On Friday, 31 August 2018 02:53:12 PDT VanCutsem, Geoffroy wrote:
> Hi folks,
>
> Some of the folder in https://download.clearlinux.org/releases/ have what
> looks like odd content to me. Many do not have the Clear Linux images
> anymore and instead there are files which names look like a truncated
> version of the original image filename but are instead log files that
> contain a single line that reads: "ok 1 sha verified".
>
> This is an example of such folder:
> https://download.clearlinux.org/releases/24270/clear/ This is an example of
> what is expected: https://download.clearlinux.org/releases/24750/clear/
>
> Do you know what happened?
Yeah, we had a big discussion yesterday about having stable URLs linking to
the latest files so one could write scripts and automation tools to download
the latest Clear Linux image and install it.
I guess we messed up.
--
Thiago Macieira - thiago.macieira (AT) intel.com
Software Architect - Intel Open Source Technology Center
On Thursday, 30 August 2018 12:17:59 PDT Kevin Walker wrote:
> No, in short, I am surprised by your response though, I was really just
> curious if you guys were considering an RT image.
We are, but RT requires proper tuning for the workload you want to run and
whitelisting drivers that you know won't overrun their time budgets. we don't
know yet if the RT build will be generic, which means we don't know it will
work for you. We're looking for it to help us with industrial workloads, not
audio.
Moreover, as Arjan said, we don't believe you need it for the numbers you
showed.
> Given the pedigree and knowledge you guys have, I expect you know a lot
> more about the RT patch than I, so should know what changes and why. It
> isn't necessarily about providing low latency but providing proper
> deterministic behaviour, something a generic operating system built for
> serving websites or running office apps isn't designed for. The whole point
> of the RT patch for audio is about protecting the audio buffers and
> ensuring nothing can interrupt the data passing from sound card through to
> userland etc and back the other way. Whilst I am sure the clear kernel is a
> huge improvement, it doesn't stop other processes in user land or greedy
> interrupts causing problems.
Even if you run those processes with one of the two real-time priority levels
(SCHED_FIFO and SCHED_RR)? You don't need an RT kernel for those.
> Your Kernel patches don't really change that
> much from the generic kernel (although they will go a long way in improving
> audio performance) other than reducing some of the interrupt noise that
> happens with the default kernel, plus a few others like Intel FPGA support
> and TCP buffer sizes. The RT patch, is a major patch of the default kernel
> behaviour, allowing IRQ scheduling priorities through threading, increased
> scheduler granularity and real-time scheduling for processes that support
> it and request it - Jack for example will make use of the RT system calls
> if available.
We are aware of what it does and its limitations. But our point is that we're
skeptical that you need it for your use-cases, not at 5 ms granularity. Even
at 1 ms, a non-RT kernel can suffice, with proper tuning and care. It's only
below that threshold that RT becomes necessary.
--
Thiago Macieira - thiago.macieira (AT) intel.com
Software Architect - Intel Open Source Technology Center
On Thursday, 30 August 2018 01:53:00 PDT Kevin Walker wrote:
> Hmmm I disagree, the RT patch is applicable for Audio latencies of 5-7ms as
> well.
So this goes back to Auke's question: do you have evidence that the current
non-RT kernel we have will not be able to cope with your 5 ms latency
requirements?
--
Thiago Macieira - thiago.macieira (AT) intel.com
Software Architect - Intel Open Source Technology Center
On 08/29/2018 02:31 PM, Kevin Walker wrote:
> Could you provide a RT Kernel bundle for Audio applications?
Hey, I'm curious -
before we go and consider this, is there any solid data showing that the
default native kernel in Clear Linux is incapable of doing what you want
with it, in this perspective?
in other words, is there any meaningful, real-world data?
Thanks,
Auke
---
After creating this bundle, I realized that we already have a lot of
education apps, particularly from the import of KDE apps.
bundles/education | 17 +++++++++++++++++
1 file changed, 17 insertions(+)
diff --git a/bundles/education b/bundles/education
index 8e1b4bf..dd125dd 100644
--- a/bundles/education
+++ b/bundles/education
@@ -6,4 +6,21 @@
include(libX11client)
include(qt-basic)
+
+blinken
gcompris-qt
+gimp
+gnome-tweak-tool
+kanagram
+kbruch
+kgeography
+khangman
+kig
+klettres
+kmplot
+kolourpaint
+ktuberling
+kturtle
+kwordquiz
+marble
+parley
--
2.18.0
On Wed, Aug 29, 2018 at 3:20 PM, Warden, Brett T
<brett.t.warden(a)intel.com> wrote:
> I'll add qt-basic. Anything else come to mind? libX11client?
libX11client would be the only one I could think of.