* [PATCH 0/2] arm64: perf: Add L2 cache events for ARMv8 PMuv3 and A53 @ 2017-04-20 19:05 ` Florian Fainelli 0 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-20 19:05 UTC (permalink / raw) To: linux-arm-kernel Hi, While investigating why there were no L2 cache events generated for a Cortex A53-like PMU, it turned out that none of the L2 cache events were mapped. This is also the case for ARMv8 PMUv3, which seems a little odd considering they are defined. Thanks! Florian Fainelli (2): arm64: perf: Wire-up Cortex A53 L2 cache events and DTLB refills arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 arch/arm64/kernel/perf_event.c | 11 +++++++++++ 1 file changed, 11 insertions(+) -- 2.9.3 ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH 0/2] arm64: perf: Add L2 cache events for ARMv8 PMuv3 and A53 @ 2017-04-20 19:05 ` Florian Fainelli 0 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-20 19:05 UTC (permalink / raw) To: linux-arm-kernel Cc: alcooperx, opendmb, Florian Fainelli, Will Deacon, Mark Rutland, Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo, Alexander Shishkin, Catalin Marinas, open list:PERFORMANCE EVENTS SUBSYSTEM Hi, While investigating why there were no L2 cache events generated for a Cortex A53-like PMU, it turned out that none of the L2 cache events were mapped. This is also the case for ARMv8 PMUv3, which seems a little odd considering they are defined. Thanks! Florian Fainelli (2): arm64: perf: Wire-up Cortex A53 L2 cache events and DTLB refills arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 arch/arm64/kernel/perf_event.c | 11 +++++++++++ 1 file changed, 11 insertions(+) -- 2.9.3 ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH 1/2] arm64: perf: Wire-up Cortex A53 L2 cache events and DTLB refills 2017-04-20 19:05 ` Florian Fainelli @ 2017-04-20 19:05 ` Florian Fainelli -1 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-20 19:05 UTC (permalink / raw) To: linux-arm-kernel Add missing L2 cache events: read/write accesses and misses, as well as the DTLB refills. Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> --- arch/arm64/kernel/perf_event.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index 57ae9d9ed9bb..4f011cdd756d 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -290,6 +290,12 @@ static const unsigned armv8_a53_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + + [C(DTLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1D_TLB_REFILL, [C(ITLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_TLB_REFILL, [C(BPU)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_BR_PRED, -- 2.9.3 ^ permalink raw reply related [flat|nested] 18+ messages in thread
* [PATCH 1/2] arm64: perf: Wire-up Cortex A53 L2 cache events and DTLB refills @ 2017-04-20 19:05 ` Florian Fainelli 0 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-20 19:05 UTC (permalink / raw) To: linux-arm-kernel Cc: alcooperx, opendmb, Florian Fainelli, Will Deacon, Mark Rutland, Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo, Alexander Shishkin, Catalin Marinas, open list:PERFORMANCE EVENTS SUBSYSTEM Add missing L2 cache events: read/write accesses and misses, as well as the DTLB refills. Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> --- arch/arm64/kernel/perf_event.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index 57ae9d9ed9bb..4f011cdd756d 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -290,6 +290,12 @@ static const unsigned armv8_a53_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + + [C(DTLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1D_TLB_REFILL, [C(ITLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_TLB_REFILL, [C(BPU)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_BR_PRED, -- 2.9.3 ^ permalink raw reply related [flat|nested] 18+ messages in thread
* [PATCH 1/2] arm64: pmu: Wire-up Cortex A53 L2 cache events and DTLB refills 2017-04-20 19:05 ` Florian Fainelli @ 2017-04-20 19:05 ` Florian Fainelli -1 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-20 19:05 UTC (permalink / raw) To: linux-arm-kernel Add missing L2 cache events: read/write accesses and misses, as well as the DTLB refills. Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> --- arch/arm64/kernel/perf_event.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index 57ae9d9ed9bb..4f011cdd756d 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -290,6 +290,12 @@ static const unsigned armv8_a53_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + + [C(DTLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1D_TLB_REFILL, [C(ITLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_TLB_REFILL, [C(BPU)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_BR_PRED, -- 2.9.3 ^ permalink raw reply related [flat|nested] 18+ messages in thread
