[RFC][PATCH] flexible array implementation

[Andrew Morton]

On Tue, 21 Jul 2009 14:56:51 -0700
Dave Hansen <dave at linux.vnet.ibm.com> wrote:

> I've added some comments in the kerneldoc for the (newly renamed) alloc
> function:
> 
>  * The maximum number of elements is currently the number of elements
>  * that can be stored in a page times the number of page pointers
>  * that we can fit in the base structure or (using integer math):
>  *
>  *      (PAGE_SIZE/element_size) * (PAGE_SIZE-8)/sizeof(void *)
>  *
>  * Here's a table showing example capacities.  Note that the maximum
>  * index that the get/put() functions is just nr_objects-1.
>  *
>  * Element size | Objects  | Objects |
>  * PAGE_SIZE=4k |  32-bit  |  64-bit |
>  * ----------------------------------|
>  *      1 byte  |  4186112 | 2093056 |
>  *      2 bytes |  2093056 | 1046528 |
>  *      3 bytes |  1395030 |  697515 |
>  *      4 bytes |  1046528 |  523264 |
>  *     32 bytes |   130816 |   65408 |
>  *     33 bytes |   126728 |   63364 |
>  *   2048 bytes |     2044 |   10228 |
>  *   2049 bytes |     1022 |     511 |
>  *       void * |  1046528 |  261632 |

4-bytes on 32-bit and 8-bytes on 64-bit are the most interesting ones
(IMO).  So what we're basically saying is "2MB on 64-bit".

I wonder if that's enough for known likely callers.  Hopefully it is.