From: Rich Felker <dalias@libc.org>
To: "H.J. Lu" <hjl.tools@gmail.com>
Cc: Richard Henderson <rth@redhat.com>, Michael Matz <matz@suse.de>,
Jan Hubicka <hubicka@ucw.cz>,
Alexander Monakov <amonakov@ispras.ru>,
GCC Patches <gcc-patches@gcc.gnu.org>,
Uros Bizjak <ubizjak@gmail.com>
Subject: Re: [PATCH i386] Allow sibcalls in no-PLT PIC
Date: Wed, 20 May 2015 01:09:00 -0000 [thread overview]
Message-ID: <20150520010602.GR17573@brightrain.aerifal.cx> (raw)
In-Reply-To: <CAMe9rOrkmbuVkWQfD+L36etWjN4JPMakPib_No2WpUeCsdd+aA@mail.gmail.com>
On Tue, May 19, 2015 at 05:10:11PM -0700, H.J. Lu wrote:
> On Tue, May 19, 2015 at 1:54 PM, Rich Felker <dalias@libc.org> wrote:
> > On Tue, May 19, 2015 at 01:27:06PM -0700, H.J. Lu wrote:
> >> On Tue, May 19, 2015 at 1:15 PM, Rich Felker <dalias@libc.org> wrote:
> >> > On Tue, May 19, 2015 at 12:17:18PM -0700, H.J. Lu wrote:
> >> >> On Tue, May 19, 2015 at 12:11 PM, Richard Henderson <rth@redhat.com> wrote:
> >> >> > On 05/19/2015 12:06 PM, H.J. Lu wrote:
> >> >> >> On Tue, May 19, 2015 at 11:59 AM, Richard Henderson <rth@redhat.com> wrote:
> >> >> >>> On 05/19/2015 11:06 AM, Rich Felker wrote:
> >> >> >>>> I'm still mildly worried that concerns for supporting
> >> >> >>>> relaxation might lead to decisions not to optimize code in ways that
> >> >> >>>> would be difficult to relax (e.g. certain types of address load
> >> >> >>>> reordering or hoisting) but I don't understand GCC internals
> >> >> >>>> sufficiently to know if this concern is warranted or not.
> >> >> >>>
> >> >> >>> It is. The relaxation that HJ is working on requires that the reads from the
> >> >> >>> got not be hoisted. I'm not especially convinced that what he's working on is
> >> >> >>> a win.
> >> >> >>>
> >> >> >>> With LTO, the compiler can do the same job that he's attempting in the linker,
> >> >> >>> without an extra nop. Without LTO, leaving it to the linker means that you
> >> >> >>> can't hoist the load and hide the memory latency.
> >> >> >>>
> >> >> >>
> >> >> >> My relax approach won't take away any optimization done by compiler.
> >> >> >> It simply turns indirect branch into direct branch with a nop prefix at
> >> >> >> link-time. I am having a hard time to understand why we shouldn't do it.
> >> >> >
> >> >> > I well understand what you're doing.
> >> >> >
> >> >> > But my point is that the only time the compiler should present you with the
> >> >> > form of indirect branch you're looking for is when there's no place to hoist
> >> >> > the load.
> >> >> >
> >> >> > At which point, is it really worth adding a new relocation to the ABI? Is it
> >> >> > really worth adding new code to the linker that won't be exercised often?
> >> >>
> >> >> I believe there are plenty of indirect branches via GOT when compiling
> >> >> PIE/PIC with -fno-plt:
> >> >>
> >> >> [hjl@gnu-6 gcc]$ cat /tmp/x.c
> >> >> extern void foo (void);
> >> >>
> >> >> void
> >> >> bar (void)
> >> >> {
> >> >> foo ();
> >> >> }
> >> >> [hjl@gnu-6 gcc]$ ./xgcc -B./ -fPIC -O3 -S /tmp/x.c -fno-plt
> >> >> [hjl@gnu-6 gcc]$ cat x.s
> >> >> ..file "x.c"
> >> >> ..section .text.unlikely,"ax",@progbits
> >> >> ..LCOLDB0:
> >> >> ..text
> >> >> ..LHOTB0:
> >> >> ..p2align 4,,15
> >> >> ..globl bar
> >> >> ..type bar, @function
> >> >> bar:
> >> >> ..LFB0:
> >> >> ..cfi_startproc
> >> >> jmp *foo@GOTPCREL(%rip)
> >> >> ..cfi_endproc
> >> >> ..LFE0:
> >> >> ..size bar, .-bar
> >> >
> >> > I agree these exist. What I question is whether the savings from the
> >> > linker being able to relax this to a direct call in the case where the
> >> > programmer failed to let the compiler make it a direct call to begin
> >> > with (by using hidden or protected visibility) are worth the cost of
> >> > not being able to hoist the load out of loops or schedule it earlier
> >> > in cases where relaxation is not possible because the call target is
> >> > not defined in the same DSO.