* [PATCH 1/2] arm64: pmu: Wire-up Cortex A53 L2 cache events and DTLB refills @ 2017-04-20 19:05 ` Florian Fainelli 0 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-20 19:05 UTC (permalink / raw) To: linux-arm-kernel Cc: alcooperx, opendmb, Florian Fainelli, Will Deacon, Mark Rutland, Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo, Alexander Shishkin, Catalin Marinas, open list:PERFORMANCE EVENTS SUBSYSTEM Add missing L2 cache events: read/write accesses and misses, as well as the DTLB refills. Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> --- arch/arm64/kernel/perf_event.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index 57ae9d9ed9bb..4f011cdd756d 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -290,6 +290,12 @@ static const unsigned armv8_a53_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + + [C(DTLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1D_TLB_REFILL, [C(ITLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_TLB_REFILL, [C(BPU)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_BR_PRED, -- 2.9.3 ^ permalink raw reply related [flat|nested] 18+ messages in thread
* [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 2017-04-20 19:05 ` Florian Fainelli @ 2017-04-20 19:05 ` Florian Fainelli -1 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-20 19:05 UTC (permalink / raw) To: linux-arm-kernel The ARMv8 PMUv3 cache map did not include the L2 cache events, add them. Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> --- arch/arm64/kernel/perf_event.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index 4f011cdd756d..a664c575f3fd 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + [C(DTLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1D_TLB_REFILL, [C(DTLB)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1D_TLB, -- 2.9.3 ^ permalink raw reply related [flat|nested] 18+ messages in thread
* [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 @ 2017-04-20 19:05 ` Florian Fainelli 0 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-20 19:05 UTC (permalink / raw) To: linux-arm-kernel Cc: alcooperx, opendmb, Florian Fainelli, Will Deacon, Mark Rutland, Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo, Alexander Shishkin, Catalin Marinas, open list:PERFORMANCE EVENTS SUBSYSTEM The ARMv8 PMUv3 cache map did not include the L2 cache events, add them. Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> --- arch/arm64/kernel/perf_event.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c index 4f011cdd756d..a664c575f3fd 100644 --- a/arch/arm64/kernel/perf_event.c +++ b/arch/arm64/kernel/perf_event.c @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, + [C(DTLB)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1D_TLB_REFILL, [C(DTLB)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1D_TLB, -- 2.9.3 ^ permalink raw reply related [flat|nested] 18+ messages in thread
* [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 2017-04-20 19:05 ` Florian Fainelli @ 2017-04-25 12:44 ` Will Deacon -1 siblings, 0 replies; 18+ messages in thread From: Will Deacon @ 2017-04-25 12:44 UTC (permalink / raw) To: linux-arm-kernel Hi Florian, On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: > The ARMv8 PMUv3 cache map did not include the L2 cache events, add > them. > > Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> > --- > arch/arm64/kernel/perf_event.c | 5 +++++ > 1 file changed, 5 insertions(+) > > diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c > index 4f011cdd756d..a664c575f3fd 100644 > --- a/arch/arm64/kernel/perf_event.c > +++ b/arch/arm64/kernel/perf_event.c > @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] > [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, > [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, > > + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, I don't think this is correct in general. 