-   Slackware (
-   -   Slackware as an optimized sound server (

jostber 01-17-2013 11:46 AM

Slackware as an optimized sound server
Has anyone any recent experience in dedicating a Slackware box for an optimized sound server? Thinking about removing unecessary apps and services, reducing latency and more so that most of the CPU and RAM goes to the sound processing. This should be only a music server for mpd and likewise for playing and streaming, and not for audio recording, multimedia and such.

StreamThreader 01-18-2013 04:53 AM

Interesting Theme.

jostber 01-18-2013 10:46 AM

Any input? Would guess that the rock stability and performance of Slackware would make an ideal sound server.

GazL 01-18-2013 11:02 AM

I used to run mpd in the background on my workstation. I don't see why you'd have any issues. I'd be inclined to start the daemon with a negative nice priority, but other than that, I don't think there is anything special to do.

Martinus2u 01-18-2013 03:58 PM

Should you experiment in that direction, I'd be interested in a comparison between the vanilla kernel (CFQ cpu scheduler) and the BFS cpu scheduler. The latter also gives you additional scheduling classes that can be set using a program called schedtool. For info on scheduling classes cf. the policy overview in the manpage below.


SCHEDTOOL(8)                System Manager's Manual              SCHEDTOOL(8)

      schedtool - query and set CPU scheduling parameters

              [-0|-N] [-1|-F] [-2|-R] [-3|-B] [-4|-I] [-5|-D]
              [-M policy]
              [-a affinity]
              [-p prio]
              [-n nice_level]
              [-e command [arg ...]]
              [-h][LIST OF PIDs]

      schedtool  can set all CPU scheduling parameters Linux is capable of or
      display information for given processes.

      Long-running, non-interactive tasks may  benefit  from  SCHED_BATCH  as
      timeslices are longer, less system-time is wasted by computing the next
      runnable process and the caches stay stable.

      Audio/video or other near-realtime applications may run with less skip‐
      ping  if  set to SCHED_RR.  Use the static priority-switch -p to desig‐
      nate inter-process-hierarchies.

      schedtool now supports setting the  CPU-affinity  introduced  in  linux

      -N or -0
              set all PIDs to SCHED_NORMAL/OTHER

      -F or -1
              to SCHED_FIFO  root-credentials required

      -R or -2
              to SCHED_RR    root-credentials required

      -B or -3
              to SCHED_BATCH

      -I or -4
              to SCHED_ISO

      -D or -5
              to SCHED_IDLEPRIO

      -M policy
              for manual/raw mode; policy is the number of the scheduling pol‐
              icy (see above for 0-4).  This option is mostly for kernel  guys
              that want to test their new implementations.

      -p prio
              specify  static  priority  required for SCHED_FIFO and SCHED_RR.
              Usually ranged from 1-99.

      -a affinity
              set the PID's affinity to this bitmask  (hexadecimal);  alterna‐
              tively, a list mode is supported.
      -n nice_level
              set the PID's nice level; see nice(2), nice(1).
      -e command [arg ...]
              execute  command  with  given  scheduling parameters (overwrites
              schedtool's process image). See EXAMPLES.
      -r    display min and max priority for each policy.

      -v    be verbose.

      -h    help

      To query the $SHELL's policies:

          #> schedtool $$

      To query some PIDs, namely 1 2 and 3:

          #> schedtool 1 2 3

      To execute mplayer in SCHED_RR with priority 20. The  priority  arg  is
      needed for both SCHED_RR and SCHED_FIFO.

          #> schedtool -R -p 20 -e mplayer -quiet some_file.avi

      To  set  current  shell  to  SCHED_BATCH,  which all programs the shell
      starts will inherit:

          #> schedtool -3 $$
      To set all processes with the name 'cpu_hog' to SCHED_BATCH:

          #> schedtool -3 `pidof cpu_hog`

      To set a process' affinity to only the first CPU (CPU0):

          #> schedtool -a 0x1 <PID>

      Using the list mode and affinty of CPU0 and CPU3:

          #> schedtool -a 0,3 <PID>
      A combination of an affinity and a policy-argument is  -  of  course  -
      always possible.

