Hi Geoff
On Sun, 16 Jun 2013 20:49:26 +1000 you wrote:
> Going to finally build a new machine. I'ts going to be Intel this time -
> AMD for 15 years or so - can any one here give some advice as to how
> many cores are optimal given current kernel >3.8 performance. Any
> install/operational issues? Any pitfalls ?
I can't provide much in the way of scientific evidence; there are others who
know far more about the technical realities of this - and push their
machines much harder - than I do.
Personally, I have always run with the idea that so long as one has N
independent processes whose links to other processes are limited to the
consumption or production of streams, then one could theoretically max out N
CPU cores. However in most practical cases involving audio and video one is
running the streams at real world speed, and this obviously limits the
extent that each process requires a CPU. The only time a single process
would have the chance to max out a single CPU core would be when
freewheeling with jack for example, or doing a final render of a video.
If one spends most of the time interacting with their AV software, this
means that there's no simple answer to the "how many cores are optimal"
questions. It depends on the precise mix of software you're running, what
each process requires of a CPU core, how much I/O each process instigates,
and so on. Another caveat is jackd: my current understanding is that jack2
can better utilise multiple cores, but I'm happy to be corrected on this
point by anyone with more knowledge than I (I really haven't looked into
this recently because for my current situation it's academic).
My system has been based on a first-generation i7 for the last couple of
years and I've noticed no major issues. However, in terms of audio work I'm
not really pushing the system all that hard during real work (I don't have
soft-synths running generally, and the plugins I use tend to be fairly
frugal with CPU requirements). This gives me 4 cores with hypertheading
and when I've done tests to see what it could handle, an audio-like workload
is able to push well above the 400% loading (so the hyperthreading seems to
be doing something useful).
Having said all that and knowing the sort of work you do, I would probably
err on the side of getting as many cores as you can reasonably afford. As
time goes on they won't go astray; you'll have the flexibility to experiment
with new ways of doing things without being too constrained by the number of
cores at your disposal.
A final comment is that with the release of Intel's Haswell-based CPUs we
are at an interesting point in time. These new cores are certainly a big
win for mobile computing due to their lower power consumption for a given
performance level. However, whether the increase in outright performance -
the primary metric for a desktop - justifies the "new product" premium these
will attract for the next 6-12 months remains to be seen, especially since
one would also expect some runout discounting on the 4th generation CPUs in
the coming months.
Regards
jonathan
Anyone else noticed this:
Ardour 3.1-3-g1606996: changing the meter source (in/pre/post/custom)
in a mixer strip generates clicks in the audio output.
??
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
hi LADs
I installed linux on a new macbook retina and have getting some trouble
running jalv.
it gives me the error
gian@gian-MacBookPro:~/mod/LV2/jalv-1.4.0$ jalv.gtk
http://guitarix.sourceforge.net/plugins/gx_amp#GUITARIX
Plugin: http://guitarix.sourceforge.net/plugins/gx_amp#GUITARIX
UI: http://guitarix.sourceforge.net/plugins/gx_amp#gui
JACK Name: GxAmplifier-X
Cannot connect to server socket err = No such file or directory
Cannot connect to server request channel
jackdmp 1.9.10
Copyright 2001-2005 Paul Davis and others.
Copyright 2004-2013 Grame.
jackdmp comes with ABSOLUTELY NO WARRANTY
This is free software, and you are welcome to redistribute it
under certain conditions; see the file COPYING for details
no message buffer overruns
no message buffer overruns
no message buffer overruns
JACK server starting in realtime mode with priority 10
audio_reservation_init
Acquire audio card Audio1
creating alsa driver ... hw:1,0|hw:1,0|256|2|44100|0|0|nomon|swmeter|-|32bit
configuring for 44100Hz, period = 256 frames (5.8 ms), buffer = 2 periods
ALSA: final selected sample format for capture: 24bit little-endian
ALSA: use 2 periods for capture
ALSA: final selected sample format for playback: 24bit little-endian
ALSA: use 2 periods for playback
Block length: 256 frames
MIDI buffers: 32768 bytes
Comm buffers: 131072 bytes
Update rate: 2 Hz
using block size: 256
MasterGain = -15,000000
PreGain = 0,000000
Distortion = 20,000000
Drive = 0,250000
Middle = 0,500000
Bass = 0,500000
Treble = 0,500000
Cabinet = 10,000000
Presence = 5,000000
model = 0,000000
t_model = 1,000000
c_model = 0,000000
Inconsistency detected by ld.so: dl-open.c: 684: _dl_open: Assertion
`_dl_debug_initialize (0, args.nsid)->r_state == RT_CONSISTENT' failed!
JackEngine::XRun: client = GxAmplifier-X was not finished, state = Triggered
JackAudioDriver::ProcessGraphAsyncMaster: Process error
gian@gian-MacBookPro:~/mod/LV2/jalv-1.4.0$ Unknown error...
terminate called after throwing an instance of
'Jack::JackTemporaryException'
what():
due to the new retina display I've installed both KDE and GNOME. I wonder
if the problems com from that.
