News for hackers. Stuff that matters.
Re:How Taalas "prints" LLM onto a chip? by amelius on Sunday February 22, 2026 @06:45AM (#47110250)
I'm looking forward to the model.toVHDL() method in PyTorch.
Re:How Taalas "prints" LLM onto a chip? by androiddrew on Sunday February 22, 2026 @09:33AM (#47111322)
Is this a thing?
Re:How Taalas "prints" LLM onto a chip? by mikeurbach on Sunday February 22, 2026 @02:35PM (#47113910)
I gave a short talk about compiling PyTorch to Verilog at Latte '22. Back then we were just looking at a simple dot product operation, but the approach could theoretically scale up to whole models.

https://capra.cs.cornell.edu/latte22/paper/2.pdf

https://www.youtube.com/watch?v=QxwZpYfD60g

Re:How Taalas "prints" LLM onto a chip? by Simboo on Sunday February 22, 2026 @10:07AM (#47111592)
Deep Differentiable Logic Gate Networks
Re:How Taalas "prints" LLM onto a chip? by thesz on Sunday February 22, 2026 @05:28PM (#47115435)
I see you and I raise approximate logic synthesis [1] [2].

[1] https://www.sciencedirect.com/science/article/pii/S138376212...

[2] https://arxiv.org/abs/2506.22772

You can synthesize a logic circuit that is as complex as it gets to have a certain accuracy.

Deep differentiable logic networks, in my experience, do not scale well for larger (more inputs) logic elements. One still has to apply logic optimization and synthesis afterwards. So why not to synthesize ones own approximate circuit to the accuracy one's desire?

Re:How Taalas "prints" LLM onto a chip? by sowbug on Sunday February 22, 2026 @12:50PM (#47113035)
Ugh, quick, everyone start panic-buying FPGAs now.
Re:How Taalas "prints" LLM onto a chip? by throwup238 on Sunday February 22, 2026 @02:50PM (#47114026)
largest FPGAs have on the order of tens of millions of logic cells/elements. They’re not even remotely big enough to emulate these designs except to validate small parts of it at a time and unlike memory chips or GPUs, companies don’t need millions of them to scale infrastructure.

(The chips also cost tens of thousands of dollars each)

Re:How Taalas "prints" LLM onto a chip? by 8note on Sunday February 22, 2026 @03:10PM (#47114190)
they also arent power friendly
Re:How Taalas "prints" LLM onto a chip? by p0u4a on Sunday February 22, 2026 @11:34PM (#47118187)
Pretty close to what you describe: https://github.com/fastmachinelearning/hls4ml