From: Segher Boessenkool <segher@kernel.crashing.org>
To: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Paul Mackerras <paulus@samba.org>,
linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org,
Nicholas Piggin <npiggin@gmail.com>
Subject: Re: [PATCH v1 1/2] powerpc/bitops: Use immediate operand when possible
Date: Wed, 14 Apr 2021 10:19:21 -0500 [thread overview]
Message-ID: <20210414151921.GW26583@gate.crashing.org> (raw)
In-Reply-To: <daacce9f-1900-1034-980b-be5a58d6be09@csgroup.eu>
On Wed, Apr 14, 2021 at 02:42:51PM +0200, Christophe Leroy wrote:
> Le 14/04/2021 à 14:24, Segher Boessenkool a écrit :
> >On Wed, Apr 14, 2021 at 12:01:21PM +1000, Nicholas Piggin wrote:
> >>Would be nice if we could let the compiler deal with it all...
> >>
> >>static inline unsigned long lr(unsigned long *mem)
> >>{
> >> unsigned long val;
> >>
> >> /*
> >> * This doesn't clobber memory but want to avoid memory
> >> operations
> >> * moving ahead of it
> >> */
> >> asm volatile("ldarx %0, %y1" : "=r"(val) : "Z"(*mem) :
> >> "memory");
> >>
> >> return val;
> >>}
> >
> >(etc.)
> >
> >That can not work reliably: the compiler can put random instructions
> >between the larx and stcx. this way, and you then do not have guaranteed
> >forward progress anymore. It can put the two in different routines
> >(after inlining and other interprocedural optimisations), duplicate
> >them, make a different number of copies of them, etc.
> >
> >Nothing of that is okay if you want to guarantee forward progress on all
> >implementations, and also not if you want to have good performance
> >everywhere (or anywhere even). Unfortunately you have to write all
> >larx/stcx. loops as one block of assembler, so that you know exactly
> >what instructions will end up in your binary.
> >
> >If you don't, it will fail mysteriously after random recompilations, or
> >have performance degradations, etc. You don't want to go there :-)
> >
>
> Could the kernel use GCC builtin atomic functions instead ?
>
> https://gcc.gnu.org/onlinedocs/gcc/_005f_005fatomic-Builtins.html
Certainly that should work fine for the simpler cases that the atomic
operations are meant to provide. But esp. for not-so-simple cases the
kernel may require some behaviour provided by the existing assembler
implementation, and not by the atomic builtins.
I'm not saying this cannot work, just that some serious testing will be
needed. If it works it should be the best of all worlds, so then it is
a really good idea yes :-)
Segher
WARNING: multiple messages have this Message-ID (diff)
From: Segher Boessenkool <segher@kernel.crashing.org>
To: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Nicholas Piggin <npiggin@gmail.com>,
linux-kernel@vger.kernel.org, linuxppc-dev@lists.ozlabs.org,
Paul Mackerras <paulus@samba.org>
Subject: Re: [PATCH v1 1/2] powerpc/bitops: Use immediate operand when possible
Date: Wed, 14 Apr 2021 10:19:21 -0500 [thread overview]
Message-ID: <20210414151921.GW26583@gate.crashing.org> (raw)
In-Reply-To: <daacce9f-1900-1034-980b-be5a58d6be09@csgroup.eu>
On Wed, Apr 14, 2021 at 02:42:51PM +0200, Christophe Leroy wrote:
> Le 14/04/2021 à 14:24, Segher Boessenkool a écrit :
> >On Wed, Apr 14, 2021 at 12:01:21PM +1000, Nicholas Piggin wrote:
> >>Would be nice if we could let the compiler deal with it all...
> >>
> >>static inline unsigned long lr(unsigned long *mem)
> >>{
> >> unsigned long val;
> >>
> >> /*
> >> * This doesn't clobber memory but want to avoid memory
> >> operations
> >> * moving ahead of it
> >> */
> >> asm volatile("ldarx %0, %y1" : "=r"(val) : "Z"(*mem) :
> >> "memory");
> >>
> >> return val;
> >>}
> >
> >(etc.)
> >
> >That can not work reliably: the compiler can put random instructions
> >between the larx and stcx. this way, and you then do not have guaranteed
> >forward progress anymore. It can put the two in different routines
> >(after inlining and other interprocedural optimisations), duplicate
> >them, make a different number of copies of them, etc.
> >
> >Nothing of that is okay if you want to guarantee forward progress on all
> >implementations, and also not if you want to have good performance
> >everywhere (or anywhere even). Unfortunately you have to write all
> >larx/stcx. loops as one block of assembler, so that you know exactly
> >what instructions will end up in your binary.
> >
> >If you don't, it will fail mysteriously after random recompilations, or
> >have performance degradations, etc. You don't want to go there :-)
> >
>
> Could the kernel use GCC builtin atomic functions instead ?
>
> https://gcc.gnu.org/onlinedocs/gcc/_005f_005fatomic-Builtins.html
Certainly that should work fine for the simpler cases that the atomic
operations are meant to provide. But esp. for not-so-simple cases the
kernel may require some behaviour provided by the existing assembler
implementation, and not by the atomic builtins.
I'm not saying this cannot work, just that some serious testing will be
needed. If it works it should be the best of all worlds, so then it is
a really good idea yes :-)
Segher
next prev parent reply other threads:[~2021-04-14 15:21 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-04-08 15:33 [PATCH v1 1/2] powerpc/bitops: Use immediate operand when possible Christophe Leroy
2021-04-08 15:33 ` Christophe Leroy
2021-04-08 15:33 ` [PATCH v1 2/2] powerpc/atomics: " Christophe Leroy
2021-04-08 15:33 ` Christophe Leroy
2021-04-12 22:08 ` Segher Boessenkool
2021-04-12 22:08 ` Segher Boessenkool
2021-04-13 16:36 ` Christophe Leroy
2021-04-13 16:36 ` Christophe Leroy
2021-04-12 21:54 ` [PATCH v1 1/2] powerpc/bitops: " Segher Boessenkool
2021-04-12 21:54 ` Segher Boessenkool
2021-04-13 16:33 ` Christophe Leroy
2021-04-13 16:33 ` Christophe Leroy
2021-04-13 21:58 ` Segher Boessenkool
2021-04-13 21:58 ` Segher Boessenkool
2021-04-14 2:01 ` Nicholas Piggin
2021-04-14 2:01 ` Nicholas Piggin
2021-04-14 12:24 ` Segher Boessenkool
2021-04-14 12:24 ` Segher Boessenkool
2021-04-14 12:42 ` Christophe Leroy
2021-04-14 12:42 ` Christophe Leroy
2021-04-14 15:19 ` Segher Boessenkool [this message]
2021-04-14 15:19 ` Segher Boessenkool
2021-04-14 15:32 ` David Laight
2021-04-14 15:32 ` David Laight
2021-04-14 17:20 ` Segher Boessenkool
2021-04-14 17:20 ` Segher Boessenkool
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=20210414151921.GW26583@gate.crashing.org \
--to=segher@kernel.crashing.org \
--cc=christophe.leroy@csgroup.eu \
--cc=linux-kernel@vger.kernel.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=npiggin@gmail.com \
--cc=paulus@samba.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.