Mereni rychlosti [em|pc}TeXu na ruznych platformach

[Petr Sojka]

Archive-Date: Wed, 18 May 1994 13:28:12 CDT
From: gwhite at trevnx.bio.dfo.ca (George White 6-8509)
Subject: Windows for Workgroups (Win 3.11?) and tex+dvips performance
Date: 18 May 1994 11:32:14 -0300
Reply-To: gwhite at bionet.bio.dfo.ca
To: tex-news at SHSU.EDU

Several people have asked me if the 32-bit file access in Windows 3.11
has any effect on DOS programs.  Subjectively, 32-bit file access
makes a big improvement.  Since I was interested to see if there would
be an impact, I made some "before and after" measurements to provide
some hard numbers to back up the subjective impression.  Times are
from the POSIX "time" command (using the MKS Toolkit version in a DOS
window for the Windows results).

Results for tex (PC-TEX tex386 and emtex tex386), and  dvips 5.55 on
25 mhz NeXT '040 (OS version 3.0) and Intel 486DX2/66.  Note that the
real time on the NeXT is larger than the sum of user and system
times (there was one use logged in, but background processes such as
mail, system logging, etc. were running).

$ time tex dvips

System                real  user   sys
NeXT                  10.1   7.0   0.7
pctex - Win 3.1       10.2   2.8   7.4
        WfWG           6.0   2.5   3.6
emtex - Win 3.1        6.9   1.7   5.2
        WfWG           4.9   1.7   3.2

$ time dvips dvips

System                real  user   sys
NeXT                   8.1   6.3   1.1
PC    - Win 3.1       15.0   2.7  12.3
        WfWG           5.7   1.0   4.7

The PC was an EISA bus DEC MTE with 32M bytes of RAM and a fast
Addstor SCSI disk (much better disk than the one on the NeXT) and
Adaptec 2742 EISA SCSI controller.  TeX and dvips both require a lot
of disk activity.  Clearly the old DOS file system put TeX at a great
disadvantage, and those who may have felt that TeX on DOS was too slow
to merit consideration may want to take another look.
--

--
 George White <GWhite at BIOnet.BIO.DFO.ca> Bedford Inst. of Oceanography