SPEC introduces "Baseline Results" for the CINT92 & CFP92 benchmark suites.

PART A: Baseline results are those that adhere to the folloing restrictions
for the use of optimization flags for compilers and/or preprocessors.

1. The same compiler and the same set of flags is used for all benchmarks of
a given language within a benchmark suite, except for portability flags
(e.g. - DSYSV and library flags, e.g. -lcurses)

2. No assertion flags may be used. An assertion flag is one that supplies
semantic information not present in the source statement of the benchmarks.

3. Deleted

4. No compilation unit or variable or subroutine name may be used within an
optimization flag except for explicit portability requirements.

5. Explicit feedback directed optimizations (or similar technology) is
prohibited.

Part A.1: Baseline results published anywhere will carry new metrics for the 
speed and rate methods. No other new reporting requirements are proposed.

Part B: Baseline results can be published starting from the June 1994 
Newsletter. Pages in this issue are free for baseline results.

Part C: Starting with the June 1994 Newsletter, non-baseline results can
be published only if a baseline result is published for the same system. 
Non-newsletter peak results published after June 30 will require the addition 
of a baseline disclosure.

Part D: For publication outside the SPEC Newsletter, SPEC members, licensees
and the press are notified of the presence of the new metrics (base appended)
based on an article from Larry Gray (see part E). Baseline results will carry 
the word base appended to the original metric name.

part E: SPEC will write an article for the June Newsletter (and a press
release) announcing and explaining baseline results- what, how, why, etc.

Copyright (c) 1995 Standard Performance Evaluation Corporation