Clarification on expert_execution policies and CPU affinity mapping

[ALISROURR]

Clarification on expert_execution policies and CPU affinity mapping

Hello srsRAN Community,

While reviewing the expert_execution configuration, I came across several points I’d like to clarify-- particularly around pinning policies, Radio Unit tasks, and L1/L2 CPU mappings.

expert_execution:
  cell_affinities:
    -
      l1_dl_cpus:       # Optional TEXT. CPU core(s) for L1 downlink tasks. Supported: [1, 2, 3, ..., N].
      l1_ul_cpus:       # Optional TEXT. CPU core(s) for L1 uplink tasks.   Supported: [1, 2, 3, ..., N].
      l1_dl_pinning:    # Optional TEXT. Policy used for assigning CPU cores to L1 downlink tasks.
      l1_ul_pinning:    # Optional TEXT. Policy used for assigning CPU cores to L1 uplink tasks.
      l2_cell_cpus:     # Optional TEXT. CPU core(s) for L2 cell tasks.     Supported: [1, 2, 3, ..., N].
      l2_cell_pinning:  # Optional TEXT. Policy used for assigning CPU cores to L2 cell tasks.
      ru_cpus:          # Optional TEXT. CPU core(s) for Radio Unit tasks.  Supported: [1, 2, 3, ..., N].
      ru_pinning:       # Optional TEXT. Policy used for assigning CPU cores to Radio Unit tasks.
  affinities:
    isolated_cpus:        # Optional TEXT. CPU core(s) isolated for the gNB application.
    low_priority_cpus:    # Optional TEXT. CPU core(s) for low-priority tasks.
    low_priority_pinning: # Optional TEXT. Policy used for assigning CPU cores to low-priority tasks.
    ru_timing_cpu:        # Optional TEXT. CPU used for timing in the Radio Unit.
    ofh:

Questions

Default pinning policies

• What is the default policy for each *_pinning field if not explicitly set?
• What are the allowed values/options for these pinning policies?

Definition of Radio Unit tasks

• What exactly falls under Radio Unit tasks here?Does this correspond to lower-PHY processing or another functionality?

L1 CPU mappings (DL/UL)

• Which L1 sublayers/functions are covered by:
  • l1_dl_cpus ?
  • l1_ul_cpus ?

Code references

• Which files implement these configuration keys and pinning behaviors?
• Is there a way to control execution control at a more detailed level of control(e.g., per sublayer function within L1/L2)?