The topic of benchmarking keeps coming up. We're working on making
the next FSF release better, but it's a good idea to track how the
current Linaro GCC stacks up against other releases. The summary is
at:
https://wiki.linaro.org/Internal/ToolChain/Now
Included is how our current 4.6 release does against FSF 4.6, the
change over six months, and how the upcoming 4.7 release fairs.
There's also a comparison against other compilers including the Google
4.6 and Android 4.4 branches.
A PDF version is attached to the page. The SPEC 2000 results are
still coming in so I'll update the page once they arrive. Everything
is generated so we'll update this with each monthly release.
-- Michael
This patch series moves vaious functionality duplicated in platform
cpuidle drivers to the core cpuidle driver. Also, the platform irq
disabling was removed as it appears that all calls into
cpuidle_call_idle will have already called local_irq_disable().
Based on 3.3-rc5 plus recent exynos cpuidle patch:
http://www.spinics.net/lists/linux-samsung-soc/msg09467.html
v4 submission can be found here:
http://lists.infradead.org/pipermail/linux-arm-kernel/2012-January/082742.h…
Changes since v4:
* Added common cpu_do_idle function to core cpuidle
* Added time keep irq en wrapper to core cpuidle
* Removed pre/post enter
* Re-added platforms that can use new common code.
v3 submission can be found here:
http://www.spinics.net/lists/arm-kernel/msg156751.html
Changes since v3:
* Removed drivers/cpuidle/common.c
** Removed the initialization helper functions
** Removed the wrapper used to consolidate time keeping and irq enable/disable
* Add time keeping and local_irq_disable handling in cpuidle_call_idle().
* Made necessary modifications to a few platforms that required the most changes
** Note on omap3: changed structure of omap3_idle_drvdata and added
per_next_state and per_saved_state vars to accomodate new framework.
v2 submission can be found here:
http://comments.gmane.org/gmane.linux.ports.arm.kernel/144199
Changes since v2:
* Made various code organization and style changes as suggested in v1 review.
* Removed at91 use of common code. A separate effort is underway to clean
at91 code and the author has offered to convert to common interface as part
of those changes (if this common interface is accepted in time).
* Made platform cpuidle_driver objects __initdata and dynamically added one
persistent instance of this object in common code.
* Removed imx5 pm usage of gpc_dvfs clock as it is no longer needed after
being enabled during clock initialization.
* Re-organized patches.
v1 submission can be found here:
http://comments.gmane.org/gmane.linux.ports.arm.kernel/142791
Changes since v1:
* Common interface moved to drivers/cpuidle and made non arch-specific.
* Made various fixes and suggested additions to the common cpuidle
code from v1 review.
* Added callback for filling in driver_data field as needed.
* Modified the various platforms with these changes.
Robert Lee (9):
cpuidle: Add commonly used functionality for consolidation
SH: shmobile: cpuidle consolidation
ARM: omap: Consolidate OMAP3 cpuidle time keeping and irq enable
ARM: omap: Consolidate OMAP4 cpuidle time keeping and irq enable
ARM: shmobile: Consolidate cpuidle functionality
ARM: davinci: Consolidate cpuidle functionality
ARM: exynos: Consolidate cpuidle functionality
ARM: kirkwood: Consolidate cpuidle functionality
ARM: at91: Consolidate cpuidle functionality
arch/arm/mach-at91/cpuidle.c | 64 +++++++++------------------
arch/arm/mach-davinci/cpuidle.c | 77 +++++++++++++--------------------
arch/arm/mach-exynos/cpuidle.c | 52 ++--------------------
arch/arm/mach-kirkwood/cpuidle.c | 71 +++++++++----------------------
arch/arm/mach-omap2/cpuidle34xx.c | 43 ++++++++-----------
arch/arm/mach-omap2/cpuidle44xx.c | 21 +--------
arch/arm/mach-shmobile/cpuidle.c | 22 +---------
arch/sh/kernel/cpu/shmobile/cpuidle.c | 10 +---
drivers/cpuidle/cpuidle.c | 37 +++++++++------
include/linux/cpuidle.h | 55 +++++++++++++++++++++++
10 files changed, 180 insertions(+), 272 deletions(-)
Re-sending on top of Tero's changes which are already
in -next and the one which adds support for twl6030
vdd1/2/3 SMPS [1].
v3 has mainly 2 differences from v2
-1- TWL driver now uses just one table for both dt and
non-dt based lookup for driver data.
