[fptex] pre-0.6 binaries
Hans Hagen
pragma@wxs.nl
Mon, 28 Jan 2002 12:04:07 +0100
At 09:29 PM 1/27/2002 +0100, Fabrice Popineau wrote:
>I have made a set of binaries using web2c-7.3.7 _plus some extension
>of mine_ . The zip file is available from :
>
>ftp://ftp.dante.de/pub/fptex/0.6/binaries-win32-20020127.zip
>
>It contains the 'bin/' and 'texmf/web2c/' directories. Be careful to
>save your current versions before using this. You need to rebuild the
>format files.
>
>I'd like to hear about stability, performance (better ?). I hope that
>it still runs on every platform.
when processing a 5000 page "hello world" document 30/31 sec pdf and 19/20
sec dvi using pdfetex there is no significant difference between old and
new binaries (will try later with more complex doc that opens more files).
since i guess that you have implemented shared kpsea database objects, i
also did some test with runtime metapost graphics:
100 simple graphics, i.e. 100 times system call for mp:
old programs : 44 sec
new programs : 7-8 sec
when graphics are processed between runs, 5 sec is needed for tex+mp but
each change in a graphic forces an additional run so for this doc we're
talking about n*5 sec which is larger than the 7-8 sec
1500 simple graphics, i.e. 1500 times mp call from within tex (indirect
texexec call):
old programs : 107 sec
new programs : 39 sec
btw, batch mode gives us another 6 seconds; when the dos box (4NT) runs on
top of an open window, or in full screen mode, writing to the screen (log
info) significantly slows down processesing (not related to new programs,
was always the case).
100 simple btex etec graphics, i.e. tex calls mp 100 times with each mp run
calling tex in turn
old program : 279 sec
new program : 98 sec
--automp : 7 sec
So, there is a significant improvement in running mp from within tex (and
tex from within mp)! For everyday jobs (where the number of mp graphics is
relatively low, say in the order of 100, runtime drops to 20%, which is
great, and even for simple test runs (or during style development) the gain
in speed is great (now running between jobs only makes sense for big docs,
like the metafun manual, with over 1500 graphics; and even there: running
from scratch in the direct mp processing takes some 3 runs, while in
between runs will take 5 (due to the fact that the new graphics start
contributing to the document); so even there, direct processing becomes an
option.
You did a great job!
Hans
PS. I use a 1000Mhz pentium laptop with 512 meg of mem, os: win2000
-------------------------------------------------------------------------
Hans Hagen | PRAGMA ADE | pragma@wxs.nl
Ridderstraat 27 | 8061 GH Hasselt | The Netherlands
tel: +31 (0)38 477 53 69 | fax: +31 (0)38 477 53 74 | www.pragma-ade.com
-------------------------------------------------------------------------
fall-back web server:
www.pragma-pod.nl
-------------------------------------------------------------------------