perf vendor events intel: Alderlake/rocketlake metric fixes
Ian Rogers
reviewerPATCH 1/4 perf vendor events intel: Alderlake/rocketlake metric fixes
Fix that the core PMU is being specified for 2 uncore events. Specify a PMU for the alderlake UNCORE_FREQ metric.
Conversion script updated in:
https://github.com/intel/perfmon/pull/126
--- .../arch/x86/alderlake/adl-metrics.json | 15 ++++++++------- .../arch/x86/rocketlake/rkl-metrics.json | 2 +- 2 files changed, 9 insertions(+), 8 deletions(-)
tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json +8 -7
@@ -69,12 +69,6 @@ "MetricName": "C9_Pkg_Residency", "ScaleUnit": "100%" }, - { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, { "BriefDescription": "Percentage of cycles spent in System Management Interrupts.", "MetricExpr": "((msr@aperf@ - cycles) / msr@aperf@ if msr@smi@ > 0 else 0)",
@@ -809,6 +803,13 @@ "ScaleUnit": "100%", "Unit": "cpu_atom" }, + { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ", + "Unit": "cpu_core" + }, { "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", "MetricExpr": "(cpu_core@UOPS_DISPATCHED.PORT_0@ + cpu_core@UOPS_DISPATCHED.PORT_1@ + cpu_core@UOPS_DISPATCHED.PORT_5_11@ + cpu_core@UOPS_DISPATCHED.PORT_6@) / (5 * tma_info_core_core_clks)",
@@ -1838,7 +1839,7 @@ }, { "BriefDescription": "Average number of parallel data read requests to external memory", - "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu_core@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@", + "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@", "MetricGroup": "Mem;MemoryBW;SoC", "MetricName": "tma_info_system_mem_parallel_reads", "PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches",
tools/perf/pmu-events/arch/x86/rocketlake/rkl-metrics.json +1 -1
@@ -985,7 +985,7 @@ }, { "BriefDescription": "Average number of parallel data read requests to external memory", - "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@", + "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@", "MetricGroup": "Mem;MemoryBW;SoC", "MetricName": "tma_info_system_mem_parallel_reads", "PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches"
Signature · 1 line
2.43.0.472.g3155946c3a-goog
Ian Rogers
reviewerPATCH 2/4 perf vendor events intel: Update emeraldrapids events to v1.02
Update to v1.02 released in:
https://github.com/intel/perfmon/pull/123
Removes events AMX_OPS_RETIRED.BF16 and AMX_OPS_RETIRED.INT8. Add events FP_ARITH_DISPATCHED.V0, FP_ARITH_DISPATCHED.V1, FP_ARITH_DISPATCHED.V2, UNC_IIO_IOMMU0.1G_HITS, UNC_IIO_IOMMU0.2M_HITS and UNC_IIO_IOMMU0.4K_HITS. Description updates.
--- .../x86/emeraldrapids/floating-point.json | 27 +++++++++++++++-- .../arch/x86/emeraldrapids/pipeline.json | 18 +---------- .../emeraldrapids/uncore-interconnect.json | 8 ++--- .../arch/x86/emeraldrapids/uncore-io.json | 30 +++++++++++++++++++ tools/perf/pmu-events/arch/x86/mapfile.csv | 2 +- 5 files changed, 60 insertions(+), 25 deletions(-)
tools/perf/pmu-events/arch/x86/emeraldrapids/floating-point.json +24 -3
@@ -23,26 +23,47 @@ "UMask": "0x10" }, { - "BriefDescription": "FP_ARITH_DISPATCHED.PORT_0", + "BriefDescription": "FP_ARITH_DISPATCHED.PORT_0 [This event is alias to FP_ARITH_DISPATCHED.V0]", "EventCode": "0xb3", "EventName": "FP_ARITH_DISPATCHED.PORT_0", "SampleAfterValue": "2000003", "UMask": "0x1" }, { - "BriefDescription": "FP_ARITH_DISPATCHED.PORT_1", + "BriefDescription": "FP_ARITH_DISPATCHED.PORT_1 [This event is alias to FP_ARITH_DISPATCHED.V1]", "EventCode": "0xb3", "EventName": "FP_ARITH_DISPATCHED.PORT_1", "SampleAfterValue": "2000003", "UMask": "0x2" }, { - "BriefDescription": "FP_ARITH_DISPATCHED.PORT_5", + "BriefDescription": "FP_ARITH_DISPATCHED.PORT_5 [This event is alias to FP_ARITH_DISPATCHED.V2]", "EventCode": "0xb3", "EventName": "FP_ARITH_DISPATCHED.PORT_5", "SampleAfterValue": "2000003", "UMask": "0x4" }, + { + "BriefDescription": "FP_ARITH_DISPATCHED.V0 [This event is alias to FP_ARITH_DISPATCHED.PORT_0]", + "EventCode": "0xb3", + "EventName": "FP_ARITH_DISPATCHED.V0", + "SampleAfterValue": "2000003", + "UMask": "0x1" + }, + { + "BriefDescription": "FP_ARITH_DISPATCHED.V1 [This event is alias to FP_ARITH_DISPATCHED.PORT_1]", + "EventCode": "0xb3", + "EventName": "FP_ARITH_DISPATCHED.V1", + "SampleAfterValue": "2000003", + "UMask": "0x2" + }, + { + "BriefDescription": "FP_ARITH_DISPATCHED.V2 [This event is alias to FP_ARITH_DISPATCHED.PORT_5]", + "EventCode": "0xb3", + "EventName": "FP_ARITH_DISPATCHED.V2", + "SampleAfterValue": "2000003", + "UMask": "0x4" + }, { "BriefDescription": "Counts number of SSE/AVX computational 128-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below. Each count represents 2 computation operations, one for each element. Applies to SSE* and AVX* packed double precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP FM(N)ADD/SUB. DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element.", "EventCode": "0xc7",
tools/perf/pmu-events/arch/x86/emeraldrapids/pipeline.json +1 -17
@@ -1,20 +1,4 @@ [ - { - "BriefDescription": "AMX retired arithmetic BF16 operations.", - "EventCode": "0xce", - "EventName": "AMX_OPS_RETIRED.BF16", - "PublicDescription": "Number of AMX-based retired arithmetic bfloat16 (BF16) floating-point operations. Counts TDPBF16PS FP instructions. SW to use operation multiplier of 4", - "SampleAfterValue": "1000003", - "UMask": "0x2" - }, - { - "BriefDescription": "AMX retired arithmetic integer 8-bit operations.", - "EventCode": "0xce", - "EventName": "AMX_OPS_RETIRED.INT8", - "PublicDescription": "Number of AMX-based retired arithmetic integer operations of 8-bit width source operands. Counts TDPB[SS,UU,US,SU]D instructions. SW should use operation multiplier of 8.", - "SampleAfterValue": "1000003", - "UMask": "0x1" - }, { "BriefDescription": "This event is deprecated. Refer to new event ARITH.DIV_ACTIVE", "CounterMask": "1",
@@ -505,7 +489,7 @@ "UMask": "0x1" }, { - "BriefDescription": "INT_MISC.UNKNOWN_BRANCH_CYCLES", + "BriefDescription": "Bubble cycles of BAClear (Unknown Branch).", "EventCode": "0xad", "EventName": "INT_MISC.UNKNOWN_BRANCH_CYCLES", "MSRIndex": "0x3F7",
tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-interconnect.json +4 -4
@@ -4825,11 +4825,11 @@ "Unit": "M3UPI" }, { - "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (AD Bouncable)", + "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (AD Bounceable)", "EventCode": "0x47", "EventName": "UNC_MDF_CRS_TxR_INSERTS.AD_BNC", "PerPkg": "1", - "PublicDescription": "AD Bouncable : Number of allocations into the CRS Egress", + "PublicDescription": "AD Bounceable : Number of allocations into the CRS Egress", "UMask": "0x1", "Unit": "MDF" },
@@ -4861,11 +4861,11 @@ "Unit": "MDF" }, { - "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (BL Bouncable)", + "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (BL Bounceable)", "EventCode": "0x47", "EventName": "UNC_MDF_CRS_TxR_INSERTS.BL_BNC", "PerPkg": "1", - "PublicDescription": "BL Bouncable : Number of allocations into the CRS Egress", + "PublicDescription": "BL Bounceable : Number of allocations into the CRS Egress", "UMask": "0x4", "Unit": "MDF" },
tools/perf/pmu-events/arch/x86/emeraldrapids/uncore-io.json +30 -0
@@ -1185,6 +1185,36 @@ "UMask": "0x70ff010", "Unit": "IIO" }, + { + "BriefDescription": ": IOTLB Hits to a 1G Page", + "EventCode": "0x40", + "EventName": "UNC_IIO_IOMMU0.1G_HITS", + "PerPkg": "1", + "PortMask": "0x0000", + "PublicDescription": ": IOTLB Hits to a 1G Page : Counts if a transaction to a 1G page, on its first lookup, hits the IOTLB.", + "UMask": "0x10", + "Unit": "IIO" + }, + { + "BriefDescription": ": IOTLB Hits to a 2M Page", + "EventCode": "0x40", + "EventName": "UNC_IIO_IOMMU0.2M_HITS", + "PerPkg": "1", + "PortMask": "0x0000", + "PublicDescription": ": IOTLB Hits to a 2M Page : Counts if a transaction to a 2M page, on its first lookup, hits the IOTLB.", + "UMask": "0x8", + "Unit": "IIO" + }, + { + "BriefDescription": ": IOTLB Hits to a 4K Page", + "EventCode": "0x40", + "EventName": "UNC_IIO_IOMMU0.4K_HITS", + "PerPkg": "1", + "PortMask": "0x0000", + "PublicDescription": ": IOTLB Hits to a 4K Page : Counts if a transaction to a 4K page, on its first lookup, hits the IOTLB.", + "UMask": "0x4", + "Unit": "IIO" + }, { "BriefDescription": ": Context cache hits", "EventCode": "0x40",
tools/perf/pmu-events/arch/x86/mapfile.csv +1 -1
@@ -7,7 +7,7 @@ GenuineIntel-6-56,v11,broadwellde,core GenuineIntel-6-4F,v22,broadwellx,core GenuineIntel-6-55-[56789ABCDEF],v1.20,cascadelakex,core GenuineIntel-6-9[6C],v1.04,elkhartlake,core -GenuineIntel-6-CF,v1.01,emeraldrapids,core +GenuineIntel-6-CF,v1.02,emeraldrapids,core GenuineIntel-6-5[CF],v13,goldmont,core GenuineIntel-6-7A,v1.01,goldmontplus,core GenuineIntel-6-B6,v1.00,grandridge,core
Signature · 1 line
2.43.0.472.g3155946c3a-goog
Ian Rogers
reviewerPATCH 3/4 perf vendor events intel: Update icelakex events to v1.23
In reply to Ian RogersUpdate to v1.23 released in:
https://github.com/intel/perfmon/pull/123
Updates to event descriptions.
