Re: [PATCH v3 3/3] rev-list: add commit object support in `--missing` option

From: Junio C Hamano <gitster@pobox.com>
Date: Oct 19, 2023, 22:05 UTC
Message-ID: <xmqq4jimuv26.fsf@gitster.g>
In-Reply-To: <20231019121024.194317-4-karthik.188@gmail.com>

Karthik Nayak <karthik.188@gmail.com> writes:

Show 21 quoted lines

> diff --git a/object.h b/object.h
> index 114d45954d..b76830fce1 100644
> --- a/object.h
> +++ b/object.h
> @@ -62,7 +62,7 @@ void object_array_init(struct object_array *array);
>  
>  /*
>   * object flag allocation:
> - * revision.h:               0---------10         15             23------27
> + * revision.h:               0---------10         15             22------28
>   * fetch-pack.c:             01    67
>   * negotiator/default.c:       2--5
>   * walker.c:                 0-2
> @@ -82,7 +82,7 @@ void object_array_init(struct object_array *array);
>   * builtin/show-branch.c:    0-------------------------------------------26
>   * builtin/unpack-objects.c:                                 2021
>   */
> -#define FLAG_BITS  28
> +#define FLAG_BITS  29
>  
>  #define TYPE_BITS 3

I am afraid that this is not a good direction to go, given that the way FLAG_BITS is used is like this:

    /*
     * The object type is stored in 3 bits.
     */
    struct object {
            unsigned parsed : 1;
            unsigned type : TYPE_BITS;
            unsigned flags : FLAG_BITS;
            struct object_id oid;
    };

28 was there, not as a random number of bits we happen to be using. It was derived by (32 - 3 - 1), i.e. ensure the bitfields above are stored within a single word.

sizeof(struct object) is 40 bytes on x86-64, with offsetof(oid) being 4 (i.e. the bitfields fit in a single 4-byte word). If we make FLAG_BITS 29, we will add 4 bytes to the structure and waste 31-bit per each and every in-core objects.

Do we really need to allocate a new bit in the object flags, which is already a scarce resource?