hi PDrs
I can't figure out how to avoid PD to freeze when using kernel 2.6
I've just downloaded and compiled pd-0.38-4, and trying to use the compiled externals from previous version (0.38-2), ok, some don't load, but basically all the ones I need, they do.
and the point is that very randomly PD freezes with the "watchdog" warn on the console.
I saw around some discussions about freezes with pd and kernel 2.6 and heard about the scheduling code from kernel not being so smooth, and -rt as not recommended using 2.6, but I even recompiled the kernel 2.6.11.8 without scheduling as well as without preemptive features.
but lockups continue exactly on the same way.
I've read a post here from Tom Schouten saying the problem could be solved running X and the window manager, and he posted a link to a "setrt" tool, but I didn't have much success with all this stuff yet.
beside those, I read Günter patched the memlock with Tim's code, as the memory seems to be quite different (from Günter's words), and that could solve the -rt flag problem.
However I wonder if this fix applies just to the debian package or to the PD source available from Miller's site.
note: I'm just using an audio patch, and any kind of video processing in this case. and tried both with and without -rt (should force using -nrt?).
last info: seems like the upcoming kernel 2.6.12 will have important fixes to the low latency stuff, and I don't know much more than that, just found it here:
http://lalists.stanford.edu/lau/2005/04/0147.html
any tips to bring PD to stable in kernel 2.6.x is very welcome.
many thx! rama.
On Mon, 2 May 2005, rama wrote:
beside those, I read G�nter patched the memlock with Tim's code, as the memory seems to be quite different (from G�nter's words), and that could solve the -rt flag problem.
However I wonder if this fix applies just to the debian package or to the PD source available from Miller's site.
Hi Rama,
It only applies to the Debian package as of yet. The mlock issue is also fixed in the pd devel version from CVS. But it is easy to patch pd from Miller with the diff provided by Tim:
Download: http://tinyurl.com/bezzj
Apply: cd pd-0.38-4/src patch < ../../mlock_fix
and rebuild and report back.
Greetings,
Guenter
note: I'm just using an audio patch, and any kind of video processing in this case. and tried both with and without -rt (should force using -nrt?).
last info: seems like the upcoming kernel 2.6.12 will have important fixes to the low latency stuff, and I don't know much more than that, just found it here:
http://lalists.stanford.edu/lau/2005/04/0147.html
any tips to bring PD to stable in kernel 2.6.x is very welcome.
many thx! rama.
--
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
hi guenter wie geht's?, all, thx for the feedback
I've just did make clean, patched, and now compiling spits something I
can't figure out:
...........................................................................
cc -g -O2 -DPD -Wall -W -Wstrict-prototypes -Wno-unused
-Wno-parentheses -Wno-switch -O6 -funroll-loops -fomit-frame-pointer
-DDL_OPEN -DPA_USE_OSS -DPA_LITTLE_ENDIAN -DUNIX -DUNISTD
-DUSEAPI_OSS -I../portaudio/pa_common -I../portaudio/pablio
-I../portaudio/portmidi-macosx -fno-strict-aliasing -DPA_USE_ALSA
-DUSEAPI_ALSA -DUSEAPI_JACK -DINSTALL_PREFIX="/usr/local" -I. -c
-o ../obj/s_inter.o s_inter.c
s_inter.c: In function sys_set_priority': s_inter.c:272: error: storage size of 'mlock_limit' isn't known s_inter.c:275: warning: implicit declaration of functionsetrlimit'
s_inter.c:275: error: `RLIMIT_MEMLOCK' undeclared (first use in this
function)
s_inter.c:275: error: (Each undeclared identifier is reported only once
s_inter.c:275: error: for each function it appears in.)
make: *** [s_inter.o] Error 1
...........................................................................
something missing?
btw, sounds sexy that about kernel 2.6.12 fixing some low latency stuff, can anyone confirm that more closer?
danke, hasta luego, r
PD: I insist in the fact about setting default "reply to:" to this ML posts as: _this list_ even if mailman strongly recommends to "original sender". but I think this forces an off-list exchange pulling out of it some mails unless the responder figures out he wants to keep it "public" and fixed the "To: " field to this list... no¿
El lun, 02-05-2005 a las 18:44 +0200, günter geiger escribió:
On Mon, 2 May 2005, rama wrote:
beside those, I read Gnter patched the memlock with Tim's code, as the memory seems to be quite different (from Gnter's words), and that could solve the -rt flag problem.
