Hi,
Thank you for the comments!
Based on the suggestions, I tested the patch for single core L3Fwd performance
with increased number of routes/flows (maximum 8K) to increase cache footprint.
However, I don’t see much improvement with the patch.
> On Jan 21, 2022, at 5:25 AM, Ananyev, Konstantin
>
>
> Hi Dharmik,
> > >
> > >>>
> > Current mempool per core cache implementation stores pointers to mbufs
> > On 64b architectures, each pointer consumes 8B
> > This patch replaces it with index-based implementation,
> > where in each buffer is addressed by (pool base addre
Hi Dharmik,
> >
> >>>
> Current mempool per core cache implementation stores pointers to mbufs
> On 64b architectures, each pointer consumes 8B
> This patch replaces it with index-based implementation,
> where in each buffer is addressed by (pool base address + index)
> I
Hi Konstatin,
> On Jan 13, 2022, at 4:37 AM, Ananyev, Konstantin
> wrote:
>
>
> Hi Dharmik,
>
>>>
Current mempool per core cache implementation stores pointers to mbufs
On 64b architectures, each pointer consumes 8B
This patch replaces it with index-based implementation,
Hi Dharmik,
> >
> >> Current mempool per core cache implementation stores pointers to mbufs
> >> On 64b architectures, each pointer consumes 8B
> >> This patch replaces it with index-based implementation,
> >> where in each buffer is addressed by (pool base address + index)
> >> It reduces the a
Hi Konstatin,
Thank you for your comments and the test report!
> On Jan 10, 2022, at 8:26 PM, Ananyev, Konstantin
> wrote:
>
>
>
>
>> Current mempool per core cache implementation stores pointers to mbufs
>> On 64b architectures, each pointer consumes 8B
>> This patch replaces it with index
> Current mempool per core cache implementation stores pointers to mbufs
> On 64b architectures, each pointer consumes 8B
> This patch replaces it with index-based implementation,
> where in each buffer is addressed by (pool base address + index)
> It reduces the amount of memory/cache require
Current mempool per core cache implementation stores pointers to mbufs
On 64b architectures, each pointer consumes 8B
This patch replaces it with index-based implementation,
where in each buffer is addressed by (pool base address + index)
It reduces the amount of memory/cache required for per core
8 matches
Mail list logo