Vorbrodt's C++ Blog

Examples based on this talk.

Below is the old and rusty way of generating random numbers. Don’t do it!

Hilarious and informative talk on C and C++ random number generation:

The following code is a race condition:

The following program, despite looking correct, is a race condition and has unpredictable behaviour:

In the previous post I described Dekker’s algorithm. Its limitation is that it only works for 2 threads. I wanted it to work for N threads, but I can’t atomically check N-1 flags belonging to other threads 🙁 now what?

Dekker’s algorithm is a way of synchronizing two threads around a critical section. It is a dance where thread 1 sets its flag declaring its intent to enter the critical section, then it checks the flag of thread 2.

Long but worth watching. This will take your C++ to the next level 😉

We will jump straight to the code. This innocent looking little program has a major issue (when compiled for release build with optimizations on my Mac using GCC, Apple’s CLANG, and LLVM, as well as on Windows using Visual Studio 2017, and ran on a multicore machine). Can you spot the problem?

As I learn about atomics and memory model I decided to take a stab at rewriting my blocking queue using atomic operations and eliminate the mutex around the critical section of code responsible for pushing and popping elements, effectively creating a fast path

I’ve been reading and learning about the C++ memory model and relaxed atomics and decided to rewrite my Interview question, part 1 using atomic‘s to synchronize the threads.

I’ve been playing around with unordered_map and unordered_set to see how they grow when elements are inserted. Here’s what I learned so far: the default load factor is 1; this means the container will grow to accommodate at most 1 element per bucket. You can change the load factor with max_load_factor call followed by

Docendo discimus.

Seneca the Younger

Did you know that a destructor can be pure virtual? It can be defined as such, but it still needs a body declaration. It is a way of making a class abstract (instances of it cannot be created) without having to make any pure virtual methods. So use it if you have a base

Designed and implemented by Chris M. Thomasson. Complete implementation can be found on GitHub.

All the code from this blog is now on GitHub under MIT License: mvorbrodt/blog

Storing template classes in STL containers is tricky 🙂 If v is a std::vector and TC is a template class, how can we do the following:

C++17 gives us an elegant way to return multiple values from a function using structured binding.

Jonathan Boccara over at Fluent{C++} made a post a while ago titled A Simple Timer in C++. I felt

Given a set of integer ranges defined as [LO, HI) and a value P, find which range P falls into.

My approach to this programming puzzle was to first define a range as a struct that can be sorted (thanks to operator < ), then perform binary search on a vector of sorted ranges. The code

Many thanks to Chris M. Thomasson for rewriting POSIX Threads for Win32 mutex into standard C++ implementation. Using auto_event class from