On Mon, Nov 12, 2018 at 5:06 PM Juliusz Chroboczek <[email protected]> wrote: > > >> Yeah, we should just include an implementation of SHA-256 in the code. > > > There's also the option... > > Given that the main selling point of HMAC vs. DTLS is that it has no > dependencies, it wouldn't be particularly wise to make the reference > implementation depend on a Linux-specific library. > > Of course, we'll make it easy for distributors to replace our bundled > implementation of SHA-256 with whatever it is they find convenient. (I > expect Dave to replace it with his hand-written implementation that > interleaves MMX with AVX2 while simultaneously performing modular > multiplication in the U pipe to hash 27.7 bytes per cycle per core.) > > (If you know what the U pipe is, you're old.)
I haven't had fun like that in ages! :) And you forgot that you can still use the old FPU stack for stuff, if you are careful. But I'd rather see babel scale past 64k routes, and I really have come to understand that it's algorithmic first, asm tweaks second. > > -- Juliusz > > _______________________________________________ > Babel-users mailing list > [email protected] > https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/babel-users -- Dave Täht CTO, TekLibre, LLC http://www.teklibre.com Tel: 1-831-205-9740 _______________________________________________ Babel-users mailing list [email protected] https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/babel-users
