Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 14 May 2015 08:39:30 +0200
From:      Andreas Andersson <aandersson@tappsi.co>
To:        Warner Losh <imp@bsdimp.com>
Cc:        freebsd-arm@freebsd.org
Subject:   Re: Performance issues with raspberry pi 2
Message-ID:  <CAOx9_Zgg7X_-hdA6PJKWHD-zSPVYvnhq8kEuNmHxytQULVMgmA@mail.gmail.com>
In-Reply-To: <1EE88AD4-2B4A-496B-9366-7F9A89F57D6D@bsdimp.com>
References:  <CAOx9_ZiUw5nkBK=Bsumb=K2W1hv0P%2BNAwT7OXE7VAz_96awqOw@mail.gmail.com> <1EE88AD4-2B4A-496B-9366-7F9A89F57D6D@bsdimp.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Just built an image for the raspberry pi without the witness options.. Will
try to build with TARGET_ARCH=3Darmv6hf later on. Eager to test the
performance of tornadoweb on the rpi first. :)

Will come back with benchmarks later. What else would be nice to benchmark?
Is there a standard set of benchmarks that I can run and provide results
for?

2015-05-13 17:04 GMT+02:00 Warner Losh <imp@bsdimp.com>:

>
> > On May 13, 2015, at 5:27 AM, Andreas Andersson <aandersson@tappsi.co>
> wrote:
> >
> > I do understand this is still very new and not ready for "producton" us=
e.
> >
> > But freebsd 11 on rpi2 suffers permance issues. Severely.
> >
> > For instance, the same code I am running (tornado project which inserts
> > stuff to rabbitmq) is running at ~190 req/s or 190 publishments/sec (to
> > rabbitmq) on raspbian.
> >
> > On FreeBSD 11 this is at around 85/s.
> >
> > With my consumer (using all the cores) I am seeing a high get rate from
> > rabbitmq. But ACKing those messages are painstakingly behind. How can I
> > help you debug this.
> >
> > What information would you need?
> >
> >
> >
> >
> > My own theory on this is that on FreeBSD we are doing stuff with softwa=
re
> > floating point, while on raspbian we are doing stuff with hardware
> floating
> > point.
>
> FreeBSD is still using soft float. We=E2=80=99ll be transitioning that to=
 hard
> float
> for 11.0, and should have the code for that pushed into head by the end
> of BSDCan 2015.
>
> In the mean time, you can try building an image with TARGET_ARCH=3Darmv6h=
f
> to see if that is the cause of the problems.
>
> Warner
>
>
>



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAOx9_Zgg7X_-hdA6PJKWHD-zSPVYvnhq8kEuNmHxytQULVMgmA>