Next: Continuous Compilation vs. Traditional
Up: Experiments
Previous: Experiments
As mentioned earlier, we used both real and synthetic
programs in our experiments. This section describes the real programs
that were used.
We studied the following programs: rcc, gzip,
gnuchess, ghostview, pico, and render. (See
Table 6.1.)
Table 6.1: Real Programs Studied
Lcc, gzip, gnuchess, and ghostview are all
under the GNU General Public License. Pico is copyrighted by
the University of Washington. Render was developed as a class
project at Washington University.
The performance of each program is, of course, dependent on its input.
The input for each program is as follows.
- rcc
- The input to rcc was the source file x86.c, which is
lcc's code generation module for the Intel x86 architecture.
- gzip
- The input provided to gzip was a 206 Kbyte directory.
- gnuchess
- The chess program was set up to play against itself
with a very shallow search depth.
- ghostview
- Ghostview was run with the file ``ghostview.ps''
specified on the command line. This 92 Kbyte Postscript file is a
documentation file included with ghostview. Each of the 14
pages was viewed quickly before exiting the program.
- pico
- The editor was started without an input file. Text was
entered for about a minute, the file was saved, and then the program
was exited.
- render
- The input to the render program was a batch file which
instructed the program to raytrace a very simple scene.
Next: Continuous Compilation vs. Traditional
Up: Experiments
Previous: Experiments