From: Dorit Nuzman <DORIT@il.ibm.com>
To: Ira Rosen <IRAR@il.ibm.com>
Cc: gcc-patches@gnu.org, dnovillo@google.com, dberlin@dberlin.org,
dvorakz@suse.cz
Subject: Re: [patch] Remove redundant calls to copy_virtual_operands
Date: Tue, 07 Aug 2007 13:06:00 -0000 [thread overview]
Message-ID: <OFEC28E137.8C9E6009-ONC2257330.004675DC-C2257330.00481FC7@il.ibm.com> (raw)
In-Reply-To: <OF930B00A8.8262D5E2-ONC2257330.0032AB21-C2257330.0042B670@il.ibm.com>
>
> Currently the vectorizer copies virtual operands from scalar memory
> operations to the corresponding vector operations (that the vectorizer
> creates). This seems to be redundant, since we set memory tags for the
> memory operations that we create and call update_ssa at the end of loop
> vectorization. This patch removes these redundant calls to
> copy_virtual_operands().
>
If I'm not mistaken this patch also fixes an ICE when applied on top of
gcc4.1.1, so not only are these calls apparently redundant, but it seems
that they can also be harmful.
> Bootstrapped and tested on ppc-linux and x86_64-linux.
on x86 it also passed bootstrap with vectorization enabled, right?
> O.K. for mainline?
>
it looks ok to me, but it would be much better if one of the aliasing/vops
experts took a look
thanks,
dorit
> Ira
>
> :ADDPATCH SSA:
>
> ChangeLog entry:
>
> * tree-vect-transform.c (vectorizable_store): Remove call to
> copy_virtual_operands() and call mark_symbols_for_renaming() for
> the created vector store.
> (vect_setup_realignment): Don't call copy_virtual_operands() and
> update_vuses_to_preheader().
> (vectorizable_load): Don't call copy_virtual_operands().
> (update_vuses_to_preheader): Remove.
>
> (See attached file: copy_vops.txt)[attachment "copy_vops.txt"
> deleted by Dorit Nuzman/Haifa/IBM]
next prev parent reply other threads:[~2007-08-07 13:06 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-08-07 12:10 Ira Rosen
2007-08-07 13:06 ` Dorit Nuzman [this message]
2007-08-07 13:23 ` Diego Novillo
2007-08-08 8:02 ` Ira Rosen
2007-08-08 12:06 ` Diego Novillo
2007-08-08 9:41 ` Zdenek Dvorak
2007-08-08 11:04 ` Ira Rosen
2007-08-12 19:40 ` Zdenek Dvorak
2007-08-07 13:22 ` Diego Novillo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=OFEC28E137.8C9E6009-ONC2257330.004675DC-C2257330.00481FC7@il.ibm.com \
--to=dorit@il.ibm.com \
--cc=IRAR@il.ibm.com \
--cc=dberlin@dberlin.org \
--cc=dnovillo@google.com \
--cc=dvorakz@suse.cz \
--cc=gcc-patches@gnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).