--- tools/perf/pmu-events/arch/x86/icelakex/other.json | 2 +- tools/perf/pmu-events/arch/x86/icelakex/pipeline.json | 2 +- .../pmu-events/arch/x86/icelakex/uncore-interconnect.json | 6 +++--- tools/perf/pmu-events/arch/x86/mapfile.csv | 2 +- 4 files changed, 6 insertions(+), 6 deletions(-)
tools/perf/pmu-events/arch/x86/icelakex/other.json +1 -1
@@ -19,7 +19,7 @@ "BriefDescription": "Core cycles where the core was running in a manner where Turbo may be clipped to the AVX512 turbo schedule.", "EventCode": "0x28", "EventName": "CORE_POWER.LVL2_TURBO_LICENSE", - "PublicDescription": "Core cycles where the core was running with power-delivery for license level 2 (introduced in Skylake Server microarchtecture). This includes high current AVX 512-bit instructions.", + "PublicDescription": "Core cycles where the core was running with power-delivery for license level 2 (introduced in Skylake Server microarchitecture). This includes high current AVX 512-bit instructions.", "SampleAfterValue": "200003", "UMask": "0x20" },
tools/perf/pmu-events/arch/x86/icelakex/pipeline.json +1 -1
@@ -519,7 +519,7 @@ "BriefDescription": "Cycles when Reservation Station (RS) is empty for the thread", "EventCode": "0x5e", "EventName": "RS_EVENTS.EMPTY_CYCLES", - "PublicDescription": "Counts cycles during which the reservation station (RS) is empty for this logical processor. This is usually caused when the front-end pipeline runs into stravation periods (e.g. branch mispredictions or i-cache misses)", + "PublicDescription": "Counts cycles during which the reservation station (RS) is empty for this logical processor. This is usually caused when the front-end pipeline runs into starvation periods (e.g. branch mispredictions or i-cache misses)", "SampleAfterValue": "1000003", "UMask": "0x1" },
tools/perf/pmu-events/arch/x86/icelakex/uncore-interconnect.json +3 -3
@@ -38,7 +38,7 @@ "EventCode": "0x10", "EventName": "UNC_I_COHERENT_OPS.CLFLUSH", "PerPkg": "1", - "PublicDescription": "Coherent Ops : CLFlush : Counts the number of coherency related operations servied by the IRP", + "PublicDescription": "Coherent Ops : CLFlush : Counts the number of coherency related operations serviced by the IRP", "UMask": "0x80", "Unit": "IRP" },
@@ -65,7 +65,7 @@ "EventCode": "0x10", "EventName": "UNC_I_COHERENT_OPS.WBMTOI", "PerPkg": "1", - "PublicDescription": "Coherent Ops : WbMtoI : Counts the number of coherency related operations servied by the IRP", + "PublicDescription": "Coherent Ops : WbMtoI : Counts the number of coherency related operations serviced by the IRP", "UMask": "0x40", "Unit": "IRP" },
@@ -454,7 +454,7 @@ "EventCode": "0x11", "EventName": "UNC_I_TRANSACTIONS.WRITES", "PerPkg": "1", - "PublicDescription": "Inbound Transaction Count : Writes : Counts the number of Inbound transactions from the IRP to the Uncore. This can be filtered based on request type in addition to the source queue. Note the special filtering equation. We do OR-reduction on the request type. If the SOURCE bit is set, then we also do AND qualification based on the source portID. : Trackes only write requests. Each write request should have a prefetch, so there is no need to explicitly track these requests. For writes that are tickled and have to retry, the counter will be incremented for each retry.", + "PublicDescription": "Inbound Transaction Count : Writes : Counts the number of Inbound transactions from the IRP to the Uncore. This can be filtered based on request type in addition to the source queue. Note the special filtering equation. We do OR-reduction on the request type. If the SOURCE bit is set, then we also do AND qualification based on the source portID. : Tracks only write requests. Each write request should have a prefetch, so there is no need to explicitly track these requests. For writes that are tickled and have to retry, the counter will be incremented for each retry.", "UMask": "0x2", "Unit": "IRP" },
tools/perf/pmu-events/arch/x86/mapfile.csv +1 -1
@@ -15,7 +15,7 @@ GenuineIntel-6-A[DE],v1.01,graniterapids,core GenuineIntel-6-(3C|45|46),v33,haswell,core GenuineIntel-6-3F,v28,haswellx,core GenuineIntel-6-7[DE],v1.19,icelake,core -GenuineIntel-6-6[AC],v1.21,icelakex,core +GenuineIntel-6-6[AC],v1.23,icelakex,core GenuineIntel-6-3A,v24,ivybridge,core GenuineIntel-6-3E,v24,ivytown,core GenuineIntel-6-2D,v24,jaketown,core
Signature · 1 line
2.43.0.472.g3155946c3a-goog
Ian Rogers
reviewerPATCH 4/4 perf vendor events intel: Update sapphirerapids events to v1.17
In reply to Ian RogersUpdate to v1.17 released in:
https://github.com/intel/perfmon/pull/123
Add events FP_ARITH_DISPATCHED.V0, FP_ARITH_DISPATCHED.V1, FP_ARITH_DISPATCHED.V2, UNC_IIO_IOMMU0.1G_HITS, UNC_IIO_IOMMU0.2M_HITS and UNC_IIO_IOMMU0.4K_HITS. Description updates.
--- tools/perf/pmu-events/arch/x86/mapfile.csv | 2 +- .../x86/sapphirerapids/floating-point.json | 27 +++++++++++++++-- .../arch/x86/sapphirerapids/pipeline.json | 2 +- .../sapphirerapids/uncore-interconnect.json | 8 ++--- .../arch/x86/sapphirerapids/uncore-io.json | 30 +++++++++++++++++++ 5 files changed, 60 insertions(+), 9 deletions(-)
tools/perf/pmu-events/arch/x86/mapfile.csv +1 -1
@@ -26,7 +26,7 @@ GenuineIntel-6-1[AEF],v4,nehalemep,core GenuineIntel-6-2E,v4,nehalemex,core GenuineIntel-6-A7,v1.01,rocketlake,core GenuineIntel-6-2A,v19,sandybridge,core -GenuineIntel-6-8F,v1.16,sapphirerapids,core +GenuineIntel-6-8F,v1.17,sapphirerapids,core GenuineIntel-6-AF,v1.00,sierraforest,core GenuineIntel-6-(37|4A|4C|4D|5A),v15,silvermont,core GenuineIntel-6-(4E|5E|8E|9E|A5|A6),v57,skylake,core
tools/perf/pmu-events/arch/x86/sapphirerapids/floating-point.json +24 -3
@@ -23,26 +23,47 @@ "UMask": "0x10" }, { - "BriefDescription": "FP_ARITH_DISPATCHED.PORT_0", + "BriefDescription": "FP_ARITH_DISPATCHED.PORT_0 [This event is alias to FP_ARITH_DISPATCHED.V0]", "EventCode": "0xb3", "EventName": "FP_ARITH_DISPATCHED.PORT_0", "SampleAfterValue": "2000003", "UMask": "0x1" }, { - "BriefDescription": "FP_ARITH_DISPATCHED.PORT_1", + "BriefDescription": "FP_ARITH_DISPATCHED.PORT_1 [This event is alias to FP_ARITH_DISPATCHED.V1]", "EventCode": "0xb3", "EventName": "FP_ARITH_DISPATCHED.PORT_1", "SampleAfterValue": "2000003", "UMask": "0x2" }, { - "BriefDescription": "FP_ARITH_DISPATCHED.PORT_5", + "BriefDescription": "FP_ARITH_DISPATCHED.PORT_5 [This event is alias to FP_ARITH_DISPATCHED.V2]", "EventCode": "0xb3", "EventName": "FP_ARITH_DISPATCHED.PORT_5", "SampleAfterValue": "2000003", "UMask": "0x4" }, + { + "BriefDescription": "FP_ARITH_DISPATCHED.V0 [This event is alias to FP_ARITH_DISPATCHED.PORT_0]", + "EventCode": "0xb3", + "EventName": "FP_ARITH_DISPATCHED.V0", + "SampleAfterValue": "2000003", + "UMask": "0x1" + }, + { + "BriefDescription": "FP_ARITH_DISPATCHED.V1 [This event is alias to FP_ARITH_DISPATCHED.PORT_1]", + "EventCode": "0xb3", + "EventName": "FP_ARITH_DISPATCHED.V1", + "SampleAfterValue": "2000003", + "UMask": "0x2" + }, + { + "BriefDescription": "FP_ARITH_DISPATCHED.V2 [This event is alias to FP_ARITH_DISPATCHED.PORT_5]", + "EventCode": "0xb3", + "EventName": "FP_ARITH_DISPATCHED.V2", + "SampleAfterValue": "2000003", + "UMask": "0x4" + }, { "BriefDescription": "Counts number of SSE/AVX computational 128-bit packed double precision floating-point instructions retired; some instructions will count twice as noted below. Each count represents 2 computation operations, one for each element. Applies to SSE* and AVX* packed double precision floating-point instructions: ADD SUB HADD HSUB SUBADD MUL DIV MIN MAX SQRT DPP FM(N)ADD/SUB. DPP and FM(N)ADD/SUB instructions count twice as they perform 2 calculations per element.", "EventCode": "0xc7",
tools/perf/pmu-events/arch/x86/sapphirerapids/pipeline.json +1 -1
@@ -505,7 +505,7 @@ "UMask": "0x1" }, { - "BriefDescription": "INT_MISC.UNKNOWN_BRANCH_CYCLES", + "BriefDescription": "Bubble cycles of BAClear (Unknown Branch).", "EventCode": "0xad", "EventName": "INT_MISC.UNKNOWN_BRANCH_CYCLES", "MSRIndex": "0x3F7",
tools/perf/pmu-events/arch/x86/sapphirerapids/uncore-interconnect.json +4 -4
