* [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case
@ 2019-10-27 15:19 Vitaly Kuznetsov
2019-10-27 16:49 ` Michael Kelley
` (3 more replies)
0 siblings, 4 replies; 6+ messages in thread
From: Vitaly Kuznetsov @ 2019-10-27 15:19 UTC (permalink / raw)
To: linux-hyperv
Cc: linux-kernel, x86, K. Y. Srinivasan, Haiyang Zhang,
Stephen Hemminger, Sasha Levin, Thomas Gleixner, Ingo Molnar,
Borislav Petkov, H. Peter Anvin, Roman Kagan, Michael Kelley,
Joe Perches
When sending an IPI to a single CPU there is no need to deal with cpumasks.
With 2 CPU guest on WS2019 I'm seeing a minor (like 3%, 8043 -> 7761 CPU
cycles) improvement with smp_call_function_single() loop benchmark. The
optimization, however, is tiny and straitforward. Also, send_ipi_one() is
important for PV spinlock kick.
I was also wondering if it would make sense to switch to using regular
APIC IPI send for CPU > 64 case but no, it is twice as expesive (12650 CPU
cycles for __send_ipi_mask_ex() call, 26000 for orig_apic.send_IPI(cpu,
vector)).
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
---
Changes since v2:
- Check VP number instead of CPU number against >= 64 [Michael]
- Check for VP_INVAL
---
arch/x86/hyperv/hv_apic.c | 16 +++++++++++++---
arch/x86/include/asm/trace/hyperv.h | 15 +++++++++++++++
2 files changed, 28 insertions(+), 3 deletions(-)
diff --git a/arch/x86/hyperv/hv_apic.c b/arch/x86/hyperv/hv_apic.c
index e01078e93dd3..40e0e322161d 100644
--- a/arch/x86/hyperv/hv_apic.c
+++ b/arch/x86/hyperv/hv_apic.c
@@ -194,10 +194,20 @@ static bool __send_ipi_mask(const struct cpumask *mask, int vector)
static bool __send_ipi_one(int cpu, int vector)
{
- struct cpumask mask = CPU_MASK_NONE;
+ int vp = hv_cpu_number_to_vp_number(cpu);
- cpumask_set_cpu(cpu, &mask);
- return __send_ipi_mask(&mask, vector);
+ trace_hyperv_send_ipi_one(cpu, vector);
+
+ if (!hv_hypercall_pg || (vp == VP_INVAL))
+ return false;
+
+ if ((vector < HV_IPI_LOW_VECTOR) || (vector > HV_IPI_HIGH_VECTOR))
+ return false;
+
+ if (vp >= 64)
+ return __send_ipi_mask_ex(cpumask_of(cpu), vector);
+
+ return !hv_do_fast_hypercall16(HVCALL_SEND_IPI, vector, BIT_ULL(vp));
}
static void hv_send_ipi(int cpu, int vector)
diff --git a/arch/x86/include/asm/trace/hyperv.h b/arch/x86/include/asm/trace/hyperv.h
index ace464f09681..4d705cb4d63b 100644
--- a/arch/x86/include/asm/trace/hyperv.h
+++ b/arch/x86/include/asm/trace/hyperv.h
@@ -71,6 +71,21 @@ TRACE_EVENT(hyperv_send_ipi_mask,
__entry->ncpus, __entry->vector)
);
+TRACE_EVENT(hyperv_send_ipi_one,
+ TP_PROTO(int cpu,
+ int vector),
+ TP_ARGS(cpu, vector),
+ TP_STRUCT__entry(
+ __field(int, cpu)
+ __field(int, vector)
+ ),
+ TP_fast_assign(__entry->cpu = cpu;
+ __entry->vector = vector;
+ ),
+ TP_printk("cpu %d vector %x",
+ __entry->cpu, __entry->vector)
+ );
+
#endif /* CONFIG_HYPERV */
#undef TRACE_INCLUDE_PATH
--
2.20.1
^ permalink raw reply related [flat|nested] 6+ messages in thread
* RE: [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case
2019-10-27 15:19 [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case Vitaly Kuznetsov
@ 2019-10-27 16:49 ` Michael Kelley
2019-10-28 9:35 ` Roman Kagan
` (2 subsequent siblings)
3 siblings, 0 replies; 6+ messages in thread
From: Michael Kelley @ 2019-10-27 16:49 UTC (permalink / raw)
To: vkuznets, linux-hyperv
Cc: linux-kernel, x86, KY Srinivasan, Haiyang Zhang,
Stephen Hemminger, Sasha Levin, Thomas Gleixner, Ingo Molnar,
Borislav Petkov, H. Peter Anvin, Roman Kagan, Joe Perches
From: Vitaly Kuznetsov <vkuznets@redhat.com> Sent: Sunday, October 27, 2019 8:20 AM
>
> When sending an IPI to a single CPU there is no need to deal with cpumasks.
