EQNTOTT
                        -------

1.0: GENERAL

1.1: Classification EQNTOTT is  an  integer-intensive  benchmark.
The primary kind of computation performed is sorting.

1.2:  Description The benchmark translates a logical  representa-
tion of a boolean equation to a truth table.

1.3:  Source/Author The source is from the Industrial Liason Pro-
gram of the University of California at Berkely.

1.4:  Version/Date The version used by SPEC is version 9, precise
date unknown, but thought to be 1985.

2.0:  PERFORMANCE

2.1:  Metrics: SPEC uses total elapsed time as  reported  by  the
shell, /bin/time, as the performance metric.

2.2:  Elapsed time: EQNTOTT runs in under 3 minutes  on  Apollo's
DN10000. The reference time (to 3 sig. fig) is 1100.

2.3: Reports: The benchmark produces a  truth  table  as  output.
When  the  benchmark is run, the output is re-directed to a file,
so the only visible output is the output from /bin/time.

2.4: Other performance characteristics: EQNTOTT dynamically allo-
cates  about  1.8 megabytes of memory in 450 calls to the library
routine malloc, when run with the int_pri_3.eqn input  file.   In
addition,  the benchmark spends nearly 95% of its time in the li-
brary routine qsort.

3.0:  SOFTWARE

3.1:  Language EQNTOTT is written in the C language and  requires
the YACC parser generator in order to build the executable image.

3.2:  Operating System and Portability The program was originally
developed  under  BSD4.2  UNIX  and will port easily to other BSD
derived versions of UNIX.  It ports with some greater  difficulty
to System V versions.  It calls on very few system services so it
should port relatively easily to non-UNIX systems that support C.

3.3:  Vectorizability, Parallelizability EQNTOTT  spends  all  of
its time sorting.  To the extent that the host's system sort rou-
tine is parallelized, the benchmark will benefit.

4.0  HARDWARE

4.1: Memory EQNTOTT does not page when run on  a  system  with  8
megabytes  of  physical memory.  Although not tested, it probably
requires less than 2.5 mega- bytes to run without paging.

4.2:  No other special hardware considerations.

5.0 OPERATIONAL

5.1: Disk About 2.5 megabytes of disk are required  to  hold  the
source code, the run time files and the executable image.

5.2:  Installation  EQNTOTT  is  distributed  in  the   directory
eqntott.   That  directory  contains all of the source code, run-
time files and a Makefile  for  building  the  executable  image.
There  is  a  README  file in the directory that describes how to
build and run the benchmark.  The Makefile passes CFLAGS to the C
compiler and you can specify appropriate flags for your host sys-
tem.

Since the benchmark uses the shell's time command,  there  is  no
timing routine in the source that needs local modification.

5.3:  Execution Once the executable image is built, the benchmark
is run in the following manner:

        time eqntott -s -.ioplte int_pri_3.eqn > int_pri_3.new
        cmp int_pri_3.new int_pri_3.esp

If things work properly, the only output produced is  the  output
from 'time'.

5.4:  Correctness Verification The 'cmp' command in the Execution
section  above  compares the output against an known good result.
No differences should be reported.