Skip navigation

Standard Performance Evaluation Corporation

 
 

176.gcc
SPEC CPU2000 Benchmark Description File


Benchmark Name

176.gcc


Benchmark Author

Richard Stallman and a large cast of helpers. For a full list, see http://www.fsf.org/software/gcc/thanks.html


Benchmark Program General Category

C Language optimizing compiler


Benchmark Description

176.gcc is based on gcc Version 2.7.2.2. It generates code for a Motorola 88100 processor. The benchmark runs as a compiler with many of its optimization flags enabled. 176.gcc has had its inlining heuristics altered slightly, so as to inline more code than would be typical on a Unix system in 1997. It is expected that this effect will be more typical of compiler usage in 2002. This was done so that 176.gcc would spend more time analyzing it's source code inputs, and use more memory. Without this effect, 176.gcc would have done less analysis, and needed more input workloads to achieve the run times required for SPECint2000.


Input Description

There are 5 input workloads in 176.gcc (compared with 57 in the 126.gcc of SPECint95). All these files are preprocessed C code (.i files); integrate.i and expr.i come from the source files of gcc itself. 166.i is made by concatenating the Fortran source files of a SPECint2000 candidate benchmark, then using the f2c translator to produce C code, and then pre-processing. 200.i comes via the same method from a previous version of the SPECfp2000 benchmark 200.sixtrack, and scilab.i comes via the same method from a version of the Scilab program.


Output Description

All output files are 88100 assembly code files.


Programming Language

The code is in C.


Known portability issues

  1. The code needs to be aware of the endianness of the host it runs on. 176.gcc will default to little endian. To run correctly on a big- endian machine, the flag HOST_WORDS_BIG_ENDIAN must be defined when the benchmark is compiled (eg -DHOST_WORDS_BIG_ENDIAN).
  2. Some of the optimizations 176.gcc performs require constant propagation of floating point constants. These form an insignificant amount of computation time, yet may depend on IEEE floating point format to produce a correct result.
  3. 176.gcc is not an ANSI C program. It uses GNU extensions.
  4. 176.gcc is inherently a 32-bit program. It has been successfully ported by SPEC to many 64-bit UNIX implementations. However use of high optimization levels with a 64 bit system in conjunction with inlining of procedures from different source files may reveal issues with 64-bit cleanliness with 176.gcc.
  5. 176.gcc has been changed slightly by SPEC to build properly with newer versions of GCC. If you're using an old gcc (~2.6 or older) to build 176.gcc, you should define the __OLDANDBUGGY__GNUC__ flag.

Reference

See the GCC home page at http://www.fsf.org/software/gcc/gcc.html


Last Updated: 25 April 2000