Difference between bump vs account.bump in anchor

Question

When declaring a PDA in anchor, what is the difference between

#[account(
    seeds = [SomeSeed],
    bump
)]
pub my_acc: Account<'info, MyAccount>,

and

#[account(
    seeds = [SomeSeed],
    bump = my_acc.bump
)]
pub my_acc: Account<'info, MyAccount>,

I assumed that both of these tell anchor to use the canonical bump, but does one of them just store it in the account/pass it as extra instruction data? Or is the difference something else?

Answer 1

Depending on the seed you can save a lot of CU saving the bumb in an account.

This is because to find a PDA that is off curve internally anchor calls create_program_address repedadly with decreasing bumbs until it finds a PDA thats not on the curve.

Here is one example:

    // 12,136 CUs
    compute_fn! { "Find PDA" =>
        Pubkey::find_program_address(&[b"counter"], ctx.program_id);
    }

    // 1,651 CUs
    compute_fn! { "Find PDA" =>
        Pubkey::create_program_address(&[b"counter", &[248_u8]], &program_id).unwrap();
    }

In anchor:

    // 12,136 CUs when not defining the bump
    #[account(
        seeds = [b"counter"],
        bump
    )]

    // only 1600 if using the bump that is saved in the counter_checked account
    #[account(
        seeds = [b"counter"],
        bump = counter_checked.bump
    )]

So here you would have a u8 field in the counter_checked account and set it when you initialize it. Then you can just always use that cached bump when you call an instruction.

Here is a guide explaining the optimization: https://solana.com/de/developers/guides/advanced/how-to-optimize-compute and here a test you can run to try it yourself: https://github.com/solana-developers/cu_optimizations/blob/main/anchor/counter/anchor/programs/counter/src/lib.rs#L18