-2- All common regulator nodes for twl4030 and twl6030 are
now defined in the twl4030.dtsi and twl6030.dtsi instead of
repeating the nodes in all board files, which also means
the patch ('arm/dts: twl: Pass regulator data from dt')
has a dependency with the series from Benoit which adds the
twl4030.dtsi and twl6030.dtsi files[2].
I have tested the patches on omap4panda and omap3beagle boards.
[1] http://marc.info/?l=linux-omap&m=133009693905864&w=2
[2] git://git.kernel.org/pub/scm/linux/kernel/git/bcousson/linux-omap-dt.git
for_3.4/dt_i2c_twl
Rajendra Nayak (2):
regulator: twl: adapt twl-regulator driver to dt
arm/dts: twl: Pass regulator data from dt
.../bindings/regulator/twl-regulator.txt | 66 +++++
arch/arm/boot/dts/omap3-beagle.dts | 6 +
arch/arm/boot/dts/twl4030.dtsi | 18 ++
arch/arm/boot/dts/twl6030.dtsi | 60 +++++
drivers/regulator/twl-regulator.c | 259 +++++++++++++-------
5 files changed, 322 insertions(+), 87 deletions(-)
create mode 100644 Documentation/devicetree/bindings/regulator/twl-regulator.txt
Hello Booher,
I was previously using Angstrom Gnome OS given for Gumstix by Sakoman, this
turned upto be a nightmare, just as you have told, with adhoc WiFi. I gave
up.. I started with Linaro (Ubuntu for embedded systems by Canonical). In
Linaro WiFi and bluetooth both work very well including the AdHoc mode in
WiFi. In Linaro I haven't figured out how to sense ADC pins, which I was
able to do using Angstrom-Gnome. So if you don't require ADC sensing for
the moment, you can start with Linaro.
Regards,
---------------------------------
Sudhangathan BS
Ph:(+91) 9731-905-205
---------------------------------
Generalize CONFIG_IRQ_TIME_ACCOUNTING between X86 and
ARM, move "noirqtime=" option to common debugging code.
For a bit of backward compatibility, X86-specific option
"tsc=noirqtime" is preserved, but issues a warning.
Suggested-by: Yong Zhang <yong.zhang0(a)gmail.com>
Suggested-by: Russell King <rmk+kernel(a)arm.linux.org.uk>
Suggested-by: Ingo Molnar <mingo(a)elte.hu>
Suggested-by: Peter Zijlstra <peterz(a)infradead.org>
Acked-by: Venkatesh Pallipadi <venki(a)google.com>
Signed-off-by: Dmitry Antipov <dmitry.antipov(a)linaro.org>
---
Documentation/kernel-parameters.txt | 9 +++++----
arch/arm/kernel/sched_clock.c | 2 ++
arch/x86/Kconfig | 11 -----------
arch/x86/kernel/tsc.c | 12 ++++++------
include/linux/sched.h | 6 +-----
kernel/sched/core.c | 24 +++++++++++++++++++-----
lib/Kconfig.debug | 12 ++++++++++++
7 files changed, 45 insertions(+), 31 deletions(-)
diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 033d4e6..a5da255 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -1790,6 +1790,9 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
noirqdebug [X86-32] Disables the code which attempts to detect and
disable unhandled interrupt sources.
+ noirqtime [X86,ARM] Run time disables IRQ_TIME_ACCOUNTING and
+ eliminates the timestamping on irq/softirq entry/exit.
+
no_timer_check [X86,APIC] Disables the code which tests for
broken timer IRQ sources.
@@ -2636,10 +2639,8 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
as the stability checks done at bootup. Used to enable
high-resolution timer mode on older hardware, and in
virtualized environment.
- [x86] noirqtime: Do not use TSC to do irq accounting.
- Used to run time disable IRQ_TIME_ACCOUNTING on any
- platforms where RDTSC is slow and this accounting
- can add overhead.
+ [x86] noirqtime: obsoleted by "noirqtime" generic option,
+ see it's documentation for details.
turbografx.map[2|3]= [HW,JOY]
TurboGraFX parallel port interface
diff --git a/arch/arm/kernel/sched_clock.c b/arch/arm/kernel/sched_clock.c
index 5416c7c..30b5f89 100644
--- a/arch/arm/kernel/sched_clock.c
+++ b/arch/arm/kernel/sched_clock.c
@@ -144,6 +144,8 @@ void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate)
*/
cd.epoch_ns = 0;
+ enable_sched_clock_irqtime();
+
pr_debug("Registered %pF as sched_clock source\n", read);
}
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 5bed94e..4759676 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -805,17 +805,6 @@ config SCHED_MC
making when dealing with multi-core CPU chips at a cost of slightly
increased overhead in some places. If unsure say N here.
