initial_step_examples

@(initial_step) / variable-initializer examples (version8 / Enhancement-7)

Self-contained correctness example for the Enhancement-7 fix to Verilog-A @(initial_step) event gating, covering DC, AC, and Transient analysis. Everything here uses the version8 toolchain:

• compiler : ../OpenVAF-master-20260610/target/opt/openvaf-r
• simulator: ../ngspice-46/build/src/ngspice

See ../Enhancement-7.md for the full writeup. Before this fix, @(initial_step)/@(final_step) parsed but never actually gated execution — the event was silently discarded and the guarded statement ran on every evaluation, forever. Fixing this properly also surfaced (and required fixing) a deeper, separate bug: ordinary real/integer variables didn't persist their value across evaluations at all (see ../variable_persistence_examples/).

The model: a parameterized seed that's applied exactly once

module initial_step_demo(in, out);
    inout in, out;
    electrical in, out;
    parameter real seed = 100.0 from [0:inf);
    real accum = seed;
    analog begin
        accum = accum + 1.0;
        V(out) <+ accum + V(in);
    end
endmodule

accum starts at the declared initializer seed and is then self-referentially incremented every evaluation. This exercises the exact same ParamKind::IsInitialStep gating machinery as an explicit @(initial_step) accum = seed; statement would, but via the implicit declared-initializer path — see the note on a currently-open, separate crash below for why this example deliberately avoids the explicit form.

If the pre-Enhancement-7 bug were still present, accum would be reset to seed on every evaluation (not just the first), so V(out) would track seed + 1 + V(in) flat forever and never accumulate.

Results

Analysis	Sweep	Expected	Observed
DC	V(in) −1V to 1V	slope 1 in V(in), offset dominated by seed=100 applied once (plus small growth from Newton-iteration increments across the sweep)	matches (dc_plot.png)
AC	1Hz–100MHz	exact unity gain (0dB), 0° phase — accum has zero small-signal sensitivity to V(in)	flat 0dB/0° (ac_plot.png)
Transient	100kHz sine	input sine rides on a slowly-rising, persistent baseline starting near seed, not resetting to seed every evaluation	matches (tran_plot.png, dual-axis)

Known limitation (why this example avoids explicit @(initial_step) syntax)

An explicit @(initial_step) accum = seed; statement (as opposed to the plain declared initializer real accum = seed; used here) currently crashes the compiler — two related bugs, both in how insert_var_init's function-editing pass interacts with pre-existing branch structure from an explicit event-control statement (one is a CFG dominance violation, the other a panic in a separate sim_back::init cache-building pass). See ../Enhancement-7.md's "Known limitation" section for repro snippets and root-cause notes. Both patterns are redundant now anyway, since plain declared initializers already get the correct once-only gating automatically.

Layout

initial_step_examples/
  initial_step_demo.va    seed-initializer + persistence demo
  initial_step_demo.osdi  compiled with version8 openvaf-r
  dc_sim.cir / ac_sim.cir / tran_sim.cir
  dc_result.txt / ac_result.txt / tran_result.txt
  dc_plot.png / ac_plot.png / tran_plot.png

Reproduce

OPENVAF=../OpenVAF-master-20260610/target/opt/openvaf-r
NGSPICE=../ngspice-46/build/src/ngspice

$OPENVAF initial_step_demo.va -o initial_step_demo.osdi
$NGSPICE -b dc_sim.cir
$NGSPICE -b ac_sim.cir
$NGSPICE -b tran_sim.cir