and since multiple people asked me "whaaat"
http://adrianchadd.blogspot.com/2015/03/cache-line-aliasing-2-or-what-happens.html
-adrian
On 23 October 2017 at 12:40, Adrian Chadd wrote:
> Sigh,
>
> It's likely more memory layout nonsense.
>
> It'd be nic
Sigh,
It's likely more memory layout nonsense.
It'd be nice to see their benchmarks run linux binaries on freebsd.
The last time I dug into this I found it was 95% due to memory layout
and 5% due to ULE being silly.
-adrian
___
freebsd-performance@fr
On 3 June 2016 at 11:27, Adrian Chadd wrote:
> That and the other NUMA stuff is something to address in -12.
And, I completely welcome continued development in NUMA scaling in
combination with discussion. The iterator changes I committed are a
more generic version of a patch people w
On 3 June 2016 at 10:55, Konstantin Belousov wrote:
> On Fri, Jun 03, 2016 at 11:29:13AM -0600, Alan Somers wrote:
>> On Fri, Jun 3, 2016 at 11:26 AM, Konstantin Belousov
>> wrote:
>> > On Fri, Jun 03, 2016 at 09:29:16AM -0600, Alan Somers wrote:
>> >> I notice that, with the exception of the VM_
To: freebsd-performance@freebsd.org
>> Subject: Re: ixgbe: Network performance tuning (#TCP connections)
>>
>> On 2016-02-03 16:34, Adrian Chadd wrote:
>> > hi,
>> >
>> > can you share your testing program source?
>> >
>> >
>> > -a
>
hi,
can you share your testing program source?
-a
On 3 February 2016 at 05:37, Meyer, Wolfgang wrote:
> Hello,
>
> we are evaluating network performance on a DELL-Server (PowerEdge R930 with 4
> Sockets, hw.model: Intel(R) Xeon(R) CPU E7-8891 v3 @ 2.80GHz) with 10
> GbE-Cards. We use progra
What's their benchmark call?
What about clock_gettime() versus gettimeofday() ?
-adrian
On 2 January 2015 at 06:09, grarpamp wrote:
> Some recent FreeBSD related questions in this app area.
>
>
> -- Forwarded message --
> From: Yawning Angel
> Date: Fri, Jan 2, 2015 at 5:20
I've done some basic experimenting with SMT on network loads. For the
most part, as long as you don't fill up one of the ports on the
execution engine that's doing SMT, you're okay.
I've found that a memcpy heavy load (read: normal, non-zero copy
network traffic) brings SMT threads to their knees.
Hi,
You can file a bug (https://bugs.freebsd.org/submit/) - and include
the contents of /var/log/Xorg.0.log .
It may be something really simple, like the nvidia driver isn't
properly working with your setup.
Thanks!
-adrian
On 28 October 2014 11:50, John wrote:
> Hello list,
>
> Is there
On 12 August 2014 11:09, John Baldwin wrote:
> On Wednesday, July 16, 2014 1:52:45 pm Adrian Chadd wrote:
>> Hi!
>>
>>
>> On 16 July 2014 06:29, Konstantin Belousov wrote:
>> > On Fri, Jun 27, 2014 at 03:56:13PM +0300, Konstantin Belousov wrote:
>> >
>
>
> - 原始邮件 -
>
> 发件人:Adrian Chadd
> 收件人:zjuso...@sina.com
> 抄送人:freebsd-performance
> 主题:Re: about network interrupt
> 日期:2014年07月18日 12点21分
>
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailm
Hi!
So em(4) doesn't use multiple CPUs on the traffic receive path, so
things won't scale all that well. can you retry with an igb(4) NIC?
I do have a TODO item to implement generic receive load balancing
based on a GSoC project and some of the RSS stuff that is in -HEAD now
but it's going to be
Hi!
On 16 July 2014 06:29, Konstantin Belousov wrote:
> On Fri, Jun 27, 2014 at 03:56:13PM +0300, Konstantin Belousov wrote:
>> Hi,
>> I did some measurements and hacks to see about the performance and
>> scalability of PostgreSQL 9.3 on FreeBSD, sponsored by The FreeBSD
>> Foundation.