-config IRQ_TIME_ACCOUNTING
- bool "Fine granularity task level IRQ time accounting"
- default n
- ---help---
- Select this option to enable fine granularity task irq time
- accounting. This is done by reading a timestamp on each
- transitions between softirq and hardirq state, so there can be a
- small performance impact.
-
- If in doubt, say N here.
-
source "kernel/Kconfig.preempt"
config X86_UP_APIC
diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index a62c201..f1b2b63 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -103,14 +103,15 @@ int __init notsc_setup(char *str)
__setup("notsc", notsc_setup);
-static int no_sched_irq_time;
-
static int __init tsc_setup(char *str)
{
if (!strcmp(str, "reliable"))
tsc_clocksource_reliable = 1;
- if (!strncmp(str, "noirqtime", 9))
- no_sched_irq_time = 1;
+ if (!strncmp(str, "noirqtime", 9)) {
+ printk(KERN_WARNING "tsc: tsc=noirqtime is "
+ "obsolete, use noirqtime instead\n");
+ disable_sched_clock_irqtime();
+ }
return 1;
}
@@ -978,8 +979,7 @@ void __init tsc_init(void)
/* now allow native_sched_clock() to use rdtsc */
tsc_disabled = 0;
- if (!no_sched_irq_time)
- enable_sched_clock_irqtime();
+ enable_sched_clock_irqtime();
lpj = ((u64)tsc_khz * 1000);
do_div(lpj, HZ);
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 7d379a6..ea4019c 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1961,11 +1961,7 @@ extern void sched_clock_idle_wakeup_event(u64 delta_ns);
#endif
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
-/*
- * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
- * The reason for this explicit opt-in is not to have perf penalty with
- * slow sched_clocks.
- */
+extern int sched_clock_irqtime;
extern void enable_sched_clock_irqtime(void);
extern void disable_sched_clock_irqtime(void);
#else
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index b342f57..4693509 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -757,11 +757,17 @@ static DEFINE_PER_CPU(u64, cpu_hardirq_time);
static DEFINE_PER_CPU(u64, cpu_softirq_time);
static DEFINE_PER_CPU(u64, irq_start_time);
-static int sched_clock_irqtime;
+
+/*
+ * -1 if not initialized, 0 if disabled with "noirqtime" kernel option
+ * or after unstable clock was detected, 1 if enabled and active.
+ */
+__read_mostly int sched_clock_irqtime = -1;
void enable_sched_clock_irqtime(void)
{
- sched_clock_irqtime = 1;
+ if (sched_clock_irqtime == -1)
+ sched_clock_irqtime = 1;
}
void disable_sched_clock_irqtime(void)
@@ -769,6 +775,14 @@ void disable_sched_clock_irqtime(void)
sched_clock_irqtime = 0;
}
+static int __init irqtime_setup(char *str)
+{
+ sched_clock_irqtime = 0;
+ return 1;
+}
+
+__setup("noirqtime", irqtime_setup);
+
#ifndef CONFIG_64BIT
static DEFINE_PER_CPU(seqcount_t, irq_time_seq);
@@ -822,7 +836,7 @@ void account_system_vtime(struct task_struct *curr)
s64 delta;
int cpu;
- if (!sched_clock_irqtime)
+ if (sched_clock_irqtime < 1)
return;
local_irq_save(flags);
@@ -2852,7 +2866,7 @@ void account_process_tick(struct task_struct *p, int user_tick)
cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
struct rq *rq = this_rq();
- if (sched_clock_irqtime) {
+ if (sched_clock_irqtime > 0) {
irqtime_account_process_tick(p, user_tick, rq);
return;
}
@@ -2886,7 +2900,7 @@ void account_steal_ticks(unsigned long ticks)
void account_idle_ticks(unsigned long ticks)
{
- if (sched_clock_irqtime) {
+ if (sched_clock_irqtime > 0) {
irqtime_account_idle_ticks(ticks);
return;
}
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 8745ac7..236e814 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -299,6 +299,18 @@ config SCHEDSTATS
application, you can say N to avoid the very slight overhead
this adds.
