9 ms·
>Requirements >A will to live (optional but recommended) >LLVM is NOT required. BarraCUDA does its own instruction encoding like an adult. >Open an issue if
by h4kunamata 8mo ago
>Requirements
>A will to live (optional but recommended)
>LLVM is NOT required. BarraCUDA does its own instruction encoding like an adult.
>Open an issue if theres anything you want to discuss. Or don't. I'm not your mum.
>Based in New Zealand
Oceania sense of humor is like no other haha
The project owner strongly emphasize the no LLM dependency, in a world of AI slope this is so refreshing.
The cheer amount of knowledge required to even start such project, is really something else, and prove the manual wrong on the machine language level is something else entirely.
When it comes to AMD, "no CUDA support" is the biggest "excuse" to join NVIDIA's walled garden.
Godspeed to this project, the more competition the less NVIDIA can continue destroying the PC parts pricing.
- querez 8mo ago> The project owner strongly emphasize the no LLM dependency, in a world of AI slope this is so refreshing. The project owner is talking about LLVM,a compiler toolkit, not an LLM.
- kmaitreys 8mo agoIt's actually quite easy to spot if LLMs were used or not. Very few total number of commits, AI like documentation and code comments. But even if LLMs were used, the overall project does feel steered by a human, given some decisions like not using bloated build systems. If this actually works then that's great.
- butvacuum 8mo agoSince when is squashing noisesum commits an AI activity instead of good manners?
- kmaitreys 8mo agoCan you prove that this is what happened?
- sigmoid10 8mo agoThe first commit was 17k lines. So this was either developed without using version control or at least without using this gh repo. Either way I have to say certain sections do feel like they would have been prime targets for having an LLM write them. You could do all of this by hand in 2026, but you wouldn't have to. In fact it would probably take forever to do this by hand as a single dev. But then again there are people who spend 2000 hours building a cpu in minecraft, so why not. The result speaks for itself.
- saidnooneever 8mo agoa lot of ppl dont use git. and just chuck stuff in there willynilly when they want to share it. people are to keen to say something was produced with an LLM if they feel its something they cannot produce themselves readily..
- kmaitreys 8mo agoI would be very concerned about someone working on a 16k loc codebase without a VCS.
- embedding-shape 8mo ago> The first commit was 17k lines. So this was either developed without using version control or at least without using this gh repo. Most of my free-time projects are developed either by my shooting the shit with code on disk for a couple of months, until it's in a working state, then I make one first commit. Alternatively, I commit a bunch iteratively, but before making it public I fold it all into one commit, which would be the init. 20K lines in the initial commit is not that uncommon, depends a lot on the type of project though. I'm sure I'm not alone with this sort of workflow(s).
- kmaitreys 8mo agoCan you explain the philosophy behind this? Why do this, what is the advantage? Genuinely asking, as I'm not a programmer by profession. I commit often irrespective of the state of the code (it may not even compile). I understand git commit as a snapshot system. I don't expect each commit to be pristine, working version. Lot of people in this thread have argued for squashing but I don't see why one would do that for a personal project. In large scale open source or corporate projects I can imagine they would like to have clean commit histories but why for a personal project?
- natvert 8mo agoSays the clawdbot
- kmaitreys 8mo agoIt's quite amusing the one time I did not make an anti-AI comment, I got called a clanker myself. I'm glad the mood here is shifting towards the right side.
- luckydata 8mo agothis type of project is the perfect project for an llm, llvm and cuda work as harnesses, easy to compare.
- kmaitreys 8mo agoWhat do you mean by harnesses?
- formerly_proven 8mo agoagentic ai harness for harness (ai)
- nurettin 8mo agoThey also said "hand written", implying that no LLMs whirred, slopped and moonwalked all over the project.
- jorvi 8mo agoI mean.. I'm one of the staunchest skeptics of LLMs as agents, but they're amazing as supercharged autocomplete and I don't see anything wrong with them in that role. There's a position between handwritten and slopped that's pareto.
- wild_egg 8mo agoThis project very most definitely has significant AI contributions. Don't care though. AI can work wonders in skilled hands and I'm looking forward to using this project
- ZaneHam 8mo agoHello! I didn't realise my project was posted here but I can actually answer this. I do use LLM's (specifically Ollama) particularly for test summarisation, writing up some boilerplate and also I've used Claude/Chatgpt on the web when my free tier allows. It's good for when I hit problems such as AMD SOP prefixes being different than I expected.
- bigyabai 8mo ago> and prove the manual wrong on the machine language level I'll be the party pooper here, I guess. The manual is still right, and no amount of reverse-engineering will fix the architecture AMD chose for their silicon. It's absolutely possible to implement a subset of CUDA features on a raster GPU, but we've been doing that since OpenCL and CUDA is still king. The best thing the industry can do is converge on a GPGPU compute standard that doesn't suck. But Intel, AMD and Apple are all at-odds with one another so CUDA's hedged bet on industry hostility will keep paying dividends.
- magicalhippo 8mo ago> Oceania sense of humor is like no other haha Reminded me of the beached whale animated shorts[1]. [1]: https://www.youtube.com/watch?v=ezJG0QrkCTA&list=PLeKsajfbDpk3aqfpDB7_6bXL-ZS3UGKNC https://www.youtube.com/watch?v=ezJG0QrkCTA&list=PLeKsajfbDp...
