This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: pre egcs-1.1 testing and Linux 2.1.x
- To: Raul Miller <rdm at test dot legislate dot com>
- Subject: Re: pre egcs-1.1 testing and Linux 2.1.x
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Mon, 24 Aug 1998 11:50:45 -0600
- cc: "David S. Miller" <davem at dm dot cobaltmicro dot com>, torvalds at transmeta dot com, dkelson at inconnect dot com, egcs at cygnus dot com
- Reply-To: law at cygnus dot com
In message <19980824134853.W30542@test.legislate.com>you write:
> Jeffrey A Law <law@hurl.cygnus.com> wrote:
> > Unfortunately, we can not detect when it is going to generate incorrect
> > code.
>
> So "register pressure" isn't anything quantifiable?
You can make guesses about register pressure, but that still does not
tell you when you're going to generate incorrect code for regparm.
You have to have accurate life information in the reload pass (which
we do not currently have) and you have to tie it into the selection
of reload registers to detect when bad things are going to happen.
Most other ports don't have this problem because reload can never use
argument registers for reloads (if it tries, then it aborts).
jeff