@@ -4825,11 +4825,11 @@ "Unit": "M3UPI" }, { - "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (AD Bouncable)", + "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (AD Bounceable)", "EventCode": "0x47", "EventName": "UNC_MDF_CRS_TxR_INSERTS.AD_BNC", "PerPkg": "1", - "PublicDescription": "AD Bouncable : Number of allocations into the CRS Egress", + "PublicDescription": "AD Bounceable : Number of allocations into the CRS Egress", "UMask": "0x1", "Unit": "MDF" },
@@ -4861,11 +4861,11 @@ "Unit": "MDF" }, { - "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (BL Bouncable)", + "BriefDescription": "Number of allocations into the CRS Egress used to queue up requests destined to the mesh (BL Bounceable)", "EventCode": "0x47", "EventName": "UNC_MDF_CRS_TxR_INSERTS.BL_BNC", "PerPkg": "1", - "PublicDescription": "BL Bouncable : Number of allocations into the CRS Egress", + "PublicDescription": "BL Bounceable : Number of allocations into the CRS Egress", "UMask": "0x4", "Unit": "MDF" },
tools/perf/pmu-events/arch/x86/sapphirerapids/uncore-io.json +30 -0
@@ -1249,6 +1249,36 @@ "UMask": "0x70ff010", "Unit": "IIO" }, + { + "BriefDescription": ": IOTLB Hits to a 1G Page", + "EventCode": "0x40", + "EventName": "UNC_IIO_IOMMU0.1G_HITS", + "PerPkg": "1", + "PortMask": "0x0000", + "PublicDescription": ": IOTLB Hits to a 1G Page : Counts if a transaction to a 1G page, on its first lookup, hits the IOTLB.", + "UMask": "0x10", + "Unit": "IIO" + }, + { + "BriefDescription": ": IOTLB Hits to a 2M Page", + "EventCode": "0x40", + "EventName": "UNC_IIO_IOMMU0.2M_HITS", + "PerPkg": "1", + "PortMask": "0x0000", + "PublicDescription": ": IOTLB Hits to a 2M Page : Counts if a transaction to a 2M page, on its first lookup, hits the IOTLB.", + "UMask": "0x8", + "Unit": "IIO" + }, + { + "BriefDescription": ": IOTLB Hits to a 4K Page", + "EventCode": "0x40", + "EventName": "UNC_IIO_IOMMU0.4K_HITS", + "PerPkg": "1", + "PortMask": "0x0000", + "PublicDescription": ": IOTLB Hits to a 4K Page : Counts if a transaction to a 4K page, on its first lookup, hits the IOTLB.", + "UMask": "0x4", + "Unit": "IIO" + }, { "BriefDescription": ": Context cache hits", "EventCode": "0x40",
Signature · 1 line
2.43.0.472.g3155946c3a-goog
Arnaldo Carvalho de Melo
maintainerEm Wed, Jan 03, 2024 at 11:42:56PM -0800, Ian Rogers escreveu:
Quoted 9 lines
Fix that the core PMU is being specified for 2 uncore events. Specify a PMU for the alderlake
UNCORE_FREQmetric.Conversion script updated in:
https://github.com/intel/perfmon/pull/126Reported-by Arnaldo Carvalho de Melo <[email protected]>Signed-off-by Ian Rogers <[email protected]>
After this first patch:
101: perf all metricgroups test : Ok 102: perf all metrics test : FAILED! 107: perf metrics value validation : Ok
102 is now failing due to some other problem:
root@number:~# perf test -v 102 102: perf all metrics test : --- start --- test child forked, pid 2701034 Testing tma_core_bound Testing tma_info_core_ilp Testing tma_info_memory_l2mpki Testing tma_memory_bound Testing tma_info_bad_spec_branch_misprediction_cost Testing tma_info_bad_spec_ipmisp_cond_ntaken Testing tma_info_bad_spec_ipmisp_cond_taken Testing tma_info_bad_spec_ipmisp_indirect Testing tma_info_bad_spec_ipmisp_ret Testing tma_info_bad_spec_ipmispredict Testing tma_info_bottleneck_mispredictions Testing tma_info_branches_callret Testing tma_info_branches_cond_nt Testing tma_info_branches_cond_tk Testing tma_info_branches_jump Testing tma_info_branches_other_branches Testing tma_branch_mispredicts Testing tma_clears_resteers Testing tma_machine_clears Testing tma_mispredicts_resteers Testing tma_icache_misses Testing tma_info_bottleneck_big_code Testing tma_itlb_misses Testing tma_unknown_branches Testing tma_info_inst_mix_bptkbranch Testing tma_info_inst_mix_ipbranch Testing tma_info_inst_mix_ipcall Testing tma_info_inst_mix_iptb Testing tma_info_system_ipfarbranch Testing tma_info_thread_uptb Testing tma_info_memory_fb_hpki Testing tma_info_memory_l1mpki Testing tma_info_memory_l1mpki_load Testing tma_info_memory_l2hpki_all Testing tma_info_memory_l2hpki_load Testing tma_info_memory_l2mpki_all Testing tma_info_memory_l2mpki_load Testing tma_info_memory_l3mpki Testing tma_l1_bound Testing tma_l2_bound Testing tma_l3_bound Testing tma_fp_scalar Testing tma_fp_vector Testing tma_fp_vector_128b Testing tma_fp_vector_256b Testing tma_int_vector_128b Testing tma_int_vector_256b Testing tma_port_0 Testing tma_x87_use Testing tma_info_botlnk_l0_core_bound_likely Testing tma_info_core_fp_arith_utilization Testing tma_info_pipeline_execute Testing tma_info_system_gflops Testing tma_info_thread_execute_per_issue Testing tma_dsb Testing tma_info_frontend_dsb_coverage Testing tma_decoder0_alone Testing tma_dsb_switches Testing tma_info_botlnk_l2_dsb_misses Testing tma_info_frontend_dsb_switch_cost Testing tma_info_frontend_ipdsb_miss_ret Testing tma_mite Testing tma_contested_accesses Testing tma_false_sharing Testing tma_backend_bound Testing tma_backend_bound_aux Testing tma_bad_speculation Testing tma_frontend_bound Testing tma_retiring Testing tma_info_botlnk_l2_ic_misses Testing tma_info_bottleneck_instruction_fetch_bw Testing tma_info_frontend_fetch_upc Testing tma_info_frontend_icache_miss_latency Testing tma_info_frontend_ipunknown_branch Testing tma_info_frontend_lsd_coverage Testing tma_info_memory_tlb_code_stlb_mpki Testing tma_fetch_bandwidth Testing tma_lsd Testing tma_branch_resteers Testing tma_lcp Testing tma_ms_switches Testing tma_info_core_flopc Testing tma_info_inst_mix_iparith Testing tma_info_inst_mix_iparith_avx128 Testing tma_info_inst_mix_iparith_avx256 Testing tma_info_inst_mix_iparith_scalar_dp Testing tma_info_inst_mix_iparith_scalar_sp Testing tma_info_inst_mix_ipflop Testing tma_fetch_latency Testing tma_avx_assists Testing tma_fp_arith Testing tma_fp_assists Testing tma_info_system_cpu_utilization Testing tma_info_system_dram_bw_use Testing tma_shuffles Testing tma_info_frontend_l2mpki_code Testing tma_info_frontend_l2mpki_code_all Testing tma_info_inst_mix_ipload Testing tma_info_inst_mix_ipstore Testing tma_info_bottleneck_memory_bandwidth Testing tma_info_bottleneck_memory_data_tlbs Testing tma_info_bottleneck_memory_latency Testing tma_info_memory_core_l1d_cache_fill_bw Testing tma_info_memory_core_l2_cache_fill_bw Testing tma_info_memory_core_l3_cache_access_bw Testing tma_info_memory_core_l3_cache_fill_bw Testing tma_info_memory_load_miss_real_latency Testing tma_info_memory_mlp Testing tma_info_memory_thread_l1d_cache_fill_bw_1t Testing tma_info_memory_thread_l2_cache_fill_bw_1t Testing tma_info_memory_thread_l3_cache_access_bw_1t Testing tma_info_memory_thread_l3_cache_fill_bw_1t Testing tma_info_memory_tlb_load_stlb_mpki Testing tma_info_memory_tlb_page_walks_utilization Testing tma_info_memory_tlb_store_stlb_mpki Testing tma_info_system_mem_parallel_reads Testing tma_info_system_mem_read_latency Testing tma_info_system_mem_request_latency Testing tma_info_thread_cpi Testing tma_fb_full Testing tma_mem_bandwidth Testing tma_sq_full Testing tma_streaming_stores Testing tma_dram_bound Testing tma_store_bound Testing tma_l3_hit_latency Testing tma_mem_latency Testing tma_store_latency Testing tma_dtlb_load Testing tma_dtlb_store Testing tma_load_stlb_hit Testing tma_load_stlb_miss Testing tma_store_stlb_hit Testing tma_store_stlb_miss Testing tma_info_memory_oro_data_l2_mlp Testing tma_info_memory_oro_load_l2_mlp Testing tma_info_memory_oro_load_l2_miss_latency Testing tma_info_memory_oro_load_l3_miss_latency Testing tma_microcode_sequencer Testing tma_info_core_clks Testing tma_info_core_clks_p Testing tma_info_core_cpi Testing tma_info_core_ipc Testing tma_info_core_slots Testing tma_info_core_upi Testing tma_info_frontend_inst_miss_cost_dramhit_percent Testing tma_info_frontend_inst_miss_cost_l2hit_percent Testing tma_info_frontend_inst_miss_cost_l3hit_percent Testing tma_info_inst_mix_branch_mispredict_ratio Testing tma_info_inst_mix_branch_mispredict_to_unknown_branch_ratio Testing tma_info_inst_mix_fpdiv_uop_ratio Testing