| Message ID | 20190515082324.112755-1-borntraeger@de.ibm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | KVM: s390: change default halt poll to 50000 | expand |
"KVM: s390: change default halt poll time to 50us Recent measurements indicate that using 50us results in a reduced CPU consumption, while still providing the benefit of halt polling. Let's use 50us instead." Acked-by: David Hildenbrand <david@redhat.com> On 15.05.19 10:23, Christian Borntraeger wrote: > older performance measurements indicated that 50000 vs 80000 reduces cpu > consumption while still providing the benefit of halt polling. We had > this change in the IBM KVM product, but it got lost so it never went > upstream. Recent re-measurement indicate that 50k is still better than > 80k.> > Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> > --- > arch/s390/include/asm/kvm_host.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h > index dbe254847e0d..cb63cc7bbf06 100644 > --- a/arch/s390/include/asm/kvm_host.h > +++ b/arch/s390/include/asm/kvm_host.h > @@ -36,7 +36,7 @@ > */ > #define KVM_NR_IRQCHIPS 1 > #define KVM_IRQCHIP_NUM_PINS 4096 > -#define KVM_HALT_POLL_NS_DEFAULT 80000 > +#define KVM_HALT_POLL_NS_DEFAULT 50000 > > /* s390-specific vcpu->requests bit members */ > #define KVM_REQ_ENABLE_IBS KVM_ARCH_REQ(0) >
On 15.05.19 12:22, David Hildenbrand wrote: > "KVM: s390: change default halt poll time to 50us > > Recent measurements indicate that using 50us results in a reduced CPU > consumption, while still providing the benefit of halt polling. Let's > use 50us instead." Yes, sounds better. > > Acked-by: David Hildenbrand <david@redhat.com> > > > > On 15.05.19 10:23, Christian Borntraeger wrote: >> older performance measurements indicated that 50000 vs 80000 reduces cpu >> consumption while still providing the benefit of halt polling. We had >> this change in the IBM KVM product, but it got lost so it never went >> upstream. Recent re-measurement indicate that 50k is still better than >> 80k.> >> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> >> --- >> arch/s390/include/asm/kvm_host.h | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h >> index dbe254847e0d..cb63cc7bbf06 100644 >> --- a/arch/s390/include/asm/kvm_host.h >> +++ b/arch/s390/include/asm/kvm_host.h >> @@ -36,7 +36,7 @@ >> */ >> #define KVM_NR_IRQCHIPS 1 >> #define KVM_IRQCHIP_NUM_PINS 4096 >> -#define KVM_HALT_POLL_NS_DEFAULT 80000 >> +#define KVM_HALT_POLL_NS_DEFAULT 50000 >> >> /* s390-specific vcpu->requests bit members */ >> #define KVM_REQ_ENABLE_IBS KVM_ARCH_REQ(0) >> > >
On 5/15/19 12:22 PM, David Hildenbrand wrote: > "KVM: s390: change default halt poll time to 50us > > Recent measurements indicate that using 50us results in a reduced CPU > consumption, while still providing the benefit of halt polling. Let's > use 50us instead." > > Acked-by: David Hildenbrand <david@redhat.com> With that change: Acked-by: Janosch Frank <frankja@linux.vnet.ibm.com> > > > > On 15.05.19 10:23, Christian Borntraeger wrote: >> older performance measurements indicated that 50000 vs 80000 reduces cpu >> consumption while still providing the benefit of halt polling. We had >> this change in the IBM KVM product, but it got lost so it never went >> upstream. Recent re-measurement indicate that 50k is still better than >> 80k.> >> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> >> --- >> arch/s390/include/asm/kvm_host.h | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h >> index dbe254847e0d..cb63cc7bbf06 100644 >> --- a/arch/s390/include/asm/kvm_host.h >> +++ b/arch/s390/include/asm/kvm_host.h >> @@ -36,7 +36,7 @@ >> */ >> #define KVM_NR_IRQCHIPS 1 >> #define KVM_IRQCHIP_NUM_PINS 4096 >> -#define KVM_HALT_POLL_NS_DEFAULT 80000 >> +#define KVM_HALT_POLL_NS_DEFAULT 50000 >> >> /* s390-specific vcpu->requests bit members */ >> #define KVM_REQ_ENABLE_IBS KVM_ARCH_REQ(0) >> > >
On Wed, 15 May 2019 10:23:24 +0200 Christian Borntraeger <borntraeger@de.ibm.com> wrote: > older performance measurements indicated that 50000 vs 80000 reduces cpu > consumption while still providing the benefit of halt polling. We had > this change in the IBM KVM product, but it got lost so it never went > upstream. Recent re-measurement indicate that 50k is still better than > 80k. > > Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> > --- > arch/s390/include/asm/kvm_host.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h > index dbe254847e0d..cb63cc7bbf06 100644 > --- a/arch/s390/include/asm/kvm_host.h > +++ b/arch/s390/include/asm/kvm_host.h > @@ -36,7 +36,7 @@ > */ > #define KVM_NR_IRQCHIPS 1 > #define KVM_IRQCHIP_NUM_PINS 4096 > -#define KVM_HALT_POLL_NS_DEFAULT 80000 > +#define KVM_HALT_POLL_NS_DEFAULT 50000 > > /* s390-specific vcpu->requests bit members */ > #define KVM_REQ_ENABLE_IBS KVM_ARCH_REQ(0) I trust your tests :) Acked-by: Cornelia Huck <cohuck@redhat.com>
diff --git a/arch/s390/include/asm/kvm_host.h b/arch/s390/include/asm/kvm_host.h index dbe254847e0d..cb63cc7bbf06 100644 --- a/arch/s390/include/asm/kvm_host.h +++ b/arch/s390/include/asm/kvm_host.h @@ -36,7 +36,7 @@ */ #define KVM_NR_IRQCHIPS 1 #define KVM_IRQCHIP_NUM_PINS 4096 -#define KVM_HALT_POLL_NS_DEFAULT 80000 +#define KVM_HALT_POLL_NS_DEFAULT 50000 /* s390-specific vcpu->requests bit members */ #define KVM_REQ_ENABLE_IBS KVM_ARCH_REQ(0)
older performance measurements indicated that 50000 vs 80000 reduces cpu consumption while still providing the benefit of halt polling. We had this change in the IBM KVM product, but it got lost so it never went upstream. Recent re-measurement indicate that 50k is still better than 80k. Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> --- arch/s390/include/asm/kvm_host.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)