profiling on sh

[mike stump]

> Date: Sun, 17 Mar 2002 23:26:07 -0800
> From: Mark Mitchell <mark@codesourcery.com>
> To: mike stump <mrs@windriver.com>, "aoliva@redhat.com" <aoliva@redhat.com>
> cc: "gcc@gcc.gnu.org" <gcc@gcc.gnu.org>

> I no longer remember the details here.

Luckily they are documented in the testcase you referenced in the cvs
checkin comment with the original work...  Right?  :-)

> I believe that someone (likely Jason or Benjamin) quoted something
> that suggested that it was safe to make this assumption, but that we
> never acted on that.  If you file a PR, and send me email noting
> this performance regression, we'll get it resolved before 3.1.

Ok, here is the requested email, and the bug is c++/5998.