From: Christophe Leroy <christophe.leroy@c-s.fr>
To: Thomas Gleixner <tglx@linutronix.de>, Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>,
the arch/x86 maintainers <x86@kernel.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"open list:BROADCOM NVRAM DRIVER" <linux-mips@vger.kernel.org>,
Paul Mackerras <paulus@samba.org>,
Andy Lutomirski <luto@kernel.org>,
Michael Ellerman <mpe@ellerman.id.au>,
Vincenzo Frascino <vincenzo.frascino@arm.com>,
linuxppc-dev <linuxppc-dev@lists.ozlabs.org>,
Linux ARM <linux-arm-kernel@lists.infradead.org>
Subject: Re: [RFC PATCH v2 05/10] lib: vdso: inline do_hres()
Date: Sat, 11 Jan 2020 09:06:49 +0000 [thread overview]
Message-ID: <8cdb8a09-b1b6-c0a4-8b30-da095a9a660c@c-s.fr> (raw)
In-Reply-To: <87o8vbrpej.fsf@nanos.tec.linutronix.de>
On 01/10/2020 09:07 PM, Thomas Gleixner wrote:
> Arnd Bergmann <arnd@arndb.de> writes:
>> On Mon, Dec 23, 2019 at 3:31 PM Christophe Leroy
>> <christophe.leroy@c-s.fr> wrote:
>>>
>>> do_hres() is called from several places, so GCC doesn't inline
>>> it at first.
>>>
>>> do_hres() takes a struct __kernel_timespec * parameter for
>>> passing the result. In the 32 bits case, this parameter corresponds
>>> to a local var in the caller. In order to provide a pointer
>>> to this structure, the caller has to put it in its stack and
>>> do_hres() has to write the result in the stack. This is suboptimal,
>>> especially on RISC processor like powerpc.
>>>
>>> By making GCC inline the function, the struct __kernel_timespec
>>> remains a local var using registers, avoiding the need to write and
>>> read stack.
>>>
>>> The improvement is significant on powerpc.
>>>
>>> Signed-off-by: Christophe Leroy <christophe.leroy@c-s.fr>
>>
>> Good idea, I can see how this ends up being an improvement
>> for most of the callers.
>>
>> Acked-by: Arnd Bergmann <arnd@arndb.de>
>
> https://lore.kernel.org/r/20191112012724.250792-3-dima@arista.com
>
> On the way to be applied.
>
Oh nice, I get even better result with the way it is done by Dmitry
compared to my own first patch.
On an mpc8xx at 132Mhz (32bits powerpc), before the patch I have
gettimeofday: vdso: 1256 nsec/call
clock-gettime-monotonic-raw: vdso: 1449 nsec/call
clock-gettime-monotonic-coarse: vdso: 768 nsec/call
clock-gettime-monotonic: vdso: 1390 nsec/call
With the patch I have:
gettimeofday: vdso: 947 nsec/call
clock-gettime-monotonic-raw: vdso: 1156 nsec/call
clock-gettime-monotonic-coarse: vdso: 638 nsec/call
clock-gettime-monotonic: vdso: 1094 nsec/call
So that's a 20-25% improvement.
I modified it slightly as follows:
diff --git a/lib/vdso/gettimeofday.c b/lib/vdso/gettimeofday.c
index 9e474d54814f..b793f211bca8 100644
--- a/lib/vdso/gettimeofday.c
+++ b/lib/vdso/gettimeofday.c
@@ -37,8 +37,8 @@ u64 vdso_calc_delta(u64 cycles, u64 last, u64 mask,
u32 mult)
}
#endif
-static int do_hres(const struct vdso_data *vd, clockid_t clk,
- struct __kernel_timespec *ts)
+static __always_inline int do_hres(const struct vdso_data *vd,
clockid_t clk,
+ struct __kernel_timespec *ts)
{
const struct vdso_timestamp *vdso_ts = &vd->basetime[clk];
u64 cycles, last, sec, ns;
@@ -67,8 +67,8 @@ static int do_hres(const struct vdso_data *vd,
clockid_t clk,
return 0;
}
-static void do_coarse(const struct vdso_data *vd, clockid_t clk,
- struct __kernel_timespec *ts)
+static __always_inline int do_coarse(const struct vdso_data *vd,
clockid_t clk,
+ struct __kernel_timespec *ts)
{
const struct vdso_timestamp *vdso_ts = &vd->basetime[clk];
u32 seq;
@@ -78,6 +78,8 @@ static void do_coarse(const struct vdso_data *vd,
clockid_t clk,
ts->tv_sec = vdso_ts->sec;
ts->tv_nsec = vdso_ts->nsec;
} while (unlikely(vdso_read_retry(vd, seq)));
+
+ return 0;
}
static __maybe_unused int
@@ -95,15 +97,16 @@ __cvdso_clock_gettime_common(const struct vdso_data
*vd, clockid_t clock,
* clocks are handled in the VDSO directly.
