[TUHS] eqn

arnold at skeeve.com arnold at skeeve.com
Sun Oct 6 18:03:26 AEST 2019

Jon Forrest <nobozo at gmail.com> wrote:

> One slightly OT fact about TeX. On my 16GB, Core i7, SATA SSD
> Lenovo T430s laptop running Fedora 30, it takes ~3 seconds to run TeX on
> the ~900 page TeXBook. That's pretty fast.

You can thank Moore's Law for this. I remember trying to run TeX on
BSD 4.[12] vax with 4 Meg of memory and it taking many minutes to format
a single page.

The first time it became easy to run TeX, for me, was on sparcstation
class systems in the early 1990s.

> TeX contains all kinds of
> code to make it fit in the constraints of a 1980s computer. I wonder
> whether a redesign for a 2020 computer would be faster or slower.

I think it's just compute-intensive code. Moern versions of TeX use
WebToC to translate Knuth's web/pascal code to C, and that has been
the case for a long time.

(As an aside, everyone here who's read "TeX: The Program", raise
you hand.  [I have, but only once.])

> I suspect, but can't prove, that classic [nt]roff might also
> benefit in the same way. groff was written latter, so it might
> suffer less.

I don't think classic [nt]roff suffers at all. I remember (boy do I sound
like an old f*art) circa 1991, having both nroff and groff on a '486 class
system.  nroff was noticeably faster at formatting man pages than
groff was.  (Groff, of course, was ditroff and gave me PostScript output,
but comparing the two versions of nroff for text output, there
was a noticeable difference.)

Again, today, it doesn't really matter.


More information about the TUHS mailing list