Date: Mon, 17 Jul 2000 19:42:18 -0400
Message-Id: <200007172342.TAA16962@envy.delorie.com>
From: DJ Delorie <dj AT delorie DOT com>
To: djgpp-workers AT delorie DOT com
In-reply-to: <39735F47.4999.174596@localhost> (snowball3@bigfoot.com)
Subject: Re: Fw: x86: Not clobbering %edx at function entry when profiling
References:  <39735F47 DOT 4999 DOT 174596 AT localhost>
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk


> This patch has been installed in gcc. Should the DJGPP take advantage of 
> this change in its target macro file?

No, because...

> > argument in it when -mregparm is used.  Modern implementations of
> > mcount do not need the value we load into %edx; it's for the ancient

DJGPP *does* use the value in %dx, for *both* -pg and -p.  It stores a
pointer through it, as a sort-of LRU cache for the structures it has
created in its own memory.  This cached pointer is used to speed up
profiling.

We probably should modify our profiling stuff to save %edx around that
call, if needed and other solutions aren't available.

DJGPP also has (or at least had, last time I checked) support for both
profile types.