This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [dataflow] PATCH COMMITTED to fix storage and performance issues
- From: "Steven Bosscher" <stevenb dot gcc at gmail dot com>
- To: "Kenneth Zadeck" <zadeck at naturalbridge dot com>
- Cc: gcc-patches <gcc-patches at gcc dot gnu dot org>, "Christopher, Eric" <echristo at gmail dot com>, "Park, Seongbae" <seongbae dot park at gmail dot com>, "Zdenek Dvorak" <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- Date: Tue, 24 Apr 2007 16:10:40 +0200
- Subject: Re: [dataflow] PATCH COMMITTED to fix storage and performance issues
- References: <462E0D0E.5090000@naturalbridge.com>
On 4/24/07, Kenneth Zadeck <zadeck@naturalbridge.com> wrote:
This patch fixes a number of issue that have shown up.
1) This is the first part of a patch to remove unnecessary loads that
occur after stores.
Eric Christopher has volunteered to finish this part of the patch
because it requires a level of rtl manipulation that is above my skill
level. This patch does the hard part of identifying when the store
should be removed and does the easy case of where the load has the same
offset and mode as the previous store. The part that is missing is the
code that handles the case where the modes of the load and store do not
match and/or the load is loading only a part of the value stored.
For this part, I wonder how much it would help if you would re-enable
-fgcse-las (load-after-store). This was buggy at one point, but I
think all bugs are fixed now. It may be worth trying this, Eric...
Gr.
Steven