Make poly1305 faster

[jedisct1]

This makes our poly1305 implementation slightly faster

crypto/benchmark (release-fast):

• Before: poly1305: 1851 MiB/s
• After: poly1305: 2119 MiB/s

[andrewrk]

Btw @jedisct1 are you aware of Zig's SIMD language features?

[jedisct1]

Btw @jedisct1 are you aware of Zig's SIMD language features?

Hi Andrew,

I'm aware that @Vector exists, but I haven't used it yet. But I really want to :)

Are shuffle instructions available? How about AESNI/CLMUL? Is there a way to access intrinsics without using inline assembly?

[ifreund]

Are shuffle instructions available? How about AESNI/CLMUL? Is there a way to access intrinsics without using inline assembly?

We've got @shuffle for shuffle instructions. Don't think there's any support for AESNI/CLMUL aside from inline assembly yet. Perhaps this example Protty made a while back would help clarify how SIMD vectors work in zig: https://zig.godbolt.org/z/oSpjR8. You may also want to check out #903

[andrewrk]

Are shuffle instructions available?

shuffle

How about AESNI/CLMUL?

This is the first I've heard of those, I will have to look into this and get back to you.

Is there a way to access intrinsics without using inline assembly?

Yes. Zig's SIMD is a higher level abstraction than inline assembly intrinsics.