Part 8 · Checking & Coverage · Intermediate

Message IDs and Verbosity Control

Stable ID naming, verbosity levels, plusargs, and per-component set_verbosity.

Message IDs — the filter key

Every UVM message carries a string ID — the primary key for selective verbosity control. Good IDs are stable, specific, and grep-friendly. They enable turning up debug on one component without flooding the log from the entire environment.

systemverilog

// GOOD IDs — specific, stable, grep-friendly
`uvm_error("SCB_MISMATCH",  "addr 0x%0h exp=%0h act=%0h", addr, exp, act)
`uvm_error("SCB_UNEXPECTED","id %0h no expected entry", id)
`uvm_fatal("NOVIF_APB",     "apb_vif not set in config_db")
`uvm_info("SCB",           "matched id=%0h", id, UVM_HIGH)
`uvm_info("RAL_PRED",      "predict addr=0x%0h data=0x%0h", addr, data, UVM_FULL)

// BAD IDs — useless for filtering
`uvm_error("ERR",  "something failed")
`uvm_info("DBG",  "value = %0d", val, UVM_NONE)

Verbosity ladder

diagram

Legend: [UVM]

  VERBOSITY LADDER (lowest → highest)

  UVM_NONE    ──► Always prints (used for PASS/FAIL result messages)
       │
  UVM_LOW     ──► Phase transitions, report_phase summaries, coverage %
       │
  UVM_MEDIUM  ──► Standard regression level — key events, error context
       │
  UVM_HIGH    ──► Per-transaction debug — scoreboard match/mismatch
       │
  UVM_FULL    ──► Detailed field dumps — register writes, TLM connections
       │
  UVM_DEBUG   ──► Everything — factory overrides, config_db sets

  Print rule: message prints if message_verbosity <= component threshold
  Example: UVM_HIGH message prints at UVM_MEDIUM threshold? NO (HIGH > MEDIUM)
           UVM_HIGH message prints at UVM_FULL threshold?  YES (HIGH <= FULL)

Controlling verbosity

systemverilog

// In component build_phase — set default for this component
set_report_verbosity_level(UVM_MEDIUM);

// Per-ID override — only SCB messages at HIGH
set_report_id_verbosity("SCB", UVM_HIGH);

// Per-ID suppress — silence VIP noise entirely
set_report_id_verbosity("VIP_INTERNAL", UVM_NONE);

bash

# Global verbosity via plusarg
simv +UVM_VERBOSITY=UVM_MEDIUM

# Per-component at time 0 (before run_phase)
simv +uvm_set_verbosity=uvm_test_top.env.scb,_ALL_,UVM_HIGH,time

# Per-component + per-ID
simv +uvm_set_verbosity=uvm_test_top.env.scb,SCB,UVM_FULL,run

# Suppress all, then enable one component
simv +uvm_set_verbosity=*,_ALL_,UVM_NONE,time \
     +uvm_set_verbosity=uvm_test_top.env.scb,_ALL_,UVM_HIGH,time

Regression vs debug verbosity guide

diagram

Use case                    Recommended verbosity
──────────────────────────────────────────────────────────
Regression batch              UVM_MEDIUM (default)
Single test debug             UVM_HIGH on failing component only
Register/RAL debug            UVM_FULL on reg predictor component
Factory/config debug          UVM_DEBUG (brief, targeted)
PASS/FAIL result line         UVM_NONE (always visible)

Key takeaways

IDs enable selective debug; verbosity controls uvm_info volume.
Regression at UVM_MEDIUM; raise to UVM_HIGH on failing component only.
UVM_NONE for result messages — always visible regardless of threshold.

Common pitfalls

Generic ID "ERR" everywhere — cannot filter selectively in regression.
UVM_FULL globally in regressions — log files grow to gigabytes.
UVM_NONE on debug messages — floods log even at UVM_MEDIUM threshold.

Practice this lessonQuestions tagged for this topic in the bank