[tex-live] TeXLive 2005 is slower than TeXLive2003

[Hans Hagen]

Jerzy B. Ludwichowski wrote:
> David Kastrup wrote:
>   
>> karl at freefriends.org (Karl Berry) writes:
>>
>>   
>>     
>>>     And the directory of TeXLive 2003 is much bigger than that of TeXLive 2005.
>>>
>>> That doesn't actually mean much, given ls-R files.  The question (well,
>>> one question) is whether the configuration files are telling the
>>> binaries to search the disk in the 05 case and not the 03 case.  You
>>> might diff the texmf.cnf files and see if anything looks suspicious,
>>> like a missing !!.  You might also run kpsewhich --debug story.tex (say)
>>> and compare the output to see if there are major differences in what
>>> gets searched.
>>>     
>>>       
>> System calls can easily be tracked by running
>>
>> strace tex ...
>>
>> on a typical Unix-like system.
>>
>>   
>>     
> There  is a system performance monitoring tool (control panel > 
> administrative tools > performance)  or started from the command line as 
> perfmon.msc) which can trace various characteristics of system activity. 
> It has got a Help. I'd think that with a bit of luck one could find out 
> what's going on. I do not say that this will be easy.
>   
the speed of a tex run under windows (but probably any system) is to 
some extend determined by the file lookup but after the first tex run 
file caching normally deals with that

fabrice has optimized his kpse driver to deal with the way windows deals 
with filenames (unicode) [since  the ls database is just a text file 
conversion needs to take place] / this is also related to sub-runs, like 
mp ones, which demand reloading of the kpse data, which is taken care of 
by shared memory mechanisms

personally i have not noticed a speed difference; factors that influence 
things are memory (pdftex may allocate more memory, macro packages may 
load more patterns, map file handling, 'tex' suffix appended lookup, 
etc); actually map file lookup has been improved much the last few 
years;  since pdftex uses some mem of itself, it may be that it crosses 
the system mem boundary and swapping starts taking place

Hans

-- 

-----------------------------------------------------------------
                                          Hans Hagen | PRAGMA ADE
              Ridderstraat 27 | 8061 GH Hasselt | The Netherlands
     tel: 038 477 53 69 | fax: 038 477 53 74 | www.pragma-ade.com
                                             | www.pragma-pod.nl
-----------------------------------------------------------------