C++ Performance & Best Practices

Performance Optimization

• Cache Locality: Access data sequentially — std::vector (contiguous) is 10-100x faster than std::list (scattered) for iteration due to CPU cache lines
• Compile-Time Computation: constexpr functions compute at compile time — lookup tables, math constants, hash functions. Zero runtime cost
• Small Buffer Optimization (SBO): std::string stores short strings (< 16-24 chars) directly in the object, no heap allocation. Same for std::function
• Reserve & Emplace: vector.reserve(n) avoids reallocations. emplace_back(args) constructs in-place instead of constructing + moving
• Branch Prediction: Keep hot paths in if (likely) branches. Use [[likely]]/[[unlikely]] attributes. Avoid unpredictable branches in tight loops
• Benchmarking: Use Google Benchmark — prevents compiler dead-code elimination, measures nanoseconds accurately, statistical analysis of results

Try It Yourself: Modern C++ Features

#include <iostream>
#include <optional>
#include <string>
#include <vector>
#include <algorithm>
using namespace std;

// C++17: std::optional
optional<string> findUser(int id) {
    if (id == 1) return "Alice";
    if (id == 2) return "Bob";
    return nullopt;
}

// C++17: Structured bindings
struct Point { double x, y; };

int main() {
    // std::optional
    auto user = findUser(1);
    if (user) cout << "Found: " << *user << endl;
    else cout << "User not found" << endl;

    // Structured bindings
    Point p{3.0, 4.0};
    auto [x, y] = p;
    cout << "Point: (" << x << ", " << y << ")" << endl;

    // Range-based for with init
    vector<int> nums = {5, 2, 8, 1, 9};
    sort(nums.begin(), nums.end());
    cout << "Sorted: ";
    for (auto n : nums) cout << n << " ";
    cout << endl;
    return 0;
}

Quick Quiz — Modern C++