> >>
> >> Just for fun. I compiled binutils as PIE with -fno-plt -flto:
> >>
> >> [hjl@gnu-mic-2 gas]$ file as-new
> >> as-new: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV),
> >> dynamically linked (uses shared libs), for GNU/Linux 2.6.32, not
> >> stripped
> >> [hjl@gnu-mic-2 gas]$
> >>
> >> There are 43:
> >>
> >> ff 25 21 93 2d 00 jmpq *0x2d9321(%rip) # 3d5f58 <_DYNAMIC+0x1e8>
> >>
> >> and 1983
> >>
> >> ff 15 eb f4 38 00 callq *0x38f4eb(%rip) # 3d60e0 <_DYNAMIC+0x370>
> >
> > How many of those would be relaxed? I suspect it depends a lot on
> > whether libbfd is static or shared.
>
> When shared libraries are enabled, there are 177 indirect branches
> to locally defined functions. Call to any locally defined functions,
> which aren't compiled with LTO, is indirect.
And are the above indirect calls/jumps (1983+43) candidates for
scheduling/hoisting the address load (that's not being done yet), or
are they the ones the compiler opted not to schedule/hoist? The win
from relaxation seems small here, but as long as you're not going to
block optimizations that would preclude relaxing, I don't see any
disadvantages to doing it.
Rich
next prev parent reply other threads:[~2015-05-20 1:06 UTC|newest]
Thread overview: 106+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-05-04 16:38 PIC calls without PLT, generic implementation Alexander Monakov
2015-05-04 16:38 ` [PATCH] Expand PIC calls without PLT with -fno-plt Alexander Monakov
2015-05-04 17:34 ` Jeff Law
2015-05-04 17:40 ` Jakub Jelinek
2015-05-04 17:42 ` Jeff Law
2015-05-06 3:08 ` Rich Felker
2015-05-10 17:07 ` Jan Hubicka
2015-05-06 15:25 ` Alexander Monakov
2015-05-06 15:46 ` Jakub Jelinek
2015-05-06 15:55 ` Jeff Law
2015-05-06 16:44 ` Alexander Monakov
2015-05-06 17:35 ` Rich Felker
2015-05-06 18:26 ` H.J. Lu
2015-05-06 18:37 ` Rich Felker
2015-05-06 18:45 ` H.J. Lu
2015-05-06 19:01 ` Rich Felker
2015-05-06 19:05 ` H.J. Lu
2015-05-06 19:18 ` Rich Felker
2015-05-06 19:24 ` H.J. Lu
2015-05-11 11:48 ` Michael Matz
2015-05-11 14:20 ` Rich Felker
2015-05-07 18:22 ` Jeff Law
2015-05-07 19:13 ` H.J. Lu
2015-05-10 16:59 ` Jan Hubicka
2015-05-11 20:36 ` Jeff Law
2015-05-11 20:55 ` H.J. Lu
2015-05-11 22:13 ` Jan Hubicka
2015-06-22 15:52 ` Jiong Wang
2015-06-22 18:18 ` Alexander Monakov
2015-06-23 8:41 ` Ramana Radhakrishnan
2015-06-23 10:43 ` Alexander Monakov
2015-06-23 13:28 ` Jeff Law
2015-07-16 10:37 ` [AArch64] Tighten direct call pattern to repair -fno-plt Jiong Wang
2015-07-16 10:47 ` Alexander Monakov
2015-07-16 10:48 ` Jiong Wang
2015-07-21 12:52 ` [AArch64][sibcall]Tighten " Jiong Wang
2015-08-04 9:50 ` James Greenhalgh
2015-08-06 16:18 ` [COMMITTED][AArch64][sibcall]Tighten " Jiong Wang
2015-08-07 8:22 ` James Greenhalgh
2015-08-07 13:28 ` Jiong Wang
2015-08-04 9:50 ` [AArch64] Tighten " James Greenhalgh
