summary.yaml

Tells cicsim summary which result sets go into the README table.

description: TB_NCM
simulations:
   tran:
    name: Transient analysis
    description: |
      Check transient operation
    data:
      - name: Sch_typ
        src: results/tran_Sch_typical
        method: typical
      - name: Sch_etc
        src: results/tran_Sch_etc
        method: minmax
      - name: Sch_3std
        src: results/tran_Sch_mc
        method: 3std

The key under simulations is the testbench name, and it is load bearing: cicsim reads the specification from <key>.yaml, so tran: here is what makes tran.yaml the spec file. Rename the testbench and this key has to follow, or the table comes out empty.

Under data, one entry per column of the table. src is the result directory a corner run wrote, and method says how to reduce the runs in it:

Method	Typical column	Min / max columns
contains `typ`	median	none
contains `3std`	mean	mean ± 3σ
contains `std`	mean	mean ± 1σ
anything else	median	the actual min and max of the set

The match is a substring test in the order above, so typical selects the first row and minmax falls through to the last. Two consequences worth knowing: typical reports the median of whatever runs are in that directory, not “the one typical run” — it only looks like a single value because the set usually holds one — and minmax reports a real minimum and maximum, so it is the honest worst case rather than a σ estimate.

The three defaults line up with the typical, etc and mc targets of the Makefile, which is why plain make runs exactly those before summary. Add a Lay_typ entry pointing at results/tran_Lay_typical to get the post-layout column beside the schematic one.