Re: [Ngspice-devel] Bison grammar

[Jim M. <jim...@gm...>]

Hi Evan and Alan,

Here are a few comments related to Alan's recent post about model
processing.

There is indeed special processing for each letter/component type. They all
call the common whitespace skipping function INPgetNetTok() I mentioned
earlier, which is called in a bit over 100 places.

Currently, the "ngspice" format of SPICE is essentially being used as an
intermediate language. Ngspice supports many flavors of SPICE used by
various software implementations of SPICE such as LTspice, and it
accomplishes much of this support by a preprocessing that translates the
relatively small differences into the format that is native ngspice. If a
more general format is selected, a good approach may be to decouple the
parsing problem into an independent step where any SPICE format first gets
transformed to it and then the parsing of the new format. For any problem
that takes a long time to run, the input processing should not account for
a significant amount of the time, and the added translation should not even
be noticeable. The choice of the new format should be considered carefully.

Something to be aware of is that the current input processing is complex
with many steps, so major changes like a new format will not be at all
trivial to implement. However, a clean modular unit to do the parsing would
be helpful in making the ngspice source code clearer and easier to maintain
and modify.

Jim

[image: Mailtrack]
<https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&>
Sender
notified by
Mailtrack
<https://mailtrack.io?utm_source=gmail&utm_medium=signature&utm_campaign=signaturevirality5&>
10/02/20,
10:55:40 AM