'LL' stands for "last-level", which may be L3 or even a system cache in the interconnect. Tying that to L2 is the wrong thing to do from perf's generic event perspective. I'm ok with what you're proposing for A53 (where the PMU can only count events out to the L2), but I'm reluctant to make this change for the generic PMUv3 events. Will ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 @ 2017-04-25 12:44 ` Will Deacon 0 siblings, 0 replies; 18+ messages in thread From: Will Deacon @ 2017-04-25 12:44 UTC (permalink / raw) To: Florian Fainelli Cc: linux-arm-kernel, alcooperx, opendmb, Mark Rutland, Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo, Alexander Shishkin, Catalin Marinas, open list:PERFORMANCE EVENTS SUBSYSTEM Hi Florian, On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: > The ARMv8 PMUv3 cache map did not include the L2 cache events, add > them. > > Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> > --- > arch/arm64/kernel/perf_event.c | 5 +++++ > 1 file changed, 5 insertions(+) > > diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c > index 4f011cdd756d..a664c575f3fd 100644 > --- a/arch/arm64/kernel/perf_event.c > +++ b/arch/arm64/kernel/perf_event.c > @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] > [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, > [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, > > + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, I don't think this is correct in general. 'LL' stands for "last-level", which may be L3 or even a system cache in the interconnect. Tying that to L2 is the wrong thing to do from perf's generic event perspective. I'm ok with what you're proposing for A53 (where the PMU can only count events out to the L2), but I'm reluctant to make this change for the generic PMUv3 events. Will ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 2017-04-25 12:44 ` Will Deacon @ 2017-04-25 17:13 ` Florian Fainelli -1 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-25 17:13 UTC (permalink / raw) To: linux-arm-kernel On 04/25/2017 05:44 AM, Will Deacon wrote: > Hi Florian, > > On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: >> The ARMv8 PMUv3 cache map did not include the L2 cache events, add >> them. >> >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> >> --- >> arch/arm64/kernel/perf_event.c | 5 +++++ >> 1 file changed, 5 insertions(+) >> >> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c >> index 4f011cdd756d..a664c575f3fd 100644 >> --- a/arch/arm64/kernel/perf_event.c >> +++ b/arch/arm64/kernel/perf_event.c >> @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] >> [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, >> [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, >> >> + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, >> + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, >> + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, >> + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > I don't think this is correct in general. 'LL' stands for "last-level", > which may be L3 or even a system cache in the interconnect. Tying that to L2 > is the wrong thing to do from perf's generic event perspective. > > I'm ok with what you're proposing for A53 (where the PMU can only count > events out to the L2), but I'm reluctant to make this change for the generic > PMUv3 events. That makes sense, shall I resubmit the first patch by itself or can you or Catalin take it as-is? Thanks! -- Florian ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 @ 2017-04-25 17:13 ` Florian Fainelli 0 siblings, 0 replies; 18+ messages in thread From: Florian Fainelli @ 2017-04-25 17:13 UTC (permalink / raw) To: Will Deacon Cc: linux-arm-kernel, alcooperx, opendmb, Mark Rutland, Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo, Alexander Shishkin, Catalin Marinas, open list:PERFORMANCE EVENTS SUBSYSTEM On 04/25/2017 05:44 AM, Will Deacon wrote: > Hi Florian, > > On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: >> The ARMv8 PMUv3 cache map did not include the L2 cache events, add >> them. >> >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> >> --- >> arch/arm64/kernel/perf_event.c | 5 +++++ >> 1 file changed, 5 insertions(+) >> >> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c >> index 4f011cdd756d..a664c575f3fd 100644 >> --- a/arch/arm64/kernel/perf_event.c >> +++ b/arch/arm64/kernel/perf_event.c >> @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] >> [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, >> [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, >> >> + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, >> + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, >> + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, >> + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > I don't think this is correct in general. 