fn salsaXor(tmp: *align(16) [16]u32, in: []align(16) const u32, out: []align(16) u32) void
fn salsaXor(tmp: *align(16) [16]u32, in: []align(16) const u32, out: []align(16) u32) void { blockXor(tmp, in, 1); salsa8core(tmp); blockCopy(out, tmp, 1); }