Hi Rich
Le 30/11/2020 à 17:22, Rich Gattie a écrit :
Greetings list..
I just upgraded the machine I have been running 2.x on to the latest
v3 of Rivendell. After an annoying issue with the NIC and driver no
longer in the distro of CentOS 7, I got the machine up and running as
it was before with one small exception.
While testing the main log, (I use Aux Log 1 for a 24/7 stream), I add
a cart, hit START, then it takes about 5 seconds to start playing.
Segues go smoothly between carts, no delay.
There was no issue like this with 2.19.x, and no hardware changes have
been made to the machine.
I've also hard time migrating a machine from 2.X to 3.X. My primary
tests on my dev machine were just fine, but when I tried on the target
machine, the CPU load was huge, and any operation took tens on
seconds... It looks like your issue.
I've spotted caed taking 100% of a core, on the target machine, compared
to ~30% of a core on my dev machine. The memory was fine, IOs were quiet
too. Since then, I'm digging into the source code of caed (mostly in
cae_alsa.cpp) to try to understand where CPU cycles are vanishing. Using
valgrind profiler, the mostly called function is AlsaPlayCallback, but
this is quite obvious and not giving so many clues on what is time
consuming in there.
Maybe you could verify you CPU load with the "top" oh "htop" commands in
a terminal in order to see if it points to caed also. What type of CPU
are you using ?
Until I find a solution to keep caed CPU consumption low on my atom
machines, I was forced to use a bigger one instead (any old Intel
Core-i3 I has were just perfect)
If anyone has an idea on why caed is taking so much on the CPU... I've
tried not to run rdvairplayd but it had no effect
Best regards
Florent
_______________________________________________
Rivendell-dev mailing list
[email protected]
http://caspian.paravelsystems.com/mailman/listinfo/rivendell-dev