+config IRQ_TIME_ACCOUNTING
+ bool "Fine granularity task level IRQ time accounting"
+ depends on X86 || ARM
+ default n
+ ---help---
+ Select this option to enable fine granularity task irq time
+ accounting. This is done by reading a timestamp on each
+ transitions between softirq and hardirq state, so there can be a
+ small performance impact.
+
+ If in doubt, say N here.
+
config TIMER_STATS
bool "Collect kernel timers statistics"
depends on DEBUG_KERNEL && PROC_FS
--
1.7.7.6
Generalize CONFIG_IRQ_TIME_ACCOUNTING between X86 and
ARM, move "noirqtime=" option to common debugging code.
For a bit of backward compatibility, X86-specific option
"tsc=noirqtime" is preserved, but issues a warning.
Suggested-by: Yong Zhang <yong.zhang0(a)gmail.com>
Suggested-by: Russell King <rmk+kernel(a)arm.linux.org.uk>
Suggested-by: Ingo Molnar <mingo(a)elte.hu>
Acked-by: Venkatesh Pallipadi <venki(a)google.com>
Signed-off-by: Dmitry Antipov <dmitry.antipov(a)linaro.org>
---
Documentation/kernel-parameters.txt | 9 +++++----
arch/arm/kernel/sched_clock.c | 2 ++
arch/x86/Kconfig | 11 -----------
arch/x86/kernel/tsc.c | 12 ++++++------
include/linux/sched.h | 17 ++++++++++-------
kernel/sched/core.c | 21 ++++++++++++---------
lib/Kconfig.debug | 12 ++++++++++++
7 files changed, 47 insertions(+), 37 deletions(-)
diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 033d4e6..a5da255 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -1790,6 +1790,9 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
noirqdebug [X86-32] Disables the code which attempts to detect and
disable unhandled interrupt sources.
+ noirqtime [X86,ARM] Run time disables IRQ_TIME_ACCOUNTING and
+ eliminates the timestamping on irq/softirq entry/exit.
+
no_timer_check [X86,APIC] Disables the code which tests for
broken timer IRQ sources.
@@ -2636,10 +2639,8 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
as the stability checks done at bootup. Used to enable
high-resolution timer mode on older hardware, and in
virtualized environment.
- [x86] noirqtime: Do not use TSC to do irq accounting.
- Used to run time disable IRQ_TIME_ACCOUNTING on any
- platforms where RDTSC is slow and this accounting
- can add overhead.
+ [x86] noirqtime: obsoleted by "noirqtime" generic option,
+ see it's documentation for details.
turbografx.map[2|3]= [HW,JOY]
TurboGraFX parallel port interface
diff --git a/arch/arm/kernel/sched_clock.c b/arch/arm/kernel/sched_clock.c
index 5416c7c..30b5f89 100644
--- a/arch/arm/kernel/sched_clock.c
+++ b/arch/arm/kernel/sched_clock.c
@@ -144,6 +144,8 @@ void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate)
*/
cd.epoch_ns = 0;
+ enable_sched_clock_irqtime();
+
pr_debug("Registered %pF as sched_clock source\n", read);
}
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 5bed94e..4759676 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -805,17 +805,6 @@ config SCHED_MC
making when dealing with multi-core CPU chips at a cost of slightly
increased overhead in some places. If unsure say N here.
-config IRQ_TIME_ACCOUNTING
- bool "Fine granularity task level IRQ time accounting"
- default n
- ---help---
- Select this option to enable fine granularity task irq time
- accounting. This is done by reading a timestamp on each
- transitions between softirq and hardirq state, so there can be a
- small performance impact.
-
- If in doubt, say N here.
-
source "kernel/Kconfig.preempt"
config X86_UP_APIC
diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
index a62c201..f1b2b63 100644
--- a/arch/x86/kernel/tsc.c
+++ b/arch/x86/kernel/tsc.c
@@ -103,14 +103,15 @@ int __init notsc_setup(char *str)
__setup("notsc", notsc_setup);
-static int no_sched_irq_time;
-
static int __init tsc_setup(char *str)
{
if (!strcmp(str, "reliable"))
tsc_clocksource_reliable = 1;
- if (!strncmp(str, "noirqtime", 9))
- no_sched_irq_time = 1;
+ if (!strncmp(str, "noirqtime", 9)) {
+ printk(KERN_WARNING "tsc: tsc=noirqtime is "
+ "obsolete, use noirqtime instead\n");
+ disable_sched_clock_irqtime();
+ }
return 1;
}
@@ -978,8 +979,7 @@ void __init tsc_init(void)
/* now allow native_sched_clock() to use rdtsc */
tsc_disabled = 0;
- if (!no_sched_irq_time)
- enable_sched_clock_irqtime();
+ enable_sched_clock_irqtime();
lpj = ((u64)tsc_khz * 1000);
do_div(lpj, HZ);
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 7d379a6..9b13f79 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1961,13 +1961,16 @@ extern void sched_clock_idle_wakeup_event(u64 delta_ns);
#endif
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
-/*
- * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
- * The reason for this explicit opt-in is not to have perf penalty with
- * slow sched_clocks.