'LL' stands for "last-level", > which may be L3 or even a system cache in the interconnect. Tying that to L2 > is the wrong thing to do from perf's generic event perspective. > > I'm ok with what you're proposing for A53 (where the PMU can only count > events out to the L2), but I'm reluctant to make this change for the generic > PMUv3 events. That makes sense, shall I resubmit the first patch by itself or can you or Catalin take it as-is? Thanks! -- Florian ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 2017-04-25 17:13 ` Florian Fainelli @ 2017-04-27 17:36 ` Will Deacon -1 siblings, 0 replies; 18+ messages in thread From: Will Deacon @ 2017-04-27 17:36 UTC (permalink / raw) To: linux-arm-kernel On Tue, Apr 25, 2017 at 10:13:51AM -0700, Florian Fainelli wrote: > On 04/25/2017 05:44 AM, Will Deacon wrote: > > Hi Florian, > > > > On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: > >> The ARMv8 PMUv3 cache map did not include the L2 cache events, add > >> them. > >> > >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> > >> --- > >> arch/arm64/kernel/perf_event.c | 5 +++++ > >> 1 file changed, 5 insertions(+) > >> > >> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c > >> index 4f011cdd756d..a664c575f3fd 100644 > >> --- a/arch/arm64/kernel/perf_event.c > >> +++ b/arch/arm64/kernel/perf_event.c > >> @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] > >> [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, > >> [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, > >> > >> + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > >> + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > >> + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > >> + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > > > I don't think this is correct in general. 'LL' stands for "last-level", > > which may be L3 or even a system cache in the interconnect. Tying that to L2 > > is the wrong thing to do from perf's generic event perspective. > > > > I'm ok with what you're proposing for A53 (where the PMU can only count > > events out to the L2), but I'm reluctant to make this change for the generic > > PMUv3 events. > > That makes sense, shall I resubmit the first patch by itself or can you > or Catalin take it as-is? I'll talk to Catalin tomorrow and try to get the A53 bit queued. Will ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 @ 2017-04-27 17:36 ` Will Deacon 0 siblings, 0 replies; 18+ messages in thread From: Will Deacon @ 2017-04-27 17:36 UTC (permalink / raw) To: Florian Fainelli Cc: linux-arm-kernel, alcooperx, opendmb, Mark Rutland, Peter Zijlstra, Ingo Molnar, Arnaldo Carvalho de Melo, Alexander Shishkin, Catalin Marinas, open list:PERFORMANCE EVENTS SUBSYSTEM On Tue, Apr 25, 2017 at 10:13:51AM -0700, Florian Fainelli wrote: > On 04/25/2017 05:44 AM, Will Deacon wrote: > > Hi Florian, > > > > On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: > >> The ARMv8 PMUv3 cache map did not include the L2 cache events, add > >> them. > >> > >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> > >> --- > >> arch/arm64/kernel/perf_event.c | 5 +++++ > >> 1 file changed, 5 insertions(+) > >> > >> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c > >> index 4f011cdd756d..a664c575f3fd 100644 > >> --- a/arch/arm64/kernel/perf_event.c > >> +++ b/arch/arm64/kernel/perf_event.c > >> @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] > >> [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, > >> [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, > >> > >> + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > >> + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > >> + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > >> + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > > > I don't think this is correct in general. 'LL' stands for "last-level", > > which may be L3 or even a system cache in the interconnect. Tying that to L2 > > is the wrong thing to do from perf's generic event perspective. > > > > I'm ok with what you're proposing for A53 (where the PMU can only count > > events out to the L2), but I'm reluctant to make this change for the generic > > PMUv3 events. > > That makes sense, shall I resubmit the first patch by itself or can you > or Catalin take it as-is? I'll talk to Catalin tomorrow and try to get the A53 bit queued. Will ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 2017-04-27 17:36 ` Will Deacon @ 2017-04-28 14:15 ` Catalin Marinas -1 siblings, 0 replies; 18+ messages in thread From: Catalin Marinas @ 2017-04-28 14:15 UTC (permalink / raw) To: linux-arm-kernel On Thu, Apr 27, 2017 at 06:36:42PM +0100, Will Deacon wrote: > On Tue, Apr 25, 2017 at 10:13:51AM -0700, Florian Fainelli wrote: > > On 04/25/2017 05:44 AM, Will Deacon wrote: > > > On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: > > >> The ARMv8 PMUv3 cache map did not include the L2 cache events, add > > >> them. > > >> > > >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> > > >> --- > > >> arch/arm64/kernel/perf_event.c | 5 +++++ > > >> 1 file changed, 5 insertions(+) > > >> > > >> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c > > >> index 4f011cdd756d..a664c575f3fd 100644 > > >> --- a/arch/arm64/kernel/perf_event.c > > >> +++ b/arch/arm64/kernel/perf_event.c > > >> @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] > > >> [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, > > >> [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, > > >> > > >> + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > > >> + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > >> + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > > >> + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > > > > > I don't think this is correct in general. 