> With 2 CPU guest on WS2019 I'm seeing a minor (like 3%, 8043 -> 7761 CPU
> cycles) improvement with smp_call_function_single() loop benchmark. The
> optimization, however, is tiny and straitforward. Also, send_ipi_one() is
> important for PV spinlock kick.
>
> I was also wondering if it would make sense to switch to using regular
> APIC IPI send for CPU > 64 case but no, it is twice as expesive (12650 CPU
> cycles for __send_ipi_mask_ex() call, 26000 for orig_apic.send_IPI(cpu,
> vector)).
>
> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
> ---
> Changes since v2:
> - Check VP number instead of CPU number against >= 64 [Michael]
> - Check for VP_INVAL
>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case
2019-10-27 15:19 [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case Vitaly Kuznetsov
2019-10-27 16:49 ` Michael Kelley
@ 2019-10-28 9:35 ` Roman Kagan
2019-11-07 13:26 ` Vitaly Kuznetsov
2019-11-12 10:50 ` [tip: x86/hyperv] x86/hyperv: Micro-optimize send_ipi_one() tip-bot2 for Vitaly Kuznetsov
3 siblings, 0 replies; 6+ messages in thread
From: Roman Kagan @ 2019-10-28 9:35 UTC (permalink / raw)
To: Vitaly Kuznetsov
Cc: linux-hyperv, linux-kernel, x86, K. Y. Srinivasan, Haiyang Zhang,
Stephen Hemminger, Sasha Levin, Thomas Gleixner, Ingo Molnar,
Borislav Petkov, H. Peter Anvin, Michael Kelley, Joe Perches
On Sun, Oct 27, 2019 at 04:19:38PM +0100, Vitaly Kuznetsov wrote:
> When sending an IPI to a single CPU there is no need to deal with cpumasks.
> With 2 CPU guest on WS2019 I'm seeing a minor (like 3%, 8043 -> 7761 CPU
> cycles) improvement with smp_call_function_single() loop benchmark. The
> optimization, however, is tiny and straitforward. Also, send_ipi_one() is
> important for PV spinlock kick.
>
> I was also wondering if it would make sense to switch to using regular
> APIC IPI send for CPU > 64 case but no, it is twice as expesive (12650 CPU
> cycles for __send_ipi_mask_ex() call, 26000 for orig_apic.send_IPI(cpu,
> vector)).
>
> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
> ---
> Changes since v2:
> - Check VP number instead of CPU number against >= 64 [Michael]
> - Check for VP_INVAL
> ---
> arch/x86/hyperv/hv_apic.c | 16 +++++++++++++---
> arch/x86/include/asm/trace/hyperv.h | 15 +++++++++++++++
> 2 files changed, 28 insertions(+), 3 deletions(-)
Reviewed-by: Roman Kagan <rkagan@virtuozzo.com>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case
2019-10-27 15:19 [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case Vitaly Kuznetsov
2019-10-27 16:49 ` Michael Kelley
2019-10-28 9:35 ` Roman Kagan
@ 2019-11-07 13:26 ` Vitaly Kuznetsov
2019-11-07 21:21 ` Thomas Gleixner
2019-11-12 10:50 ` [tip: x86/hyperv] x86/hyperv: Micro-optimize send_ipi_one() tip-bot2 for Vitaly Kuznetsov
3 siblings, 1 reply; 6+ messages in thread
From: Vitaly Kuznetsov @ 2019-11-07 13:26 UTC (permalink / raw)
To: Sasha Levin
Cc: linux-hyperv, linux-kernel, x86, K. Y. Srinivasan, Haiyang Zhang,
Stephen Hemminger, Thomas Gleixner, Ingo Molnar, Borislav Petkov,
H. Peter Anvin, Roman Kagan, Michael Kelley, Joe Perches
Vitaly Kuznetsov <vkuznets@redhat.com> writes:
> When sending an IPI to a single CPU there is no need to deal with cpumasks.