          #> schedtool -B -a 0x1 <PID>

      The  affinity-argument determines on which CPUs a process is allowed to
      run. It consists of a simple bitmask represented in hexadecimal.  CPU0
      is  denoted by the least-significant bit, CPU1 by the second least-sig‐
      nificant and so on, thus giving:
          0x1 -> only run on CPU0

          0x2 -> only run on CPU1
          0x4 -> only run on CPU2

          0x8 -> only run on CPU3 ... and so on.
      Multi-target CPUs may be specified using bitwise OR of the values:

          0x7 -> run on CPUs 0, 1, 2 but NOT on 4

          0x3 -> run only on CPUs 0 and 1

      The default is to run a process on all CPUs, giving a mask of
          0xf for all 4 CPUs

          0xff for all 8 CPUs

      Alternatively a list mode is supported where you can specify  the  CPUs
      delimited  by  ",".  The following runs <PID> on CPU0 and CPU1 (equiva‐
      lent to 0x3):

          #> schedtool -a 0,1 <PID>

      SCHED_NORMAL / SCHED_OTHER This is the default policy and for the aver‐
      age program with some interaction. Does preemption of other processes.

      SCHED_FIFO  First-In, First Out Scheduler, used only for real-time con‐
      traints.  Processes in this class are usually not preempted by  others,
      they need to free themselves from the CPU via sched_yield() and as such
      you need special designed applications. Use with extreme  care.  ROOT-
      credentials required.

      SCHED_RR  Round-Robin  Scheduler,  also used for real-time constraints.
      CPU-time is assigned in an round-robin  fashion  with  a  much  smaller
      timeslice  than  with  SCHED_NORMAL  and  processes  in  this group are
      favoured over SCHED_NORMAL. Usable for  audio/video  applications  near
      peak rate of the system.  ROOT-credentials required.
      SCHED_BATCH  [  since 2.6.16 in mainline ] SCHED_BATCH was designed for
      non-interactive, CPU-bound applications.  It uses longer timeslices (to
      better exploit the cache), but can be interrupted anytime by other pro‐
      cesses in other classes to guaratee interaction  of  the  system.  Pro‐
      cesses in this class are selected last but may result in a considerable
      speed-up (up to 300%). No interactive boosting is done.

      SCHED_ISO [ patch needed ] SCHED_ISO  was  designed  to  give  users  a
      SCHED_RR-similar  class.  To quote Con Kolivas: "This is a non-expiring
      scheduler policy designed to guarantee a timeslice within a  reasonable
      latency  while  preventing  starvation.  Good for gaming, video at the
      limits of hardware, video capture etc."

      SCHED_IDLEPRIO  [  patch  needed  ]  SCHED_IDLEPRIO  is  similar  to
      SCHED_BATCH,  but was explicitely designed to consume only the time the
      CPU is idle. No interactive boosting is done.  If you used  SCHED_BATCH
      in the -ck kernels this is what you want since 2.6.16

      PID  0  means "current process", in our case, schedtool. May occur when
      using the -e switch.

      Be careful with SCHED_FIFO! You may lock out other processes  from  the
      CPU, including your shell.
      For SCHED_BATCH you certainly need the a recent 2.6 kernel.

      A  short  overview  is  given  in  SCHED_DESIGN and the README contains
      thourough discussion. The INSTALL file also lists all prerequisites and
      where you can get patches.

      Affinity 0x0 should never be used.

      sched_setscheduler(2),    sched_setaffinity(2),    nice(2),    nice(1),

      You need some knowledge about the kernel and scheduling. The author  is
      a grumpy little elitist.


      Please  contact  me  via's  "contact  author"-feature

                                1 November 2006                  SCHEDTOOL(8)

jostber 01-23-2013 01:54 PM

Thanks. Has anyone tried out SlaXBMC? Does this work well?

dugan 01-23-2013 02:43 PM

You might consider switching from Alsa to OSS4 (Slackbuilds are on SBo). OSS4 is reputed to have lower latency.

All times are GMT -5. The time now is 08:47 PM.