLST8wqpqHU7L9PKcE4L9AG6A1ThE96YW9F1dY3qSuq+wmL05p791PkG+wuHcXictx/83xwKle+RCETsBKVc30uX1HwTvv5F0DoKo/4hTBkQHTLz8qWvNU9gNIwtXi/lNKuNW4GssO37EHr0WA++g34NVEPLdX8WrwuBbBRJ9fXVxNqFWEe6Xh1U6BG4CZnvs4X3rPAH8pEm8ZcKJP3nCLYCVwqi/c1RzYIrjUd/1/wLPe+QBcgVnH5/8arjstEVcD6ujzuw7I8M5HAj+V8lxr4Vobc4EcYAUl1Nhwtbqu3vko4DOf31DvHSV61/W859jAu84AHvaF74j7IyZSeotghS9ebS9sc1zBlgvU9vmPI0KLoOi79a4vA2YVCTMDGOmT48EA3+ZvcDXhh3AK/TPv+8lvEXhy78JrgeGU5LveeUsvbIci38EOXCvmvmLuPSpSXoHrgeXFxEkHtvuu8591kPdexD3INxjx3RVzn77ed9jEu14K3B4hXKQWwWV4/1Wf22jgJe98Cl7rwOf/CzCgpLxH47AWgQ+vv/UVXGFwk8/9I29GR5aIXOKLcraqNggfuEI6zKHAz6oa8rmtwdVMw/wcPlHVLFx3yaGliPkckCoiQ4u4twHu9Ab0dojIDqBVMekd6r93kfMwG3zne3CKKsx2Vd3tu17jpdkESKZwLbXYPEdCVfeq6t9UtQeuRvwarvbXCEBE7hSRJd7g3A5czc8/KLrRd74X2KKuRhq+pkhe/PKswVUECg2yFkP+81HVPb50DwW2+dyK3qM0DuXAWn6ZnqGPV4CLcYXfyxH8h+GU1ofe9XjgNK8PezsQAvL79VX19953/jZOqZSFw3DfNyJSW0T+4w2S7sJ1ozXwWi4RCfDewwT5Bot7d5EYAXyqqlu86wkUGUspgSzgkCJuh+C6DIP4VximCDy8puELuEGbc1U1J+ynbsZIXe8orfsmzDqgVZHBvNY4jR+mle/+dXHN3nUlJerJ9Wdcl5K/OfszMNqvmFS1tqq+GiGZ9bga3wFyBKSh13UQprUn9xZc7alNET9/njXoTVR1F64ftQ7QVkT64cZJLgAaeoXSTgo/h7Liz3trnPxbigkbhPVAIxGpXcw9SmMdhZ9fWK4yP0NVXYPrAz8deCtCkBG4AvAnEdmAG19JAi7yFP03wDllkD0i3n9gKK7LB+BO4Cigt6oeghungIL3qEXil+W9B/kGg8pdy7vnid4stg3A7UBXEekaIIlFQDsRqedz60pBl/MifIPuItIOqAn8UFZZfy2mCAp4Bje4O1RV95YWOADf4JrQvxeRJHFz+IcCE31hTheRviKSjCvYv1HVILW9V3AfzGCf23PA9SLS25tNUkdEzijyEYZ5DbhbRBqKyGH4Wj9l4M8ikuz9SYcAr3s179dwffz1RKQNcAeR+6YjIiJ/EpFjvbRTgFtx3RHLcF07ubhBwBoicj8H1qjKyqUi0tEruB8E3vC1IMqMV/jOAUZ5eeiDe+9B+RA3DnSxiNQQkeG4Lqv3D1Kkq4CTirTg8N77INy7S/eOrsAjFNR4fw9cKSJ/FJFmXryWuHGjUvG++6NxXaLNceMd4N7jXmCH19J7oEjUjbj+fXzhA7338vgGfZyN667tSMEzOhqn0C738pjgfadJ7lJSvP8z6saA5gMPeO7DcONt4XGX8cBQEennVaweBN7SwmMKFYIpAsD7WK7DvegNxXQDlQlVzcZNNTsNV0t5GrhcVZf6gk3A/Qm2AT1wg8dB0s7z4jXyuc0BrsENum3H9a2PLCaJB3Ezf1bhxkPewPWrBmWDd491uI/5el++bsYpwJW42RkTcIPYQVHcQOIWL/2TgTO8rrNPgI9wNaY1uJklZel2icQruPGaDbhpnLf8yvTAvcc+uFk4f8UNxAZ6vurWEQzB1Zq34grjIb6uiTKhqj9630ZRLgPmq+qnqrohfOAGqY8Rkc6qOg04CVdj/8HrkvkY14f/RAm3HS4i4fGzyV4+eqhquLX7L9xY0BbcoPbHReI/Dpwnbt7+vyn7e/+132CYEcB/VfWnIs/oSeAScdOfw7OuPqRg4PtTXxoXAj1x/5eHgfNUdTOAqi7CjZ2Mx81qq0fh7uUKQ7wBCqOCEZGXcAO091UCWX4LXKiqJ8ZalopERDJwA5vPR/k+k4Clqlq05msYlQJrEcQhItJCRE7wmrVH4Wqfb8daruqC17XV3nu+g4GzgHdiLJZhFEtZR/6N6kEybqpgW1zzfSKxWtFYPWmOG5xtjOuC+62qfhtbkQyjeKxryDAMI86xriHDMIw4p8p1DTVp0kTT0tICh9+9ezd16tQpPWA1I17zDfGbd8t3fFHWfM+dO3eLqjaN5FflFEFaWhpz5kSaDReZjIwMBgwYED2BKinxmm+I37xbvuOLsuZbRIqzS2VdQ4ZhGPGOKQLDMIw4xxSBYRhGnGOKwDAMI84xRWAYhhHnmCIwDMOIc0wRGIZhxDmmCAzDMKoCGQ/Dyi+ikrQpAsMwjMrOnm1OEfz8TVSSN0VgGIZR2VnzNaCQ1i8qyZsiMAzDqOys/gpq1ILDekQleVMEhmEYlZ1VX0Hr3lAjOSrJmyIwDMOozOzeApsWRa1bCEwRGIZhVG5WT3O/bftH7RamCAzDMCozq7+CpDpwaLeo3cIUgWEYRmVm1VfQpg8kJkXtFqYIDMMwKitZm2DLsqiOD4ApAsMwjIonFIJtq0oPt/or9xtLRSAifUTkKRFZICKbReQnEflQRG4UkfpRlcwwDKO68vEf4N/psOjtksOt+gqS60GLrlEVp1hFICIfAVcDnwCDgRZAR+A+IAV4V0TOjKp0hmEY1Y1lH8OsMa6Af+cG2LCw+LCrv4I2x0NidLeXL6lFcJmqXqWqk1V1narmqmqWqs5T1X+o6gDg66hKZxiGUZ3I3ADv3gCpXeC306HmITDxYmdLqCgznoKtK6D9wKiLVawiUNUtACJSR0QSvPMjReRMEUnyhzEMwzBKIRSCt6+D7D1w3gvQsA0MHweZ6+GNK2DnLy6cKvxvFHxyD3Q8C3peGXXRgrQ3vgT6iUhDYAowBxgOXBJNwQzDMKoVc16AlRkw9HFoepRza3UsnPEYTL4J/tkRDu0OdZrC8k+gxxVwxj8gITHqogWZNSSqugc4B3hCVYfhxgoMwzCMIOzPgi8ecbN/uo8o7Nf9MrhpDgy6313/OAX6/x6G/LNClAAEaxGIiPTBtQCuKkM8wzCM+ODTP0G95tDnxsj+s8bA7s1w4QQQOdC/yRHQ70535OVGfXC4KEFaBLcCdwNvq+oiEWkHTI2uWIZhGFWE7N3wzbMw/d9uHKAo+3bC9MfhiFOhVa/S06tgJQABavaq+iVunCB8vRK4JZpCGYZhVBnWfA152ZC1ATZ8d6BNoBlPwb4dMPCemIgXhJLWEYwRkS7F+NURkStFpNgBYxFpJSJTRWSJiCwSkVsjhBER+beIrPAWrXU/uGwYhmHEiB8/h8RkQOCHTwr77d4KM56Go8+EQ9NjIV0gSmoRPA38yVMGC4HNuIVkRwCHAC8C40uInwvcqarzRKQeMFdEPlPVxb4wp3npHQH0Bp7xfg3DMKoGP06FNie4LqJlH8GAPxb4ff1vyM6CgffGTr4AFKsIVHU+cIGI1AV64lYW7wWWqOqy0hJW1fXAeu88U0SWAIcBfkVwFvCyqiowU0QaiEgLL65hGEblZtc62LwE0i+GUA5MeRB2rYdDWrhFYrOfh87nQLMOsZa0RIKMEWQBGb/mJiKSBnQDvinidRjws+96redWSBGIyLXAtQCpqalkZAQXJysrq0zhqwvxmm+I37xbviue5uun0AGYvf0QAI4Flr3/BOsPPYU2qyfSNjuL2Sn92R0F+co136oa1QOoC8wFzong9wHQ13c9BehRUno9evTQsjB16tQyha8uxGu+VeM375bvGPD6lar/d7hqKOSOxzqrTrhIdd8u1Ydaq064MGq3Lmu+gTlaTLkaVTPUnimKN4HxqvpWhCBrgVa+65bAumjKZBiGUS6EQrByqrMFJOKOI091bjOfcTOF+t0ZaykDEVgRiEidsiQsIgK8gBtTeKyYYJOBy73ZQ8cBO9XGBwzDqApsWAB7tkL7kwrcjhoMOXsg42FoeyK07Bk7+cpAqYpARI4XkcXAEu+6q4g8HSDtE4DLgJNEZL53nC4i14vI9V6YD4GVwArgOeCGg8qFYRhGRfPj5+633YACtzZ93f7Cmgf974qJWAdDkCVs/wROxdXeUdXvRKR/aZFUdRoQYS11oTAKFLMm2zAMoxLz4+eQ2tmZlgiTlOJmCe1cG/VdxcqTQGuZVfVnKWwfIy864hiGYVQyVA+0DzR3rNs05sQ/Hhj+rCcjx6nEBBkj+FlEjgdURJJF5C68biLDMIxqzcK34OE2MPlmZzMI3A5j79/uxgaK6/6pQkoAgrUIrgcex83vXwt8inXnGIZRnQmFnNnoLx6GxofDt+Ng+f+cddHP/wrNu8AFL0NiUqwlLReCLCjbgm1CYxhGvJCXA29eDYvfgfRL3L4AGxfCuzfBp/dCwzS45HWoWS/WkpYbpSoCEWkL3Ayk+cOrqm1cbxhG9WP5p04JnHQf9LvLdfMc1gOu/QIWTHSzhOo2i7WU5UqQrqF3cOsB3gMiGNs2DMOoRqyeDjVS4PhbCvf110iG7pfHTq4oEkQR7FPVf0ddEsMwjMrAmmnQ8lioUTPWklQYQWYNPS4iD4hIHxHpHj6iLplhGEZFs28nbPge2hwfa0kqlCAtgi54K4Qp6BpS79owDKP68NM3oCG3v0AcEUQRDAPaqWp2tIUxDMOoMFZPd3sNn/t8QTfQmumQkOS6huKIIF1D3wENoiyHYRhGxfLFw7Bksls0FmbNdDisOyTXjp1cMSCIIkgFlorIJyIyOXxEWzDDMIyosfVHWPWlO5/5tDMJkb0b1n0bd+MDEKxr6IGoS2EYhlGRzP0vJNSA/r+HjL/Bmq8hLxtCuc6CaJwRZGXxFxUhiGEYRrmxbSUg0KjtgX65+2H+BDjqNDjhFjdOMPNpaHY0SAK06lXh4saaYruGRGSa95spIrt8R6aI7Ko4EQ3DMMrIm9fAC6dA5oYD/Za85zaU6XEFJNWCnlfA0g/cWEGLrpBySMXLG2NKGiP4HYCq1lPVQ3xHPVWNvydlGEbVIJTnbAPt3gRvXAl5uYX95/zX2QtqN9BdH3sNJCTCth/jbtpomJIUwVMVJoVhGEZ5sW0V5O6Dw092s4A+/0uB3+Yf3Mrh7iMgwSv+DmkBnc5x53GqCEoaI6haBrUNwzAANi12vwPvgfotYfq/3PXWFW4zmYQk6HZp4Tgn/sFtL9m21M0XqyUlKYK2JU0TNeujhmFUSjYtBgSadoDBD8O6eU4Z1G8FHYZC1+EHWg9tcjic92IspK0UlKQINgP/qChBDMMwiiV3f3AjcBsXudlC4UVhIz+APdugQesqt3NYRVHSGEGmqn5R3FFhEhqGER+snQMTLoT9mUXc58IjaTDv5cLue7bBmAGkbvi8sPumxdCsY8F1zXrQsI0pgRIoSRGsrighDMOIc1Thoz/ADx/BjKcL+00dDTl74MPfu8HecPh3b4R139J8w9SCsDl73RqC1E4VJ3s1oFhFoKrnVKQghmHEMcs/hV/mQJ1m8PUTsHurc/95Fvw4BY670XX1vHml6yaa+TQs+xAaplF/52JnHgJg8zJnPbTZ0bHLSxUkiK0hwzCM6KHqav0N0+CytyFnN0x7zPllPAS1m8BJ98KZT7q9Al6/Aj57AI46A854jATNdSYioGDGUDNrEZQFUwSGYcSWpR/A+u/cFM7mnaHrxTDrOVj4Jvz4OZxwKyTXgQ6nw7FXw7IPoF4LOPspaHM8eQnJLhw4RZBYExq1i22eqhhBNq8X4BLcngQPikhroLmqzoq6dIZhVG9CIZj6N2h8OHS5wLkN+CN8/5ozE1GnKRx7VUH4U/4KicmQfjHUagjAzvodaRRWBBsXQ9OjIDGIPU0jTJAWwdNAH+Ai7zoTW3VsGEZ5sOgt2LQIBtxdUHg3aOVq/ppX0BoIk1QLBj8EzbvkO21r1A02L4Wdvxw4Y8gIRBC12VtVu4vItwCqul1EkqMsl2EY1Z19O+GTe12h3mlYYb8T/+AWfR17TanJbG+Y7k4WvQWZ6yHVFEFZCaIIckQkEbdPMSLSlIK9iw3DMA6O/41yhuEumuCMvvmp1QD63h4omd112kDdVJj5rHOwgeIyE6Rr6N/A20AzERkNTAMeiqpUhmFUb9bMgDkvQu/r4bAevy4tEWh/Euxa666tRVBmgmxMM15E5gKDcIbozlbVJVGXzDCM6knufnjvVqjfGgbeWz5ptj8JvnsVUhq4GUVGmQgya+gVVb0MWBrBzTAMo2zMeBK2LIOLX4eadcsnzXYD3G+zjmZK4iAIMkZQqMPNGy/4lW05wzDikpx9zoTEEafAkaeUX7p1m8Exw+GwnuWXZhxRrCIQkbuBe4Ba3taUYTWbDYypANkMw6huLHwT9myBPjeWf9rnWLF0sJRka+ghVa0H/N23RWU9VW2sqndXoIyGYVQHVOGbZ1z3TdsTYy2N4SNI19BHInLAtj2q+mVJkUTkRWAIsElVO0fwHwC8C6zynN5S1QcDyGMYRlVkzdfOVtDQx60fv5IRRBH8zneeAvQC5gInlRLvJeBJ4OUSwnylqkMCyGAYRlXnm2egViPXl29UKoJMHx3qvxaRVsD/BYj3pYikHbxohmFUWVTh9RGQuRG6X+7WCiz9AE64zZmJMCoVoqpli+CM0C1Q1S4BwqYB75fQNfQmsBZYB9ylqouKSeda4FqA1NTUHhMnTgwsb1ZWFnXrltMUtSpEvOYb4jfvlSnfDbd9S9cFo8hOqk9yzk4UAYSZxz3H/pQm5XqvypTviqSs+R44cOBcVY04rapURSAiT+CZl8ANLqcDq1X10tJuXIoiOAQIqWqWiJwOPK6qR5SWZs+ePXXOnDmlBcsnIyODAQMGBA5fXYjXfEP85j1m+c7LgcSkgmtVeH4QZG2Cm+fC2tlum8nGR8CJvys+nYPE3ncwRKRYRRBkjMBf6uYCr6rq9MB3LwZV3eU7/1BEnhaRJqq65dembRhGBbH1R3juJOh8DpzxmBsE/uET+GUuDP2323A+ra87jEpLkDGCsZ610Q64lsGy8rixiDQHNqqqikgvXGtja3mkbRhGBZCXC29dC9lZzm5QQhIMfrhgt7H0i2MtoRGQICYmTgf+A/yIW1TWVkSuU9WPSon3KjAAaCIia4EHgCQAVX0WOA/4rYjkAnuBC7WsAxaGYcSOaY+5fYbPfQHWfetMR2xeChsWwNnPFu4uMio1QbqGHgMGquoKABFpD3wAlKgIVPWiUvyfxE0vNQyjqvHLPMh4GDqfB13Og87nQs4e1zJofAR0OT/WEhplIIgi2BRWAh4rgU1RkscwjMpO7n54+zq3B8AZjzo3ETj9H26v4DbH21aRVYwgb2uRiHwIvIYbIzgfmC0i5wCo6ltRlM8wjMrGgkmw5Qe4+LX8fYMBSEiA42+OnVzGQRNEEaQAG4GwcZDNQCNgKE4xmCIwjHhBFWY8BaldnAVRo1oQZNbQFRUhiGEYlRDVwnaBVkxxA8JnP2v2gqoRQWYNtQVuBtL84VX1zOiJZRhGTFGF6Y+72v/5/y1YBzDjSbcDWOdzYyufUa4E6Rp6B3gBeA/btN4wqj+hEHx6H8x8CpJqw4QLYeR7bp3Ayqkw6AGokRxrKY1yJIgi2Keq/466JIZhxJ68HHj3Rjcg3Os6N/j739Nh3LnQ/BinGHqMjLWURjlT7MY0Ph4XkQdEpI+IdA8fUZfMMIyK58u/OyVw0n1w2iPQoBVc/g5IomsNdLsUajeKtZRGOROkRdAFuAy3/0C4a0gpfT8CwzCqEvt2wsxn4eih0N9nHK5xe6cMMh6CE26NmXhG9AiiCIYB7VQ1O9rCGIYRQ2Y9B/t3FlYCYVI7wfBxFS+TUSEE6Rr6DmgQZTkMw4gl2bvdDKEjToEWXWMtjVHBBGkRpAJLRWQ2sD/saNNHDaN8ycnJYe3atezbt++g4tevX58lS5Yc3M33Z0L/Z53ZiINNI0b8qnxXYYrLd0pKCi1btiQpKbjRvyCK4IEyyGYYxkGydu1a6tWrR1paGnIQi7UyMzOpV69e6QFVYc8WZzOoZj1IrgOblkCNhtCk1L2hKh2B813NiJRvVWXr1q2sXbuWtm3bBk4ryMriL0QkFTjWc5qlqmZ0zjDKmX379h20EghMbjbsWOP2EEBg9+YCv3rNo3dfo0IQERo3bszmzZtLD+wjyMriC4C/Axm4/QieEJHfqeobByOoYRjFE1UlsGcb7FwLKNRvBbUaOYWwfxcgkBx/+/5WRw7mGwoyWHwvcKyqjlDVy4FewJ/KfCfDMGJDKA+2r3EtgRop0LQD1GnirIWmHAL1W0L9ww7KdtDy5csZMmQI7du3p0ePHgwcOJAvv/yy3EQPujn7xx9/TK9evejQoQPp6ekMHz6cn376CYCRI0fStm1b0tPT6d69OzNmzABgwIAB+Pc/X716NZ07H7C9OqFQiFtuuYXOnTvTpUsXjj32WFatWlUOuSuetLQ0tmypuF17g4wRJBTpCtpKMAViGEasyd4D21dD3n6o29x1/5RTq2Pfvn2cccYZPProo5x5pps7snDhQubMmUP//v0Lhc3NzaVGjejsUbB48WJuvvlmJk+ezNFHHw3A5MmTWb16Na1btwbg73//O+eddx6ffvop1113HQsWLAic/qRJk1i3bh0LFiwgISGBtWvXUqdOnajk5WDJy8sjMTHxoOMHKdA/FpFPRGSkiIwkwO5khmFUAvJyYesK0BA0PhwOaVGqEnj55Zc55phj6Nq1K5dddhngatRvvFHQExyupY8fP54+ffrkKwGAzp07M3LkSABGjRrFtddeyymnnMLll1/O6tWr6devH927d6d79+58/fXXAGRkZNC/f3+GDRtGx44duf766wmFCsya3XvvvXTt2pXjjjuOjRs3HiDzP//5T+655558JQBw5plnHqCMAPr378+KFSsOcC+J9evX06JFCxISXHHZsmVLGjZ0+zD89re/pWfPnnTq1IkHHiiYV5OWlsY999xDnz596NmzJ/PmzePUU0+lffv2PPvss4HyHWbcuHH06tWL9PR0rrvuOvLy8gBo0aIF999/P717985v5RwsQQaLf+dtQtMXN0YwRlXf/lV3NQyjRP783iIWr9sVPIKGyAtp4Vph3n5nOyipNshCOh56CA8M7VRsEosWLWL06NFMnz6dJk2asG3bthJvuWjRIrp3L9nazNy5c5k2bRq1atViz549fPbZZ6SkpLB8+XIuuuii/K6ZWbNmsXjxYtq0acPgwYN56623OO+889i9ezfHHXcco0eP5ve//z3PPfcc9913X6F7LF26lLvvvruUB+R477336NKlS6CwYS644AL69u3LV199xaBBg7j00kvp1q0bAKNHj6ZRo0bk5eUxaNAgFixYwDHHHANAq1atmDFjBrfffjsjR45k+vTp7Nu3j06dOnH99deXmO8wS5YsYdKkSUyfPp2kpCRuuOEGxo8fz+WXX87u3bvp3LkzDz74YJnyE4liWwQicriInABuFzJVvUNVbwe2evsWG4ZRGdAQ5OwhIZRd2C0vx1kMlWA9uZ9//jnnnXceTZo0AaBRo7LZFBo2bBidO3fmnHPOyXc788wzqVWrFuDWSVxzzTV06dKF888/n8WLF+eH69WrF+3atSMxMZGLLrqIadOmAZCcnMyQIUMA6NGjB6tXry5Rhq1bt5Kens6RRx7Jo48+mu/+u9/9jvT0dMaMGcMLL7wARB5UjeTWsmVLli1bxkMPPURCQgKDBg1iypQpALz22mt0796dbt26sWjRokJ5CreUunTpQu/evalXrx5NmzYlJSWFHTt2lJjvMFOmTGHu3Lkce+yxpKenM2XKFFauXAlAYmIi555bPubAS2oR/Au4J4L7Hs9vaLlIYBjGAZRUcz+AHWvcjCCAOk3hkMNg20q3WrjZ0ZAYbGGRqkYsCGvUqJHfZaGqZGc7hdOpU6dCA8Nvv/02c+bM4a677sp38/el//Of/yQ1NZXvvvuOUChESkpKvl/R+4avk5KS8s8TExPJzc09QL4OHTowb948unbtSuPGjZk/fz6PPvooWVlZ+WHCYwR+GjduzPbt2/Ovt23blq8Ei1KzZk1OO+00TjvtNFJTU3nnnXdo164djz76KLNnz6Zhw4aMHDmy0GLAmjVrApCQkJB/Hr4O56O4fIdRVUaMGMFDDz10gEwpKSm/alzAT0lVhTRVPWBERVXn4DapMQwj1uTlwJ7tULsJ2Un13bqA7WvclNC6qYGVAMCgQYN47bXX2Lp1K0B+11BaWhpz584F4N133yUnJweAiy++mOnTpzN58uT8NPbs2VNs+jt37szva3/llVfy+7rBdZGsWrWKUCjEpEmT6Nu3b2C5b7vtNkaPHl1olW1JcoQZMGAA48aNQ1UBGDt2LAMHDjwg3Lx581i3bh3gZhAtWLCANm3asGvXLurUqUP9+vXZuHEjH31U9qHT0vI9aNAg3njjDTZtcvN1tm3bxpo1a8p8n9IoqUWQUoJfrfIWxDCMg2DPFkChTlP2780mOVFg33ZITIa6TcuUVKdOnbj33ns58cQTSUxMpFu3brz00ktcc801nHXWWfTq1YtBgwbl1/Jr1arF+++/zx133MFtt91Gamoq9erVO6APP8wNN9zAueeey+uvv87AgQMLtRb69OnDH//4R77//vv8AdSyyP34449z+eWXk5mZSePGjWndujV//vOfS4x37bXXsnTpUrp27YqI0LNnz4g1702bNnHNNdewf7+zsNOrVy9uuukmUlJS6NatG506daJdu3accMIJgWUOU1q+O3bsyF//+ldOOeUUQqEQSUlJPPXUU7Rp06bM9yoRVY14AK8C10RwvwqYVFy8aB89evTQsjB16tQyha8uxGu+Vatu3hcvXly2CHl5qusXqG5Zoaqqu3btcm47flbdlxkFCaPD1KlT9Ywzzjjo+Lt27SpHaSqOaOY70rcEzNFiytWSWgS3AW+LyCXAXM+tJ5CMM01tGEYs2bcdQrlQt1mBW0KCWyBmGGWgWEWgqhuB40VkIBBebveBqn5eIZIZhlE8GoKszW6lcBU3DTFgwAAGDBgQazEqnMqU7yDrCKYCUytAFsMwSiKUC7u3OJPR2bsBhQaty22lsBG/RGfNt2EY5UteDmz9EXL3Qo1azlZQzXpQ85BYS2ZUA0wRGEZlJ3e/UwJ5OdCovTMUZxjliCkCw6hsaAhy9joFkLcfdm91bk0Od5vIGEY5U+racxHJFJFdRY6fReRtEWlXEUIaRtyQsxc2/wBbfnArhjM3QEKi2zmsApRAYmIi6enpdO7cmfPPP7/UhVllMUM9a9Ys+vfvz1FHHUWHDh24+uqr2bNnDy+99BJNmzYlPT2djh078txzzwHOaJ3fTAQUb575xRdfpEuXLhxzzDF07tyZd9999yCfQDCKGuKr6gRpETwGrAMm4IzOXQg0B5YBLwIDoiWcYcQNoZBbDbx5mSv4G7SGpDpuYVhCxVl9r1WrFvPnzwfgkksu4dlnn+WOO+6IGLYsZqg3btzI+eefz8SJE+nTpw+qyptvvklmZiYAw4cP58knn2TTpk106tSpkEXT0vjll18YPXo08+bNo379+mRlZZV5h65oE00z3OVBEMkGq2pv3/UYEZmpqg+KSCRbRIZhlIW8HHh9JLS5HFKOcLuHlcE0RLTo168fCxYs4E9/+hNNmjTh1ltvBZxZ6NTUVOrUqRPRDHWkzV2eeuopRowYQZ8+fQBnU6eo7R+AZs2a0b59+zKZUdi8eTP16tXLN49dt27d/PPnnnuOMWPGkJ2dzeGHH84rr7xC7dq1GTlyJLVq1WLp0qWsWbOG//73v4wdO5YZM2bQu3dvXnrppfy0rrvuOqZOnUrDhg2ZOHEiTZsWXrE9d+5c7rjjDrKysmjSpAkvvfQSLVq0YMCAARx//PFMnz6dM888kzvvvDNwniqaIIog5G1XGW4H+d+elr9IhhFH5OXCW9fA0vehw63QsK2bDvrRH2HD92VKqlZeLiSW8Jdu3gVOezhQWrm5uXz00UcMHjyY0047jXPOOYdbb72VUCjExIkTmTVrFqNHjy7VDHWYhQsXMmLEiFLDrVy5kpUrV3L44YcHShecdc/U1FTatm3LoEGDOOeccxg61NnEPOecc7jmmmsAuO+++3jhhRe4+eabAdi+fTuff/45kydPZujQoUyfPp3nn3+eY489lvnz55Oens7u3bvp3r07//jHP3jwwQf585//zJNPPpl/75ycHG6++WbeffddmjZtyqRJk7j33nt58cUXAdixYwdffPFF4LzEiiCK4BLgceBpXME/E7hURGoBN0VRNsOo3oRC8O6NsOhtOOWvbjpojNcE7N27l/T0dMC1CK666iqSk5Np3Lgx3377LRs3bqRbt240btz4gLjDhg1j+fLlHHnkkbz11ltluu+kSZOYNm0aNWvW5D//+Q+NGjUqdu/dou6JiYl8/PHHzJ49mylTpnD77bczd+5cRo0axcKFC7nvvvvYsWMHWVlZnHrqqfnxhg4diojkK5LwPgWdOnVi9erVpKenk5CQwPDhwwG49NJLC5nYBli2bBkLFy7k5JNPBtxOYS1atMj3D8et7ARZULaS4k1OTyvGHRF5ERgCbFLVA9qK4t7m48DpONPWI1V1XhChDaPKs3YufPEILP8ETroPjr8ZfNYzg9bc/ezNzKRevXq/Siz/GIGfq6++mpdeeokNGzZw5ZVXAsHMUIfp1KkTc+fO5ayzzop43/AYgZ/GjRuzfv36Qm6ZmZk0aNDggPgiQq9evejVqxcnn3wyV1xxBaNGjWLkyJG88847dO3alZdeeomMjIz8OEHMREe6jx9VpVOnTsXuEFbZtrQsjiCzhpqKyD0iMkZEXgwfAdJ+CRhcgv9pwBHecS3wTBCBDaNK8+Pn8OJgeP4k+Gmmawn0/12spSqVYcOG5de6w7Xqspihvummmxg7dizffPNNvtu4cePYsGFDsffs378/kydPzh9Qfuutt+jatesBNvjXr1/PvHkFdcj58+fnW+fMzMykRYsW5OTkMH78+DLm2pmdDs8OmjBhwgFmoo866ig2b96crwhycnJYtGhRme8Ta4J0Db0LfAX8D8grJWw+qvqliKSVEOQs4GXPKt5MEWkgIi1UdX0JcQyjapKXC58/CNMfdzOCBj8M3S513UFVgOTkZAYOHEiDBg3yC+KymKFOTU1l4sSJ3HXXXWzatImEhAT69+9/QFeLn2OOOYabbrqJvn37IiI0a9aM559//oBwOTk53HXXXaxbt46UlBSaNm2avy/wX/7yF3r37k2bNm3o0qVLvlIJSp06dVi0aBE9evSgfv36TJo06YDn8sYbb3DLLbewc+dOcnNzue222+jUqQwbC1UCRLXk8V4Rma+q6QeVuFME7xfTNfQ+8LCqTvOupwB/ULfxTdGw1+JaDaSmpvaYOHFiYBmysrLyZxDEE/Gab6h8eU/ev52Oix+lwc6F/HLoYFYcfjWacOCsoPr165dpkLQoeXl55bZjVVFCoRD9+vVj7Nixv0rGaBDNfLdo0eKA7qnKQkn5XrFiBTt37izkNnDgwLmq2jNS+CAtgvdF5HRV/bDMkpZMpJGgiFpJVccAYwB69uypZbHYl5GRUWks/FUk8ZpvqGR537sDnu0Le7bCsDEc1nU4hxUTdMmSJb+qjz+zHMYIIrF48WKGDBnCsGHD8jdtr0xEK99hopn2r6GkfIc3zQlKEEVwK3CPiOwHcnAFuKrqrzV4shZo5btuiVu4ZhjVh4/vhl3r4MpPoNWxsZbmoOjYsWP+hunxhn/f4+pMqYPFqlpPVRNUtZaqHuJdl4fVq8nA5eI4Dthp4wNGtWLZR/DdBOh3R5VVAkZ8UGyLQEQ6qOpSEYm4YqS0qZ4i8irO/EQTEVkLPAAkeXGfBT7ETR1dgZs+esXBZMAwKiV7tsF7t0GzTtD/94GjqWqx8+cNIwiljftGoqSuoTuBa4B/RLoXcFIpwlxUir8CN5YmoGFUCVRhzXTYtd5ZDF082W0sf/EkqJEcKImUlBS2bt1K48aNTRkYB4WqsnXrVlJSUsoUr6StKq/xfgf+StkMo/oz/XH43wOF3QbdD4emB06iZcuWrF279qANpu3bt6/MBUB1wPJdmJSUFFq2LNu+1SV1DRU/wRdQ1bKtITeM6sqC150S6HQODLgbatR0JqPrNClTMklJSbRt2/agxcjIyKiUs3qijeX711NS11BxZiXAdQ2ZIjCMVV/CO7+FNn1h2LNOCRhGFaOkriEbvDWMSOTshTVfO3MR816Gxu3hwnGmBIwqS6nrCESkMW7GT19cS2Aa8KCqbo2ybIZR+Vj1JUwYDjl7ILEmpPWFof+CWg1jLZlhHDRBFpRNBL4EzvWuLwEmAb+JllCGUSkJheCTe6BOUzjjH9DmBEiuHWupDONXE0QRNFLVv/iu/yoiZ0dJHsOovCx5120WM2wMHHFyrKUxjHIjyGaoU0XkQhFJ8I4LgA+iLZhhVCpCeTD1b9C0A3Q5cItFw6jKBFEE1+E2rs/2jonAHSKSKSK7oimcYcSMHz6Fzx6ALG9O/4LXYMsPMPAet7m8YVQjguxQVjlN7xlGtPj+DXjrWtA8mP0C9L0Vvh0HzY+Bo88sPb5hVDGCjBEgIscAaf7wtqDMqJZ8O97tI9zmeDh1NHzxd/j8r87v4tdjvqewYUSDINNHXwSOARYBIc/ZFpQZ1Yu8HPj63zDlQWg3EC6c4GYEXTQBVk+HTYttgNiotgRpERynqh2jLolhxIpVX8GHv4PNS1zXzznPQZLPhkvaCe4wjGpKEEUwQ0Q6quriqEtjGBVJ1ia3LuD7190+whe+CkedZt0/RtwRRBGMxSmDDcB+CnYoOyaqkhlGtFCFb1+BT//kVgif+Afoezsk1Yq1ZIYRE4IogheBy4DvKRgjMIyqSfYemHQp/DjFrQwe8i9oemSspTKMmBJEEfykqpOjLolhRJvc/fDaZc5Y3OmPQs+rICHIUhrDqN4EUQRLRWQC8B6uawiw6aNGFSMvF968Clb8D858ArpfHmuJDKPSEEQR1MIpgFN8bjZ91Kg6qMLkm2HJe3DqQ6YEDKMIQVYW274ERtXmi0fguwlu97A+N8RaGsOodARZUJYCXAV0AvInV6vqlVGUyzDKh+/fgIyHoOvFbnaQYRgHEGSk7BWgOXAq8AXQEsiMplCGUS78PBveuQFaH+82j7H1AYYRkSBjBIer6vkicpaqjvUGjj+JtmCGcVCoUm/XMvjkM2co7pBDYbhtI2kYJRFEEeR4vztEpDOwAWeAzjBiz861kPEwZG6Avdth1y/0yFwPCUnQ/iQY/BDUaRxrKQ2jUhNEEYwRkYbAn4DJQF3v3DBiy/5MGH8BbF8FTY+ClAaQ1pclOYdy9Fl3QK0GsZbQMKoEQWYNPe+dfgG0i644hhGQUB68eTVsXgqXvuFq/x4bMzI42pSAYQSm2MFiERkqIm181/eLyHciMllE2laMeIZRDJ/dDz98DKc9UkgJGIZRdkqaNTQa2AwgIkOAS4Ercd1Dz0ZfNMOIwPY1buOYGU9Cr2uh1zWxlsgwqjwldQ2pqu7xzs8BXlDVucBcEbFVOUbFsmWFK/y/HeemgR53A5z8l1hLZRjVgpIUgYhIXWAPMAh42ueXEjmKYZQj+7NgwST47lVYOxsSk6HHCOh7B9Q/LNbSGUa1oSRF8C9gPrALWKKqcwBEpBuwPuqSGfFN9h4YOwTWfQvNOrra/zEXQL3msZbMMKodxSoCVX1RRD4BmgHf+bw2AGZ/yIgeoRC8cz2smw/nj4WOZ9mqYMOIIiVOH1XVX4BfirhZa8CILhl/g8XvulZAp7NjLY1hVHuCLCgzjIohLxdmPw9f/h26XQrH3xxriQwjLjBFYMSe3Gw3IDztn26VcLsBcMY/rTvIMCqIQPv0iUhfEbnCO28adEGZiAwWkWUiskJE/hjBf4CI7BSR+d5xf9nEN6o0ufth1nPw727w3i2QUh+Gj4dL34YaybGWzjDihiD7ETwA9ASOAv4LJAHjgBNKiZcIPAWcDKwFZovIZFVdXCToV6o65CBkN6oqqjBvLGQ8ApnroFVvGPo4HD7IWgGGEQOCdA0NA7oB8wBUdZ2I1AsQrxewQlVXAojIROAsoKgiMOIJVfjsT/D1E9C6Dwx7BtqeaArAMGJIEEWQraoqIgogInUCpn0Y8LPvei3QO0K4PiLyHbAOuEtVFxUNICLXAtcCpKamkpGREVAEyMrKKlP46kKlzLfmceQP/+HQ9Z/wy6Gns7ztNfAT8NMX5XqbSpn3CsDyHV+Ua75VtcQDuAv4D7ASuAaYAdwcIN75wPO+68uAJ4qEOQSo652fDiwvLd0ePXpoWZg6dWqZwlcXKl2+c7NV37hK9YFDVD8bpRoKRe1WlS7vFYTlO74oa76BOVpMuRrEDPWjInIyboXxUcD9qvpZAB2zFmjlu26Jq/X7097lO/9QRJ4WkSaquiVA+kZVIWcfvD4SfvgIBj0A/e6ItUSGYfgINH3UK/iDFP5+ZgNHeDOMfgEuBC72BxCR5sBGVVUR6YWbxbS1jPcxKjP7s2DiRbDqSzj9UbMWahiVkCCzhjIBLeK8E5gD3KneYHBRVDVXRG7C7W+cCLyoqotE5HrP/1ngPOC3IpIL7AUu9JowRlVn91ZYORVmPAXr58PZz0L6RbGWyjCMCARpETyG69KZAAiuZt8cWAa8CAwoLqKqfgh8WMTtWd/5k8CTZRXaqKSEQrD4bVf4/zIPUKjdGC54GY4eGmvpDMMohiCKYLCq+mf7jBGRmar6oIjcEy3BjCqEKiz/FKb8BTZ+D007wMB7oP0gODQdEhJjLaFhGCUQRBGEROQC4A3v+jyfn3XjxDN5ubD4HZj+L9jwPTRMg3Oeg87nWuFvGFWIIIrgEuBx3MY0CswELhWRWsBNUZTNqKyowsI3YcqfYcdP0PgIOPMJ6HoRJCbFWjrDMMpIkOmjK4HiOninla84RqUnazN8cAcsmQwtusLgR+DIwZAQyGyVYRiVkCCzhlKAq4BO+LaoVNUroyiXURlZ9I5TAvsz4TejoM/NkGgGbA2jqhOkGvcKbpbQqcAXuIVhmdEUyqhk7NkGr18Br4+ABq3hui+h7+2mBAyjmhDkn3y4qp4vImep6lgRmYBbG2DEA8s+gsm3wN7tcNJ9cIIpAMOobgT5R+d4vztEpDNuz+K0qElkVA72Z8In98C8lyG1C1z2FjTvEmupDMOIAkEUwRgRaQjcB0wG6gJ/iqpURmxZPR3evQG2r3FdQAPusY1iDKMaU6IiEJEEYJeqbge+BNpViFRGbFg7120cv+J/bizgig+hzfGxlsowjChToiJQ1ZBnL+i1CpLHiAVblsNn98OyD6FWI/jNn51xuOSgW08YhlGVCdI19JmI3AVMAnaHHVV1W9SkMiqGPdsg42GY8wLUqOUGg3tfDzWDbEBnGEZ1IYgiCK8XuNHnplg3UdVl306Y+awzDpedCd1HwMB7oW7TWEtmGEYMCLKyuG1FCGJUALn7Yfq/YcYTThl0GOKMw6V2irVkhmHEkCAri2sDdwCtVfVaETkCOEpV34+6dEb5sWkpvHm1sw561Bkw4A/ORIRhGHFPkK6h/wJzgfD0kbXA64ApgqpA9h749hU3GJxcBy6aCEedFmupDMOoRARRBO1VdbiIXASgqntFRKIsl/Fr2J9Jy58nw8v/gjVfQ95+tzfA2c9AvdRYS2cYRiUjiCLI9kxOK4CItAf2R1Uq4+BZ9SW8cyOH7/zJbRDT6xo4/DfQ9kSzEGoYRkSCKIJRwMdAKxEZD5wAjIyiTMbBsD8TpjwIs8ZAo/bM6/Yw3c/6baylMgyjChBk1tCnIjIXOA63Z/Gtqrol6pIZwdi3yxX+M550huF6Xw+DHmDX17NiLZlhGFWEILOGJgOvApNVdXdp4Y0KQBXWzYOFb8H88U4BHHGqmwl0WI9YS2cYRhUjSNfQP4DhwMMiMgu3wvh9Vd0XVcmMAwnX/ue9DDvWQEISHHkq9LvDFIBhGAdNkK6hL4AvRCQROAm4BngROCTKshlhinb/tD0RTvw9dDgDajWMtXSGYVRxAu0w4s0aGoprGXQHxkZTKMNj20qY9Rx8Ow7277LuH8MwokKQMYJJQG/czKGngAxVDUVbsLjml7nw1WOw9ANISISOZ0OfG+Gw7rGWzDCMakjQlcUXq2oegIicICIXq+qNpcQzykLuflg9zXX//Pg5pDSAfnfCsVfDIS1iLZ1hGNWYIGMEH4tIureyeDiwCngr6pLFA/uzYN5YWP4p/DQTcvdBnaZuP4BjrzJz0IZhVAjFKgIRORK4ELgI2IqbLSSqOrCCZKu+5OXCty/D1Idg9yZo1hF6jHSDwO0HQlKtWEtoGEYcUVKLYCnwFTBUVVcAiMjtFSJVdUQVNi+FJe/BdxNh24/Qug9cOB5a9Yq1dIZhxDElKYJzcS2CqSLyMTARt7LYKAubf4CFb7jFX1uXAwKtesMpf3VWQM1+n2EYMaZYRaCqbwNvi0gd4GzgdiBVRJ4B3lbVTytGxCrIzrWw8E34/nXY8D0gkNYXjrvebQZTr3msJTQMw8gnyGDxbmA8MF5EGgHnA38ETBH4ydzgun0WvQNrpgPq5vsPfhg6DbPC3zCMSkugBWVhvA3r/+MdxvY1sPR9pwB+mgkoNDnSbf/Y5TxoZNs6G4ZR+SmTIjCArT/CordhyWRY/51zS+3sCv+jz4RmHWIrn2EYMUNVyc4LkZOnhFTREIRUyQ0peSElJy/k+YfIzg2xNzuPvTl57MspvEZ3f24ee7Lz2JudR24oREhdOt1aNaRP+8blLrcpgtLYn+lW+v40E5Z9WFD4tzwWTv4LHD3Eav6GUQqhkJKnrjDMDSl5eUpOKJRfOIZCeAVeQaEZ8grR8JGb5/ycf4i8EOSFQizYkMvO+b8UpB1Scr3CODfk/eaF7+/i5eaFvLTCMrh4/iM3FCI7T8nJDZGnSijk5Mjz8pKb5wr9/TmucN+XnceenDzyQhq153j9ie2rniIQkcHA40Ai8LyqPlzEXzz/04E9wEhVnRdNmUokLwe2LHcF/y9z4Zc5sHERaAgQZ+LhlNHQ8Sxo0CpmYhqVF1XNr735CyZVRdVt85cbcrXB7FxXCPnD53gFWE5eqEihpPmFZChUUBiF75cXUpasyWH5lyvJzgsRCimKm7XsCq1Qftx895ArwPbnhsgN12BxhXZEOdRzz/UK73B4DRfshe/jl1mjVzY65s8PFCwxQUgUoUaiUCNBqJGY4H4ThMREoUZCAgniwiUlJniHkCBCco0EEkRI9MInJDi3mt6RkpRI7eREaifXyI8jIiQI+fdJTBBq1kgg2Uu7dnIiKcmJ1KyRgCC4J0p+WrWSEqmRmECiCCJQIyE6swyjpgg8a6VPASfjNryfLSKTVXWxL9hpwBHe0Rt4xvuNHqqwbyfs+gW2rnDHluWuwN+8FPKyXbia9V3B3+8uaN0bDusJtRoc5C3dn3VPdi57s/PYnZ3Hvhx3hP+E/tpKnvcHioTgZpwW98dys1GFRRtz2bdwfX4hA7gPEwpqTV4B5eJJ/txgzZfZyR2+V3ima1i+kK/m4wqEA+Pl16BC6qvdOT9Vzb+XPz8hr5Byz6QgzbwQXgHlCptEcX/eRBFfzS/E1m37GLtqVr5MId/9wzLlhFxNLycvhEL+Hw1whXRe4cLRyaf58uV4/uFmuxYKF0OWLIno7Ao996zAve9EryBL9hV2Iu47SUpMIKlGAkkJBeHChWNyYgI1vPAJUpBWkleQJnpxwgVm/rUUFLwJhcKTHy8hXMhKOB5eYSr5BXaNRCExISE/3Ldz53Dccb18BXxCgTxe3sIFvG23Hplotgh6AStUdSWAiEwEzgL8iuAs4GV1pdFMEWkgIi1UdX15CzNvyms0/up+mrGNWkW2XN5EQ5Zra5YxmOW0ZpG0Z83+FsjqRHQVhDQP1W+Awgspcn21MufnfBX1FYrlnZMy8G3sGleRECkocAXv1zv3V3RECgqPBK9gSkwoqJUleQVBKOS6F0IhzS9gkhITyNqvsNsp9HBhlegrXGqIUDsxgeREFz5BJL/GK5BfONbwFY6E5fXyUSMhIb9wTPDlJ0EKCsjExMKFmksFEhMTqJlYED8xwX054RpmcriWmugvEAsKTf/zSPAK2ASBWTNnMODEfvmFr0B+jbQ6F4CbfkigfdO6sRajShNNRXAY8LPvei0H1vYjhTkMKKQIRORa4FqA1NRUMjIyAguRlZVFRkYGWRu3klszjaUJ3dme0IhtCY3ZVKM5mxNbkJ1Yq1ABnwo0U1egh//4fv9w2e7+gK6mEa4JKuQXav64CQLJiULNRKiZ6M6TE3E1LoFEwSv0yC9Qiv51w/dV1WL/2OGa6Z49e6lbp7aTwRc/LF/4fn4/P2GZ/QV0OEwC5BeQfinEFz4hnHcJhy/PwkiBvCLShnMXIisrj7p1c8vpXr+S8BigX9yckqPklB4kMtm7mTNj2sHErNKE/+PxRnnmO5qKINK/PlJ5U1oYVHUMMAagZ8+eOmDAgMBCZGRk4MIPAOJnM/eCfMcf8Zp3y3d8UZ75TiiXVCKzFvCPqLYE1h1EGMMwDCOKRFMRzAaOEJG2IpKMs1s0uUiYycDl4jgO2BmN8QHDMAyjeKLWNaSquSJyE/AJbvroi6q6SESu9/yfBT7ETR1dgZs+ekW05DEMwzAiE9V1BKr6Ia6w97s96ztXwHY6MwzDiCHR7BoyDMMwqgCmCAzDMOIcUwSGYRhxjikCwzCMOEc05sZRyoaIbAbWlCFKE2BLlMSpzMRrviF+8275ji/Kmu82qto0kkeVUwRlRUTmqGrPWMtR0cRrviF+8275ji/KM9/WNWQYhhHnmCIwDMOIc+JBEYyJtQAxIl7zDfGbd8t3fFFu+a72YwSGYRhGycRDi8AwDMMoAVMEhmEYcU61VgQiMlhElonIChH5Y6zliRYi0kpEporIEhFZJCK3eu6NROQzEVnu/TaMtazRQEQSReRbEXnfu672+fa2dX1DRJZ6771PnOT7du8bXygir4pISnXMt4i8KCKbRGShz63YfIrI3V45t0xETi3r/aqtIhCRROAp4DSgI3CRiHSMrVRRIxe4U1WPBo4DbvTy+kdgiqoeAUzxrqsjtwL+XdvjId+PAx+ragegKy7/1TrfInIYcAvQU1U748zbX0j1zPdLwOAibhHz6f3XLwQ6eXGe9sq/wFRbRQD0Alao6kpVzQYmAmfFWKaooKrrVXWed56JKxQOw+V3rBdsLHB2TASMIiLSEjgDeN7nXK3zLSKHAP2BFwBUNVtVd1DN8+1RA6glIjWA2rgdDatdvlX1S2BbEefi8nkWMFFV96vqKtz+Lr3Kcr/qrAgOA372Xa/13Ko1IpIGdAO+AVLDO755v81iKFq0+Bfwewq2iYfqn+92wGbgv16X2PMiUodqnm9V/QV4FPgJWI/b0fBTqnm+fRSXz19d1lVnRSAR3Kr1XFkRqQu8CdymqrtiLU+0EZEhwCZVnRtrWSqYGkB34BlV7Qbspnp0h5SI1yd+FtAWOBSoIyKXxlaqSsGvLuuqsyJYC7TyXbfENSOrJSKShFMC41X1Lc95o4i08PxbAJtiJV+UOAE4U0RW47r+ThKRcVT/fK8F1qrqN971GzjFUN3z/RtglapuVtUc4C3geKp/vsMUl89fXdZVZ0UwGzhCRNqKSDJuMGVyjGWKCiIiuP7iJar6mM9rMjDCOx8BvFvRskUTVb1bVVuqahru/X6uqpdS/fO9AfhZRI7ynAYBi6nm+cZ1CR0nIrW9b34Qbjysuuc7THH5nAxcKCI1RaQtcAQwq0wpq2q1PYDTgR+AH4F7Yy1PFPPZF9cUXADM947Tgca42QXLvd9GsZY1is9gAPC+d17t8w2kA3O8d/4O0DBO8v1nYCmwEHgFqFkd8w28ihsHycHV+K8qKZ/AvV45tww4raz3MxMThmEYcU517hoyDMMwAmCKwDAMI84xRWAYhhHnmCIwDMOIc0wRGIZhxDmmCOIIEckTkfm+I60C7pkuIqeX4N9NRJ6P4D7D+30nvIimHGQ5szQrtCIyIGzFNILfahFpUh6y/FpEZKSIPBlrOcqCiGSISODN1v3vQkSGiMifoyddfGOKIL7Yq6rpvmN1BdwzHbemoTjuAZ7wO4jI4cAKb9FQc/Xsq/xaVHWyqj5cHmmVFc9ImnHwfIBbRV471oJUR0wRxDEiUldEpojIPBH5XkTO8tzTPBv3z3m23z8VkVqeX7qIzBSRBSLydtgmur+2JyJNvNpzMvAgMNxrgQwvcv96wDGq+p13XUtE5gOf4xaILQGO9OKmR5A/S0RGi8h3nkypnntTEXlTRGZ7xwmee34tWkTae3Fmi8iDIpLlS7quFNj6H+8ppDC/E5FZ3nG4l1Yb7zku8H5be+4vichjIjIVeKSI7CNF5C0R+Vicffn/8/ld5L2PhSLyiM/9ChH5QUS+wJnXCLsXl98Tfa2/b73n7Zehjoh84D2/heH3IyL3e+ksFJEx4fx77/ifIvKl930c6+VhuYj81fftLBWRsd7zeCNS4S0ip4jIDO/be12cnazwHiJLRWQacE44vLoFTxnAkKJpGeVArFfQ2VFxB5BHwcrjt3HGyw7x/JrgzNcKkIbb4yDd83sNuNQ7XwCc6J0/CPzLO8/A2YkPp7XaOx8JPFmMPAOBNyO4Pw308OLeWEJ+FBjqnf8fcJ93PgHo6523xpneKCQL8D5wkXd+PZDlnQ8AduLstSQAM3xprcZboQ5cTsFK5veAEd75lcA73vlL3n0SI8g+ElgJ1AdSgDU4ezGH4kwpNPXez+c4c8MtfO7JwHRfXorL73vACd55XaBGERnOBZ7zXdf3fv0rVl/xPeMM4BHv/FacPZsWuNW9a3ErX9O89xK+74vAXf5vxPs+vgTqeO5/AO73nsPPOBMJgvvu3vfJcgnwRKz/R9XxsBZBfOHvGhqG+7P9TUQWAP/Dma5N9cKuUtX53vlcIE1E6gMNVPULz30szi7+wdICZ065KF1wJgS64JRWcWTjCtp8Gb3z3wBPeq2LycAhRWvDQB/gde98QhG/Waq6VlVD3v3TfH6v+n77+NIKp/EKzuRHmNdVNa8Y+aeo6k5V3YezFdQGOBbIUGdYLRcYj3vGvX3u2cAkXzrF5Xc68JiI3IJ7b7lF7v898BsReURE+qnqTs99oIh8IyLfAyfhNjwJM9kXd5G6vTD245Ra2PDZz6o63TsfV+R5gNs8qSMw3ZN5hJf3Drjvbrm6kn9ckXibcIrSKGes3zK+uQRXw+yhqjnirHimeH77feHygFqlpJVLQVdjSkkBfez1hxWR+3G11Pa4/RTaAaeIyMeq+rsI8XO8AiMsY/h7TgD6qOpef+DCPTwlUjTv/v+JFnNOMe67y3ifkoQs7n4R8ws8LCIf4MZoZorIb1R1aX5iqj+ISA/P/yER+RTXsnoa17r7WURGUfh9hmUOFZE/RMFzKipn0WsBPlPViwo5uu6/kmzepOC+GaOcsRZBfFMfZ88/R0QG4mplxeLVGLeLSD/P6TIg3DpYjevOATjPFy0TKFobD7MEONyX/oPA1cB/cTXg71S1SzFKoCQ+BW4KX0QaXwBm4pQOOMulQRnu+53hnX/tS+MSYFoZ0ivKN8CJ4sZZEoGLcM/4G2CAiDQWZ3L8fF+ciPkVkfaq+r2qPoIzUNfBfyMRORTYo6rjcBu+dKeg0N/i9dv732VQWotIuLV0EQc+j5nACb4xltoiciTOmFxbEWnvi+vnSFxL0ShnTBHEN+OBniIyB1eALS0lPLhm/N+97qR03DgBuILktyLyNa4POMxUoKNEGCz2aqf1i3TbnAh8hdtqb2bZswR4+9p6g5WLcWMARbkNuENEZuG6qHZGCBOJmiLyDa6P/Hbf/a7wnsllnt9BoW6G1N245/YdME9V3/XcR+GUz/+Aeb5oxeX3Nm/A9ztcTfqjIrfrAszyumfuBf6qbsvL53BdP+/gzLmXlSXACO95NAKeKZLHzbgxkle9MDOBDl4X2bXAB95g8Zoi6Q7EzR4yyhmzPmrEFBG5HchU1QPWEkT5vrVxYyYqIhfiBo6r5Z7WFYm4tSnvq9tcvjzTTQUmqOqg8kzXcNgYgRFrnqFwN0dF0QM3wCrADtxsH6Py0hq4M9ZCVFesRWAYhhHn2BiBYRhGnGOKwDAMI84xRWAYhhHnmCIwDMOIc0wRGIZhxDn/D/cKAfH7pfa/AAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "plt.plot(np.array(range(1,100)), res_cugraph, label='cuGraph GPU Sampler')\n",
    "plt.plot(np.array(range(1,100)), res_pyg_cpu, label='PyG CPU Sampler')\n",
    "plt.xlabel('Fanout (# neighbor nodes sampled)')\n",
    "plt.ylabel('Average Sampling Compute Time (s)')\n",
    "plt.title('2-Hop Neighbor Sampling for MAG Data on A100')\n",
    "plt.legend(loc='center right')\n",
    "plt.grid()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "df = pd.DataFrame({\n",
    "    'res_pyg_cpu': res_pyg_cpu,\n",
    "    'res_cugraph': res_cugraph\n",
    "})\n",
    "\n",
    "df.to_csv('./benchmark1.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.7 ('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": "3.9.13"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "f708a36acfaef0acf74ccd43dfb58100269bf08fb79032a1e0a6f35bd9856f51"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}