- */
-extern void enable_sched_clock_irqtime(void);
-extern void disable_sched_clock_irqtime(void);
+extern int sched_clock_irqtime;
+static inline void enable_sched_clock_irqtime(void)
+{
+ if (sched_clock_irqtime == -1)
+ sched_clock_irqtime = 1;
+}
+static inline void disable_sched_clock_irqtime(void)
+{
+ sched_clock_irqtime = 0;
+}
#else
static inline void enable_sched_clock_irqtime(void) {}
static inline void disable_sched_clock_irqtime(void) {}
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 5255c9d..4e7a197 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -757,18 +757,21 @@ static DEFINE_PER_CPU(u64, cpu_hardirq_time);
static DEFINE_PER_CPU(u64, cpu_softirq_time);
static DEFINE_PER_CPU(u64, irq_start_time);
-static int sched_clock_irqtime;
-void enable_sched_clock_irqtime(void)
-{
- sched_clock_irqtime = 1;
-}
+/*
+ * -1 if not initialized, 0 if disabled with "noirqtime" kernel option
+ * or after unstable clock was detected, 1 if enabled and active.
+ */
+__read_mostly int sched_clock_irqtime = -1;
-void disable_sched_clock_irqtime(void)
+static int __init irqtime_setup(char *str)
{
sched_clock_irqtime = 0;
+ return 1;
}
+__setup("noirqtime", irqtime_setup);
+
#ifndef CONFIG_64BIT
static DEFINE_PER_CPU(seqcount_t, irq_time_seq);
@@ -822,7 +825,7 @@ void account_system_vtime(struct task_struct *curr)
s64 delta;
int cpu;
- if (!sched_clock_irqtime)
+ if (sched_clock_irqtime < 1)
return;
local_irq_save(flags);
@@ -2853,7 +2856,7 @@ void account_process_tick(struct task_struct *p, int user_tick)
cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
struct rq *rq = this_rq();
- if (sched_clock_irqtime) {
+ if (sched_clock_irqtime > 0) {
irqtime_account_process_tick(p, user_tick, rq);
return;
}
@@ -2887,7 +2890,7 @@ void account_steal_ticks(unsigned long ticks)
void account_idle_ticks(unsigned long ticks)
{
- if (sched_clock_irqtime) {
+ if (sched_clock_irqtime > 0) {
irqtime_account_idle_ticks(ticks);
return;
}
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 8745ac7..236e814 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -299,6 +299,18 @@ config SCHEDSTATS
application, you can say N to avoid the very slight overhead
this adds.
+config IRQ_TIME_ACCOUNTING
+ bool "Fine granularity task level IRQ time accounting"
+ depends on X86 || ARM
+ default n
+ ---help---
+ Select this option to enable fine granularity task irq time
+ accounting. This is done by reading a timestamp on each
+ transitions between softirq and hardirq state, so there can be a
+ small performance impact.
+
+ If in doubt, say N here.
+
config TIMER_STATS
bool "Collect kernel timers statistics"
depends on DEBUG_KERNEL && PROC_FS
--
1.7.7.6
Hi, guys:
As I know linaro is working on a unified memory manager for soc world like
a similar one already exists in PC world (gem/ttm).
But I'm curious about how this manager handle the different sync notify
between various different IP vendor?
As we know, on PC the vsp+capture+graphic+display always done by one
unit--the GPU, if a piece of hardware memory buffer wanted to be
zero-copied between vsp/graphic/display there must be some sync/notify
mechanism otherwise the race condition will occur. On gpu always an
interrupt driven object fence can handle it. But in soc, there is no
integrated vsp/graphic/display, they all may come from various independent
ip vendor, so there may no unified interrupt source can be collected by
memory manager to know whether one buffer hasn't be completed by the
previous engine and the next engine should block wait but without cpu
blocked too. Thanks
Regards
