[dpdk-dev] mbuf pool and ring size...

[dpdk-dev] mbuf pool and ring size...

[Vithal Mohare]

Hi,

While creating mbuf pool, suppose if mbuf-pool size passed to DPDK API is already pow-of-2  [ rte_mempool_create()-->rte_mempool_xmem_create()].  Then, noticed that, corresponding size of the ring created for this pool will be double the size of mbuf-pool.  This is because of below code snippet:
                rte_mempool_xmem_create(...) {
                                ....
r = rte_ring_create(rg_name, rte_align32pow2(n+1), socket_id, rg_flags);  <<<<<<<< Notice that its 'n+1' here,  rte_align32pow2(n+1)
....
                }
Question: why is this 'n+1'? Shouldn't this be just 'n'?

Thanks,
-Vithal

Re: [dpdk-dev] mbuf pool and ring size...

[Olivier MATZ]

Hi Vithal,

On 09/22/2015 07:10 AM, Vithal Mohare wrote:
> Hi,
> 
> While creating mbuf pool, suppose if mbuf-pool size passed to DPDK API is already pow-of-2  [ rte_mempool_create()-->rte_mempool_xmem_create()].  Then, noticed that, corresponding size of the ring created for this pool will be double the size of mbuf-pool.  This is because of below code snippet:
>                 rte_mempool_xmem_create(...) {
>                                 ....
> r = rte_ring_create(rg_name, rte_align32pow2(n+1), socket_id, rg_flags);  <<<<<<<< Notice that its 'n+1' here,  rte_align32pow2(n+1)
> ....
>                 }
> Question: why is this 'n+1'? Shouldn't this be just 'n'?

>From rte_ring_init() documentation:

 The ring size is set to *count*, which must be a power of two.
 The real usable ring size is *count-1* instead of *count* to
 differentiate a free ring from an empty ring.

Therefore, if the user asks for a mempool with 2048 elements,
the ring size has to be 4096.

>From mempool documentation:

 The optimum size (in terms of memory usage) for a mempool is
 when n is a power of two minus 1:  n = (2^q - 1).

Regards,
Olivier