>>
>> The
Hi,
Are you able to repeat these tests (for both 9.2 and 9.3) whilst
grabbing some performance data from lock profiling and hwpmc?
The benchmarking is great but it doesn't tell us enough information as
to "why" things behave poorly compared to Linux and why the mmap drop
isn't so great.
What abo
Hi,
the pgsql testing done has been analysed by a few developers. The
TL;DR version is that there's significant lock contention in the VM /
mmap path and it sticks out like a sore thumb when one does lock
profiling.
-a
On 18 March 2014 05:00, Matthew Seaman wrote:
> On 03/18/14 03:12, Petr Ja
Hi David,
How about just picking a revision of stable/8 half way between the good and
not good version, then re-test?
It'd help to narrow down the range of commits that could've caused problems.
-adrian
___
freebsd-performance@freebsd.org mailing lis
On Oct 7, 2013 1:28 PM, "David Wolfskill" wrote:
>
> At work, we have a bunch of machines that developers use to build some
> software. The machines presently run FreeBSD/amd64 8.3-STABLE @rxx
> (with a few local patches, which have since been committed to stable/8),
> and the software is bui
Wait a sec - your system is spending most of its time in poll() ?
Have you run hwpmc to see if it's actually spinning in poll(), or if
its just doing a lot of poll calls that are very short-lived?
-adrian
___
freebsd-performance@freebsd.org mailing li
On 5 July 2013 00:45, TJ wrote:
> I thought you might say that.
> So is there nothing i can do on the OS side to make things a little faster.
> Or general performance enhancements that would show some benefit?
Where'd you acquire the above tuning parameters from?
You haven't provided any informa
outperform at what?
adrian
On 28 May 2013 00:08, O. Hartmann wrote:
> Phoronix has emitted another of its "famous" performance tests
> comparing different flavours of Linux (their obvious favorite OS):
>
> http://www.phoronix.com/scan.php?page=article&item=bsd_linux_8way&num=1
>
> It is "impre
... that's what the scheduler is for, right? dynamically assigning
interrupt threads to inactive CPUs?
Adrian
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail t
Can you please run a Linux install in a FreeBSD jail so we can see
whether it's the kernel or userland?
Thanks,
Adrian
On 25 March 2013 07:45, Davide D'Amico wrote:
> Il 25/03/13 15:00, Davide D'Amico ha scritto:
>
>> Thank you Daniel for your tests, here my tests using sysbench v0.5 MySQL
On 24 March 2013 11:45, Adam Vande More wrote:
> jemalloc also has concurrency issues when threads > areas:
>
> http://people.freebsd.org/~jasone/jemalloc/bsdcan2006/jemalloc.pdf
Right.
I still think it's worth trying the mysql test in a debian/kfreebsd
install in a jail on the same machine you
The contention is due to memory allocations being page aligned and
those pools all hitting the same cache line mappings.
Adrian
On 24 March 2013 09:09, Adam Vande More wrote:
> I think increasing the number of arenas may help the contention, eg "ln -s
> 3N /etc/malloc.conf"
>
> On Sun, Mar 24
... and how about setting up MySQL inside a Linux jail? Say,
installing debian/kfreebsd in a jail and then testing mysql in there?
Adrian
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
T
Yup, that one.
I wonder if that has anything to do here..
Adrian
On 23 March 2013 17:20, Adam Vande More wrote:
> On Sat, Mar 23, 2013 at 7:08 PM, Adrian Chadd wrote:
>>
>> Hi,
>>
>> I recall that there were significant issues with jemalloc on
>> computat
Hi,
I recall that there were significant issues with jemalloc on
computational loads, primarily because of the alignment jemalloc ends
up giving to various allocation sizes and the cache-busting behaviour
of that.
Does anyone remember the thread in which that happened? Maybe someone
posted a patc
On 12 October 2012 11:10, Miroslav Lachman <000.f...@quip.cz> wrote:
> I don't like comparing Release Candidates without any details about config,
> but the fact that DF 3.2 is much better than DF 3.0 is interesting. And they
> are very close to performance of Scientific Linux 6.2.