However I wonder if this fix applies just to the debian package or to the PD source available from Miller's site.
Hi Rama,
It only applies to the Debian package as of yet. The mlock issue is also fixed in the pd devel version from CVS. But it is easy to patch pd from Miller with the diff provided by Tim:
Download: http://tinyurl.com/bezzj
Apply: cd pd-0.38-4/src patch < ../../mlock_fix
and rebuild and report back.
Greetings,
Guenter
note: I'm just using an audio patch, and any kind of video processing in this case. and tried both with and without -rt (should force using -nrt?).
last info: seems like the upcoming kernel 2.6.12 will have important fixes to the low latency stuff, and I don't know much more than that, just found it here:
http://lalists.stanford.edu/lau/2005/04/0147.html
any tips to bring PD to stable in kernel 2.6.x is very welcome.
many thx! rama.
--
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Hi,
Hmm, yes it seems that on newer systems the patch lacks an include. You should have
#include <sys/mman.h> #include <sys/resource.h>
somewhere at the top of s_inter.c. (The sys/resource.h is missing in the patch)...
About the fixing of latency stuff I don't know. I very seldom have latency problems, because I don't do disk I/O or video together with sound on one machine.
Greetings,
Guenter
On Tue, 3 May 2005, rama wrote:
hi guenter wie geht's?, all, thx for the feedback
I've just did make clean, patched, and now compiling spits something I can't figure out: ........................................................................... cc -g -O2 -DPD -Wall -W -Wstrict-prototypes -Wno-unused -Wno-parentheses -Wno-switch -O6 -funroll-loops -fomit-frame-pointer -DDL_OPEN -DPA_USE_OSS -DPA_LITTLE_ENDIAN -DUNIX -DUNISTD -DUSEAPI_OSS -I../portaudio/pa_common -I../portaudio/pablio -I../portaudio/portmidi-macosx -fno-strict-aliasing -DPA_USE_ALSA -DUSEAPI_ALSA -DUSEAPI_JACK -DINSTALL_PREFIX="/usr/local" -I. -c -o ../obj/s_inter.o s_inter.c s_inter.c: In function
sys_set_priority': s_inter.c:272: error: storage size of 'mlock_limit' isn't known s_inter.c:275: warning: implicit declaration of functionsetrlimit' s_inter.c:275: error: `RLIMIT_MEMLOCK' undeclared (first use in this function) s_inter.c:275: error: (Each undeclared identifier is reported only once s_inter.c:275: error: for each function it appears in.) make: *** [s_inter.o] Error 1 ...........................................................................something missing?
btw, sounds sexy that about kernel 2.6.12 fixing some low latency stuff, can anyone confirm that more closer?
danke, hasta luego, r
PD: I insist in the fact about setting default "reply to:" to this ML posts as: _this list_ even if mailman strongly recommends to "original sender". but I think this forces an off-list exchange pulling out of it some mails unless the responder figures out he wants to keep it "public" and fixed the "To: " field to this list... no�
El lun, 02-05-2005 a las 18:44 +0200, g�nter geiger escribi�:
On Mon, 2 May 2005, rama wrote:
beside those, I read Gnter patched the memlock with Tim's code, as the memory seems to be quite different (from Gnter's words), and that could solve the -rt flag problem.
However I wonder if this fix applies just to the debian package or to the PD source available from Miller's site.
Hi Rama,
It only applies to the Debian package as of yet. The mlock issue is also fixed in the pd devel version from CVS. But it is easy to patch pd from Miller with the diff provided by Tim:
Download: http://tinyurl.com/bezzj
Apply: cd pd-0.38-4/src patch < ../../mlock_fix
and rebuild and report back.
Greetings,
Guenter
note: I'm just using an audio patch, and any kind of video processing in this case. and tried both with and without -rt (should force using -nrt?).
last info: seems like the upcoming kernel 2.6.12 will have important fixes to the low latency stuff, and I don't know much more than that, just found it here:
http://lalists.stanford.edu/lau/2005/04/0147.html
any tips to bring PD to stable in kernel 2.6.x is very welcome.
many thx! rama.
--
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
--
-
g�nter geiger -
rama