From: Mark Kettenis <mark.kettenis@xs4all.nl>
To: jan.kratochvil@redhat.com
Cc: gdb-patches@sourceware.org
Subject: Re: [patch 2/3] Align stack for SSE (PR tdep/14222)
Date: Wed, 13 Jun 2012 21:20:00 -0000 [thread overview]
Message-ID: <201206132119.q5DLJSI4003388@glazunov.sibelius.xs4all.nl> (raw)
In-Reply-To: <20120613155020.GB26214@host2.jankratochvil.net> (message from Jan Kratochvil on Wed, 13 Jun 2012 17:50:20 +0200)
> Date: Wed, 13 Jun 2012 17:50:20 +0200
> From: Jan Kratochvil <jan.kratochvil@redhat.com>
>
> Hi,
>
> this is mostly independent patch. But it does not make sense to discuss
> alignment in [patch 3/3] when it has been already broken.
>
> http://groups.google.com/group/ia32-abi/browse_thread/thread/4f9b3e5069943bf1
> says that gcc-4.6 aligns stack to 32 bytes. I do not see it anywhere:
> 80482e8: 83 e4 f0 and $0xfffffff0,%esp
> It is aligned just to 16 bytes. Also checked that Intel CPU documentation
> does not require any more alignment than 16 bytes (sure it may be less
> effective but traps never occur). Therefore I consider amd64-tdep.c to be
> correct and I have not changed its existing 16-bytes alignment.
>
> gdb/
> 2012-06-13 Jan Kratochvil <jan.kratochvil@redhat.com>
>
> PR tdep/14222
> * i386-tdep.c (i386_push_dummy_call): Align to 16 bytes unconditionally.
Committed a better diff (see below) that removes the now unused
have_16_byte_aligned_arg variable.
> gdb/testsuite/
> 2012-06-13 Jan Kratochvil <jan.kratochvil@redhat.com>
>
> PR tdep/14222
> * gdb.arch/i386-sse-stack-align.S: New file.
> * gdb.arch/i386-sse-stack-align.c: New file.
> * gdb.arch/i386-sse-stack-align.exp: New file.
Can you commit this testsuite bit?
2012-06-13 Mark Kettenis <kettenis@gnu.org>
Jan Kratochvil <jan.kratochvil@redhat.com>
PR tdep/14222
* i386-tdep.c (i386_push_dummy_call): Unconditionally align the
stack on a 16-byte boundary.
Index: i386-tdep.c
===================================================================
RCS file: /cvs/src/src/gdb/i386-tdep.c,v
retrieving revision 1.352
diff -u -p -r1.352 i386-tdep.c
--- i386-tdep.c 13 Jun 2012 20:29:15 -0000 1.352
+++ i386-tdep.c 13 Jun 2012 21:15:44 -0000
@@ -2351,7 +2351,6 @@ i386_push_dummy_call (struct gdbarch *gd
for (write_pass = 0; write_pass < 2; write_pass++)
{
int args_space_used = 0;
- int have_16_byte_aligned_arg = 0;
if (struct_return)
{
@@ -2389,19 +2388,20 @@ i386_push_dummy_call (struct gdbarch *gd
else
{
if (i386_16_byte_align_p (value_enclosing_type (args[i])))
- {
- args_space = align_up (args_space, 16);
- have_16_byte_aligned_arg = 1;
- }
+ args_space = align_up (args_space, 16);
args_space += align_up (len, 4);
}
}
if (!write_pass)
{
- if (have_16_byte_aligned_arg)
- args_space = align_up (args_space, 16);
sp -= args_space;
+
+ /* The original System V ABI only requires word alignment,
+ but modern incarnations need 16-byte alignment in order
+ to support SSE. Since wasting a few bytes here isn't
+ harmful we unconditionally enforce 16-byte alignment. */
+ sp &= ~0xf;
}
}
next prev parent reply other threads:[~2012-06-13 21:20 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-06-13 15:50 Jan Kratochvil
2012-06-13 21:20 ` Mark Kettenis [this message]
2012-06-13 21:31 ` Jan Kratochvil
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=201206132119.q5DLJSI4003388@glazunov.sibelius.xs4all.nl \
--to=mark.kettenis@xs4all.nl \
--cc=gdb-patches@sourceware.org \
--cc=jan.kratochvil@redhat.com \
/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