2015-08-06 16:16 ` [COMMITTED][AArch64] " Jiong Wang
2015-05-04 16:38 ` [PATCH i386] Allow sibcalls in no-PLT PIC Alexander Monakov
2015-05-15 16:37 ` Alexander Monakov
2015-05-15 16:48 ` H.J. Lu
2015-05-15 20:08 ` Jan Hubicka
2015-05-15 20:23 ` H.J. Lu
2015-05-15 20:35 ` Rich Felker
2015-05-15 20:37 ` H.J. Lu
2015-05-15 20:45 ` Rich Felker
2015-05-15 22:16 ` H.J. Lu
2015-05-15 23:14 ` Jan Hubicka
2015-05-15 23:30 ` H.J. Lu
2015-05-15 23:35 ` H.J. Lu
2015-05-15 23:44 ` H.J. Lu
2015-05-16 0:18 ` Rich Felker
2015-05-16 14:33 ` H.J. Lu
2015-05-16 19:03 ` H.J. Lu
2015-05-16 19:32 ` Rich Felker
2015-05-16 23:23 ` H.J. Lu
2015-05-15 23:49 ` Rich Felker
2015-05-19 14:48 ` Michael Matz
2015-05-19 15:11 ` Jeff Law
2015-05-19 16:03 ` Michael Matz
2015-05-19 19:11 ` Rich Felker
2015-05-19 18:08 ` Rich Felker
2015-05-19 19:03 ` Richard Henderson
2015-05-19 19:10 ` H.J. Lu
2015-05-19 19:17 ` Richard Henderson
2015-05-19 19:20 ` H.J. Lu
2015-05-19 19:54 ` Richard Henderson
2015-05-19 20:27 ` Rich Felker
2015-05-19 20:44 ` H.J. Lu
2015-05-19 21:28 ` Rich Felker
2015-05-20 0:52 ` H.J. Lu
2015-05-20 1:09 ` Rich Felker [this message]
2015-05-22 19:32 ` Richard Henderson
2015-05-19 19:48 ` Rich Felker
2015-05-19 20:16 ` Richard Henderson
2015-05-20 12:13 ` Michael Matz
2015-05-20 12:40 ` H.J. Lu
2015-05-20 14:17 ` Rich Felker
2015-05-20 14:33 ` Michael Matz
2015-05-18 18:25 ` Alexander Monakov
2015-05-18 19:03 ` Jan Hubicka
2015-05-04 16:38 ` [PATCH i386] Move CLOBBERED_REGS earlier in register class list Alexander Monakov
2015-05-10 16:44 ` Jan Hubicka
2015-05-10 17:51 ` Uros Bizjak
2015-05-10 18:09 ` Uros Bizjak
2015-05-11 16:26 ` Alexander Monakov
2015-05-11 16:30 ` Uros Bizjak
2015-05-04 16:38 ` [PATCH i386] Extend sibcall peepholes to allow source in %eax Alexander Monakov
2015-05-10 16:54 ` Jan Hubicka
2015-05-11 17:50 ` Alexander Monakov
2015-05-11 18:00 ` Jan Hubicka
2015-05-11 19:46 ` Uros Bizjak
2015-05-11 19:48 ` Jeff Law
2015-05-11 20:16 ` Jan Hubicka
2015-05-13 19:05 ` Alexander Monakov
2015-05-13 20:04 ` Jan Hubicka
2015-05-14 17:36 ` Alexander Monakov
2015-05-04 16:38 ` [RFC PATCH] ira: accept loads via argp rtx in validate_equiv_mem Alexander Monakov
2015-05-04 17:37 ` Jeff Law
2015-05-04 16:38 ` [PATCH i386] PR65753: allow PIC tail calls via function pointers Alexander Monakov
2015-05-10 16:37 ` Jan Hubicka
2015-05-11 16:11 ` Alexander Monakov
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=20150520010602.GR17573@brightrain.aerifal.cx \
--to=dalias@libc.org \
--cc=amonakov@ispras.ru \
--cc=gcc-patches@gcc.gnu.org \
--cc=hjl.tools@gmail.com \
--cc=hubicka@ucw.cz \
--cc=matz@suse.de \
--cc=rth@redhat.com \
--cc=ubizjak@gmail.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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).