This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: IA64 performance patch
- From: Richard Henderson <rth at redhat dot com>
- To: Steve Ellcey <sje at cup dot hp dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 5 Mar 2002 15:32:42 -0800
- Subject: Re: IA64 performance patch
- References: <200203042331.PAA18551@hpsje.cup.hp.com>
On Mon, Mar 04, 2002 at 03:31:15PM -0800, Steve Ellcey wrote:
> Here is an IA64 specific patch that fixes this by using different
> virtual registers to save/restore the gp for each function call. When
> not optimizing or when there are setjumps we still use a single fixed
> register.
Do they always get unified such that they only take up one
register or stack slot?
I'm not sure that this is important enough to address on the
branch, and for mainline I think it's the wrong approach.
There, we should simply not expose the gp register at all as
early as we do.
The earliest should probably be the instruction splitting
pass just before sched1. Even better would be to use
optimize_mode_switching to place the (single) copy from
the gp to the save register.
r~