That's a positive last post Thomas. Really sorry that I couldn't
resist posting another one, but there's been some factual
misinformation in this thread, and I have a feeling that it's not been
clear that it is actually misinformation.
Port gain connections does not require extra memory, extra context
switches, or extra CPU usage (except when applying the gain itself if
connection gain is not 1.0). That's fine, we all
make mistakes, but unfortunately several people just repeated these
claims afterwards. I think those who just repeated these claims should
search within themselves and think about why they did that.
For anyone reading this later who think about maybe implementing port
gain, the only things needed to do is to extend the API, add a gain
array into the connection manager shared memory (well, this requires
some memory, but it's not much compared to what's already there), plus
modify the mixing routine. The last part seems to require a little bit
work since there's some hardware optimization you either have to try
match up in performance, or to just do a plain float vector addition,
at least as a first step. Smoothing the gain values to avoid clicks
and zipper noise would be nice too, but it's not essential as a first
step. It might be slightly more complicated than this, but I'm pretty
sure that's the essential steps.
From the discussion here, it seems unlikely that such a
modification
would be merged, but I hope we one day would move away from installing
libjack and friends into the system, but that these are instead
included with a controller program (for instance qjackctl), like I do
for Radium on Windows, i.e that all libjack functions are weak-linked
and automatically links to the libraries specified by the the
controller program (i.e the controller program specify the path of the
jack libraries). Then the controller program could include a client
mixer and a connection GUI, plus interface to start and stop the
server. This requires current clients to be recompiled though, but not
only is this way much simpler for the user, I think it may also be a
way to make jack more popular on windows and osx, and even linux. This
way, it would also be much simpler to fork jack, and include features
such as jack port connections.
I wish we could have focused more on the philosophical discussion
about what jack is and what it should be used for (and in a slightly
less hostile tone), but it's important to get the facts right.
On Fri, Mar 29, 2019 at 5:03 PM Thomas Brand <tom(a)trellis.ch> wrote:
I want to post on top and have the last word. Just because it was very
exciting to see all the good lords having their say on this list. It's
still alive!
Cheers
On Fri, March 29, 2019 10:19, Orm Finnendahl wrote:
Hi,
thanks Fernando. To all: Please, can we just close this bikeshed for good?
--
Orm
Am Donnerstag, den 28. März 2019 um 18:20:26 Uhr (-0700) schrieb Fernando
Lopez-Lezcano:
On 3/28/19 5:18 PM, Robin Gareus wrote:
On 3/27/19 7:54 PM, Stéphane Letz wrote:
> As an old JACK2 developer I would say: Fons is (almost always…)
> right, listen to him ((-;
words to live by!
Indeed. I have learned over time to pay close attention and read
carefully when some users post - Fons is one of them.
To the proponents of gain control in the crosspoints of the jack
matrix:
please post code in the form of a patch for jack1 and jack2 that can be
reviewed. Not a big deal if this is so easy to do without side effects
or performance degradation in all platforms Jack supports.
-- Fernando "who has not groked the source, so should shut up"
_______________________________________________
Jack-Devel mailing list
Jack-Devel(a)lists.jackaudio.org
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org
_______________________________________________
Jack-Devel mailing list
Jack-Devel(a)lists.jackaudio.org
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org
_______________________________________________
Jack-Devel mailing list
Jack-Devel(a)lists.jackaudio.org
http://lists.jackaudio.org/listinfo.cgi/jack-devel-jackaudio.org