Jeremy Kun

Looking back on https://www.jeremykun.com/2023/11/15/mlir-a-global-optimization-and-dataflow-analysis/#an-ilp-optimization-pass, I'm not seeing any real obstacle to adapting that ILP to this problem. Basically, after each BGV op in the IR, you add a "InsertRelin" variable...

> Otherwise, you could end-up with a trivial solution that lets the degree explode throughout the computation and only relinearizes at the very end, which would be both very slow...

@ahmedshakill Any progress or help needed on this topic? Feel free to post a draft PR and I can take a look and offer some suggestions.

My rough plan here is to check in polygeist outputs for the transpiler repo examples, and map out what changes are needed to lower them to cggi, and then file...

I can certainly help answer any questions if you have them. Feel free to chime in on Discord as well (see the link at https://heir.dev/community/) for less formal discussions On...

Let's make this issue a bit more specific for anyone reading this issue who isn't familiar with the details. Cf. the Dilithium spec, section 1.2 "Implementation Considerations" (page 5) of...

> This ticket should be split into more manageable pieces. I think they might be: > > * Define appropriate IR constructs to identify polynomial representations in NTT/FFT/coefficient domains. I...

There are other cases like this where there's a longish search that could be done to fix some parameter at the outset. I'm thinking of security parameter selection, how much...

I think there should be a way to avoid tf-lite, and maybe these docs could give a head start https://discuss.tensorflow.org/t/conversion-of-tensorflow-graphdef-to-only-tosa-mlir-using-python-apis/21866/3

What do you think about waiting until we have a unified LWE/RLWE dialect? I agree that lowering to arith is weird, but if you want to go through poly now,...