PCG32 RSH-RR with vectorized 64-bit generation

pcg32.zig

const Pcg32 = struct {
  state: u64,

  fn next_u32(self: *Pcg32) u32 {
    const s = self.state;
    self.state = s *% 0x5851f42d4c957f2d +% 0x14057b7ef767814f;

    const xorshifted: u32 = @truncate((s ^ (s >> 18)) >> 27);
    const rot: u5 = @intCast(s >> 59);
    return (xorshifted >> rot) | (xorshifted << -% rot);
  }

  fn next_u64(self: *Pcg32) u64 {
    const s0 = self.state;
    const s1 = s0 *% 0x5851f42d4c957f2d +% 0x14057b7ef767814f;
    self.state = s0 *% 0x685f98a2018fade9 +% 0x1a08ee1184ba6d32;
    const s: @Vector(2, u64) = .{ s0, s1 };

    const mask: @Vector(2, u64) = comptime @splat(0xffffffff);
    const xorshifted = ((s ^ (s >> @splat(18))) >> @splat(27)) & mask;
    const rot: @Vector(2, u5) = @intCast(s >> @splat(59));
    const out = (xorshifted >> rot) | (xorshifted << -% rot);
    
    return (out[0] << 32) | @as(u32, @truncate(out[1]));
  }
};