Re: Speeding up AUCTeX's parsing

[Tassilo Horn]

Arash Esbati <arash@gnu.org> writes:

Hi Arash,

> I was playing with the regexp's in AUCTeX and before coming to the
> actual thing I wanted to test, I took an arbitraty, not so complex
> .tex file (around 1180 lines) and eval'ed these two forms:
>
>   (let ((gc-cons-threshold 800000))
>     (benchmark-run 15 (TeX-normal-mode)))
>
>   (let ((gc-cons-threshold most-positive-fixnum))
>     (benchmark-run 15 (TeX-normal-mode)))
>
> The results are:
>
>   (0.984757 13 0.517633)
>
>   (0.564401 0 0.0)
>
> Preventing GC has a large impact on parsing.  I'd like to hear from
> people with large, complex files if they could also run the test on
> them and report back their findings.

These are the results with the largest and most complex LaTeX document
I've ever written:

  (7.79938455 43 3.516914559)
  (4.881446238 0 0.0)

> And is there any major restriction why we shouldn't use this in order
> to speed up parsing?

The above effectively forbids gc during parsing and I'm not sure what
would happen when a very large file is parsed on a system with little
memory then.  Maybe it would be better to use a gc-cons-percentage value
of 0.5 or something.  With that, I get

  (5.124124372000001 6 0.5745011899999923)

which is not much slower but doesn't forbid gc altogether.

Bye,
Tassilo