tma_info_inst_mix_idiv_uop_ratio Testing tma_info_inst_mix_ipfarbranch Testing tma_info_inst_mix_ipmisp_cond_ntaken Testing tma_info_inst_mix_ipmisp_cond_taken Testing tma_info_inst_mix_ipmisp_indirect Testing tma_info_inst_mix_ipmisp_ret Testing tma_info_inst_mix_ipmispredict Testing tma_info_inst_mix_microcode_uop_ratio Testing tma_info_inst_mix_x87_uop_ratio Testing tma_info_l1_bound_address_alias_blocks Testing tma_info_l1_bound_load_splits Testing tma_info_l1_bound_store_fwd_blocks Testing tma_info_memory_cycles_per_demand_load_dram_hit Testing tma_info_memory_cycles_per_demand_load_l2_hit Testing tma_info_memory_cycles_per_demand_load_l3_hit Testing tma_info_memory_memloadpki Testing tma_info_system_kernel_cpi Testing tma_info_system_kernel_utilization Testing tma_data_sharing Testing tma_lock_latency Testing tma_fused_instructions Testing tma_info_pipeline_ipassist Testing tma_info_pipeline_retire Testing tma_info_pipeline_strings_cycles Testing tma_info_thread_clks Testing tma_info_thread_uoppi Testing tma_int_operations Testing tma_memory_operations Testing tma_non_fused_branches Testing tma_nop_instructions Testing tma_other_light_ops Testing tma_ports_utilization Testing tma_ports_utilized_0 Testing tma_ports_utilized_1 Metric 'tma_ports_utilized_1' not printed in: # Running 'internals/synthesize' benchmark: Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself: Average synthesis took: 49.581 usec (+- 0.030 usec) Average num. events: 47.000 (+- 0.000) Average time per event 1.055 usec Average data synthesis took: 53.367 usec (+- 0.032 usec) Average num. events: 246.000 (+- 0.000) Average time per event 0.217 usec
Performance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%)
<not counted> cpu_core/topdown-retiring/ (0.00%)
<not counted> cpu_core/topdown-mem-bound/ (0.00%)
<not counted> cpu_core/topdown-bad-spec/ (0.00%)
<not counted> cpu_core/topdown-fe-bound/ (0.00%)
<not counted> cpu_core/topdown-be-bound/ (0.00%)
<not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%)
<not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%)
<not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 1.180394056 seconds time elapsed
0.409881000 seconds user
0.764134000 seconds sysTesting tma_ports_utilized_2
Metric 'tma_ports_utilized_2' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 50.456 usec (+- 0.066 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.074 usec
Average data synthesis took: 52.904 usec (+- 0.030 usec) Average num. events: 246.000 (+- 0.000)
Average time per event 0.215 usec
Performance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%)
<not counted> cpu_core/topdown-retiring/ (0.00%)
<not counted> cpu_core/topdown-mem-bound/ (0.00%)
<not counted> cpu_core/topdown-bad-spec/ (0.00%)
<not counted> cpu_core/topdown-fe-bound/ (0.00%)
<not counted> cpu_core/topdown-be-bound/ (0.00%)
<not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%)
<not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%)
<not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 1.187012938 seconds time elapsed
0.397919000 seconds user
0.782854000 seconds sysTesting tma_ports_utilized_3m
Metric 'tma_ports_utilized_3m' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 48.248 usec (+- 0.028 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.027 usec
Average data synthesis took: 52.781 usec (+- 0.036 usec) Average num. events: 245.000 (+- 0.000)
Average time per event 0.215 usec
Performance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%)
<not counted> cpu_core/topdown-retiring/ (0.00%)
<not counted> cpu_core/topdown-mem-bound/ (0.00%)
<not counted> cpu_core/topdown-bad-spec/ (0.00%)
<not counted> cpu_core/topdown-fe-bound/ (0.00%)
<not counted> cpu_core/topdown-be-bound/ (0.00%)
<not counted> cpu_core/UOPS_EXECUTED.CYCLES_GE_3/ (0.00%)
<not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%)
<not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%)
<not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 1.160315379 seconds time elapsed
0.388639000 seconds user
0.765363000 seconds sysTesting tma_serializing_operation
Testing C10_Pkg_Residency
Testing C1_Core_Residency
Testing C2_Pkg_Residency
Testing C3_Pkg_Residency
Testing C6_Core_Residency
Testing C6_Pkg_Residency
Testing C7_Core_Residency
Testing C7_Pkg_Residency
Testing C8_Pkg_Residency
Testing C9_Pkg_Residency
Testing tma_info_system_average_frequency
Testing tma_info_system_turbo_utilization
Testing tma_info_inst_mix_ipswpf
Testing tma_info_bottleneck_branching_overhead
Testing tma_info_core_coreipc
Testing tma_info_thread_ipc
Testing tma_heavy_operations
Testing tma_light_operations
Testing tma_info_core_core_clks
Testing tma_info_system_smt_2t_utilization
Testing tma_info_thread_slots_utilization
Testing UNCORE_FREQ
Testing tma_info_system_socket_clks
Testing tma_info_inst_mix_instructions
Testing tma_info_thread_slots
Testing tma_base
Testing tma_ms_uops
Testing tma_resource_bound
Testing tma_alloc_restriction
Testing tma_branch_detect
Testing tma_branch_resteer
Testing tma_cisc
Testing tma_decode
Testing tma_divider
Testing tma_fast_nuke
Testing tma_few_uops_instructions
Testing tma_fpdiv_uops
Testing tma_mem_scheduler
Testing tma_non_mem_scheduler
Testing tma_nuke
Testing tma_other_fb
Testing tma_other_load_store
Testing tma_other_ret
Testing tma_predecode
Testing tma_register
Testing tma_reorder_buffer
Testing tma_serialization
Testing tma_assists
Testing tma_disambiguation
Testing tma_fp_assist
Testing tma_ld_buffer
Testing tma_memory_ordering
Testing tma_other_l1
Testing tma_page_fault
Testing tma_rsv
Testing tma_smc
Testing tma_split_loads
Testing tma_split_stores
Testing tma_st_buffer
Testing tma_stlb_hit
Testing tma_stlb_miss
Testing tma_store_fwd_blk
Testing tma_alu_op_utilization
Testing tma_load_op_utilization
Testing tma_mixing_vectors
Testing tma_page_faults
Testing tma_store_op_utilization
Testing tma_memory_fence
Metric 'tma_memory_fence' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.458 usec (+- 0.033 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.052 usec
Average data synthesis took: 53.268 usec (+- 0.027 usec) Average num. events: 244.000 (+- 0.000)
Average time per event 0.218 usec
Performance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%)
<not counted> cpu_core/topdown-retiring/ (0.00%)
<not counted> cpu_core/topdown-mem-bound/ (0.00%)
<not counted> cpu_core/topdown-bad-spec/ (0.00%)
<not counted> cpu_core/topdown-fe-bound/ (0.00%)
<not counted> cpu_core/topdown-be-bound/ (0.00%)
<not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%)
<not counted> cpu_core/MISC2_RETIRED.LFENCE/ (0.00%)
<not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%)
<not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 1.177929044 seconds time elapsed
0.434552000 seconds user
0.736874000 seconds sysTesting tma_port_1
Testing tma_port_6
Testing tma_slow_pause
Metric 'tma_slow_pause' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.987 usec (+- 0.049 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.064 usec
Average data synthesis took: 53.490 usec (+- 0.033 usec) Average num. events: 245.000 (+- 0.000)
Average time per event 0.218 usec
Performance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%)
<not counted> cpu_core/topdown-retiring/ (0.00%)
<not counted> cpu_core/topdown-mem-bound/ (0.00%)
<not counted> cpu_core/topdown-bad-spec/ (0.00%)
<not counted> cpu_core/topdown-fe-bound/ (0.00%)
<not counted> cpu_core/topdown-be-bound/ (0.00%)
<not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/ (0.00%)
<not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%)
<not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%)
<not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 1.186254766 seconds time elapsed
0.427220000 seconds user
0.752217000 seconds sysTesting smi_cycles
Testing smi_num
Testing tsx_aborted_cycles
Testing tsx_cycles_per_elision
Testing tsx_cycles_per_transaction
Testing tsx_transactional_cycles
test child finished with -1
---- end ----
perf all metrics test: FAILED!