'LL' stands for "last-level", > > > which may be L3 or even a system cache in the interconnect. Tying that to L2 > > > is the wrong thing to do from perf's generic event perspective. > > > > > > I'm ok with what you're proposing for A53 (where the PMU can only count > > > events out to the L2), but I'm reluctant to make this change for the generic > > > PMUv3 events. > > > > That makes sense, shall I resubmit the first patch by itself or can you > > or Catalin take it as-is? > > I'll talk to Catalin tomorrow and try to get the A53 bit queued. I queued patch 1/2. Shall I add your ack? -- Catalin ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 @ 2017-04-28 14:15 ` Catalin Marinas 0 siblings, 0 replies; 18+ messages in thread From: Catalin Marinas @ 2017-04-28 14:15 UTC (permalink / raw) To: Will Deacon Cc: Florian Fainelli, Mark Rutland, opendmb, Peter Zijlstra, alcooperx, open list:PERFORMANCE EVENTS SUBSYSTEM, Arnaldo Carvalho de Melo, Alexander Shishkin, Ingo Molnar, linux-arm-kernel On Thu, Apr 27, 2017 at 06:36:42PM +0100, Will Deacon wrote: > On Tue, Apr 25, 2017 at 10:13:51AM -0700, Florian Fainelli wrote: > > On 04/25/2017 05:44 AM, Will Deacon wrote: > > > On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: > > >> The ARMv8 PMUv3 cache map did not include the L2 cache events, add > > >> them. > > >> > > >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> > > >> --- > > >> arch/arm64/kernel/perf_event.c | 5 +++++ > > >> 1 file changed, 5 insertions(+) > > >> > > >> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c > > >> index 4f011cdd756d..a664c575f3fd 100644 > > >> --- a/arch/arm64/kernel/perf_event.c > > >> +++ b/arch/arm64/kernel/perf_event.c > > >> @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] > > >> [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, > > >> [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, > > >> > > >> + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > > >> + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > >> + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > > >> + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > > > > > I don't think this is correct in general. 'LL' stands for "last-level", > > > which may be L3 or even a system cache in the interconnect. Tying that to L2 > > > is the wrong thing to do from perf's generic event perspective. > > > > > > I'm ok with what you're proposing for A53 (where the PMU can only count > > > events out to the L2), but I'm reluctant to make this change for the generic > > > PMUv3 events. > > > > That makes sense, shall I resubmit the first patch by itself or can you > > or Catalin take it as-is? > > I'll talk to Catalin tomorrow and try to get the A53 bit queued. I queued patch 1/2. Shall I add your ack? -- Catalin ^ permalink raw reply [flat|nested] 18+ messages in thread
* [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 2017-04-28 14:15 ` Catalin Marinas @ 2017-04-28 14:17 ` Will Deacon -1 siblings, 0 replies; 18+ messages in thread From: Will Deacon @ 2017-04-28 14:17 UTC (permalink / raw) To: linux-arm-kernel On Fri, Apr 28, 2017 at 03:15:01PM +0100, Catalin Marinas wrote: > On Thu, Apr 27, 2017 at 06:36:42PM +0100, Will Deacon wrote: > > On Tue, Apr 25, 2017 at 10:13:51AM -0700, Florian Fainelli wrote: > > > On 04/25/2017 05:44 AM, Will Deacon wrote: > > > > On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: > > > >> The ARMv8 PMUv3 cache map did not include the L2 cache events, add > > > >> them. > > > >> > > > >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> > > > >> --- > > > >> arch/arm64/kernel/perf_event.c | 5 +++++ > > > >> 1 file changed, 5 insertions(+) > > > >> > > > >> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c > > > >> index 4f011cdd756d..a664c575f3fd 100644 > > > >> --- a/arch/arm64/kernel/perf_event.c > > > >> +++ b/arch/arm64/kernel/perf_event.c > > > >> @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] > > > >> [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, > > > >> [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, > > > >> > > > >> + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > > > >> + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > > >> + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > > > >> + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > > > > > > > I don't think this is correct in general. 