[dpdk-dev] [PATCH v2 6/7] virtio: simple tx routine

[Xie, Huawei]

On 10/19/2015 12:17 PM, Stephen Hemminger wrote:
> On Sun, 18 Oct 2015 14:29:03 +0800
> Huawei Xie <huawei.xie at intel.com> wrote:
>
>> bulk free of mbufs when clean used ring.
>> shift operation of idx could be further saved if vq_free_cnt means
>> free slots rather than free descriptors.
>>
>> Signed-off-by: Huawei Xie <huawei.xie at intel.com>
> Did you measure this. I finished my transmit optimizations and gets 25% performance improvement
> without any of these restrictions.
Which restriction do you mean?  For the ring layout optimization, this
is the core idea. For the single segment mbuf, this is what all other
PMDs assume for fastest rx/tx path.
With all vhost and virtio optimizations, it could achieve approximately
3~4 times performance improvement(depending on the workload).
Do you mean the indirect feature support or additional optimization not
submitted? Would review your patch this week.
>