Hi,
We're an Internet start-up offering online video solutions that are
inno= vative, affordable, easy to use and easy to manage. With a
background in th= e contemporary art scene as well as in web
production, we combine the lates= t in technology with an understanding
of what makes good cont
adi wrote:
> On Tue, Oct 07, 2008 at 11:20:50AM -0700, Brandeburg, Jesse wrote:
>> usually that is the reason for per packet costs going sky high. So
>> yes, it is likely locking related and should be very visible which
>> lock by using oprofile.
>
> Oprofile result shows that ipt_do_table is the
On Tue, Oct 07, 2008 at 11:20:50AM -0700, Brandeburg, Jesse wrote:
>usually that is the reason for per packet costs going sky high. So yes,
>it is likely locking related and should be very visible which lock by
>using oprofile.
Oprofile result shows that ipt_do_table is the culprit. If I do flush
adi wrote:
> On Tue, Oct 07, 2008 at 10:41:22AM -0700, Brandeburg, Jesse wrote:
>> ethtool -S statistics will show lots of rx_no_buffer_count errors if
>> the cpu cannot keep new buffers flowing to the hardware fast enough
>> (possibly meaning you might need to increase the interrupt rate)
>
> I k
On Tue, Oct 07, 2008 at 10:41:22AM -0700, Brandeburg, Jesse wrote:
>ethtool -S statistics will show lots of rx_no_buffer_count errors if the
>cpu cannot keep new buffers flowing to the hardware fast enough (possibly
>meaning you might need to increase the interrupt rate)
I keep getting this 'pro
On Tue, 7 Oct 2008, Luca Deri wrote:
> in order to accelerate the vanilla drivers I have created a
> multithreaded version of them (as of today for igb and ixgbe). I start
> one kernel thread per RX queue (I have enabled multiqueus with
> IntMode=3 in igb), and when I pass packets to upper la
Hello Jesse,
Thank you for the response. Now I know that at least this task is not
impossible.
I attempted to register on premier.intel.com, but was unsuccessful (did
not have a serial# that matched their format). I talked to an Intel
representative using the Live Chat feature, and was told that
New life!!!
http://9i2qia.blu.livefilestore.com/y1p1Up56dp8F6CrxSt3Qfnao43cOKLJd5mBaUI5ZVE1034OZOHV7XrXIJ_F7_uN95H3MEWtIXZZqUoA6yKOFpSkEA/yjn2jlhxsch.html
Roll of parchment from his bag. In response to the result
was as he had foreseen. Men, as a rule, 'it's very hot,
ain't it,' observed
See http://marc.info/?t=12221870784&r=1&w=2
On Tue, 2008-10-07 at 11:34 -0400, Frank Peters wrote:
> Hello,
>
> There has been much publicity regarding the e1000e driver Linux kernel
> bug. I am wondering if this bug also applies to the independent Intel
> e1000e driver that is available fr
Hello,
There has been much publicity regarding the e1000e driver Linux kernel
bug. I am wondering if this bug also applies to the independent Intel
e1000e driver that is available from the Intel web site at
http//downloads.intel.com .
The version number of this independent Intel driver is e1000e
Dear all,
in order to accelerate the vanilla drivers I have created a
multithreaded version of them (as of today for igb and ixgbe). I start
one kernel thread per RX queue (I have enabled multiqueus with
IntMode=3 in igb), and when I pass packets to upper layers I use my
PF_RING (http://www
New liiife!
http://yiyjow.blu.livefilestore.com/y1pYYPchdkEXWqgbeRRNlTSgz7zl3qzYAOhvps-_J5N7WuC5fXub5A_10AWTRw2USM_UTwK3I-xILEYl-klpzfGfg/uk08errdh4a1.html
Instead of one of these flibbertigibbets running cried one,
take heed how you resist lawful authority: side only. These
implements
12 matches
Mail list logo