{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Chapter 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Timing Julia\n",
    "`@time` is the simplest way to measure execution times"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "@time sqrt.(rand(1000));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`@time` can also be used for any expression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "s=0\n",
    "@time for i=1:1000\n",
    "   global s=s+sqrt(i)\n",
    "end"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`@timev` provides additoinal information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "@timev sqrt.(rand(1000));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`@time` returns the value of the expression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sum(@time sqrt.(rand(1000)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`@elapsed` returns the elapsed execution time as a value ..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "@elapsed sqrt.(rand(1000))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "... and hence can be used in automated tests"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "using Test\n",
    "@test @elapsed(sqrt.(rand(1000))) <= 10e-4"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Profiling"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the `Profile` stdlib module"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "using Profile"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We profile a simple function that creates a random matrix, and then\n",
    "computes the mean of the square of each row."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "using Statistics\n",
    "function randmsq()\n",
    "    x = rand(10000, 1000)\n",
    "    y = mean(x.^2, dims = 1)\n",
    "    return y\n",
    "end"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run the function once before profiling, to compile it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "randmsq()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Execute the function while profiling it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "@profile randmsq()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`print` displays a tree view of the execution traces"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Profile.print()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Clear saved profiles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Profile.clear()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Use the `init` function to configure the profiler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Profile.init(delay=.01)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run the profile using new setting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "@profile randmsq()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Profileview can be used to visualise the\n",
    "`] add ProfileView` to add the package"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "using Pkg\n",
    "Pkg.add(\"ProfileView\")\n",
    "\n",
    "using ProfileView\n",
    "ProfileView.view()\n",
    "ProfileView.svgwrite(\"profile_results.svg\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Statistically significant benchmarking\n",
    "`]add BenchmarkTools` to add the package"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Pkg.add(\"BenchmarkTools\")\n",
    "using BenchmarkTools\n",
    "@benchmark sqrt.(rand(1000))\n",
    "\n",
    "@btime mean(rand(1000));"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*This notebook was generated using [Literate.jl](https://github.com/fredrikekre/Literate.jl).*"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Julia 1.6.0-rc3",
   "language": "julia",
   "name": "julia-1.6"
  },
  "language_info": {
   "file_extension": ".jl",
   "mimetype": "application/julia",
   "name": "julia",
   "version": "1.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 3
}