[tex-live] TeXLive 2005 is slower than TeXLive2003

[gnwiii at gmail.com]

On 9/8/06, Hans Hagen <pragma at wxs.nl> wrote:

> David Kastrup wrote:
> > I thought that in most respects, the Windows and Unix executables
> > should be compiled from the same source code, right?

> no, one reason for fabrice quitting the dev is that this was not the
> case; this should be known by now:
>
> https://xemtex.groups.foundry.supelec.fr/xemtex-web-gb-2-5.html

Even without changing the compiler flags, using different compilers,
options, runtime libraries, etc. can have a big impact.  In an earlier
post I compared TL2005 with
RH tetex3.  TL2005 is normally built with static libraries.
Relinking with all dynamic
(as was done for tetex3) gives run times only slightly longer than the
tetex3 times:

Fabrice put the kpathsea structures in shared memory so they didn't
have to be loaded again for each run.   TL2003/fptex sometimes asks
you to restart the programs after updating the texmf trees -- I don't
recall seeing this on TL2005 (Win32), but I also don't recall TL2005
on Win32 taking 7 s. to compile sample2e.
I doubt Fabrice's tweaks make that sort of difference.  My experience
has been that most programs have similar run times on linux and Win32
(but it generally takes considerably more effort to get things to work
properly on Win32).

-- 
George N. White III <aa056 at chebucto.ns.ca>
Head of St. Margarets Bay, Nova Scotia