Re: [dpdk-stable] [dpdk-dev] [PATCH 1/2] net/qede: fix performance bottleneck in Rx path

Re: [dpdk-stable] [dpdk-dev] [PATCH 1/2] net/qede: fix performance bottleneck in Rx path

[Shahed Shaikh]

> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@intel.com>
> Sent: Friday, January 18, 2019 8:11 PM
> To: Shahed Shaikh <shshaikh@marvell.com>; dev@dpdk.org
> Cc: stable@dpdk.org; Rasesh Mody <rmody@marvell.com>; Thomas Monjalon
> <thomas@monjalon.net>
> Subject: [EXT] Re: [dpdk-dev] [PATCH 1/2] net/qede: fix performance bottleneck
> in Rx path
> 
> 
> ----------------------------------------------------------------------
> On 1/18/2019 2:41 PM, Ferruh Yigit wrote:
> > On 1/18/2019 10:29 AM, Shahed Shaikh wrote:
> >> Allocating replacement buffer per received packet is expensive.
> >> Instead, process received packets first and allocate
> >> replacement buffers in bulk later.
> >>
> >> This improves performance by ~25% in terms of PPS on AMD
> >> platforms.
> >>
> >> Fixes: 2ea6f76aff40 ("qede: add core driver")
> >> Cc: stable@dpdk.org
> >>
> >> Signed-off-by: Shahed Shaikh <shshaikh@marvell.com>
> >
> > Hi Shahed,
> >
> > This patch has been sent same day of the RC3, very close to the actual release,
> > and updating data path of the driver, so not a trivial patch.
> >
> > Although I tend to accept driver patches late in process this one was too late,
> > you won't have proper time to fix if any errors detected, but also I can see it
> > has a good amount performance effect.
> >
> > I would like to explicitly ask if you are willing to take the risk. If answer is
> > yes, please also be sure to get Rasesh's ack.

Hi Ferruh,

We have done thorough regression testing with this series and would like to get this included in 19.02.
Please consider this series.

Rasesh,
Please ack this series.

Thanks,
Shahed
> 
> cc'ed Thomas.

Re: [dpdk-stable] [dpdk-dev] [PATCH 1/2] net/qede: fix performance bottleneck in Rx path

[Thomas Monjalon]

18/01/2019 17:57, Rasesh Mody:
> >From: dev <dev-bounces@dpdk.org> On Behalf Of Shahed Shaikh
> >Sent: Friday, January 18, 2019 2:29 AM
> >
> >Allocating replacement buffer per received packet is expensive.
> >Instead, process received packets first and allocate replacement buffers in
> >bulk later.
> >
> >This improves performance by ~25% in terms of PPS on AMD platforms.
> >
> >Fixes: 2ea6f76aff40 ("qede: add core driver")
> >Cc: stable@dpdk.org
> >
> >Signed-off-by: Shahed Shaikh <shshaikh@marvell.com>

Applied, thanks

Re: [dpdk-stable] [dpdk-dev] [PATCH 1/2] net/qede: fix performance bottleneck in Rx path

[Ferruh Yigit]

On 1/18/2019 2:41 PM, Ferruh Yigit wrote:
> On 1/18/2019 10:29 AM, Shahed Shaikh wrote:
>> Allocating replacement buffer per received packet is expensive.
>> Instead, process received packets first and allocate
>> replacement buffers in bulk later.
>>
>> This improves performance by ~25% in terms of PPS on AMD
>> platforms.
>>
>> Fixes: 2ea6f76aff40 ("qede: add core driver")
>> Cc: stable@dpdk.org
>>
>> Signed-off-by: Shahed Shaikh <shshaikh@marvell.com>
> 
> Hi Shahed,
> 
> This patch has been sent same day of the RC3, very close to the actual release,
> and updating data path of the driver, so not a trivial patch.
> 
> Although I tend to accept driver patches late in process this one was too late,
> you won't have proper time to fix if any errors detected, but also I can see it
> has a good amount performance effect.
> 
> I would like to explicitly ask if you are willing to take the risk. If answer is
> yes, please also be sure to get Rasesh's ack.

cc'ed Thomas.

Re: [dpdk-stable] [dpdk-dev] [PATCH 1/2] net/qede: fix performance bottleneck in Rx path

[Ferruh Yigit]

On 1/18/2019 10:29 AM, Shahed Shaikh wrote:
> Allocating replacement buffer per received packet is expensive.
> Instead, process received packets first and allocate
> replacement buffers in bulk later.
> 
> This improves performance by ~25% in terms of PPS on AMD
> platforms.
> 
> Fixes: 2ea6f76aff40 ("qede: add core driver")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Shahed Shaikh <shshaikh@marvell.com>

Hi Shahed,

This patch has been sent same day of the RC3, very close to the actual release,
and updating data path of the driver, so not a trivial patch.

Although I tend to accept driver patches late in process this one was too late,
you won't have proper time to fix if any errors detected, but also I can see it
has a good amount performance effect.

I would like to explicitly ask if you are willing to take the risk. If answer is
yes, please also be sure to get Rasesh's ack.

Thanks,
ferruh