mbox series

[net-next,0/2] net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

Message ID 20240903192524.4158713-1-sean.anderson@linux.dev (mailing list archive)
Headers show
Series net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM | expand

Message

Sean Anderson Sept. 3, 2024, 7:25 p.m. UTC
To improve performance without sacrificing latency under low load,
enable DIM. While I appreciate not having to write the library myself, I
do think there are many unusual aspects to DIM, as detailed in the last
patch.

This series depends on [1].

[1] https://lore.kernel.org/netdev/20240903180059.4134461-1-sean.anderson@linux.dev/


Sean Anderson (2):
  net: xilinx: axienet: Support adjusting coalesce settings while
    running
  net: xilinx: axienet: Enable adaptive IRQ coalescing with DIM

 drivers/net/ethernet/xilinx/Kconfig           |   1 +
 drivers/net/ethernet/xilinx/xilinx_axienet.h  |  18 +-
 .../net/ethernet/xilinx/xilinx_axienet_main.c | 254 ++++++++++++++----
 3 files changed, 220 insertions(+), 53 deletions(-)

Comments

Simon Horman Sept. 4, 2024, 4:35 p.m. UTC | #1
On Tue, Sep 03, 2024 at 03:25:22PM -0400, Sean Anderson wrote:
> To improve performance without sacrificing latency under low load,
> enable DIM. While I appreciate not having to write the library myself, I
> do think there are many unusual aspects to DIM, as detailed in the last
> patch.
> 
> This series depends on [1].
> 
> [1] https://lore.kernel.org/netdev/20240903180059.4134461-1-sean.anderson@linux.dev/

Hi Sean,

Unfortunately the CI doesn't understand dependencies,
and so it is unable to apply this patchset :(

I would suggest bundling patches for the same driver for net-next
in a single patchset. And in any case, only having one active
at any given time.
Sean Anderson Sept. 5, 2024, 2:27 p.m. UTC | #2
On 9/4/24 12:35, Simon Horman wrote:
> On Tue, Sep 03, 2024 at 03:25:22PM -0400, Sean Anderson wrote:
>> To improve performance without sacrificing latency under low load,
>> enable DIM. While I appreciate not having to write the library myself, I
>> do think there are many unusual aspects to DIM, as detailed in the last
>> patch.
>> 
>> This series depends on [1].
>> 
>> [1] https://lore.kernel.org/netdev/20240903180059.4134461-1-sean.anderson@linux.dev/
> 
> Hi Sean,
> 
> Unfortunately the CI doesn't understand dependencies,
> and so it is unable to apply this patchset :(
> 
> I would suggest bundling patches for the same driver for net-next
> in a single patchset. And in any case, only having one active
> at any given time.
> 

Well, I would normally do so, but that patch is a fix and this series is
an improvement. So that one goes into net and this one goes into net-next.

I've been advised in the past to split up independent patches so they can be
reviewed/applied individually.

--Sean
Simon Horman Sept. 6, 2024, 7:03 a.m. UTC | #3
On Thu, Sep 05, 2024 at 10:27:00AM -0400, Sean Anderson wrote:
> On 9/4/24 12:35, Simon Horman wrote:
> > On Tue, Sep 03, 2024 at 03:25:22PM -0400, Sean Anderson wrote:
> >> To improve performance without sacrificing latency under low load,
> >> enable DIM. While I appreciate not having to write the library myself, I
> >> do think there are many unusual aspects to DIM, as detailed in the last
> >> patch.
> >> 
> >> This series depends on [1].
> >> 
> >> [1] https://lore.kernel.org/netdev/20240903180059.4134461-1-sean.anderson@linux.dev/
> > 
> > Hi Sean,
> > 
> > Unfortunately the CI doesn't understand dependencies,
> > and so it is unable to apply this patchset :(
> > 
> > I would suggest bundling patches for the same driver for net-next
> > in a single patchset. And in any case, only having one active
> > at any given time.
> > 
> 
> Well, I would normally do so, but that patch is a fix and this series is
> an improvement. So that one goes into net and this one goes into net-next.
> 
> I've been advised in the past to split up independent patches so they can be
> reviewed/applied individually.

Thanks Sean,

Understood. Given the first point, which I had missed earlier,
I'd would have suggested marking this patch-set as an RFC,
then reposting it once the dependency hits net-next (via net).