[pdftex] TeX as a composition server?

[John Dey]

One more thing--I ran the test with "time latex test.tex" and it took 12.3 seconds to create the dvi file, 5.5 seconds to go from dvi to ps (dvips) and then over 3 minutes to convert the dvi to pdf using ps2pdf (ghostscript).  Resulting pdf file was twice a large as the pdf file created from pdflatex.


On Oct 25, 2010, at 4:40 PM, Reinhard Kotucha wrote:

> On 25 October 2010 Peter Davis wrote:
> 
>> On Mon, Oct 25, 2010 at 1:41 PM, Reinhard Kotucha
>> <reinhard.kotucha at web.de>wrote:
>> 
>>> On 25 October 2010 Peter Davis wrote:
>>> 
>>>> Ok, now I have a few days to prove that some flavor of TeX is
>>>> equal to the task.  Specifically, I want to have TeX compose
>>>> and output tens of thousands of pages (or copies of the same
>>>> page).  I suppose the simplest thing would be some kind of
>>>> look that just composes the same page of text over and over
>>>> again, as many times as I want.
>>> 
>>> Maybe I misunderstand, do you want to create documents where the
>>> content of all pages is exactly the same?
>>> 
>> 
>> Basically, yes.  I just want to see how fast TeX can compose a few
>> tens of thousands of pages of text.  It could all be "Lorem ipsum
>> ..." or whatever.
> 
> Ok, then I misunderstood you indeed.  If one really needs thousands of
> pages with the same content, one would certainly create only one page
> and set references to it in the PDF file.  This way I can generate
> 43,000 pages per second on my machine, regardless of the complexity of
> the page.  I had to redirect standard output to /dev/null, otherwise
> it's slower by the factor of two.
> 
> For what you want to achieve I usually do the same as James proposed
> already.  However, you have to do the test in \batchmode, on my
> machine TeX needs 4 seconds to print the page numbers to screen (with
> James' example).
> 
> I get some interesting results.  LaTeX needs 33 seconds while plain
> TeX needs only 23 seconds (both with pdftex).  Can anybody explain?
> The text doesn't contain any active characters or control sequences.
> I never expected such a big difference.
> 
> When producing DVI, pdftex needs 11.5 seconds while Knuth's TeX needs
> only 10 seconds.
> 
> pdflatex with font expansion and character protrusion takes a bit more
> than a minute, I can't test it right now with plain TeX.
> 
> BTW, I would generate a file which can be \input by other files.  Then
> one can compare different formats.
> 
> lorem.pl:
> -------------------------------------------------------
> #! /usr/bin/env perl
> $^W=1;
> my $n = 100000;
> my $lorem = <<'TXT';
> Lorem ipsum dolor sit amet, consectetur adipisicing elit,
> sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
> Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris
> nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor
> in reprehenderit in voluptate velit esse cillum dolore eu fugiat
> nulla pariatur. Excepteur sint occaecat cupidatat non proident,
> sunt in culpa qui officia deserunt mollit anim id est laborum.
> 
> TXT
> 
> open TEXT,">","lorem.tex";
> for(1..$n) {
>   print TEXT $lorem;
> }
> close TEXT
> -------------------------------------------------------
> 
> Then you can test it with different formats:
> 
> lorem-latex.tex:
> -------------------------------------------------------
> \batchmode
> \documentclass{article}
> %\usepackage{microtype}
> \begin{document}
> \input{lorem}
> \end{document}
> -------------------------------------------------------
> 
> lorem-plain.tex:
> -------------------------------------------------------
> \batchmode
> % Uncomment in order to create DVI:
> %\ifx\pdfoutput\undefined\else\pdfoutput=0 \fi
> 
> \def\newpage{\par\vfill\eject} % allow \newpage in lorem.tex
> \input lorem.tex
> \bye
> -------------------------------------------------------
> 
> Regards,
>  Reinhard
> 
> -- 
> ----------------------------------------------------------------------------
> Reinhard Kotucha			              Phone: +49-511-3373112
> Marschnerstr. 25
> D-30167 Hannover	                      mailto:reinhard.kotucha at web.de
> ----------------------------------------------------------------------------
> Microsoft isn't the answer. Microsoft is the question, and the answer is NO.
> ----------------------------------------------------------------------------