On Tue, May 3, 2022 at 9:26 AM Niels Möller wrote:
> Maamoun TK writes:
>
> > hmm right, didn't cross my mind. I'll add 2^64 -> 2^26 procedure at
> > prologue of _nettle_poly1305_4core() and 2^26 -> 2^64 at epilogue to
> > workaround this.
>
> If possible, I think it would be nice to let
Maamoun TK writes:
> hmm right, didn't cross my mind. I'll add 2^64 -> 2^26 procedure at
> prologue of _nettle_poly1305_4core() and 2^26 -> 2^64 at epilogue to
> workaround this.
If possible, I think it would be nice to let subkeys stored in struct
poly1305_ctx stay in radix-2^64, and compute
On Tue, May 3, 2022 at 8:43 AM Niels Möller wrote:
> Maamoun TK writes:
>
> > I've added Poly1305 optimization based on radix 26 using AVX2 extension
> for
> > x86_64 architecture with fat build support, the patch yields significant
> > speedup compared to upstream.
> >
Maamoun TK writes:
> I've added Poly1305 optimization based on radix 26 using AVX2 extension for
> x86_64 architecture with fat build support, the patch yields significant
> speedup compared to upstream.
> https://git.lysator.liu.se/nettle/nettle/-/merge_requests/46
Cool. Do I get it right,