Sound Config moOde 7

[TheOldPresbyope]

Once again I'll preface my post with "I'm not here to argue for or against tweaks" but this time I want to clear up an apparent misunderstanding concerning the Linux concepts of isolcpus and taskset.

isolcpus

This is a kernel parameter and this is the definition of it from kernel.org/doc/v5.4/admin-guide/kernel-parameters.html

Code:

isolcpus=       [KNL,SMP,ISOL] Isolate a given set of CPUs from disturbance.
                       [Deprecated - use cpusets instead]
                       Format: [flag-list,]<cpu-list>

                       Specify one or more CPUs to isolate from disturbances
                       specified in the flag list (default: domain):

                        nohz
                          Disable the tick when a single task runs.

                          A residual 1Hz tick is offloaded to workqueues, which you
                          need to affine to housekeeping through the global
                          workqueue's affinity configured via the
                          /sys/devices/virtual/workqueue/cpumask sysfs file, or
                          by using the 'domain' flag described below.

                          NOTE: by default the global workqueue runs on all CPUs,
                          so to protect individual CPUs the 'cpumask' file has to
                          be configured manually after bootup.

                        domain
                          Isolate from the general SMP balancing and scheduling
                          algorithms. Note that performing domain isolation this way
                          is irreversible: it's not possible to bring back a CPU to
                          the domains once isolated through isolcpus. It's strongly
                          advised to use cpusets instead to disable scheduler load
                          balancing through the "cpuset.sched_load_balance" file.
                          It offers a much more flexible interface where CPUs can
                          move in and out of an isolated set anytime.

                          You can move a process onto or off an "isolated" CPU via
                          the CPU affinity syscalls or cpuset.
                          <cpu number> begins at 0 and the maximum value is
                          "number of CPUs in system - 1".

                        The format of <cpu-list> is described above.

Oops. Deprecated? Yup, and has been for some time. However, as can be seen in the syntax definition, the flag-list is optional and the default results in the same behavior as in older kernels, e.g., "Isolate from the general SMP balancing and scheduling algorithms". Hence it still works, at least for now.

So, when one uses isolcpus=2,3 then only cpus 0 and 1 (assuming the RPi model in question has 4 cpus) can be assigned processes in a normal way by moOde/RaspiOS. On the other hand, cpus 2 and 3 will be used if and only if processes are manually assigned to them.

It appears to me in reading this forum thread that the first part of this concept is well understood but that the implication of being isolated from the balancing and scheduling algorithms is not as well understood.

taskset

From the man page for this command (reformatted to fix this forum page)

Code:

DESCRIPTION
      taskset  is  used to set or retrieve the CPU affinity of a running process given its
      pid, or to launch a new command with a given CPU affinity.  CPU affinity is a
      scheduler property that "bonds" a process to a given set of CPUs on the system.
      The Linux scheduler will honor the given CPU affinity and the  process  will  not
      run  on  any  other  CPUs.   Note that the Linux scheduler also supports natural
      CPU affinity: the scheduler attempts to keep processes on the same CPU as long as
      practical for performance reasons.  Therefore, forcing a specific CPU affinity is
      useful only in certain applications.

As noted by another in a previous post, using this command is equivalent to specifying CPUAffinity in the service file. Either way, there is a not-subtle difference between using a cpu range here when the isolcpus= parameter is not used and when it is used.

1. not used: The scheduler will honor the taskset cpu range. Suppose taskset=2,3. Then the specified process (and all the threads it may spawn) will only ever run on cpu 2 or cpu 3. If more than one process has the same assigned cpu range, then the scheduler/balancer will distribute them across the range as their algorithms dictate.
   
2. used: Suppose, again, taskset=2,3 but this time isolcpus=2,3 is also used. Then the specified process (and all the threads it may spawn) will only ever run on cpu 2. If more than one process has the same assigned cpu range, then they will all only ever run on cpu 2. Remember, these cpus have been isolated so the scheduler/balancer algorithms never see how they are being used or not used. 
   

One practical implication of this for our example case is that cpu 3 is never used (except for the inevitable kworker processes which get started on every cpu). That's 25 percent of the total cpu capacity. Even more is lost because the other isolated cpu likely is lightly loaded by the few processes which can run on it. 

Another practical implication is that all the remaining processes are thrown onto cpu 0 and cpu 1, essentially doubling their loads. On a typical RPi4B moOde installation, I've found that each of the 4 cpus is running approximately 35 processes, albeit not necessarily all active at once. With 2 cpus isolated, on the other hand, the remaining two cpus are each running approximately 70 processes. Whether this introduces new performance issues depends on the overall usage.

Bottom line:

If you must isolate a process using this deprecated technique, then I suggest you try using isolcpus=3 and taskset=3. If you want to isolate two processes, each to its own cpu, then try using isolcpus=2,3 while using taskset=2 for one process and taskset=3 for the other.

Regards,
Kent

PS - the purpose of this post is only to clarify Linux. I don't wish to discuss fine tuning either the deprecated technique or the cpusets alternative. Consult your favorite search engine.