Compilation Speeds SUN 2/50, 2/120, 2/120 FPA

[stuart at rochester.UUCP]

From: stuart

We recently got some SUN 2/50's (diskless) and a couple of floating
point accelerators to augment our SUN 2/120's.  I have a non-trivial
project using the SunWindow environment and decided to see how three of
the configurations perform on a intensive compilation.

The project includes a large number of public include files in addition
to the .c and .h sources of the project itself, so I have shown how the
project looks after preprocessor expansion.

There are too many variables to attempt a fully controlled experiment
so I tried only two runs on each of the machines.  The environment on
each machine was quiet but not silent (demons running and so forth) and
the network load was moderate.  The only attempt at eliminating variables
I made was starting each of the runs at about the same time.

The times reported are for compiling each of the source files with -g
and then loading all the objects and three library files with -g using
make.  The sizes reported are from the Release 1.2 systems; I don't
know what the differences are for Release 2.0.

                                Lines   Words   Bytes
1 project makefile                120     245    1932
24 project source files          2782    7213   60108
28 project header files           516    1247   11171
After preprocessor expansion    50711   37062  371769

24 project object files                         60108
3 public library files                         272498
(-lsuntool -lsunwindow -lpixrect)

resulting executable                           391168
stripped executable                            167936

Configuration   Real    User    System  Comments
SUN 2/50        717.7   300.8   127.1   Everything on file server
Release 2.0     721.5   299.6   128.5

SUN 2/120       580.0   322.2    93.2   /bin, /usr on file server
Release 1.2     575.1   321.8    93.9

SUN 2/120 FPA   561.6   322.0    93.3   /bin, /usr on file server
Release 1.2     562.8   321.0    93.3

NOTE: The lack of interpretative commentary is deliberate.

Stu Friedberg  {seismo, allegra}!rochester!stuart   stuart at rochester