* [PATCH v3 0/2] clocksource/drivers/dw_apb_timer: improve performance a bit
@ 2015-11-25 15:48 ` Jisheng Zhang
0 siblings, 0 replies; 8+ messages in thread
From: Jisheng Zhang @ 2015-11-25 15:48 UTC (permalink / raw)
To: linux-arm-kernel
These two patches try to improve the dw_apb_timer clocksource/clockevent
performance.
since v2:
- only use relaxed version in critical path
- inline apbt_readl/apbt_writel
- add some performance numbers
since v1:
- correct the wrong sentence in commit msg about writel performance on
CA9 with outer L2 cache. Thank Arnd for pointing out this in another
thread.
Jisheng Zhang (2):
clocksource/drivers/dw_apb_timer: Inline apbt_readl and apbt_writel
clocksource/drivers/dw_apb_timer: Use {readl|writel}_relaxed in
critical path
drivers/clocksource/dw_apb_timer.c | 30 +++++++++++++++++++++---------
1 file changed, 21 insertions(+), 9 deletions(-)
--
2.6.2
^ permalink raw reply [flat|nested] 8+ messages in thread* [PATCH v3 0/2] clocksource/drivers/dw_apb_timer: improve performance a bit @ 2015-11-25 15:48 ` Jisheng Zhang 0 siblings, 0 replies; 8+ messages in thread From: Jisheng Zhang @ 2015-11-25 15:48 UTC (permalink / raw) To: daniel.lezcano, tglx, arnd; +Cc: linux-kernel, linux-arm-kernel, Jisheng Zhang These two patches try to improve the dw_apb_timer clocksource/clockevent performance. since v2: - only use relaxed version in critical path - inline apbt_readl/apbt_writel - add some performance numbers since v1: - correct the wrong sentence in commit msg about writel performance on CA9 with outer L2 cache. Thank Arnd for pointing out this in another thread. Jisheng Zhang (2): clocksource/drivers/dw_apb_timer: Inline apbt_readl and apbt_writel clocksource/drivers/dw_apb_timer: Use {readl|writel}_relaxed in critical path drivers/clocksource/dw_apb_timer.c | 30 +++++++++++++++++++++--------- 1 file changed, 21 insertions(+), 9 deletions(-) -- 2.6.2 ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH v3 1/2] clocksource/drivers/dw_apb_timer: Inline apbt_readl and apbt_writel 2015-11-25 15:48 ` Jisheng Zhang @ 2015-11-25 15:48 ` Jisheng Zhang -1 siblings, 0 replies; 8+ messages in thread From: Jisheng Zhang @ 2015-11-25 15:48 UTC (permalink / raw) To: linux-arm-kernel It seems gcc can automatically inline apbt_writel() for us, but apbt_real isn't inlined. This patch makes them inline to get a trivial performance improvement: the time __apbt_read_clocksource() spends on Marvell BG4CT platform: before the patch 1275240ns on average after the patch 1263240ns on average so we get 1% performance improvement. Signed-off-by: Jisheng Zhang <jszhang@marvell.com> --- drivers/clocksource/dw_apb_timer.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/clocksource/dw_apb_timer.c b/drivers/clocksource/dw_apb_timer.c index 3a6d9db..114696b 100644 --- a/drivers/clocksource/dw_apb_timer.c +++ b/drivers/clocksource/dw_apb_timer.c @@ -49,13 +49,13 @@ clocksource_to_dw_apb_clocksource(struct clocksource *cs) return container_of(cs, struct dw_apb_clocksource, cs); } -static u32 apbt_readl(struct dw_apb_timer *timer, unsigned long offs) +static inline u32 apbt_readl(struct dw_apb_timer *timer, unsigned long offs) { return readl(timer->base + offs); } -static void apbt_writel(struct dw_apb_timer *timer, u32 val, - unsigned long offs) +static inline void apbt_writel(struct dw_apb_timer *timer, u32 val, + unsigned long offs) { writel(val, timer->base + offs); } -- 2.6.2 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH v3 1/2] clocksource/drivers/dw_apb_timer: Inline apbt_readl and apbt_writel @ 2015-11-25 15:48 ` Jisheng Zhang 0 siblings, 0 replies; 8+ messages in thread From: Jisheng Zhang @ 2015-11-25 15:48 UTC (permalink / raw) To: daniel.lezcano, tglx, arnd; +Cc: linux-kernel, linux-arm-kernel, Jisheng Zhang It seems gcc can automatically inline apbt_writel() for us, but apbt_real isn't inlined. This patch makes them inline to get a trivial performance improvement: the time __apbt_read_clocksource() spends on Marvell BG4CT platform: before the patch 1275240ns on average after the patch 1263240ns on average so we get 1% performance improvement. Signed-off-by: Jisheng Zhang <jszhang@marvell.com> --- drivers/clocksource/dw_apb_timer.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/clocksource/dw_apb_timer.c b/drivers/clocksource/dw_apb_timer.c index 3a6d9db..114696b 100644 --- a/drivers/clocksource/dw_apb_timer.c +++ b/drivers/clocksource/dw_apb_timer.c @@ -49,13 +49,13 @@ clocksource_to_dw_apb_clocksource(struct clocksource *cs) return container_of(cs, struct dw_apb_clocksource, cs); } -static u32 apbt_readl(struct dw_apb_timer *timer, unsigned long offs) +static inline u32 apbt_readl(struct dw_apb_timer *timer, unsigned long offs) { return readl(timer->base + offs); } -static void apbt_writel(struct dw_apb_timer *timer, u32 val, - unsigned long offs) +static inline void apbt_writel(struct dw_apb_timer *timer, u32 val, + unsigned long offs) { writel(val, timer->base + offs); } -- 2.6.2 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH v3 2/2] clocksource/drivers/dw_apb_timer: Use {readl|writel}_relaxed in critical path 2015-11-25 15:48 ` Jisheng Zhang @ 2015-11-25 15:48 ` Jisheng Zhang -1 siblings, 0 replies; 8+ messages in thread From: Jisheng Zhang @ 2015-11-25 15:48 UTC (permalink / raw) To: linux-arm-kernel It's safe to use the relaxed version. From another side, the relaxed io accessor macros are available on all architectures now, so we can use the relaxed versions to get a trivial system performance improvement, we measured time the following functions spent on Marvell BG4CT: __apbt_read_clocksource(): before the patch: 1263240ns on average after the patch: 1250080ns on average improved by 1% apbt_eoi(): before the patch: 1290960ns on average after the patch: 1248240ns on average apbt_next_event(): before the patch: 3333660ns on average after the patch: 1322040ns on average improved by 60%! Signed-off-by: Jisheng Zhang <jszhang@marvell.com> --- drivers/clocksource/dw_apb_timer.c | 24 ++++++++++++++++++------ 1 file changed, 18 insertions(+), 6 deletions(-) diff --git a/drivers/clocksource/dw_apb_timer.c b/drivers/clocksource/dw_apb_timer.c index 114696b..6334526 100644 --- a/drivers/clocksource/dw_apb_timer.c +++ b/drivers/clocksource/dw_apb_timer.c @@ -60,6 +60,17 @@ static inline void apbt_writel(struct dw_apb_timer *timer, u32 val, writel(val, timer->base + offs); } +static inline u32 apbt_readl_relaxed(struct dw_apb_timer *timer, unsigned long offs) +{ + return readl_relaxed(timer->base + offs); +} + +static inline void apbt_writel_relaxed(struct dw_apb_timer *timer, u32 val, + unsigned long offs) +{ + writel_relaxed(val, timer->base + offs); +} + static void apbt_disable_int(struct dw_apb_timer *timer) { u32 ctrl = apbt_readl(timer, APBTMR_N_CONTROL); @@ -81,7 +92,7 @@ void dw_apb_clockevent_pause(struct dw_apb_clock_event_device *dw_ced) static void apbt_eoi(struct dw_apb_timer *timer) { - apbt_readl(timer, APBTMR_N_EOI); + apbt_readl_relaxed(timer, APBTMR_N_EOI); } static irqreturn_t dw_apb_clockevent_irq(int irq, void *data) @@ -200,13 +211,13 @@ static int apbt_next_event(unsigned long delta, struct dw_apb_clock_event_device *dw_ced = ced_to_dw_apb_ced(evt); /* Disable timer */ - ctrl = apbt_readl(&dw_ced->timer, APBTMR_N_CONTROL); + ctrl = apbt_readl_relaxed(&dw_ced->timer, APBTMR_N_CONTROL); ctrl &= ~APBTMR_CONTROL_ENABLE; - apbt_writel(&dw_ced->timer, ctrl, APBTMR_N_CONTROL); + apbt_writel_relaxed(&dw_ced->timer, ctrl, APBTMR_N_CONTROL); /* write new count */ - apbt_writel(&dw_ced->timer, delta, APBTMR_N_LOAD_COUNT); + apbt_writel_relaxed(&dw_ced->timer, delta, APBTMR_N_LOAD_COUNT); ctrl |= APBTMR_CONTROL_ENABLE; - apbt_writel(&dw_ced->timer, ctrl, APBTMR_N_CONTROL); + apbt_writel_relaxed(&dw_ced->timer, ctrl, APBTMR_N_CONTROL); return 0; } @@ -342,7 +353,8 @@ static cycle_t __apbt_read_clocksource(struct clocksource *cs) struct dw_apb_clocksource *dw_cs = clocksource_to_dw_apb_clocksource(cs); - current_count = apbt_readl(&dw_cs->timer, APBTMR_N_CURRENT_VALUE); + current_count = apbt_readl_relaxed(&dw_cs->timer, + APBTMR_N_CURRENT_VALUE); return (cycle_t)~current_count; } -- 2.6.2 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH v3 2/2] clocksource/drivers/dw_apb_timer: Use {readl|writel}_relaxed in critical path @ 2015-11-25 15:48 ` Jisheng Zhang 0 siblings, 0 replies; 8+ messages in thread From: Jisheng Zhang @ 2015-11-25 15:48 UTC (permalink / raw) To: daniel.lezcano, tglx, arnd; +Cc: linux-kernel, linux-arm-kernel, Jisheng Zhang It's safe to use the relaxed version. From another side, the relaxed io accessor macros are available on all architectures now, so we can use the relaxed versions to get a trivial system performance improvement, we measured time the following functions spent on Marvell BG4CT: __apbt_read_clocksource(): before the patch: 1263240ns on average after the patch: 1250080ns on average improved by 1% apbt_eoi(): before the patch: 1290960ns on average after the patch: 1248240ns on average apbt_next_event(): before the patch: 3333660ns on average after the patch: 1322040ns on average improved by 60%! Signed-off-by: Jisheng Zhang <jszhang@marvell.com> --- drivers/clocksource/dw_apb_timer.c | 24 ++++++++++++++++++------ 1 file changed, 18 insertions(+), 6 deletions(-) diff --git a/drivers/clocksource/dw_apb_timer.c b/drivers/clocksource/dw_apb_timer.c index 114696b..6334526 100644 --- a/drivers/clocksource/dw_apb_timer.c +++ b/drivers/clocksource/dw_apb_timer.c @@ -60,6 +60,17 @@ static inline void apbt_writel(struct dw_apb_timer *timer, u32 val, writel(val, timer->base + offs); } +static inline u32 apbt_readl_relaxed(struct dw_apb_timer *timer, unsigned long offs) +{ + return readl_relaxed(timer->base + offs); +} + +static inline void apbt_writel_relaxed(struct dw_apb_timer *timer, u32 val, + unsigned long offs) +{ + writel_relaxed(val, timer->base + offs); +} + static void apbt_disable_int(struct dw_apb_timer *timer) { u32 ctrl = apbt_readl(timer, APBTMR_N_CONTROL); @@ -81,7 +92,7 @@ void dw_apb_clockevent_pause(struct dw_apb_clock_event_device *dw_ced) static void apbt_eoi(struct dw_apb_timer *timer) { - apbt_readl(timer, APBTMR_N_EOI); + apbt_readl_relaxed(timer, APBTMR_N_EOI); } static irqreturn_t dw_apb_clockevent_irq(int irq, void *data) @@ -200,13 +211,13 @@ static int apbt_next_event(unsigned long delta, struct dw_apb_clock_event_device *dw_ced = ced_to_dw_apb_ced(evt); /* Disable timer */ - ctrl = apbt_readl(&dw_ced->timer, APBTMR_N_CONTROL); + ctrl = apbt_readl_relaxed(&dw_ced->timer, APBTMR_N_CONTROL); ctrl &= ~APBTMR_CONTROL_ENABLE; - apbt_writel(&dw_ced->timer, ctrl, APBTMR_N_CONTROL); + apbt_writel_relaxed(&dw_ced->timer, ctrl, APBTMR_N_CONTROL); /* write new count */ - apbt_writel(&dw_ced->timer, delta, APBTMR_N_LOAD_COUNT); + apbt_writel_relaxed(&dw_ced->timer, delta, APBTMR_N_LOAD_COUNT); ctrl |= APBTMR_CONTROL_ENABLE; - apbt_writel(&dw_ced->timer, ctrl, APBTMR_N_CONTROL); + apbt_writel_relaxed(&dw_ced->timer, ctrl, APBTMR_N_CONTROL); return 0; } @@ -342,7 +353,8 @@ static cycle_t __apbt_read_clocksource(struct clocksource *cs) struct dw_apb_clocksource *dw_cs = clocksource_to_dw_apb_clocksource(cs); - current_count = apbt_readl(&dw_cs->timer, APBTMR_N_CURRENT_VALUE); + current_count = apbt_readl_relaxed(&dw_cs->timer, + APBTMR_N_CURRENT_VALUE); return (cycle_t)~current_count; } -- 2.6.2 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH v3 0/2] clocksource/drivers/dw_apb_timer: improve performance a bit 2015-11-25 15:48 ` Jisheng Zhang @ 2015-11-25 15:50 ` Jisheng Zhang -1 siblings, 0 replies; 8+ messages in thread From: Jisheng Zhang @ 2015-11-25 15:50 UTC (permalink / raw) To: linux-arm-kernel On Wed, 25 Nov 2015 23:48:32 +0800 Jisheng Zhang wrote: > These two patches try to improve the dw_apb_timer clocksource/clockevent > performance. These patches depends on the apbt_readl return value fix patch: http://lists.infradead.org/pipermail/linux-arm-kernel/2015-November/388250.html thanks, Jisheng > > since v2: > - only use relaxed version in critical path > - inline apbt_readl/apbt_writel > - add some performance numbers > > since v1: > - correct the wrong sentence in commit msg about writel performance on > CA9 with outer L2 cache. Thank Arnd for pointing out this in another > thread. > > Jisheng Zhang (2): > clocksource/drivers/dw_apb_timer: Inline apbt_readl and apbt_writel > clocksource/drivers/dw_apb_timer: Use {readl|writel}_relaxed in > critical path > > drivers/clocksource/dw_apb_timer.c | 30 +++++++++++++++++++++--------- > 1 file changed, 21 insertions(+), 9 deletions(-) > ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH v3 0/2] clocksource/drivers/dw_apb_timer: improve performance a bit @ 2015-11-25 15:50 ` Jisheng Zhang 0 siblings, 0 replies; 8+ messages in thread From: Jisheng Zhang @ 2015-11-25 15:50 UTC (permalink / raw) To: daniel.lezcano, tglx, arnd; +Cc: linux-kernel, linux-arm-kernel On Wed, 25 Nov 2015 23:48:32 +0800 Jisheng Zhang wrote: > These two patches try to improve the dw_apb_timer clocksource/clockevent > performance. These patches depends on the apbt_readl return value fix patch: http://lists.infradead.org/pipermail/linux-arm-kernel/2015-November/388250.html thanks, Jisheng > > since v2: > - only use relaxed version in critical path > - inline apbt_readl/apbt_writel > - add some performance numbers > > since v1: > - correct the wrong sentence in commit msg about writel performance on > CA9 with outer L2 cache. Thank Arnd for pointing out this in another > thread. > > Jisheng Zhang (2): > clocksource/drivers/dw_apb_timer: Inline apbt_readl and apbt_writel > clocksource/drivers/dw_apb_timer: Use {readl|writel}_relaxed in > critical path > > drivers/clocksource/dw_apb_timer.c | 30 +++++++++++++++++++++--------- > 1 file changed, 21 insertions(+), 9 deletions(-) > ^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2015-11-25 15:55 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-11-25 15:48 [PATCH v3 0/2] clocksource/drivers/dw_apb_timer: improve performance a bit Jisheng Zhang
2015-11-25 15:48 ` Jisheng Zhang
2015-11-25 15:48 ` [PATCH v3 1/2] clocksource/drivers/dw_apb_timer: Inline apbt_readl and apbt_writel Jisheng Zhang
2015-11-25 15:48 ` Jisheng Zhang
2015-11-25 15:48 ` [PATCH v3 2/2] clocksource/drivers/dw_apb_timer: Use {readl|writel}_relaxed in critical path Jisheng Zhang
2015-11-25 15:48 ` Jisheng Zhang
2015-11-25 15:50 ` [PATCH v3 0/2] clocksource/drivers/dw_apb_timer: improve performance a bit Jisheng Zhang
2015-11-25 15:50 ` Jisheng Zhang
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.