Hey cool! And F
I bet the answer is something like "Get FreeBSD up on it or work with
someone who can help you do that."
It's a catch-22 just like GPU - unless ${COMPANY} has customers using
it, they're not likely to dedicate resources, and no users will use it
if it doesn't work, so .. who will break the cycle.
On 16 April 2012 23:31, Richard Kojedzinszky wrote:
>
> So now reactions here, creating files with multilabel is still slow.
>
> I would like to use multilabel access control on my /tmp, for example, my
> web server places it's session files there in a subdirectory. Of course, I
> would like to as
A lot of the linux work is pushed not by hobbyists, but by large
companies with customers that request support.
Don't mislead yourself by thinking all this Linux work gets done by a
large number of unpaid volunteers. Go look at the contribution
statistics sometime.
If people would like to see CUD
On 9 March 2012 09:31, O. Hartmann wrote:
> Well, having to pick up existing ideas and incarntions of those for
> Linux is always a pain in the ass, but necessary at the moment. The
> "experts" neglected long time the need for keeping FBSD on par with KMS
> stuff or all the other development done
Can you verify that it's properly using kqueue, rather than poll?
Adrian
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "freebsd-performance-unsubscr...@fre
On 3 January 2012 00:34, Marc Olzheim wrote:
> On Tue, Jan 03, 2012 at 12:21:10AM -0800, Garrett Cooper wrote:
>> The file is 3.0GB in size. Look at all those page faults though!
>> Thanks!
>> -Garrett
>
> From usr.bin/cmp/c_regular.c:
>
> #define MMAP_CHUNK (8*1024*1024)
> ...
> for (..) {
>
Hi,
I think this thread has gone far, far off the rails.
If you're able to provide some solid debugging or willing to put in
the effort to provide said solid debugging, then great. The easier you
can make it for someone to fix for you (whether they're a FreeBSD
committer or otherwise) the more li
Hi,
I think this thread has gone far, far off the rails.
If you're able to provide some solid debugging or willing to put in
the effort to provide said solid debugging, then great. The easier you
can make it for someone to fix for you (whether they're a FreeBSD
committer or otherwise) the more li
Guys, girls, fuzzy creatures,
This is by far the best example of a constructive email in this entire thread.
If people would like to help, Erik here is exactly the kind of person
with exactly the kind of software that needs a hand.
I think enough philosophizing has been done - now we have questi
Is there a specific version of the test suite that should be used, to
compare against the published results?
Adrian
On 20 December 2011 17:18, Matthew Tippett wrote:
> For such a system, the greatest immediate value would be to attempt to
> reproduce the benchmarks in question.
>
> Install PTS
The trouble is that there's lots of anecdotal evidence, but noone's
really gone digging deep into _their_ example of why it's broken. The
developers who know this stuff don't see anything wrong. That hints to
me it may be something a little more creepy - as an example, the
interplay between netisr/
Hi,
What Attilllo and others need are KTR traces in the most stripped down
example of interactive-busting workload you can find.
Eg: if you're doing 32 concurrent buildworlds and trying to test
interactivity - fine, but that's going to result in a lot of KTR
stuff.
If you can reproduce it using a
Can someone please write up a nice, concise blog post somewhere
outlining all of this?
Extra bonus points if it's a blog that is picked up by
blogs.freebsdish.org and/or some of the other BSD sites.
Guys/girls/fuzzy things - this is 2011; people look at shiny blog
sites with graphs rather than ma
On 15 December 2011 06:49, Tony McC wrote:
> I suggest always ignoring benchmarks. They are like reading the
> astrology column in a tabloid newspaper. Instead, try FreeBSD for your
> work. Is it fast enough? Surely that is all you need to know. FreeBSD
> is quite fast enough for my needs and I
On 14 December 2011 23:32, O. Hartmann wrote:
> Just saw this shot benchmark on Phoronix dot com today:
>
> http://www.phoronix.com/scan.php?page=news_item&px=MTAyNzA
>
> It may be worth to discuss the sad performance of FBSD in some parts of
> the benchmark. A difference of a factor 10 or 100 is
On 13 December 2011 01:00, Andrey Chernov wrote:
>> If the algorithm ULE does not contain problems - it means the problem
>> has Core2Duo, or in a piece of code that uses the ULE scheduler.