root@number:~#
Quoted 61 lines
--- .../arch/x86/alderlake/adl-metrics.json | 15 ++++++++------- .../arch/x86/rocketlake/rkl-metrics.json | 2 +- 2 files changed, 9 insertions(+), 8 deletions(-)tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json +8
-7@@ -69,12 +69,6 @@ "MetricName": "C9_Pkg_Residency", "ScaleUnit": "100%" }, - { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, { "BriefDescription": "Percentage of cycles spent in System Management Interrupts.", "MetricExpr": "((msr@aperf@ - cycles) / msr@aperf@ if msr@smi@ > 0 else 0)",@@ -809,6 +803,13 @@ "ScaleUnit": "100%", "Unit": "cpu_atom" }, + { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ", + "Unit": "cpu_core" + }, { "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", "MetricExpr": "(cpu_core@UOPS_DISPATCHED.PORT_0@ + cpu_core@UOPS_DISPATCHED.PORT_1@ + cpu_core@UOPS_DISPATCHED.PORT_5_11@ + cpu_core@UOPS_DISPATCHED.PORT_6@) / (5 * tma_info_core_core_clks)",@@ -1838,7 +1839,7 @@ }, { "BriefDescription": "Average number of parallel data read requests to external memory", - "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu_core@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@", + "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@", "MetricGroup": "Mem;MemoryBW;SoC", "MetricName": "tma_info_system_mem_parallel_reads", "PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches",tools/perf/pmu-events/arch/x86/rocketlake/rkl-metrics.json +1
-1@@ -985,7 +985,7 @@ }, { "BriefDescription": "Average number of parallel data read requests to external memory", - "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@", + "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@", "MetricGroup": "Mem;MemoryBW;SoC", "MetricName": "tma_info_system_mem_parallel_reads", "PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches"Signature · 1 line
2.43.0.472.g3155946c3a-goog
Signature · 1 line
- Arnaldo
Ian Rogers
reviewerQuoted 448 lines · Arnaldo Carvalho de Melo
Em Wed, Jan 03, 2024 at 11:42:56PM -0800, Ian Rogers escreveu:
Quoted 9 lines
Fix that the core PMU is being specified for 2 uncore events. Specify a PMU for the alderlake
UNCORE_FREQmetric.Conversion script updated in:
https://github.com/intel/perfmon/pull/126Reported-by Arnaldo Carvalho de Melo <[email protected]>Signed-off-by Ian Rogers <[email protected]>After this first patch:
101: perf all metricgroups test : Ok 102: perf all metrics test : FAILED! 107: perf metrics value validation : Ok102 is now failing due to some other problem:
root@number:~# perf test -v 102 102: perf all metrics test : --- start --- test child forked, pid 2701034 Testing tma_core_bound Testing tma_info_core_ilp Testing tma_info_memory_l2mpki Testing tma_memory_bound Testing tma_info_bad_spec_branch_misprediction_cost Testing tma_info_bad_spec_ipmisp_cond_ntaken Testing tma_info_bad_spec_ipmisp_cond_taken Testing tma_info_bad_spec_ipmisp_indirect Testing tma_info_bad_spec_ipmisp_ret Testing tma_info_bad_spec_ipmispredict Testing tma_info_bottleneck_mispredictions Testing tma_info_branches_callret Testing tma_info_branches_cond_nt Testing tma_info_branches_cond_tk Testing tma_info_branches_jump Testing tma_info_branches_other_branches Testing tma_branch_mispredicts Testing tma_clears_resteers Testing tma_machine_clears Testing tma_mispredicts_resteers Testing tma_icache_misses Testing tma_info_bottleneck_big_code Testing tma_itlb_misses Testing tma_unknown_branches Testing tma_info_inst_mix_bptkbranch Testing tma_info_inst_mix_ipbranch Testing tma_info_inst_mix_ipcall Testing tma_info_inst_mix_iptb Testing tma_info_system_ipfarbranch Testing tma_info_thread_uptb Testing tma_info_memory_fb_hpki Testing tma_info_memory_l1mpki Testing tma_info_memory_l1mpki_load Testing tma_info_memory_l2hpki_all Testing tma_info_memory_l2hpki_load Testing tma_info_memory_l2mpki_all Testing tma_info_memory_l2mpki_load Testing tma_info_memory_l3mpki Testing tma_l1_bound Testing tma_l2_bound Testing tma_l3_bound Testing tma_fp_scalar Testing tma_fp_vector Testing tma_fp_vector_128b Testing tma_fp_vector_256b Testing tma_int_vector_128b Testing tma_int_vector_256b Testing tma_port_0 Testing tma_x87_use Testing tma_info_botlnk_l0_core_bound_likely Testing tma_info_core_fp_arith_utilization Testing tma_info_pipeline_execute Testing tma_info_system_gflops Testing tma_info_thread_execute_per_issue Testing tma_dsb Testing tma_info_frontend_dsb_coverage Testing tma_decoder0_alone Testing tma_dsb_switches Testing tma_info_botlnk_l2_dsb_misses Testing tma_info_frontend_dsb_switch_cost Testing tma_info_frontend_ipdsb_miss_ret Testing tma_mite Testing tma_contested_accesses Testing tma_false_sharing Testing tma_backend_bound Testing tma_backend_bound_aux Testing tma_bad_speculation Testing tma_frontend_bound Testing tma_retiring Testing tma_info_botlnk_l2_ic_misses Testing tma_info_bottleneck_instruction_fetch_bw Testing tma_info_frontend_fetch_upc Testing tma_info_frontend_icache_miss_latency Testing tma_info_frontend_ipunknown_branch Testing tma_info_frontend_lsd_coverage Testing tma_info_memory_tlb_code_stlb_mpki Testing tma_fetch_bandwidth Testing tma_lsd Testing tma_branch_resteers Testing tma_lcp Testing tma_ms_switches Testing tma_info_core_flopc Testing tma_info_inst_mix_iparith Testing tma_info_inst_mix_iparith_avx128 Testing tma_info_inst_mix_iparith_avx256 Testing tma_info_inst_mix_iparith_scalar_dp Testing tma_info_inst_mix_iparith_scalar_sp Testing tma_info_inst_mix_ipflop Testing tma_fetch_latency Testing tma_avx_assists Testing tma_fp_arith Testing tma_fp_assists Testing tma_info_system_cpu_utilization Testing tma_info_system_dram_bw_use Testing tma_shuffles Testing tma_info_frontend_l2mpki_code Testing tma_info_frontend_l2mpki_code_all Testing tma_info_inst_mix_ipload Testing tma_info_inst_mix_ipstore Testing tma_info_bottleneck_memory_bandwidth Testing tma_info_bottleneck_memory_data_tlbs Testing tma_info_bottleneck_memory_latency Testing tma_info_memory_core_l1d_cache_fill_bw Testing tma_info_memory_core_l2_cache_fill_bw Testing tma_info_memory_core_l3_cache_access_bw Testing tma_info_memory_core_l3_cache_fill_bw Testing tma_info_memory_load_miss_real_latency Testing tma_info_memory_mlp Testing tma_info_memory_thread_l1d_cache_fill_bw_1t Testing tma_info_memory_thread_l2_cache_fill_bw_1t Testing tma_info_memory_thread_l3_cache_access_bw_1t Testing tma_info_memory_thread_l3_cache_fill_bw_1t Testing tma_info_memory_tlb_load_stlb_mpki Testing tma_info_memory_tlb_page_walks_utilization Testing tma_info_memory_tlb_store_stlb_mpki Testing tma_info_system_mem_parallel_reads Testing tma_info_system_mem_read_latency Testing tma_info_system_mem_request_latency Testing tma_info_thread_cpi Testing tma_fb_full Testing tma_mem_bandwidth Testing tma_sq_full Testing tma_streaming_stores Testing tma_dram_bound Testing tma_store_bound Testing tma_l3_hit_latency Testing tma_mem_latency Testing tma_store_latency Testing tma_dtlb_load Testing tma_dtlb_store Testing tma_load_stlb_hit Testing tma_load_stlb_miss Testing tma_store_stlb_hit Testing tma_store_stlb_miss Testing tma_info_memory_oro_data_l2_mlp Testing tma_info_memory_oro_load_l2_mlp Testing tma_info_memory_oro_load_l2_miss_latency Testing tma_info_memory_oro_load_l3_miss_latency Testing tma_microcode_sequencer Testing tma_info_core_clks Testing tma_info_core_clks_p Testing tma_info_core_cpi Testing tma_info_core_ipc Testing tma_info_core_slots Testing tma_info_core_upi Testing tma_info_frontend_inst_miss_cost_dramhit_percent Testing tma_info_frontend_inst_miss_cost_l2hit_percent Testing tma_info_frontend_inst_miss_cost_l3hit_percent Testing tma_info_inst_mix_branch_mispredict_ratio Testing tma_info_inst_mix_branch_mispredict_to_unknown_branch_ratio Testing tma_info_inst_mix_fpdiv_uop_ratio Testing tma_info_inst_mix_idiv_uop_ratio Testing tma_info_inst_mix_ipfarbranch Testing tma_info_inst_mix_ipmisp_cond_ntaken Testing tma_info_inst_mix_ipmisp_cond_taken Testing tma_info_inst_mix_ipmisp_indirect Testing tma_info_inst_mix_ipmisp_ret Testing tma_info_inst_mix_ipmispredict Testing tma_info_inst_mix_microcode_uop_ratio Testing tma_info_inst_mix_x87_uop_ratio Testing tma_info_l1_bound_address_alias_blocks Testing tma_info_l1_bound_load_splits Testing tma_info_l1_bound_store_fwd_blocks Testing tma_info_memory_cycles_per_demand_load_dram_hit Testing tma_info_memory_cycles_per_demand_load_l2_hit Testing tma_info_memory_cycles_per_demand_load_l3_hit Testing tma_info_memory_memloadpki Testing tma_info_system_kernel_cpi Testing tma_info_system_kernel_utilization Testing tma_data_sharing Testing tma_lock_latency Testing tma_fused_instructions Testing tma_info_pipeline_ipassist Testing tma_info_pipeline_retire Testing tma_info_pipeline_strings_cycles Testing tma_info_thread_clks Testing tma_info_thread_uoppi Testing tma_int_operations Testing tma_memory_operations Testing tma_non_fused_branches Testing tma_nop_instructions Testing tma_other_light_ops Testing tma_ports_utilization Testing tma_ports_utilized_0 Testing tma_ports_utilized_1 Metric 'tma_ports_utilized_1' not printed in: # Running 'internals/synthesize' benchmark: Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself: Average synthesis took: 49.581 usec (+- 0.030 usec) Average num. events: 47.000 (+- 0.000) Average time per event 1.055 usec Average data synthesis took: 53.367 usec (+- 0.032 usec) Average num. events: 246.000 (+- 0.000) Average time per event 0.217 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.180394056 seconds time elapsed 0.409881000 seconds user 0.764134000 seconds sysTesting
tma_ports_utilized_2
Metric 'tma_ports_utilized_2' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 50.456 usec (+- 0.066 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.074 usec
Average data synthesis took: 52.904 usec (+- 0.030 usec) Average num. events: 246.000 (+- 0.000)
Average time per event 0.215 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.187012938 seconds time elapsed 0.397919000 seconds user 0.782854000 seconds sysTesting
