The Anatomy of an Impossible Port: Bringing Dead Cells to the R36S
1d 8h ago by lemmy.world/u/PerfectDark in games
So a few weeks back my friend klops: the lead dev of PortMaster, introduced me to a developer called bmdhacks (who was once, long ago, a dev on a couple PS2 games!). bmdhacks had a crazy plan (well by the time I got to chat to him, that plan was very close to complete!) of bringing Dead Cells to the inexpensive R36S retro handheld. Which maybe you'll know, if you're a retro handheld fan, is a device that was never designed to run games of that scale.
What started as a impossible request in the PortMaster world (they had it rated as “Low” in feasibility) turned into a stupidly difficult technical project involving a custom ARM JIT compiler for HashLink, an LLVM-based ahead-of-time compilation backend, and an entirely new decompiler pipeline capable of reconstructing structured code from bytecode.
And did I understand everything he told me when telling me what he did, how he did it, and what was next? No. Not at all.
My article I got to write on it (from all bmdhacks told me) covers the whole process, so if you're into the technical side of porting games? Then you're in for a treat!
Here's my link:
https://gardinerbryant.com/the-anatomy-of-an-impossible-port/
Little edit: bmdhacks is in the comments below, if anyone has any specific question to ask, or comment to share! Yay!
Hi... happy to answer any questions about my work here if this sorta thing interests you.
Nice to see you dropped by!
...even nicer to see you're on Lemmy, you kept this one quiet!
I have an R36s that I installed Rocknix on to give it a tiny bit of an optimization and performance boost and that thing still kinda choked on some PS1 and most PSP games. The thought of it running Dead Cells 💀
Massive props!
Smart usage for AI. (LLM)
yes indeed - I had a go at decompilation a little while ago (wanted to get a mod working on linux) and while it was hella geeky fun, it was very slow and I could tell I didn't really have a hope of achieving my goal. I can really see how an LLM could turbo-charge that process.
I really enjoyed this article and own a R36s myself which is used almost exclusively for postmaster. The variety of games available is mind blowing - I recall someone mentioning on a YT comment that their R36s has taken the place of a steam deck for them and even though that seems hard to believe, its the case for me as well. Nearly all my portable gaming I do is on my R36s playing a portmaster title. The compact form factor and cost of entry really can't be beat.
Not unbelievable at all!
The R36S achieved a strange sort of cult status a long time ago. They managed to break out of a very niche hobby and appeal to regular people. They're sold in kiosks, on TikTok, in regular phone stores around Asia and are now easily the most bought retro handheld (in many variations)
That they bring gaming to so many is a great thing, and now thanks to bmdhacks you can add one more game to the list!
Edit: Apparently my AI-detector is malfunctioning today. Keeping the original comment below for context.
Tried reading the article but it’s clearly written with AI. There might’ve been some editing but the stink of slop is all over it. Disappointing.
Fucking what?
I have not written this with A.I.
This gets so exhausting.
-
Check my post history, filter by my very first posts and go through each and every one of them. I've written (exclusively on and for Lemmy for a solid 6 months or so) for a very long time now. This might give you an indication I write myself.
-
I have each bit of progress from writing this article myself. The framework we run for the site saves each increment, so we can revert if there is an issue down the track. Each. Step. Writing. This. Is. Saved.
-
Its funny how this kind of bullshit claim gets thrown about, these LLMs are trained on writers. What I write might remind you of whatever A.I. you love to use, but the reason is that these take what we write and then replicate it. They are trained on us.
-
I've got plenty of very close friends who I can get to chime in with how trustworthy I am, and how my writing is my own. I'm sure he'll cringe at me saying it, but Jason Evangelho (former lead Linux tech writer for Forbes) is a close friend who will swear to me writing what I do. My good friend Rowan wrote for a big outlet before losing his place recently, he'll do the same. Gardiner himself whose reputation I'd hope would speak for itself after a decade plus in this space will speak for me. The fact I have to defend myself before this one lazy line from you runs away and becomes a thing just makes me sick.
-
I've been interviewing and working with a CEO of a beyond reputable PC gaming company for the last week or so. He and I have been going over the article line-by-line. Literally line-by-line, section-by-section in real time while he sees me write it up. For a feature article coming very soon. More than happy to show him this comment and get his idea on your views. Then edit this and show you what he thinks (after my article is done)
-
In fact I have done like 35-40 (at a guess) interviews with developers in this scene. From Linux to Android to handhelds and everything in-between. Go and find an interview I've done, choose a dev at random and ask them about me. They'll tell you I'm on the level.
-
In fact want me to get bmdhacks to share hus thoughts on your idea of this being generated? I chatted to him for a long time about all of this. I wrote this while asking him about each part. Would that help your shitty claim?
I'm sorry my response is so strong, but it has to be
I hate this shit. Because one lazy claim by you forces me to respond or this ends up becoming the truth, if I ignored it. As a writer that kind of thing is damaging.
Sorry, didn’t mean to cause any stress - clearly I was wrong here. Reading through I saw a lot of sentence structures typical of LLM writing, but like you say this is partly because they were trained on writers’ work.
I’m a writer myself, so I’ve seen first hand how LLMs are rotting our profession from the inside. That’s not an excuse for making false accusations, but I hope you can understand my exasperated tone when I found what seemed like slop on my feed.
We emphatically disallow AI-written articles on our site.
Here: https://gardinerbryant.com/the-bryant-review-style-guide/
Here: https://heavyelement.com/ethics/
I have personally witnessed dash's writing process through various drafts and through my editing passes in my role as Editor in Chief. Dash makes typos and other mistakes where LLMs don't.
...for once my embarrassing unending typos save me -____-
there's nothing in this article that makes me think it's LLM generated, no idea what that guy was on about. It's very well written and readable, which I don't think LLM can really achieve, not that I've ever seen anyway. And it wasn't easy but I did manage to find a minor typo - "All my thanks to bmdhacks for keeping me informed through and of every step he took" 😁
God. Damn. It.
Hahahahahah, its basically tradition that Gardiner finds a bunch, and then someone here on Lemmy points something out.
That one's on me, I'll fix it when I get back in from walking me cat (harness time!) around my yard. Thanks for pointing it out, and for all your kind words!!!