[comp.text.tex] brief comparison of 2 DOS Metafonts

teexdwu@ioe.lon.ac.uk (DOMINIK WUJASTYK) (11/17/90)

I spent some time recently comparing sbmf and emmf: 

Times are in seconds.  The tests were run on a Dell 220 (80286 at
20MHz) with VGA and about 540k available RAM.  sbMF was from sbmf11.zip
and is version 2.7(c)sb11.  Matte's MF is version 2.0 [3a].

	cminch		cminch with	cmr10
	@ 300dpi	graphics	@ 300dpi
			display
-------------------------------------------------
sb MF	54.10		59.65		170.92
em MF	64.32		71.51		193.83

In other words, both implimentations are excellent.  Wayne Sullivan's
is a little faster (between 10% and 18%).  His distribution now
includes sbmkbat.exe, a batch file writer that reads your
specifications for MF command line parameters and then writes a big
batch file to do the work (including gftopk, and putting things in the
right directories).  sbmkbat is very good, and is only 9k compiled, as
against about 140k for mfjob.  I would say the syntax of mfjob is a
little easier than the sbmkbat input, but hey, they're both okay.

My opinion:  The same as with sbTeX and emTeX: emTeX comes as a
complete distribution, with full instructions for installation etc.,
fonts, DVI drivers, the lot.  Also, Big versions of TeX and MF in case
you have to compile something colossal.  sbTeX and sbMF are just TeX
and MF (and an editor and sbmkbat).  They need a little more memory,
but if you've got it they have the edge in speed.  If you're a speed
freak, then install emTeX and slot sbMF and sbTeX in place of Matte's
versions.  (and use sbmkbat instead of mfjob.)  

Users of DOS should consider themselves lucky to have a choice of two
such excellent implementations.  Thank you, Wayne and Eberhard.  And
keep competing!

Incidentally, I find it interesting that the speeds of these
implementations are so close.  Wayne works with WEB, a normal change
file (with assembly language portions) and Turbo Pascal.  Eberhard uses
a program to translate the output of Tangle into C and then compiles
with Microsoft C 6.  Very different routes to very similar results,
no?


Dominik

mattes@azu.informatik.uni-stuttgart.de (Eberhard Mattes) (11/27/90)

In article <1990Nov16.200934.21562@ioe.lon.ac.uk> teexdwu@ioe.lon.ac.uk (DOMINIK WUJASTYK) writes:

> I spent some time recently comparing sbmf and emmf: 
> [...]
> In other words, both implimentations are excellent.  Wayne Sullivan's
> is a little faster (between 10% and 18%).  His distribution now
> includes sbmkbat.exe, a batch file writer that reads your
> specifications for MF command line parameters and then writes a big
> batch file to do the work (including gftopk, and putting things in the
> right directories).  sbmkbat is very good, and is only 9k compiled, as
> against about 140k for mfjob.  I would say the syntax of mfjob is a
> little easier than the sbmkbat input, but hey, they're both okay.
If you want to create a batch file using MFjob, you can use
    mfjob all m=lj /n | sed 1d;/^=/d >mfjob1.bat
Or (if you don't have sed) use a text editor to remove the first line
and all lines beginning with = from the MFjob output.
Using a batch file saves memory (and therefore may speed up MF) but
you don't have the error detection and log file saving features of MFjob.
You can even change the batch file to call sbmf instead of `emmf'.

BTW, Wayne Sullivan's METAFONT supports more graphics adapters:
Hercules, for instance.

--
       Eberhard Mattes   (mattes@azu.informatik.uni-stuttgart.de)