Stroustrup's statistics

Phil Pfeiffer's fine notes about the 1993 Perlis Symposium at Yale
summarizes Bjarne Stroustrup as saying:


Grady Booch's 140,000-line Ada data structures library, for
example, shrunk to 10,000 lines of code when rewritten in C++:
the run-time speed and memory consumption of the two libraries
are identical.


Stroustrup's statistics might sound impressive, but they remind me of a
talk that John Flower of ParaSoft gave at the Los Angeles chapter of the
ACM on January 8, 1992. Here's my summary of part of Flower's talk:


We had one customer who wanted to parallelize a global weather
model written in C++ by some grad student, about 8000 lines of
code. Well, parallelizing C++ is generally a waste of time,
so we recoded it into Fortran, where it was about 6000 lines.
The Fortran version parallelized OK, but the performance was
still a bit low, so we recoded it into Fortran 90, and it was
about 600 lines as a result (and much faster). Granted, it's
just one example, but for this kind of software Fortran 90
seems to work fairly well.


Every language partisan will have a story or two like this, I suppose.
But what strikes me about these two stories is how Stroustrup is still
refighting the battles of the '70s, whereas Flower is at least trying to
take advantage of newer technology.
--

Return to the comp.compilers page.
Search the comp.compilers archives again.