3 ms·
> I read about an approach a few years ago which needed hundreds of milliseconds of computation for simple operations. Has the state of the art improved? Googl
by ipsum2 5y ago
> I read about an approach a few years ago which needed hundreds of milliseconds of computation for simple operations. Has the state of the art improved?
Google still relies on the same group you read about, probably. From the docs:
> This transpiler connects Google’s XLS library to the TFHE library.
On the TFHE website (https://tfhe.github.io/tfhe/ https://tfhe.github.io/tfhe/):
> Each binary gate takes about 13 milliseconds single-core time to evaluate, which improves [DM15] by a factor 53, and the mux gate takes about 26 CPU-ms.
- disgruntledphd2 5y agoAs a minor pedantic point, this is just someone who works at Google's side project, they have disclaimers in the docs, but it's released by "Google" because of their IP policy, AFAIK.
- ipsum2 5y agoNo, probably not. They have an entire team on it, and has an official Google blog post: https://developers.googleblog.com/2021/06/our-latest-updates-on-fully-homomorphic-encryption.html https://developers.googleblog.com/2021/06/our-latest-updates...
- disgruntledphd2 5y agoYeah, I saw that post later, and was pleasantly surprised that they seem to be putting work into it.