Path: utzoo!attcan!uunet!husc6!bloom-beacon!gatech!hubcap!ncrcae!ncr-sd!ncrlnk!fenix!kcby
From: kcby@fenix.Atlanta.NCR.COM (kcby)
Newsgroups: comp.lang.c++
Subject: Estimating size, LOC and performance in C++
Keywords: estimating, c++, memory, performance
Message-ID: <326@fenix.Atlanta.NCR.COM>
Date: 23 May 88 16:13:52 GMT
Organization: NCR Systems Engineering - Retail / Atlanta
Lines: 41



We are planning to use c++ to implement an object oriented design. Since
this is our first use of both c++ and the OO approach, I have no idea how
to answer the usual questions of "How much memory will the system take?",
"How long will it take to code (i.e. how many lines of code will be
required?)" and "How fast will it run?"

It seems to me that there are certain aspects of the design which must be
worked out before estimating any of these: what classes will be needed and
what messages must each of these classes support?  Then to work out memory
and coding time, some kind of inheritance structure must be assumed, to
determine how much reuse will occur. And a guess must be made as to how
many lines of code will be required for each message (either individually,
or in general).  Finally, to estimate how fast it will run, it seems that
you need to know the particular objects which will implement a given
operation, and what messages must flow between them to accomplish the
activity, as well as how many lines of code each message will require.
And, some kind of approximation will be required regarding c++ source to
native translation for memory and speed estimates.

Since none of this information is available early in the design process,
and since approximations on source to native size/speed require previous
implementations to look at, I wonder what alternatives might be used for
"guesstimating" size, coding time and performance for a first time OO/c++
implementation.  For example, could it be done by figuring what it would
take doing a standard structured approach using C, and then applying some
factors to figure size, LOC and performance? Or would I be better off making
wild swags at the figures for the information needed to estimate an OO
design?

Can anyone suggest an approach which they used? Did you over or under
estimate based on that approach? If so, by how much?

K.C. Burgess Yakemovic     	"Some of my best friends are mammals." 
kcby@Atlanta.NCR.com          		   - Robert T. Bakker, author
				 	     The Dinosaur Heresies
-- 
K.C. Burgess Yakemovic     	"Some of my best friends are mammals." 
kcby@Atlanta.NCR.com          		   - Robert T. Bakker, author
				 	     The Dinosaur Heresies