Operation tricks with packed structs with explicit size

Help
1

I saw some complicated (new?) stories somewhere about packed structs with explicit size like:

const Pack = packed struct (u8)
{
    field1: bool,
    field2: bool,
    field3: bool,
}

What is the difference when omitting the (u8)?
What is the explicit meaning of the added (u8)?

Then some question about performance (yes I still am not able to read that assembler view in godbolt)…

Is there a possiblity to bitwise AND / OR / XOR etc. two structs? That would be nice…
If not: what is the smartest way to do it?

Is the compiler so extremely smart that when executing:

pack.field1 = true;
pack.field2 = false;
pack.field3 = true;

to immediately stuff in the binary byte instead of executing 3 statements?

2

A backing integer is inferred from the total bit size of the fields.
When you provide a backing integer–after packed struct–zig compiler enforces it and an error is displayed if the inferred bit size is different from the provided one.

3

It is possible using bitCasting to and from the backing type:

const xorResult: Pack = @bitCast(@as(u8, @bitCast(a)), @as(u8, @bitCast(b)));

Bitcasts of small values like this are free.

I suggest using godbolt to answer such questions. In this godbolt example, you can see that your assumption is correct and the compiler just inserts 5 (101₂) into the struct directly, however the compiler is doing some extra work to preserve the padding bits of the struct. In this case a direct initialization is better.

4 Likes
4

Absolutely crazy, Nice.
Now I see the trick: using export to just see the function.

1 Like