[tex-live] running programs from texlive tlc2

Tigran Aivazian tigran at aivazian.fsnet.co.uk
Thu Dec 9 12:12:33 CET 2004


just checked my email archives and found another inaccuracy --- it wasn't 
actually 1 month on Windows but a "mere" 3-4 days :)

Ok, so the absolute final truth is:

1. with no changes to the sources, switching from 2.2GHz Windows to 600MHz 
Linux results in a speed up from 3-4 days to 3.5 hours.

2. with trivial source changes it speeds up to 1 hour on Linux (or 1 day 
on Windows).

So, instead of saying "Linux is faster than Windows by hundreds" of times 
I should have checked my emails and and told only the facts --- it is only 
24 times faster (ignoring the 2.2GHz -> 600MHz, i.e. excusing Windows for 
using 2.2GHz machine as if it was only 600MHz).

Kind regards
Tigran

On Thu, 9 Dec 2004, Tigran Aivazian wrote:

> On Wed, 8 Dec 2004, Sebastian Rahtz wrote:
>
>> Tigran Aivazian wrote:
>> 
>>> I have seen some reasonably complex books typeset in LaTeX which can take 
>>> about a month of computer time to be processed under Windows Xp on a 2GHz 
>>> with 1G RAM machine and took only a few hours on a Linux Celeron 600MHz 
>>> with 256M RAM.
>>> 
>> I can't imagine what your books are like. I have been typesetting books for 
>> 20 years using LaTeX and I can't recall a LaTeX run for more than
>> an hour. I count 10 minutes these days as a really big TeX job.
>
> books with Hebrew or/and "interlinear" (bilingual) layout do take an awful 
> long time to typeset. My Hebrew OT takes 1 hour to process on Celeron 600MHz 
> 256M RAM but only 9.5 minutes on x86_64 2x4GHz 1G RAM (using natice 64bit 
> binaries).
>
>> 
>>> So, the Windows OS can make things tens if not hundreds of times slower
>> 
>> I find that a fairly extraordinary claim. What do you suggest the operating 
>> system may be doing
>> to hold up the processor, the disk, and the memory by that much? I have to 
>> assume that your LaTeX job
>> which takes so long is "computing", rather gthen vast amounts of disk 
>> access
>
> actually, there was a tiny little "white" lie in there :)
>
> i.e. in reality, it was a _combination_ of several factors. Someone (Victor 
> Zhuromsky) complained that his Interlinear Russian-Greek NT takes about a 
> month to do all three passes on a reasonably fast Windows machine (the one I 
> mentioned). He asked me to have a look. I tried his sources on my very very 
> slow Celeron 600MHz 256M RAM and it was painfully slow but the progress it 
> was making I estimated the completion at about 3-3.5 hours, which was still a 
> lot better than Windows.
>
> Then I had a look at his sources and noticed that he is not making any 
> paragraphs but only explicit line breaks \\ and manual indentation to 
> simulate paragraphs (which makes TeX very slow). So I fixed this and the 
> resulting sources passed all three passes (it has to pass three times in his 
> case) within a matter of an hour or so.
>
> So, the above story I "averaged" as "1 month -> 1 hour" conversion, skipping 
> some details :)
>
> Kind regards
> Tigran
>
> _______________________________________________
> TeX Live mailing list
> http://tug.org/mailman/listinfo/tex-live
>



More information about the tex-live mailing list