On Mon, Apr 26, 2021 at 2:03 PM Norman Wilson <norman@oclsc.org> wrote:

> Adam Thornton:
>
>   I sat in on an undergrad course from [Dave Hanson] my first year of
>   grad school (94-95) and he taught it with lcc.  I asked `why not
>   gcc' and he said, `gcc is 100,000 lines and I don't know what 90%
>   of them are doing; lcc is 10,000'.
>
> ===
>
> My copy is indeed about 10K lines, not counting the code-generator
> modules.  Those are C files generated by a utility program lburg
> from a template file.  The three architectures supplied in the
> distribution, for MIPS, SPARC, and X86, have template files of
> about 900, 1200, and 700 lines respectively.
>
> The template file for the VAX is about 2800 lines, but includes
> some metalanguage of my own, interpreted by an awk script, to
> generate extra rules for all the direct-store type-to-type
> instructions.  The C output from lburg for the other architectures
> is 5000-6000 lines; for the VAX, after expansion by my awk
> program and then by lburg, is nearly 20K.
>
> Did someone say Complex Instruction Set?
>

Indeed!

https://yarchive.net/comp/vax.html

I recall one of Mashey's posts talking about the number of page faults that
_might_ arise from execution of one instance of one of the more baroque VAX
instructions. It was something like 40 (!!).

        - Dan C.