From: sommar@enea.se (Erland Sommarskog)
Subject: Generics and disk overhead
Date: 2 Mar 88 12:51:46 GMT [thread overview]
Message-ID: <2792@enea.se> (raw)
The scene: Four generic packages, a list handler(L), a queue and
stack handler(QS), a binary-tree handler(B) and a package for
topological sorting(T). Some of the procedures in B are also generic.
Add to this a main program (M). Now M instantiates T, which instantiates
B and QS and also some of the procedures in B. Finally QS instantiates L.
All packages have a limited type parameter. The size of the code was
around 1100 lines, including comments, in kbytes around 40.
After compilation and linking this expanded to 900 kbytes! I find
that quite extreme. My environment is VADS 5.41, 4.3 BSD and VAX.
VADS uses the DIANA format, and these files are also the big
consumers of disk space, but I have no idea if it is really
necessary.
To compare I also compiled and linked the program on VMS with
DEC-Ada. This costed 430 blocks or 220 kbytes. Somewhat more
reasonable. However, DEC does cheat. Recompiling a generic body
invalidates the instantiations. On the other hand recompilation
with seven reinstantiations took a good time with VADS.
I'm not really sure whether DEC's cheating is permitted or not. The
paragraph in question would be 10.3(7-9). Anyone who knows? But
VADS seems definitely to be doing it wrong. Under some circumstances
the library acts as if a generic body was dependent of its instantiation.
--
Erland Sommarskog
ENEA Data, Stockholm
sommar@enea.UUCP "Souvent pour s'amuser les hommes d'equipages
and it's like talking to a stranger" -- H&C.
reply other threads:[~1988-03-02 12:51 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox