2014-10-29 14:38:42

by Tejun Heo

Subject: Re: [PATCH] idr: optimize ida_init() to avoid an extra memset

On Wed, Oct 29, 2014 at 05:26:34PM +0800, [email protected] wrote:
> The memset in ida_init() already handles idr, so there's some
> redundancy in the following idr_init().
>
> This patch removes the memset, and clears ida->free_bitmap instead.
>
> Signed-off-by: pang.xunlei <[email protected]>
> ---
> lib/idr.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/lib/idr.c b/lib/idr.c
> index e654aeb..bbe5779 100644
> --- a/lib/idr.c
> +++ b/lib/idr.c
> @@ -1141,8 +1141,7 @@ EXPORT_SYMBOL(ida_simple_remove);
> */
> void ida_init(struct ida *ida)
> {
> - memset(ida, 0, sizeof(struct ida));
> idr_init(&ida->idr);
> -
> + ida->free_bitmap = NULL;

I don't know. Does this matter? If this *really* matters, I'd much
rather have memset(&ida->FIRST_FIELD, 0, sizeof(struct ida) - offset
of FIRST_FIELD) to ensure that all fields get reset or implement an
internal function like __idr_init_without_zeroing(); however, given
the size of an idr and the low frequency of the operation, I'd prefer
to just leave it as-is.

Thanks.

--
tejun

2014-10-30 02:03:48

by Lai Jiangshan

Subject: Re: [PATCH] idr: optimize ida_init() to avoid an extra memset

On 10/29/2014 10:38 PM, Tejun Heo wrote:
> On Wed, Oct 29, 2014 at 05:26:34PM +0800, [email protected] wrote:
>> The memset in ida_init() already handles idr, so there's some
>> redundancy in the following idr_init().
>>
>> This patch removes the memset, and clears ida->free_bitmap instead.
>>
>> Signed-off-by: pang.xunlei <[email protected]>
>> ---
>> lib/idr.c | 3 +--
>> 1 file changed, 1 insertion(+), 2 deletions(-)
>>
>> diff --git a/lib/idr.c b/lib/idr.c
>> index e654aeb..bbe5779 100644
>> --- a/lib/idr.c
>> +++ b/lib/idr.c
>> @@ -1141,8 +1141,7 @@ EXPORT_SYMBOL(ida_simple_remove);
>> */
>> void ida_init(struct ida *ida)
>> {
>> - memset(ida, 0, sizeof(struct ida));
>> idr_init(&ida->idr);
>> -
>> + ida->free_bitmap = NULL;
>
> I don't know. Does this matter? If this *really* matters, I'd much
> rather have memset(&ida->FIRST_FIELD, 0, sizeof(struct ida) - offset
> of FIRST_FIELD) to ensure that all fields get reset or implement an
> internal function like __idr_init_without_zeroing(); however, given
> the size of an idr and the low frequency of the operation, I'd prefer
> to just leave it as-is.
>

memset(ptr, 0, sizeof()) or kzalloc() is convenient and good for buffers
but bad for structures, objects...

general way for object initialization is:

xxx_init()
{
explicitly init every field...

/* maybe complicated, over elaborate */
}