>
> I observe ULE interactivity slowness even on single core machine (Pentium
> 4) in very visible places,
What is FBFS?
Adrian
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "freebsd-performance-unsubscr...@freebsd.org"
Is this an AMD class CPU?
The way the counters work on some platforms is that they count _up_ to
an overflow value, trigger the overflow interrupt, and PMC then is
supposed to calculate what the original value was.
It sounds like maybe that isn't working correctly for a certain class
of counters?
Guys (; Girls; small fuzzy creatures);
We can talk about this forever, or someone can go over the existing 32
bit CUDA stuff in a 32 bit Linuxolator, get it all going, document it,
and post it or all to use.
Which do you think is going to be more productive?
Adrian
_
[snip]
32 bit CUDA support may be limiting to you, but it's:
* not necessarily limiting to others;
* a good starting point to demonstrate it's actually feasible;
* a potential stepping stone to getting 64 bit support of some sort in
place (whether it's the push for 64 bit linux support; or to get
Hi,
http://forums.nvidia.com/index.php?showtopic=38242 Post 18
This indicates the driver supports CUDA somehow. What's missing is a
FreeBSD runtime.
Can someone please do some legwork with this and see if it's possible
to bring the Linux CUDA SDK up in the linuxulator?
Adrian
On 29 August 201
On 22 October 2010 06:42, David Wolfskill wrote:
>> Julian's suggestion of booting the 8.1 kernel on the 7.1 OS will definitely
>> narrow down the list of suspects.
>
> I'll see about doing something along those lines, but I doubt it will be
> all that helpful, actually.
It narrows down the susp
On 10 June 2010 06:19, Nick Rogers wrote:
> FYI there is a bugfix/patch for this issue being discussed in
> freebsd-hackers:
>
> http://www.mail-archive.com/freebsd-hack...@freebsd.org/msg157097.html
Hm, that's just a simple caching mechanism, right? If it matches the
last entry, re-use it?
What
2009/10/20 István :
> therefore i like netpipe runs you can see the performance and the latency as
> well using the packet size as your "x" axis, i think it makes more sense
> then just 1 number
My point was to demonstrate that saturating gigabit ethernet is very
doable with FreeBSD, and his lim
7:55 AM, Adrian Chadd wrote:
>> FYI, I installed netperf on my local p4-D test boxes that I use for
>> other testing.
>>
>> 128 byte send/receive buffers on the client side:
>>
>> kristy# netperf -H 192.168.10.2 -p 22113 -l 10
>> TCP STREAM TEST fro
FYI, I installed netperf on my local p4-D test boxes that I use for
other testing.
128 byte send/receive buffers on the client side:
kristy# netperf -H 192.168.10.2 -p 22113 -l 10
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.10.2
(192.168.10.2) port 0 AF_INET
Recv SendSe
2009/10/16 István :
> I see.
> It shows that linux default setup is better.
.. being completely correct, it shows the linux default setup _for
netpipe_ is better on that particular hardware.
That identifies a few other variables which may need addressing. :)
Adrian
_
2009/10/15 Hongtao Yin :
> Hi,
>
>
>
> I compared TCP performance between FreeBSD and Linux by running test tools
> Netperf and Iperf with Intel NIC.
Did you compare syscalls made and time taken?
For example, do either/both of them do a lot of gettimeofday() calls?
FreeBSD and Linux have (had?) d
A few things!
* Since you've changed two things - hwpmc _AND_ the kernel version -
you can't easily conclude which one (if any!) has any influence on
Giant showing up in your top output. I suggest recompiling without
hwpmc and seeing if the behaviour changes.
* The gprof utility expects something
2008/11/25 Ivan Voras <[EMAIL PROTECTED]>:
>> I believe most of the synthetic numbers (mp3 encoding etc.) difference
>> comes from the different version of gcc the different OS uses...
