[Music] I've built a program that lets me use machine instructions directly without typing a single line of code. Instead, I can build my own abstractions using node graphs, which comes with multiple benefits such as easier code reuse and millisecond compilation times. The only downside is that I now have to implement absolutely everything from scratch. I'm not just talking about some basic utility functions here; I'm talking about concepts like the binary encoding of instructions and executable file formats. And to top it all off, I want to use this tool for my Master's thesis at the end of this year, a thesis about real-time computer graphics. So, before 2024 is over, I'll have to work my way from basic arithmetic all the way up to light simulations running on a graphics card. Perfect conditions for a weekly devlog series, if you ask me. To get started, I'll release one short video per week every Wednesday at noon. A shorter format makes consistent uploads less stressful for the next months, which also helps with faster development. The content will be very technical for now.

since I'll implement all the basics before focusing on anything more exciting. I'll obviously work on a video game eventually. It's actually the main motivation behind my thesis topic. There's no point in talking about those plans right now, so I'll get back to that once everything else is set up properly. So if you're not really interested in non-graphic stuff, I'd recommend coming back a bit later. But if you're curious how x64 instructions and the Vulkan API are used behind the scenes, this channel might be for you. Node editors are a popular choice for hiding complexity to benefit user experience. Most game engines utilize node networks to represent shaders, which are little programs that describe the appearance of a surface. They then compile the graphical representation into optimized code for the GPU. My version of such a compiler is very similar, but unlike the specific use cases other software aims for, my editor simply writes sequences of bytes to a given file. It can technically be used to build anything from image and sound files to text documents, and as I've

already mentioned I'll be using it for programming. An executable is just another sequence of bites after all, but before I dive into the first proper implementation, a surface-level introduction might help you follow along. [Music] This is a note. It has a title, a color, and an array of sockets. A socket can either be an input or an output. Inputs can have five different types, four of which allow input of a certain format. Data from other nodes can be provided to a socket through connection, which can only be created between outputs and inputs. The compiler provides built-in nodes that can be categorized into two groups: those that manipulate data and those that define a node's behavior. These built-in nodes can be used to define custom nodes, which effectively hide underlying complexity behind a simple user interface. Custom nodes can then be instantiated like any other node, allowing for easy reuse of completed work. In the end, all the actual compilation algorithm does is traverse a node graph and output the resulting bite sequence into a file. When the compiler encounters a custom node, the defining

node network is traversed recursively until the underlying built-in fixed function nodes are discovered. Since the results of built-in nodes are determined programmatically, recursive compilation eventually ends. We can achieve the millisecond compilation times mentioned earlier by determining which node inputs have changed. If no change is detected, that instance doesn't need to be recompiled, allowing us to skip a lot of work. But details about core mechanics like this are a topic for another video. I'll explain my compiler usage in other devlogs as we go along, so don't worry if this all still appears a bit confusing. So let's start by tackling the most fundamental type of data, the integer.

Writing software without a programming language

Scripts