Plusargs, Files & Runtime Configuration

Overview

A compiled simulation image is expensive to build and cheap to run. Runtime configuration is what lets one image serve a thousand tests: plusargs select the test and tune its knobs, files feed stimulus and memory images in and carry logs and dumps out, and wave-dump controls decide what gets recorded. Teams that configure at run time iterate in seconds; teams that recompile for every knob change iterate in hours.

This topic covers the whole runtime boundary: $value$plusargs and $test$plusargs for command-line knobs, $fopen/$readmemh for file traffic, the three-layer configuration architecture that keeps it all sane, and the dump/exit controls that make regressions debuggable.

Sub-topics

1. $test$plusargs & $value$plusargs — existence vs value parsing, defaults, and the knob registry idiom.

2. File I/O — $fopen/$fclose, formatted writes, parsing reads, and $readmemh memory preload.

3. Config Architecture: Files vs Plusargs vs Objects — the three-layer model and precedence design.

4. Memory Preload & Dump — hierarchical readmem, generated images, end-of-test dump and compare.

5. Wave Dumping & Runtime Control — dump scope and windows, $finish/$stop/$fatal, verbosity switching.

TOPIC MAP — THE RUNTIME BOUNDARY

  build once                         run many times
  ──────────                         ──────────────
  compile + elaborate ──► simv ◄── +TESTNAME=smoke +VERBOSITY=HIGH
                            │  ◄── +PRELOAD=image.hex +DUMP_START=12us
                            │
            ┌───────────────┼────────────────┐
            │               │                │
            ▼               ▼                ▼
        plusargs        file I/O         wave control
        (lesson 1)      (lesson 2)       (lesson 5)
            │               │
            ▼               ▼
        config object   memory preload
        architecture    & dump compare
        (lesson 3)      (lesson 4)

Key takeaways

• One compiled image, many configured runs — plusargs and files are the levers that make regressions fast.

• Parse plusargs once into a central config object; never scatter $value$plusargs calls across the testbench.

• Files carry data both ways: preload images and stimulus in, logs, memory dumps, and waves out.