>
> You're very likely right. Ubuntu 8.10 has gcc 4.3.x - it could make for
> the small difference in gzip and 7z
2% may not sound like a lot but it starts becoming measurable savings
when the number of boxes involved is ${LARGE}.
2c,
Adrian
2008/11/24 Mike Tancsa <[EMAIL PROTECTED]>:
> At 03:28 PM 11/24/2008, Steven Hartland wrote:
>>
>> http://www.phoronix.com/scan.php?page=article&item=os_threeway_2008&
2008/11/25 Mike Tancsa <[EMAIL PROTECTED]>:
> At 12:06 PM 11/25/2008, Adrian Chadd wrote:
>>
>> 2% may not sound like a lot but it starts becoming measurable savings
>> when the number of boxes involved is ${LARGE}.
>
> True, but then again is there such a th
art from intel pro 1000. I am talking about stability not
> performance, I expect a intel pro 1000 to outperform a realtek however
> I expect both to be stable in terms of connectivity. I expect a
> realtek in freebsd to perform as well as a realtek in windows and
> linux. :)
Patches
y comparisons of the sort being
> discussed and there's still lots of room for improvement.
>
> As to nsd vs bind, understand they are very different applications w/
> totally different goals. Comparing performance is not entirely fair and
> certainly is difficult. Kris investigated the performance of nsd mostly
> to understand how bind might scale if certain architectural changes were
> made to eliminate known bottlenecks in the application.
>
>
> Sam
>
> ___
> freebsd-performance@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-performance
> To unsubscribe, send any mail to "[EMAIL PROTECTED]"
>
--
Adrian Chadd - [EMAIL PROTECTED]
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "[EMAIL PROTECTED]"
locks. Under 6.x consider the mpsafenet
> ZB> sysctl. This is also a point on which 7.0 will shine.
>
> I upgrade to 7.0RC3 but still the same. 418Mbit is the roof.
What PPS is that then?
Adrian
--
Adrian Chadd - [EMAIL PROTECTED]
it's the SSL library, then php + ssl
init's the SSL library, and stuff gets funny."
--
Adrian Chadd - [EMAIL PROTECTED]
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To un
ess-specific profiling
information.
Kernel/System profiling will probably show you an interesting thing or
two. One thing I noticed was high in my high-TCP-transaction tests
(but not on hardware anywhere near as nice as yours!) was crypto calls
for, IIRC, syncookies.
Adrian
-
Userspace?
Interrupt thread/device driver? Other kernel stuff?
This is just a curiousity for me at the moment. I hear people
occasionally using 10GE cards in Linux/FreeBSD setups as routers and
I'm interested in how the degenerate case performs.
--
Adrian Chadd - [EMAIL PROTECTED]
__
G'day,
Could you run some iperf tests with varying packet sizes, right down
to the minimum packet size? I'd be interested in packet-per-second
throughput information.
Adrian
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/ma
On 22/12/06, David Xu <[EMAIL PROTECTED]> wrote:
I suspect in such a test, memory copying speed will be a key factor,
I don't have number to back up my idea, but I think Linux has lots
of tweaks, such as using MMX instruction to copy data.
I had the oppertunity to study the AMD Athlon XP Optim
On 10/24/06, Mark Bucciarelli <[EMAIL PROTECTED]> wrote:
On Mon, Oct 23, 2006 at 02:21:07PM +0800, Adrian Chadd wrote:
>
> Has anyone come across some network software which uses kqueue
> "differently" to the above ?
lighttpd uses kqueue. Don't know how "d
an
bunching them up all into larger updates.)
Adrian
--
Adrian Chadd - [EMAIL PROTECTED]
___
freebsd-performance@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-performance
To unsubscribe, send any mail to "[EMAIL PROTECTED]"
poll or
Solaris /dev/poll. Even libevent uses it pretty naively.
Has anyone come across some network software which uses kqueue
"differently" to the above ?
Thanks,
Adrian
--
Adrian Chadd - [EMAIL PROTECTED]
___
freebsd-performance@freebsd.o
71 matches
Mail list logo