Any help is appreciated.
kind regards
Gian
I'm working on a project using NTK as the gui toolkit, and am trying to
decide if I need to distribute NTK with it or not.
Does anyone have any suggestions on this front? Do most distributions
have NTK packages by now, or most not?
Thanks for any pointers,
Nick
Hi
Did anyone here know if the GPL+ v2.0 /v3.0 is compatible with the CC-BY
v3.0 (unported)
http://creativecommons.org/licenses/by/3.0/
I only found here
http://wiki.debian.org/DFSGLicenses#Creative_Commons_Attribution_Share-Alik…
that the CC-BY-SA v3.0 is compatible, but no mention of the CC-BY v3.0
My understanding is that the CC-BY v3.0 has less restrictions then the
CC-BY-SA version, but I'm a bit unsure.
Background: I would include some work which is under the CC-BY v3.0 to
my project, which is under the GPL+ v2.0 (or later). I wouldn't violate
the DFSG, so I would make sure there is no issue at all when I'm do so.
The Author of the CC-BY v3.0 files is fine with my wishes.
any hints?
hermann
Going to finally build a new machine. I'ts going to be Intel this time -
AMD for 15 years or so - can any one here give some advice as to how
many cores are optimal given current kernel >3.8 performance. Any
install/operational issues? Any pitfalls ?
any advice very welcome.
cheers
g.
Hello all,
I wonder if any other users have experienced this problem and
how they handled it.
This has occured three times when doing an fresh Archlinux install
on a system using the RME MADI cards.
There seems to be something in the combination of recent versions
of the driver and alsactl that leads to alsactl freezing when the
configured (external) clock source for the card is not available.
The 'freeze' seems to be quite deep: it's impossible to kill the
process (even while that process is still a child of e.g. the
xterm from which it was launched, and not of PID 1). Any other
process trying to access the sound card (e.g. jackd) hangs in
the same way. This also means that when doing a poweroff or reboot
systemd will hang on the 'alsactl store' service, and the only
option is a power cycle.
An added difficulty when trying to resolve this (things will be
OK once you have the correct /var/lib/alsa/asound.state) is that
recent systemd doesn't allow to disable or enable the alsa store/
restore services easily (why not ?), you have to manually edit
some symlinks in order to do that.
Note: if this happens to be a driver problem, please do NOT revert
to the ancient behaviour of silently changing the clock source to
'internal' when the external clock is not available. I DO still
expect to see opening the device fail if the external clock isn't
present, as has been the case for some time. The thing that shouldn't
happen is that alsactl chokes on this condition - it didn't before
so it shouldn't have to.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
Hi,
I started working with Linux audio very recently. Primarily, I wanted to
understand the Linux audio stack and audio processing flow. As a simple
exercise,
- I write a simple capture/playback application for PCM interface and store
the captured audio data in a .wav or .raw file
- The device parameters I played with are: hw:2,0/plughw:2,0 (for USB
headset) , 1/2 channel, 44100 sample rate, SND_PCM_ACCESS_RW_INTERLEAVED
mode, period size 32, S16_LE format.
- Use Ubuntu, kernel 3.9.4 and Logitech USB headset for development/testing
To understand the control flow, I inspected the ALSA driver source code and
got an understanding of the flow in kernel space. However, I am kind of lost
in the user-space ALSA library.
- What happens after ALSA handovers the data i.e. the audio data from mic is
copied from USB transfer buffer to userspace buffer? What are the data
processing steps?
- Where is Pulseaudio working here?
- Is there any software mixing happening?
- In the USB sound driver, I changed the data in urb->transferbuffer to a
pattern like 'abcdefghijklmnop', but in the capture application when I store
the audio data in a .raw file, I don't completely get the data back.
If I set 1 channel in the hw params, I get almost 'abcdefghijklmnop' pattern
but sometimes get 'abcdefghijklm' i.e. the pattern incomplete and another
pattern starts over. For 2 channels, the data interleaves in a pattern like
'ababcdcd...' but there are also some incomplete patterns and also I see
some unexpected characters.
I know it's a long post, but even if you can help with a part of it, I'll be
greatly benefited. Thanks!
--
View this message in context: http://linux-audio.4202.n7.nabble.com/Linux-Audio-Architecture-tp85571.html
Sent from the linux-audio-dev mailing list archive at Nabble.com.
Hello all,
Returning home late, on the way from the car parking to my
door I was greeted by a nebula of hundreds of fireflies doing
their social thing.
A lovely thing to see, but it also reminded me that I should
really post the following:
The last months I'm receiving lots of invitations to join
Circles, Friends, Contacts etc. etc. on Google+, Facebook,
LinkedIn etc. etc, many of which from members of this list.
While I do appreciate the motivation behind such requests,
I will never accept them, and from now on I will also stop
responding to any such invitations. If you want to discuss
anything (Linux) audio you're welcome to get in contact via
private email or the LAU or LAD mailing lists.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