tma_ports_utilized_3m
Metric 'tma_ports_utilized_3m' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 48.248 usec (+- 0.028 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.027 usec
Average data synthesis took: 52.781 usec (+- 0.036 usec) Average num. events: 245.000 (+- 0.000)
Average time per event 0.215 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/UOPS_EXECUTED.CYCLES_GE_3/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.160315379 seconds time elapsed 0.388639000 seconds user 0.765363000 seconds sysTesting
tma_serializing_operation
TestingC10_Pkg_Residency
TestingC1_Core_Residency
TestingC2_Pkg_Residency
TestingC3_Pkg_Residency
TestingC6_Core_Residency
TestingC6_Pkg_Residency
TestingC7_Core_Residency
TestingC7_Pkg_Residency
TestingC8_Pkg_Residency
TestingC9_Pkg_Residency
Testingtma_info_system_average_frequency
Testingtma_info_system_turbo_utilization
Testingtma_info_inst_mix_ipswpf
Testingtma_info_bottleneck_branching_overhead
Testingtma_info_core_coreipc
Testingtma_info_thread_ipc
Testingtma_heavy_operations
Testingtma_light_operations
Testingtma_info_core_core_clks
Testingtma_info_system_smt_2t_utilization
Testingtma_info_thread_slots_utilization
TestingUNCORE_FREQ
Testingtma_info_system_socket_clks
Testingtma_info_inst_mix_instructions
Testingtma_info_thread_slots
Testingtma_base
Testingtma_ms_uops
Testingtma_resource_bound
Testingtma_alloc_restriction
Testingtma_branch_detect
Testingtma_branch_resteer
Testingtma_cisc
Testingtma_decode
Testingtma_divider
Testingtma_fast_nuke
Testingtma_few_uops_instructions
Testingtma_fpdiv_uops
Testingtma_mem_scheduler
Testingtma_non_mem_scheduler
Testingtma_nuke
Testingtma_other_fb
Testingtma_other_load_store
Testingtma_other_ret
Testingtma_predecode
Testingtma_register
Testingtma_reorder_buffer
Testingtma_serialization
Testingtma_assists
Testingtma_disambiguation
Testingtma_fp_assist
Testingtma_ld_buffer
Testingtma_memory_ordering
Testingtma_other_l1
Testingtma_page_fault
Testingtma_rsv
Testingtma_smc
Testingtma_split_loads
Testingtma_split_stores
Testingtma_st_buffer
Testingtma_stlb_hit
Testingtma_stlb_miss
Testingtma_store_fwd_blk
Testingtma_alu_op_utilization
Testingtma_load_op_utilization
Testingtma_mixing_vectors
Testingtma_page_faults
Testingtma_store_op_utilization
Testingtma_memory_fence
Metric 'tma_memory_fence' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.458 usec (+- 0.033 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.052 usec
Average data synthesis took: 53.268 usec (+- 0.027 usec) Average num. events: 244.000 (+- 0.000)
Average time per event 0.218 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/MISC2_RETIRED.LFENCE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.177929044 seconds time elapsed 0.434552000 seconds user 0.736874000 seconds sysTesting
tma_port_1
Testingtma_port_6
Testingtma_slow_pause
Metric 'tma_slow_pause' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.987 usec (+- 0.049 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.064 usec
Average data synthesis took: 53.490 usec (+- 0.033 usec) Average num. events: 245.000 (+- 0.000)
Average time per event 0.218 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.186254766 seconds time elapsed 0.427220000 seconds user 0.752217000 seconds sys
Quoted 10 lines
Testing
smi_cycles
Testingsmi_num
Testingtsx_aborted_cycles
Testingtsx_cycles_per_elision
Testingtsx_cycles_per_transaction
Testingtsx_transactional_cycles
test child finished with -1
---- end ----
perf all metrics test: FAILED!
root@number:~#
Have a try disabling the NMI watchdog. Agreed that there is more to fix here but I think the PMU driver is in part to blame because manually breaking the weak group of events is a fix. Fwiw, if we switch to the buddy watchdog mechanism then we'll no longer need to disable the NMI watchdog:
https://lore.kernel.org/lkml/20230421155255.1.I6bf789d21d0c3d75d382e7e51a804a7a51315f2c@changeid/
Thanks,
Ian
Quoted 66 lines
Quoted 59 lines
--- .../arch/x86/alderlake/adl-metrics.json | 15 ++++++++------- .../arch/x86/rocketlake/rkl-metrics.json | 2 +- 2 files changed, 9 insertions(+), 8 deletions(-)tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json +8
-7@@ -69,12 +69,6 @@ "MetricName": "C9_Pkg_Residency", "ScaleUnit": "100%" }, - { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, { "BriefDescription": "Percentage of cycles spent in System Management Interrupts.", "MetricExpr": "((msr@aperf@ - cycles) / msr@aperf@ if msr@smi@ > 0 else 0)",@@ -809,6 +803,13 @@ "ScaleUnit": "100%", "Unit": "cpu_atom" }, + { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ", + "Unit": "cpu_core" + }, { "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", "MetricExpr": "(cpu_core@UOPS_DISPATCHED.PORT_0@ + cpu_core@UOPS_DISPATCHED.PORT_1@ + cpu_core@UOPS_DISPATCHED.PORT_5_11@ + cpu_core@UOPS_DISPATCHED.PORT_6@) / (5 * tma_info_core_core_clks)",@@ -1838,7 +1839,7 @@ }, { "BriefDescription": "Average number of parallel data read requests to external memory", - "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu_core@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@", + "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@", "MetricGroup": "Mem;MemoryBW;SoC", "MetricName": "tma_info_system_mem_parallel_reads", "PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches",tools/perf/pmu-events/arch/x86/rocketlake/rkl-metrics.json +1
-2@@ -985,7 +985,7 @@ }, { "BriefDescription": "Average number of parallel data read requests to external memory", - "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@", + "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@", "MetricGroup": "Mem;MemoryBW;SoC", "MetricName": "tma_info_system_mem_parallel_reads", "PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches" --Quoted 2 lines
2.43.0.472.g3155946c3a-goog
--
- Arnaldo
Quoted 44 lines · Ian Rogers
Quoted 34 lines
Testing
tma_slow_pause
Metric 'tma_slow_pause' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.987 usec (+- 0.049 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.064 usec
Average data synthesis took: 53.490 usec (+- 0.033 usec) Average num. events: 245.000 (+- 0.000)
Average time per event 0.218 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.186254766 seconds time elapsed 0.427220000 seconds user 0.752217000 seconds sysQuoted 10 lines
Testing
smi_cycles
Testingsmi_num
Testingtsx_aborted_cycles
Testingtsx_cycles_per_elision
Testingtsx_cycles_per_transaction
Testingtsx_transactional_cycles
test child finished with -1
---- end ----
perf all metrics test: FAILED!
root@number:~#
Quoted 3 lines
Have a try disabling the NMI watchdog. Agreed that there is more to fix here but I think the PMU driver is in part to blame because manually breaking the weak group of events is a fix.
I think we have a NO_GROUP_EVENTS_NMI metric constraint to mark a group which require disabling of the NMI watchdog.
Maybe we should mark the group a NO_GROUP_EVENTS_NMI metric.
Thanks,
Kan
Quoted 4 lines
Fwiw, if we
switch to the buddy watchdog mechanism then we'll no longer need to disable the NMI watchdog:
https://lore.kernel.org/lkml/20230421155255.1.I6bf789d21d0c3d75d382e7e51a804a7a51315f2c@changeid/
Ian Rogers
reviewerQuoted 48 lines · Kan
Quoted 44 lines · Ian Rogers
Quoted 44 lines
Testing
tma_slow_pause
Metric 'tma_slow_pause' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.987 usec (+- 0.049 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.064 usec
Average data synthesis took: 53.490 usec (+- 0.033 usec) Average num. events: 245.000 (+- 0.000)
Average time per event 0.218 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.186254766 seconds time elapsed 0.427220000 seconds user 0.752217000 seconds sysTesting
smi_cycles
Testingsmi_num
Testingtsx_aborted_cycles
Testingtsx_cycles_per_elision
Testingtsx_cycles_per_transaction
Testingtsx_transactional_cycles
test child finished with -1
---- end ----
perf all metrics test: FAILED!
root@number:~#
Quoted 7 lines
Quoted 3 lines
Have a try disabling the NMI watchdog. Agreed that there is more to fix here but I think the PMU driver is in part to blame because manually breaking the weak group of events is a fix.
I think we have a
NO_GROUP_EVENTS_NMImetric constraint to mark a group which require disabling of the NMI watchdog.
Maybe we should mark the group aNO_GROUP_EVENTS_NMImetric.
+Weilin due to the affects of event grouping.
Thanks Kan, NO_GROUP_EVENTS_NMI would be good. Something I see for tma_ports_utilized_1 that may be worsening things is:
```
Testing tma_ports_utilized_1
Metric 'tma_ports_utilized_1' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by
synthesizing events on the perf process itself:
Average synthesis took: 49.581 usec (+- 0.030 usec)
Average num. events: 47.000 (+- 0.000)
Average time per event 1.055 usec
Average data synthesis took: 53.367 usec (+- 0.032 usec)
Average num. events: 246.000 (+- 0.000)
Average time per event 0.217 usec
Performance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/
(0.00%)
<not counted> cpu_core/topdown-retiring/
(0.00%)
<not counted> cpu_core/topdown-mem-bound/
(0.00%)
<not counted> cpu_core/topdown-bad-spec/
(0.00%)
<not counted> cpu_core/topdown-fe-bound/
(0.00%)
<not counted> cpu_core/topdown-be-bound/
(0.00%)
<not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/
(0.00%)
<not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/
(0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/
(0.00%)
<not counted> cpu_core/ARITH.DIV_ACTIVE/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/
(0.00%)
1.180394056 seconds time elapsed
0.409881000 seconds user
0.764134000 seconds sys
```The event EXE_ACTIVITY.1_PORTS_UTIL is repeated, this is because the metric code deduplicates events based purely on their name and so doesn't realize EXE_ACTIVITY.1_PORTS_UTIL is the same as cpu_core@EXE_ACTIVITY.1_PORTS_UTIL@. This is a hybrid only glitch as we only prefix with a PMU for hybrid metrics, and I should find and remove why there's no PMU for the 1 case of EXE_ACTIVITY.1_PORTS_UTIL.
