It used to be that building a game also meant building all the authoring tools to go along with it. With the advent and spread of game engines like Unity, Unreal, Godot (and literally hundreds of others) along with amazing tools like Photoshop and Blender, the need to make your own tooling has dramatically decreased. Almost to the point that in a majority of cases, you probably don’t need to write tools.
Even if you do find you can’t use an existing tool, others suggest using chatGPT to either extend an existing tool or a tool in the engine you’re using via their SDK. Let AI do the work for you since tools are not shipping code nor need to be overly performant.
Strict_Bench_6264 wrote up a whole blog article to describe what he learned:
After analyzing nearly 10 years of CVEs, Google researchers calculated that at least 40% of safety exploits in C++ were related to spatial memory exploits like writing to an out-of-bounds memory location.
Google researchers showed they were able to “retrofit” spatial safety onto their C++ codebases, and to do it with a surprisingly low impact on performance. They used straightforward strategies such as bounds checking buffers and data structures – as is done in other languages and released a new, safer Hardened libc++.
The results show up in this chart of segfaults across the entire fleet of computers before and after using the improvements. Their internal red team testing results were also much improved, uncovered over 1000 bugs and likely prevent 1000-2000 new bugs each year based on current development rate.
Andreas from Insomniac Games made a Amiga 500 demo in 2019 as part of this work with The Black Lotus demo group. He presented not only the Eon Amiga 500 demo, but tons of great technical information about the 4 years it took to develop it.
Old demo scene programmers hold amazing amounts of wisdom. When solving the core pieces of logic, I found this is true (but when doing larger, complete system development, these don’t work)
Work backwards from desired outcome to discover your constraints. Don’t just brute force. Instead, ask, what must be in place for us to get the peak performance from the key component we’re dependent on (render, disk load, etc). Then work from that constraint.
Do everything you can at compile time, not run time. Pre-compute tons of things – even the build-up of the data structures in memory. Just run it and then save and reload that blob automatically.
Over-generalizing early is a trap many devs fall into. Solve the problem in front of you. Trust that you can delete the code and do something else if it sucks. It’s cheaper and faster than trying to anticipate things ahead of time. Do the simplest thing that will work and if it sucks come back and delete it.
If you end up with a small runtime table/code that doesn’t require runtime checks because you can prove it can’t go wrong, you’re doing something right.
When developing, the actual Amiga is super slow and limited. They took an Amiga emulator and hacked it up so they could debug on it instead. Using calltraps to trigger the emulator, they added memory protection, fast forward, trigger debug, loading symbols, cycle accurate profiling, single step, high-resolution timers, etc. Also allows perfect input playback.
Modern threading and consumer/producer components (disk loading, data transfer, decompressors, etc) often just throw things in buffers and YOLO. There’s no clear backpressure to show you where you’re wasting time/space. Running on this kind of hardware/simulator shows you how much time the design is wasting by poorly and inefficiently designed algorithms/constraints.
When teaching myself to program as a kid, my first language was type-in BASIC programs. After that, I made the very un-orthodox choice to learn assembly. I wrote a small database, a TSR (Terminate and stay resident program), and a couple other small creations.
Looks like GreatCorn did one better by writing his own game in x86 assembly.
Using AI as a character animation compression algorithm
Daniel Holden from Ubisoft gave this great talk at GDC 2018 on how data-driven analysis of their character animation control system turned into a AI system that vastly reduced the complexity and manpower involved in building an animation system for character control.
The world has gotten very familiar to retro hardware re-creations, game emulation, re-releases, speed runs, creating new games for old platforms, as well as new exploits, tools, and discoveries. The nitty gritty work of doing all of this, however, is a labor of love. For those that dig into the binary, there’s tricky copyright concerns that need to be managed, only scraps of information about old hardware and software, highly optimized/tricky code that is tough to read, and almost no financial gain – except for commercial re-releases.
Made Up of Wires walks us through a live bit of decompiling of the PS1 classic: Castlevania: Symphony of the Night to give you a taste of the work involved in this kind of work. Not really that different than any other reverse engineering but surprisingly accessible as these old games were relatively small and simple.
Back in the day, I worked on this little project called Larrabee – which later turned into the Intel Xeon Phi coprocessor. It was an ambitious and exciting platform. It consisted of a ton of 512 bit wide instructions to operate like a lot of streaming GPU architectures, yet was fully general purpose x86.
It turned out that getting performance out of this hardware was difficult. In order to get the full potential of the hardware, you simply had to utilize the vector units. Without that, it is like writing a single threaded app on a 8 core system. Single SIMD lane operation just wasn’t going to cut it as was written about in 2017 International Journal of Parallel Programming article:
“Our results show that, although the Xeon Phi delivers a relatively good speedup in comparison with a shared-memory architecture in terms of scalability, the relatively low computing power of its computational units when specific vectorization and SIMD instructions are not fully exploited makes this first generation of Xeon Phi architectures not competitive”
The paper, and the host of others linked on the page as references, are a good read and gives some hints why fixed-function GPUs have an advantage when it comes to raw streaming throughput. Hint: cache and data flow behavior is as, if not more, important as utilizing vectorization in such architectures.
Ever want to know what it’s like to work in a game studio? Double Fine has released a 33 episode series called PsychOdyssey which shows them developing Psychonauts 2 over 7 years.
Acerola created a graphics to text shader. He talks about a number of interesting techniques beyond just ASCII lookups such as: edge detection, depth color falloff, blooming and tone mapping, color tones, color quantization, and other filters, etc. Definitely worth a watch and you can check out a good amount of his code on his github repro.