*/
msk = 1U << clock;
- if (likely(msk & VDSO_HRES)) {
- return do_hres(&vd[CS_HRES_COARSE], clock, ts);
- } else if (msk & VDSO_COARSE) {
- do_coarse(&vd[CS_HRES_COARSE], clock, ts);
- return 0;
- } else if (msk & VDSO_RAW) {
- return do_hres(&vd[CS_RAW], clock, ts);
- }
- return -1;
+ if (likely(msk & VDSO_HRES))
+ vd += CS_HRES_COARSE;
+ else if (msk & VDSO_COARSE)
+ return do_coarse(&vd[CS_HRES_COARSE], clock, ts);
+ else if (msk & VDSO_RAW)
+ vd += CS_RAW;
+ else
+ return -1;
+
+ return do_hres(vd, clock, ts);
}
static __maybe_unused int
---
Christophe
_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
next prev parent reply other threads:[~2020-01-11 9:07 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-12-23 14:31 [RFC PATCH v2 00/10] powerpc/32: switch VDSO to C implementation Christophe Leroy
2019-12-23 14:31 ` [RFC PATCH v2 01/10] lib: vdso: ensure all arches have 32bit fallback Christophe Leroy
2019-12-24 2:07 ` Andy Lutomirski
2020-01-10 20:56 ` Thomas Gleixner
2020-01-10 21:02 ` Andy Lutomirski
2019-12-30 12:27 ` Arnd Bergmann
2020-01-02 11:29 ` Arnd Bergmann
2020-01-09 15:43 ` Christophe Leroy
2019-12-23 14:31 ` [RFC PATCH v2 02/10] lib: vdso: move call to fallback out of common code Christophe Leroy
2019-12-24 2:24 ` Andy Lutomirski
2019-12-24 11:41 ` christophe leroy
2019-12-24 12:09 ` Andy Lutomirski
2019-12-23 14:31 ` [RFC PATCH v2 03/10] lib: vdso: Change __cvdso_clock_gettime/getres_common() to __cvdso_clock_gettime/getres() Christophe Leroy
2019-12-23 14:31 ` [RFC PATCH v2 04/10] lib: vdso: get pointer to vdso data from the arch Christophe Leroy
2019-12-24 2:27 ` Andy Lutomirski
2019-12-24 11:53 ` christophe leroy
2019-12-24 12:15 ` Andy Lutomirski
2019-12-24 12:41 ` Andy Lutomirski
2019-12-24 14:46 ` Segher Boessenkool
2019-12-23 14:31 ` [RFC PATCH v2 05/10] lib: vdso: inline do_hres() Christophe Leroy
2019-12-24 2:29 ` Andy Lutomirski
2019-12-30 12:07 ` Arnd Bergmann
2020-01-10 21:07 ` Thomas Gleixner
2020-01-11 9:06 ` Christophe Leroy [this message]
2019-12-23 14:31 ` [RFC PATCH v2 06/10] lib: vdso: make do_coarse() return 0 Christophe Leroy
2019-12-23 14:31 ` [RFC PATCH v2 07/10] lib: vdso: don't use READ_ONCE() in __c_kernel_time() Christophe Leroy
2019-12-24 1:58 ` Andy Lutomirski
2019-12-24 11:12 ` christophe leroy
2019-12-24 12:04 ` Andy Lutomirski
2020-01-10 21:12 ` Thomas Gleixner
2020-01-11 8:05 ` Christophe Leroy
2020-01-11 11:07 ` Thomas Gleixner
2020-01-13 6:52 ` Christophe Leroy
2019-12-23 14:31 ` [RFC PATCH v2 08/10] lib: vdso: Avoid duplication in __cvdso_clock_getres() Christophe Leroy
2019-12-24 1:59 ` Andy Lutomirski
2019-12-23 14:31 ` [RFC PATCH v2 09/10] powerpc/vdso32: inline __get_datapage() Christophe Leroy
2019-12-23 14:31 ` [RFC PATCH v2 10/10] powerpc/32: Switch VDSO to C implementation Christophe Leroy
[not found] ` <bd4557a7-9715-59aa-5d8e-488c5e516a98@c-s.fr>
2020-01-09 20:07 ` Surprising code generated for vdso_read_begin() Segher Boessenkool
2020-01-10 6:45 ` Christophe Leroy
2020-01-11 11:33 ` Segher Boessenkool
2020-02-16 18:10 ` Arnd Bergmann
2020-02-19 8:45 ` Christophe Leroy
2020-02-19 9:52 ` Arnd Bergmann
2020-02-19 13:08 ` 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=8cdb8a09-b1b6-c0a4-8b30-da095a9a660c@c-s.fr \
--to=christophe.leroy@c-s.fr \
--cc=arnd@arndb.de \
--cc=benh@kernel.crashing.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mips@vger.kernel.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=luto@kernel.org \
--cc=mpe@ellerman.id.au \
--cc=paulus@samba.org \
--cc=tglx@linutronix.de \
--cc=vincenzo.frascino@arm.com \
--cc=x86@kernel.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 NNTP newsgroup(s).