Uriopass
Joined 6 years ago
Joined 6 years ago
I've worked with ranking algorithms at scale (5k reqs/s) and hand tuned, manually crafting features, searching through user logs to see what went wrong. And now we can just pop an LLM on top of that pile of data and it's just better... End of an era, to me.
They do once you restart the server. Unsure what the "phase 3 monitoring" shows where the goroutines go down gradually. If you have new code to deploy you must recompile and therefore restart and those goroutines are gone anyway.
This feels like an AI made-up story but I'm not sure I understand the point of making this up.
However, goroutine leak is interesting! I hope what I learned from this post isn't hallucination. For example, how could the subscriber send messages/heartbeats to the closed goroutine without an error...
It doesn't pass the white furnace test.
That made me realize I didn't understand much of the math I implemented. Random multiplications by pi and 2 as the author stated didn't work out.
In practice however I didn't see any real good generated 3D asset for now. The problem is probably much harder as there are far more 2D assets than 3D, especially with PBR support.
Every so often, citizens and companies (I call the abstraction a "Soul") look through everything they can do and pick the on with the highest utility score.
If they're hungry, they'll get food. If it's the time of day where they should work, they'll go to work, except if they're too hungry!
All the blocks are there and the language is really well suited to games.
On top of my head:
The pros:
- The crate ecosystem and the package manager makes it really easy to integrate any useful component such as pathfinding, spatial partitioning, graphics backend, audio system.. Most crates take a lot of effort to be cross-platform so I can develop on linux and not spend too much time debugging windows releases.
- The strong typing and algebraic data types makes expressing the game state very pleasant. I also found I was able to develop a very big game without too many bugs even though I don't write many tests.
- Ahead of time compilation + LLVM guarantees you won't have to optimise for weird things around a virtual machine. Rust gives you more control to optimise hot loops as you can go low-level.
- I find wgpu to be the perfect balance between ergonomics and power compared to Vulkan. OpenGL support through wgpu is also a nice addition for lower end devices.
- The Rust community is very helpful, you can often talk directly to crate maintainers
The cons:
- Compilation times, when compared to JITed languages such as C# can be very painful. It can be alleviated by buying a 3950X but I still often get 10-30s iteration times.
- The static nature of Rust means you often need a dynamism layer above to tweak stuff that can be awkward to manage. I made inline_tweak for this purpose but it's really far from how easy Unity makes it. https://github.com/Uriopass/inline_tweak
- Since Rust feels very ergonomic, you are tempted to write almost all game logic within it, so mod support feels very backwards to implement as you cannot really tweak "everything" like in Unity games. Thankfully "Systems" game like Factorio or Egregoria can be theoretically split into the "simulation" and the "entities" so mod can still have a great impact. Factorio is built in C++ so has the same problematic. Their Lua API surface is quite insane to be able to hook into everything. https://lua-api.factorio.com/latest/
Now, I have to talk about Bevy: https://bevyengine.org/. It did not exist when I started but it is a revolution in the Rust gamedev space. It is a very powerful 100% Rust game engine that makes you write game code in Rust too. It has incredible energy behind it and I feel like if I'd used Bevy from the start I wouldn't have had to develop many core engine systems. Its modular design is also incredibly pleasant as you can just replace any part you don't like with your own.