[OS X TeX] Re: The microtype mystery

[Martin Costabel]

Josep Maria Font wrote:
[]
> Again, a very large increase factor! The preamble of the document is:
> 
> \documentclass[a4paper]{article}
> \usepackage{fixltx2e}
> \usepackage{amssymb,latexsym}                         
> \usepackage{amsmath,amsthm}
> \usepackage{bm}                 
> \usepackage{enumerate}
> \usepackage{mathtools}
> \usepackage{url}
> 
> I will e-mail the files privately to Robert. 

Here is another data point, obtained by playing with your preamble.

 From my own experiment, I would suspect the interaction of microtype 
with the latexsym or similar packages:

I took some random article of mine (46 pages; if others want to repeat 
this experiment, the latex sources, two files, are publicly available 
here: <http://arxiv.org/format/1002.1772v1>). Then I inserted after the 
\documentclass line

\usepackage{microtype}

and then \usepackage{latexsym}, or alternatively just the following two 
lines from latexsym.sty:

   \DeclareSymbolFont{lasy}{U}{lasy}{m}{n}
   \SetSymbolFont{lasy}{bold}{U}{lasy}{b}{n}

These two lines do absolutely nothing visible, as far as I can tell. But 
here are the timings from

  time pdflatex CoDaNi_2010_art1

on my MbookPro core2duo:

Original file:
user	0m1.835s

With just the two lines from latexsym.sty:
user	0m1.909s

With \usepackage{microtype} alone:
user	0m8.591s

With \usepackage{microtype} plus the two lines from latexsym.sty:
user	0m49.910s

This is a factor of roughly 5 for microtype alone, and of 25 for 
microtype+latexsym. The extra time is spent in raw computing cpu time, 
no system calls or disk activity.

-- 
Martin