> With 2 CPU guest on WS2019 I'm seeing a minor (like 3%, 8043 -> 7761 CPU
> cycles) improvement with smp_call_function_single() loop benchmark. The
> optimization, however, is tiny and straitforward. Also, send_ipi_one() is
> important for PV spinlock kick.
>
> I was also wondering if it would make sense to switch to using regular
> APIC IPI send for CPU > 64 case but no, it is twice as expesive (12650 CPU
> cycles for __send_ipi_mask_ex() call, 26000 for orig_apic.send_IPI(cpu,
> vector)).
>
> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
> ---
> Changes since v2:
> - Check VP number instead of CPU number against >= 64 [Michael]
> - Check for VP_INVAL
Hi Sasha,
do you have plans to pick this up for hyperv-next or should we ask x86
folks to?
Thanks!
--
Vitaly
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case
2019-11-07 13:26 ` Vitaly Kuznetsov
@ 2019-11-07 21:21 ` Thomas Gleixner
0 siblings, 0 replies; 6+ messages in thread
From: Thomas Gleixner @ 2019-11-07 21:21 UTC (permalink / raw)
To: Vitaly Kuznetsov
Cc: Sasha Levin, linux-hyperv, linux-kernel, x86, K. Y. Srinivasan,
Haiyang Zhang, Stephen Hemminger, Ingo Molnar, Borislav Petkov,
H. Peter Anvin, Roman Kagan, Michael Kelley, Joe Perches
On Thu, 7 Nov 2019, Vitaly Kuznetsov wrote:
> Vitaly Kuznetsov <vkuznets@redhat.com> writes:
>
> > When sending an IPI to a single CPU there is no need to deal with cpumasks.
> > With 2 CPU guest on WS2019 I'm seeing a minor (like 3%, 8043 -> 7761 CPU
> > cycles) improvement with smp_call_function_single() loop benchmark. The
> > optimization, however, is tiny and straitforward. Also, send_ipi_one() is
> > important for PV spinlock kick.
> >
> > I was also wondering if it would make sense to switch to using regular
> > APIC IPI send for CPU > 64 case but no, it is twice as expesive (12650 CPU
> > cycles for __send_ipi_mask_ex() call, 26000 for orig_apic.send_IPI(cpu,
> > vector)).
> >
> > Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
> > ---
> > Changes since v2:
> > - Check VP number instead of CPU number against >= 64 [Michael]
> > - Check for VP_INVAL
>
> Hi Sasha,
>
> do you have plans to pick this up for hyperv-next or should we ask x86
> folks to?
I'm picking up the constant TSC one anyway, so I can just throw that in as
well.
Thanks,
tglx
^ permalink raw reply [flat|nested] 6+ messages in thread
* [tip: x86/hyperv] x86/hyperv: Micro-optimize send_ipi_one()
2019-10-27 15:19 [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case Vitaly Kuznetsov
` (2 preceding siblings ...)
2019-11-07 13:26 ` Vitaly Kuznetsov
@ 2019-11-12 10:50 ` tip-bot2 for Vitaly Kuznetsov
3 siblings, 0 replies; 6+ messages in thread
From: tip-bot2 for Vitaly Kuznetsov @ 2019-11-12 10:50 UTC (permalink / raw)
To: linux-tip-commits
Cc: Vitaly Kuznetsov, Thomas Gleixner, Michael Kelley, Roman Kagan,
Ingo Molnar, Borislav Petkov, linux-kernel
The following commit has been merged into the x86/hyperv branch of tip:
Commit-ID: b264f57fde0c686c5c1dfdd0c21992c49196bb87
Gitweb: https://git.kernel.org/tip/b264f57fde0c686c5c1dfdd0c21992c49196bb87
Author: Vitaly Kuznetsov <vkuznets@redhat.com>
AuthorDate: Sun, 27 Oct 2019 16:19:38 +01:00
Committer: Thomas Gleixner <tglx@linutronix.de>
CommitterDate: Tue, 12 Nov 2019 11:44:20 +01:00
x86/hyperv: Micro-optimize send_ipi_one()
When sending an IPI to a single CPU there is no need to deal with cpumasks.
