Re: [patch][rfc] acpi: do not use kmem caches

["Pekka Enberg" <penberg@cs.helsinki.fi>]

On Mon, Dec 1, 2008 at 7:36 PM, Alexey Starikovskiy <aystarik@gmail.com> wrote:
>> [ The size of ACPI kmem caches with wasted bytes per object in
>> parenthesis. ]
>>
>>                 32-bit size  64-bit size
>>  Acpi-Namespace  24 (8)       32 (0)
>>  Acpi-Operand    40 (24)      72 (24)
>>  Acpi-Parse      32 (0)       48 (16)
>>  Acpi-ParseExt   44 (20)      72 (24)
>>  Acpi-State      44 (20)      80 (16)
>>
>> Though I suspect this situation could be improved by avoiding those
>> fairly big unions ACPI does (like union acpi_operand_object).
>
> No, last time I checked, operand may get down to 16 bytes in 32-bit case --
> save byte by having 3 types of operands... and making 2 more caches :)

I'm not sure what you mean. I wasn't suggesting adding new caches but
instead, avoid big unions and allocate plain structs with kmalloc()
instead. If you look at union acpi_operand_object, for example, it's
such a bad fit on 64-bit (72 bytes) only because of struct
acpi_object_mutex. Other structs in that union fit in a kmalloc-64
cache just fine.

So really, ACPI should probably be fixing the unions rather than paper
over the problem by adding new kmem caches.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>