In case you haven't seen it yet, there is a article titled "The Bench++ Benchmark Suite" by Joseph M. Orost in the October issue of Dr. Dobb's Journal.
In the article he compares the performance of his benchmark compiled under GNU g++ v2.7.2 and Sun CC in SC 4.0 on a Ultra-2 (even listing the compilation flags.)
Later he compares his benchmark compiled under Sun CC (no flags listed this time) on a Ultra-2 against the Ada PIWG suite compiled underVAX/Ada 2.1-28 (no compiler options listed) on a VAX 8650.
Why he didn't compare g++ vs gnat on the Ultra-2 (with and without error checking enabled) is something of a mystery to me. Why the editors of DDJ allowed such a meaningless comparison in their magazine is too depressing to contemplate...
Ron Skoog
rons@aonix