With 2 CPU guest on WS2019 a minor (like 3%, 8043 -> 7761 CPU cycles)
improvement with smp_call_function_single() loop benchmark can be seeb. The
optimization, however, is tiny and straitforward. Also, send_ipi_one() is
important for PV spinlock kick.
Switching to the regular APIC IPI send for CPU > 64 case does not make
sense as it is twice as expesive (12650 CPU cycles for __send_ipi_mask_ex()
call, 26000 for orig_apic.send_IPI(cpu, vector)).
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Reviewed-by: Roman Kagan <rkagan@virtuozzo.com>
Link: https://lkml.kernel.org/r/20191027151938.7296-1-vkuznets@redhat.com
---
arch/x86/hyperv/hv_apic.c | 16 +++++++++++++---
arch/x86/include/asm/trace/hyperv.h | 15 +++++++++++++++
2 files changed, 28 insertions(+), 3 deletions(-)
diff --git a/arch/x86/hyperv/hv_apic.c b/arch/x86/hyperv/hv_apic.c
index 5c056b8..86c8674 100644
--- a/arch/x86/hyperv/hv_apic.c
+++ b/arch/x86/hyperv/hv_apic.c
@@ -194,10 +194,20 @@ do_ex_hypercall:
static bool __send_ipi_one(int cpu, int vector)
{
- struct cpumask mask = CPU_MASK_NONE;
+ int vp = hv_cpu_number_to_vp_number(cpu);
- cpumask_set_cpu(cpu, &mask);
- return __send_ipi_mask(&mask, vector);
+ trace_hyperv_send_ipi_one(cpu, vector);
+
+ if (!hv_hypercall_pg || (vp == VP_INVAL))
+ return false;
+
+ if ((vector < HV_IPI_LOW_VECTOR) || (vector > HV_IPI_HIGH_VECTOR))
+ return false;
+
+ if (vp >= 64)
+ return __send_ipi_mask_ex(cpumask_of(cpu), vector);
+
+ return !hv_do_fast_hypercall16(HVCALL_SEND_IPI, vector, BIT_ULL(vp));
}
static void hv_send_ipi(int cpu, int vector)
diff --git a/arch/x86/include/asm/trace/hyperv.h b/arch/x86/include/asm/trace/hyperv.h
index ace464f..4d705cb 100644
--- a/arch/x86/include/asm/trace/hyperv.h
+++ b/arch/x86/include/asm/trace/hyperv.h
@@ -71,6 +71,21 @@ TRACE_EVENT(hyperv_send_ipi_mask,
__entry->ncpus, __entry->vector)
);
+TRACE_EVENT(hyperv_send_ipi_one,
+ TP_PROTO(int cpu,
+ int vector),
+ TP_ARGS(cpu, vector),
+ TP_STRUCT__entry(
+ __field(int, cpu)
+ __field(int, vector)
+ ),
+ TP_fast_assign(__entry->cpu = cpu;
+ __entry->vector = vector;
+ ),
+ TP_printk("cpu %d vector %x",
+ __entry->cpu, __entry->vector)
+ );
+
#endif /* CONFIG_HYPERV */
#undef TRACE_INCLUDE_PATH
^ permalink raw reply related [flat|nested] 6+ messages in thread
end of thread, other threads:[~2019-11-12 10:50 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-10-27 15:19 [PATCH v3] x86/hyper-v: micro-optimize send_ipi_one case Vitaly Kuznetsov
2019-10-27 16:49 ` Michael Kelley
2019-10-28 9:35 ` Roman Kagan
2019-11-07 13:26 ` Vitaly Kuznetsov
2019-11-07 21:21 ` Thomas Gleixner
2019-11-12 10:50 ` [tip: x86/hyperv] x86/hyperv: Micro-optimize send_ipi_one() tip-bot2 for Vitaly Kuznetsov
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).