This problem doesn't occur for tma_slow_pause and I wondered if you could give insight. That metric has the counters below:
```
$ perf stat -M tma_slow_pause -a sleep 0.1
Performance counter stats for 'system wide':
<not counted> cpu_core/TOPDOWN.SLOTS/
(0.00%)
<not counted> cpu_core/topdown-retiring/
(0.00%)
<not counted> cpu_core/topdown-mem-bound/
(0.00%)
<not counted> cpu_core/topdown-bad-spec/
(0.00%)
<not counted> cpu_core/topdown-fe-bound/
(0.00%)
<not counted> cpu_core/topdown-be-bound/
(0.00%)
<not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/
(0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/
(0.00%)
<not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/
(0.00%)
<not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/
(0.00%)
<not counted> cpu_core/ARITH.DIV_ACTIVE/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/
(0.00%)
<not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/
(0.00%)
0.102074888 seconds time elapsed
```With -vv I see the event string is:
'{RESOURCE_STALLS.SCOREBOARD/metric-id=RESOURCE_STALLS.SCOREBOARD/,cpu_core/EXE_ACTIVITY.1_PORTS_UTIL,metric-id=cpu_core!3EXE_ACTIVITY.1_PORTS_UTIL!3/,cpu_core/TOPDOWN.SLOTS,metric-id=cpu_core!3TOPDOWN.SLOTS!3/,cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS,metric-id=cpu_core!3EXE_ACTIVITY.BOUND_ON_LOADS!3/,cpu_core/topdown-retiring,metric-id=cpu_core!3topdown!1retiring!3/,cpu_core/topdown-mem-bound,metric-id=cpu_core!3topdown!1mem!1bound!3/,cpu_core/topdown-bad-spec,metric-id=cpu_core!3topdown!1bad!1spec!3/,CPU_CLK_UNHALTED.PAUSE/metric-id=CPU_CLK_UNHALTED.PAUSE/,cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL,metric-id=cpu_core!3CYCLE_ACTIVITY.STALLS_TOTAL!3/,cpu_core/CPU_CLK_UNHALTED.THREAD,metric-id=cpu_core!3CPU_CLK_UNHALTED.THREAD!3/,cpu_core/ARITH.DIV_ACTIVE,metric-id=cpu_core!3ARITH.DIV_ACTIVE!3/,cpu_core/topdown-fe-bound,metric-id=cpu_core!3topdown!1fe!1bound!3/,cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc,metric-id=cpu_core!3EXE_ACTIVITY.2_PORTS_UTIL!0umask!20xc!3/,cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80,metric-id=cpu_core!3EXE_ACTIVITY.3_PORTS_UTIL!0umask!20x80!3/,cpu_core/topdown-be-bound,metric-id=cpu_core!3topdown!1be!1bound!3/}:W'
which without the metric-ids becomes:
'{RESOURCE_STALLS.SCOREBOARD,cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/,cpu_core/TOPDOWN.SLOTS/,cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/,cpu_core/topdown-retiring/,cpu_core/topdown-mem-bound/,cpu_core/topdown-bad-spec/,CPU_CLK_UNHALTED.PAUSE,cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/,cpu_core/CPU_CLK_UNHALTED.THREAD/,cpu_core/ARITH.DIV_ACTIVE/,cpu_core/topdown-fe-bound/,cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/,cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/,cpu_core/topdown-be-bound/}:W'
I count 9 none slots/top-down counters there, but I see CPU_CLK_UNHALTED.THREAD can use fixed counter 1. Should perf_event_open fail for a CPU that has a pinned use of a fixed counter and the group needs the fixed counter? I'm guessing you don't want this as CPU_CLK_UNHALTED.THREAD can also go on a generic counter and the driver doesn't want to count counter usage, it seems feasible to add it though. I guess we need a NO_GROUP_EVENTS_NMI whenever CPU_CLK_UNHALTED.THREAD is an event and 8 generic counters are in use.
Checking on Tigerlake I see:
```
$ perf stat -M tma_slow_pause -a sleep 0.1
Performance counter stats for 'system wide':
105,210,913 TOPDOWN.SLOTS # 0.1 %
tma_slow_pause (72.65%)
6,701,129 topdown-retiring
(72.65%)
52,359,712 topdown-fe-bound
(72.65%)
32,904,532 topdown-be-bound
(72.65%)
14,117,814 topdown-bad-spec
(72.65%)
6,602,391 RESOURCE_STALLS.SCOREBOARD
(76.17%)
4,220,773 cpu/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/
(76.73%)
421,812 EXE_ACTIVITY.BOUND_ON_STORES
(76.69%)
5,164,088 EXE_ACTIVITY.1_PORTS_UTIL
(76.70%)
299,681 cpu/INT_MISC.RECOVERY_CYCLES,cmask=1,edge/
(76.69%)
245 MISC_RETIRED.PAUSE_INST
(76.67%)
58,403,687 CPU_CLK_UNHALTED.THREAD
(76.72%)
25,297,841 CYCLE_ACTIVITY.STALLS_MEM_ANY
(76.67%)
3,788,772 EXE_ACTIVITY.2_PORTS_UTIL
(62.69%)
20,973,875 CYCLE_ACTIVITY.STALLS_TOTAL
(62.16%)
68,053 ARITH.DIVIDER_ACTIVE
(62.18%)
0.102624327 seconds time elapsed
```so 10 generic counters which would never fit and the weak group is broken - the difference in the metric explaining why I've not been seeing the issue. I think I need to add alderlake/sapphirerapids constraints here:
https://github.com/captain5050/perfmon/blob/main/scripts/create_perf_json.py#L1382
Ideally we'd automate the constraint generation (or the PMU driver would help us out by failing to open the weak group).
Thanks,
Ian
Quoted 8 lines
Thanks,
KanQuoted 4 lines
Fwiw, if we
switch to the buddy watchdog mechanism then we'll no longer need to disable the NMI watchdog:
https://lore.kernel.org/lkml/20230421155255.1.I6bf789d21d0c3d75d382e7e51a804a7a51315f2c@changeid/
Quoted 165 lines · Ian Rogers
Quoted 48 lines · Kan
Quoted 44 lines · Ian Rogers
Quoted 44 lines
Testing
tma_slow_pause
Metric 'tma_slow_pause' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.987 usec (+- 0.049 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.064 usec
Average data synthesis took: 53.490 usec (+- 0.033 usec) Average num. events: 245.000 (+- 0.000)
Average time per event 0.218 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.186254766 seconds time elapsed 0.427220000 seconds user 0.752217000 seconds sysTesting
smi_cycles
Testingsmi_num
Testingtsx_aborted_cycles
Testingtsx_cycles_per_elision
Testingtsx_cycles_per_transaction
Testingtsx_transactional_cycles
test child finished with -1
---- end ----
perf all metrics test: FAILED!
root@number:~#Quoted 7 lines
Quoted 3 lines
Have a try disabling the NMI watchdog. Agreed that there is more to fix here but I think the PMU driver is in part to blame because manually breaking the weak group of events is a fix.
I think we have a
NO_GROUP_EVENTS_NMImetric constraint to mark a group which require disabling of the NMI watchdog.
Maybe we should mark the group aNO_GROUP_EVENTS_NMImetric.+Weilin due to the affects of event grouping.
Thanks Kan,
NO_GROUP_EVENTS_NMIwould be good. Something I see fortma_ports_utilized_1that may be worsening things is:``` Testing tma_ports_utilized_1 Metric 'tma_ports_utilized_1' not printed in: # Running 'internals/synthesize' benchmark: Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself: Average synthesis took: 49.581 usec (+- 0.030 usec) Average num. events: 47.000 (+- 0.000) Average time per event 1.055 usec Average data synthesis took: 53.367 usec (+- 0.032 usec) Average num. events: 246.000 (+- 0.000) Average time per event 0.217 usec Performance counter stats for 'perf bench internals synthesize': <not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 1.180394056 seconds time elapsed 0.409881000 seconds user 0.764134000 seconds sys ```The event EXE_ACTIVITY.1_PORTS_UTIL is repeated, this is because the metric code deduplicates events based purely on their name and so doesn't realize EXE_ACTIVITY.1_PORTS_UTIL is the same as cpu_core@EXE_ACTIVITY.1_PORTS_UTIL@. This is a hybrid only glitch as we only prefix with a PMU for hybrid metrics, and I should find and remove why there's no PMU for the 1 case of EXE_ACTIVITY.1_PORTS_UTIL.
This problem doesn't occur for
tma_slow_pauseand I wondered if you could give insight. That metric has the counters below:``` $ perf stat -M tma_slow_pause -a sleep 0.1 Performance counter stats for 'system wide': <not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 0.102074888 seconds time elapsed ```With -vv I see the event string is:
'{RESOURCE_STALLS.SCOREBOARD/metric-id=RESOURCE_STALLS.SCOREBOARD/,cpu_core/EXE_ACTIVITY.1_PORTS_UTIL,metric-id=cpu_core!3EXE_ACTIVITY.1_PORTS_UTIL!3/,cpu_core/TOPDOWN.SLOTS,metric-id=cpu_core!3TOPDOWN.SLOTS!3/,cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS,metric-id=cpu_core!3EXE_ACTIVITY.BOUND_ON_LOADS!3/,cpu_core/topdown-retiring,metric-id=cpu_core!3topdown!1retiring!3/,cpu_core/topdown-mem-bound,metric-id=cpu_core!3topdown!1mem!1bound!3/,cpu_core/topdown-bad-spec,metric-id=cpu_core!3topdown!1bad!1spec!3/,CPU_CLK_UNHALTED.PAUSE/metric-id=CPU_CLK_UNHALTED.PAUSE/,cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL,metric-id=cpu_core!3CYCLE_ACTIVITY.STALLS_TOTAL!3/,cpu_core/CPU_CLK_UNHALTED.THREAD,metric-id=cpu_core!3CPU_CLK_UNHALTED.THREAD!3/,cpu_core/ARITH.DIV_ACTIVE,metric-id=cpu_core!3ARITH.DIV_ACTIVE!3/,cpu_core/topdown-fe-bound,metric-id=cpu_core!3topdown!1fe!1bound!3/,cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc,metric-id=cpu_core!3EXE_ACTIVITY.2_PORTS_UTIL!0umask!20xc!3/,cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80,metric-id=cpu_core!3EXE_ACTIVITY.3_PORTS_UTIL!0umask!20x80!3/,cpu_core/topdown-be-bound,metric-id=cpu_core!3topdown!1be!1bound!3/}:W'
Quoted 7 lines
which without the metric-ids becomes:
'{RESOURCE_STALLS.SCOREBOARD,cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/,cpu_core/TOPDOWN.SLOTS/,cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/,cpu_core/topdown-retiring/,cpu_core/topdown-mem-bound/,cpu_core/topdown-bad-spec/,CPU_CLK_UNHALTED.PAUSE,cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/,cpu_core/CPU_CLK_UNHALTED.THREAD/,cpu_core/ARITH.DIV_ACTIVE/,cpu_core/topdown-fe-bound/,cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/,cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/,cpu_core/topdown-be-bound/}:W'I count 9 none slots/top-down counters there, but I see CPU_CLK_UNHALTED.THREAD can use fixed counter 1. Should
perf_event_openfail for a CPU that has a pinned use of a fixed counter and the group needs the fixed counter?