'LL' stands for "last-level", > > > > which may be L3 or even a system cache in the interconnect. Tying that to L2 > > > > is the wrong thing to do from perf's generic event perspective. > > > > > > > > I'm ok with what you're proposing for A53 (where the PMU can only count > > > > events out to the L2), but I'm reluctant to make this change for the generic > > > > PMUv3 events. > > > > > > That makes sense, shall I resubmit the first patch by itself or can you > > > or Catalin take it as-is? > > > > I'll talk to Catalin tomorrow and try to get the A53 bit queued. > > I queued patch 1/2. Shall I add your ack? Yes, please. Will ^ permalink raw reply [flat|nested] 18+ messages in thread
* Re: [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 @ 2017-04-28 14:17 ` Will Deacon 0 siblings, 0 replies; 18+ messages in thread From: Will Deacon @ 2017-04-28 14:17 UTC (permalink / raw) To: Catalin Marinas Cc: Florian Fainelli, Mark Rutland, opendmb, Peter Zijlstra, alcooperx, open list:PERFORMANCE EVENTS SUBSYSTEM, Arnaldo Carvalho de Melo, Alexander Shishkin, Ingo Molnar, linux-arm-kernel On Fri, Apr 28, 2017 at 03:15:01PM +0100, Catalin Marinas wrote: > On Thu, Apr 27, 2017 at 06:36:42PM +0100, Will Deacon wrote: > > On Tue, Apr 25, 2017 at 10:13:51AM -0700, Florian Fainelli wrote: > > > On 04/25/2017 05:44 AM, Will Deacon wrote: > > > > On Thu, Apr 20, 2017 at 12:05:46PM -0700, Florian Fainelli wrote: > > > >> The ARMv8 PMUv3 cache map did not include the L2 cache events, add > > > >> them. > > > >> > > > >> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com> > > > >> --- > > > >> arch/arm64/kernel/perf_event.c | 5 +++++ > > > >> 1 file changed, 5 insertions(+) > > > >> > > > >> diff --git a/arch/arm64/kernel/perf_event.c b/arch/arm64/kernel/perf_event.c > > > >> index 4f011cdd756d..a664c575f3fd 100644 > > > >> --- a/arch/arm64/kernel/perf_event.c > > > >> +++ b/arch/arm64/kernel/perf_event.c > > > >> @@ -264,6 +264,11 @@ static const unsigned armv8_pmuv3_perf_cache_map[PERF_COUNT_HW_CACHE_MAX] > > > >> [C(L1I)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE, > > > >> [C(L1I)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L1I_CACHE_REFILL, > > > >> > > > >> + [C(LL)][C(OP_READ)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > > > >> + [C(LL)][C(OP_READ)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > > >> + [C(LL)][C(OP_WRITE)][C(RESULT_ACCESS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE, > > > >> + [C(LL)][C(OP_WRITE)][C(RESULT_MISS)] = ARMV8_PMUV3_PERFCTR_L2D_CACHE_REFILL, > > > > > > > > I don't think this is correct in general. 'LL' stands for "last-level", > > > > which may be L3 or even a system cache in the interconnect. Tying that to L2 > > > > is the wrong thing to do from perf's generic event perspective. > > > > > > > > I'm ok with what you're proposing for A53 (where the PMU can only count > > > > events out to the L2), but I'm reluctant to make this change for the generic > > > > PMUv3 events. > > > > > > That makes sense, shall I resubmit the first patch by itself or can you > > > or Catalin take it as-is? > > > > I'll talk to Catalin tomorrow and try to get the A53 bit queued. > > I queued patch 1/2. Shall I add your ack? Yes, please. Will ^ permalink raw reply [flat|nested] 18+ messages in thread
end of thread, other threads:[~2017-04-28 14:17 UTC | newest] Thread overview: 18+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2017-04-20 19:05 [PATCH 0/2] arm64: perf: Add L2 cache events for ARMv8 PMuv3 and A53 Florian Fainelli 2017-04-20 19:05 ` Florian Fainelli 2017-04-20 19:05 ` [PATCH 1/2] arm64: perf: Wire-up Cortex A53 L2 cache events and DTLB refills Florian Fainelli 2017-04-20 19:05 ` Florian Fainelli 2017-04-20 19:05 ` [PATCH 1/2] arm64: pmu: " Florian Fainelli 2017-04-20 19:05 ` Florian Fainelli 2017-04-20 19:05 ` [PATCH 2/2] arm64: pmu: Wire-up L2 cache events for ARMv8 PMUv3 Florian Fainelli 2017-04-20 19:05 ` Florian Fainelli 2017-04-25 12:44 ` Will Deacon 2017-04-25 12:44 ` Will Deacon 2017-04-25 17:13 ` Florian Fainelli 2017-04-25 17:13 ` Florian Fainelli 2017-04-27 17:36 ` Will Deacon 2017-04-27 17:36 ` Will Deacon 2017-04-28 14:15 ` Catalin Marinas 2017-04-28 14:15 ` Catalin Marinas 2017-04-28 14:17 ` Will Deacon 2017-04-28 14:17 ` Will Deacon
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.