- samrus 8mo ago> >LLVM is NOT required. BarraCUDA does its own instruction encoding like an adult. > The project owner strongly emphasize the no LLM dependency, in a world of AI slope this is so refreshing. "Has tech literacy deserted the tech insider websites of silicon valley? I will not beleove it is so. ARE THERE NO TRUE ENGINEERS AMONG YOU?!"
- typh00n 8mo agoi loled hard in public transport
- dboreham 8mo agoUnrelated: just returned from a month in NZ. Amazing people.
- ZaneHam 8mo agoHope you enjoyed it!!
- colordrops 8mo agoI'm still blown away that AMD hasn't made it their top priority. I've said this for years. If I was AMD I would spend billions upon billions if necessary to make a CUDA compatibility layer for AMD. It would certainly still pay off, and it almost certainly wouldn't cost that much.
- andy_ppp 8mo agoMoving target, honestly just get PyTorch working fully (loads of stuff just doesn’t work on AMD hardware) and also make it work on all graphics cards from a certain generation. The matrix of support needed GFX cards, architectures and software together is quite astounding but still yes that should have at least that working and equivalent custom kernels.
- colordrops 8mo agoThat would be a great start.
- spacebanana7 8mo agoThe headline that PyTorch has full compatibility on all AMD GPUs would increase their stock by > $50 billion overnight. They should do it even if it takes 500 engineers and 2 years.
- DonThomasitos 8mo agoDoes anybody really understand why this hasn‘t been done? I know about ongoing efforts but is it really THAT difficult?
- andy_ppp 8mo agoYou know it’s probably a combination of things but mostly that AMD do not have a capable software team… probably not the individuals but the managers likely don’t have a clue.
- ekianjo 8mo agoLLVM, nothing to do with LLMs
- freakynit 8mo ago"If this doesn't work, your gcc is broken, not the Makefile." ... bruh.. the confidence.
- lambda 8mo ago> The project owner strongly emphasize the no LLM dependency, in a world of AI slope this is so refreshing. Huh? This is obvious AI slop from the readme. Look at that "ASCII art" diagram with misaligned "|" at the end of the lines. That's a very clear AI slop tell, anyone editing by hand would instinctively delete the extra spaces to align those.
- RockRobotRock 8mo ago>No LLVM. No HIP translation layer. No "convert your CUDA to something else first." Just ...... Another obvious tell. https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing#Rule_of_three https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing#... https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing#Negative_parallelisms https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing#...
- ZaneHam 8mo agoOh gosh, Emdashes are already ruined for me and now I can't use that to? I've already had to drop boldface in some of my writings because it's become prolific too. This is also just what I intentionally avoided when making this by the way. I don't really know how else to phrase this because LLVM and HIP are quite prolific in the compiler/GPU world it seems.
- cmdr2 8mo agoFor what it's worth - for people who're in this space - this project is awesome, and I hope you keep going with it! The compiler space for GPUs really needs truly open-source efforts like this. Code/doc generators are just another tool. A carpenter uses power tools to cut or drill things quickly, instead of screwing everything manually. That doesn't mean they're doing a sloppy job, because they're still going to obsessively pore over every detail of the finished product. A sloppy carpenter will be sloppy even without power tools. So yeah, I don't think it's worth spending extra effort to please random HN commenters, because the people who face the problem that you're trying to solve will find it valuable regardless. An errant bold or pipe symbol doesn't matter to people who actually need what you're building.
- dirasieb 8mo agoit's so funny to watch the people who pearl clutch over AI expose that they don't even know the difference between LLVM and LLM rofl
- amelius 8mo ago> LLVM is NOT required. BarraCUDA does its own instruction encoding like an adult. This is not an advantage since you will now not benefit from any improvements in LLVM.
- viraptor 8mo agoNor will they be restricted by the LLVM design. That project is huge and generic trying to be everything and target everything (and takes ages to rebuild if you need some changes). Sometimes it's better to go simple and targeted - time will tell if that's the right choice. Zluda used LLVM and ended up bundling a patched version to achieve what they wanted https://vosen.github.io/ZLUDA/blog/zluda-update-q4-2025/ https://vosen.github.io/ZLUDA/blog/zluda-update-q4-2025/ > Although we strive to emit the best possible LLVM bitcode, the ZLUDA compiler simply is not an optimizing, SSA-based compiler. There are certain optimizations relevant to machine learning workloads that are beyond our reach without custom LLVM optimization passes.
- amelius 8mo agoThat's a much better argumentation than "we did it because we are adults". (except that it applies to Zluda, not necessarily this project)
- pezgrande 8mo agoI think Zig is trying to get rid of it as well, much harder to debug iirc.
- deeringc 8mo ago> /* 80 keywords walk into a sorted bar */ https://github.com/Zaneham/BarraCUDA/blob/master/src/lexer.c#L4 https://github.com/Zaneham/BarraCUDA/blob/master/src/lexer.c...
- renewiltord 8mo ago> in a world of AI slope The scientific term for this is “gradient descent”.
- HPsquared 8mo agoThe Descent of (artificial) Man.
- kittbuilds 8mo ago[dead]
- moffkalast 8mo ago>A will to live (optional but recommended) Ah I'm glad it's just optional, I was concerned for a second.