I tried, but the idea was rejected.
Quoted 5 lines
I'm guessing you don't
want this as CPU_CLK_UNHALTED.THREAD can also go on a generic counter and the driver doesn't want to count counter usage, it seems feasible to add it though. I guess we need aNO_GROUP_EVENTS_NMIwhenever CPU_CLK_UNHALTED.THREAD is an event and 8 generic counters are in use.
Yes, it looks good to me.
Quoted 42 lines
Checking on Tigerlake I see:
``` $ perf stat -M tma_slow_pause -a sleep 0.1 Performance counter stats for 'system wide': 105,210,913 TOPDOWN.SLOTS # 0.1 % tma_slow_pause (72.65%) 6,701,129 topdown-retiring (72.65%) 52,359,712 topdown-fe-bound (72.65%) 32,904,532 topdown-be-bound (72.65%) 14,117,814 topdown-bad-spec (72.65%) 6,602,391 RESOURCE_STALLS.SCOREBOARD (76.17%) 4,220,773 cpu/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (76.73%) 421,812 EXE_ACTIVITY.BOUND_ON_STORES (76.69%) 5,164,088 EXE_ACTIVITY.1_PORTS_UTIL (76.70%) 299,681 cpu/INT_MISC.RECOVERY_CYCLES,cmask=1,edge/ (76.69%) 245 MISC_RETIRED.PAUSE_INST (76.67%) 58,403,687 CPU_CLK_UNHALTED.THREAD (76.72%) 25,297,841 CYCLE_ACTIVITY.STALLS_MEM_ANY (76.67%) 3,788,772 EXE_ACTIVITY.2_PORTS_UTIL (62.69%) 20,973,875 CYCLE_ACTIVITY.STALLS_TOTAL (62.16%) 68,053 ARITH.DIVIDER_ACTIVE (62.18%) 0.102624327 seconds time elapsed ```
Quoted 7 lines
so 10 generic counters which would never fit and the weak group is broken - the difference in the metric explaining why I've not been seeing the issue. I think I need to add alderlake/sapphirerapids constraints here:
https://github.com/captain5050/perfmon/blob/main/scripts/create_perf_json.py#L1382
Ideally we'd automate the constraint generation (or the PMU driver would help us out by failing to open the weak group).
Yes, an automation will be great. The NO_GROUP_EVENTS_NMI can be set for a group which has CPU_CLK_UNHALTED.THREAD and the number of core events (expect topdown) == the max number of GP counters + 1.
Thanks,
Kan
Quoted 14 lines
Thanks,
IanQuoted 8 lines
Thanks,
KanQuoted 4 lines
Fwiw, if we
switch to the buddy watchdog mechanism then we'll no longer need to disable the NMI watchdog:
https://lore.kernel.org/lkml/20230421155255.1.I6bf789d21d0c3d75d382e7e51a804a7a51315f2c@changeid/
Ian Rogers
reviewerQuoted 179 lines · Kan
Quoted 165 lines · Ian Rogers
Quoted 48 lines · Kan
Quoted 44 lines · Ian Rogers
Quoted 44 lines
Testing
tma_slow_pause
Metric 'tma_slow_pause' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.987 usec (+- 0.049 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.064 usec
Average data synthesis took: 53.490 usec (+- 0.033 usec) Average num. events: 245.000 (+- 0.000)
Average time per event 0.218 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.186254766 seconds time elapsed 0.427220000 seconds user 0.752217000 seconds sysTesting
smi_cycles
Testingsmi_num
Testingtsx_aborted_cycles
Testingtsx_cycles_per_elision
Testingtsx_cycles_per_transaction
Testingtsx_transactional_cycles
test child finished with -1
---- end ----
perf all metrics test: FAILED!
root@number:~#Quoted 7 lines
Quoted 3 lines
Have a try disabling the NMI watchdog. Agreed that there is more to fix here but I think the PMU driver is in part to blame because manually breaking the weak group of events is a fix.
I think we have a
NO_GROUP_EVENTS_NMImetric constraint to mark a group which require disabling of the NMI watchdog.
Maybe we should mark the group aNO_GROUP_EVENTS_NMImetric.+Weilin due to the affects of event grouping.
Thanks Kan,
NO_GROUP_EVENTS_NMIwould be good. Something I see fortma_ports_utilized_1that may be worsening things is:``` Testing tma_ports_utilized_1 Metric 'tma_ports_utilized_1' not printed in: # Running 'internals/synthesize' benchmark: Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself: Average synthesis took: 49.581 usec (+- 0.030 usec) Average num. events: 47.000 (+- 0.000) Average time per event 1.055 usec Average data synthesis took: 53.367 usec (+- 0.032 usec) Average num. events: 246.000 (+- 0.000) Average time per event 0.217 usec Performance counter stats for 'perf bench internals synthesize': <not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 1.180394056 seconds time elapsed 0.409881000 seconds user 0.764134000 seconds sys ```The event EXE_ACTIVITY.1_PORTS_UTIL is repeated, this is because the metric code deduplicates events based purely on their name and so doesn't realize EXE_ACTIVITY.1_PORTS_UTIL is the same as cpu_core@EXE_ACTIVITY.1_PORTS_UTIL@. This is a hybrid only glitch as we only prefix with a PMU for hybrid metrics, and I should find and remove why there's no PMU for the 1 case of EXE_ACTIVITY.1_PORTS_UTIL.
This problem doesn't occur for
tma_slow_pauseand I wondered if you could give insight. That metric has the counters below:``` $ perf stat -M tma_slow_pause -a sleep 0.1 Performance counter stats for 'system wide': <not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.PAUSE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%) 0.102074888 seconds time elapsed ```With -vv I see the event string is:
'{RESOURCE_STALLS.SCOREBOARD/metric-id=RESOURCE_STALLS.SCOREBOARD/,cpu_core/EXE_ACTIVITY.1_PORTS_UTIL,metric-id=cpu_core!3EXE_ACTIVITY.1_PORTS_UTIL!3/,cpu_core/TOPDOWN.SLOTS,metric-id=cpu_core!3TOPDOWN.SLOTS!3/,cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS,metric-id=cpu_core!3EXE_ACTIVITY.BOUND_ON_LOADS!3/,cpu_core/topdown-retiring,metric-id=cpu_core!3topdown!1retiring!3/,cpu_core/topdown-mem-bound,metric-id=cpu_core!3topdown!1mem!1bound!3/,cpu_core/topdown-bad-spec,metric-id=cpu_core!3topdown!1bad!1spec!3/,CPU_CLK_UNHALTED.PAUSE/metric-id=CPU_CLK_UNHALTED.PAUSE/,cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL,metric-id=cpu_core!3CYCLE_ACTIVITY.STALLS_TOTAL!3/,cpu_core/CPU_CLK_UNHALTED.THREAD,metric-id=cpu_core!3CPU_CLK_UNHALTED.THREAD!3/,cpu_core/ARITH.DIV_ACTIVE,metric-id=cpu_core!3ARITH.DIV_ACTIVE!3/,cpu_core/topdown-fe-bound,metric-id=cpu_core!3topdown!1fe!1bound!3/,cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc,metric-id=cpu_core!3EXE_ACTIVITY.2_PORTS_UTIL!0umask!20xc!3/,cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80,metric-id=cpu_core!3EXE_ACTIVITY.3_PORTS_UTIL!0umask!20x80!3/,cpu_core/topdown-be-bound,metric-id=cpu_core!3topdown!1be!1bound!3/}:W'Quoted 7 lines
which without the metric-ids becomes:
'{RESOURCE_STALLS.SCOREBOARD,cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/,cpu_core/TOPDOWN.SLOTS/,cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/,cpu_core/topdown-retiring/,cpu_core/topdown-mem-bound/,cpu_core/topdown-bad-spec/,CPU_CLK_UNHALTED.PAUSE,cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/,cpu_core/CPU_CLK_UNHALTED.THREAD/,cpu_core/ARITH.DIV_ACTIVE/,cpu_core/topdown-fe-bound/,cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/,cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/,cpu_core/topdown-be-bound/}:W'I count 9 none slots/top-down counters there, but I see CPU_CLK_UNHALTED.THREAD can use fixed counter 1. Should
perf_event_openfail for a CPU that has a pinned use of a fixed counter and the group needs the fixed counter?I tried, but the idea was rejected.
Quoted 7 lines
Quoted 5 lines
I'm guessing you don't
want this as CPU_CLK_UNHALTED.THREAD can also go on a generic counter and the driver doesn't want to count counter usage, it seems feasible to add it though. I guess we need aNO_GROUP_EVENTS_NMIwhenever CPU_CLK_UNHALTED.THREAD is an event and 8 generic counters are in use.Yes, it looks good to me.
Fixes all sent out, see and its links:
https://lore.kernel.org/lkml/[email protected]/
Thanks,
Ian
Quoted 70 lines
Quoted 42 lines
Checking on Tigerlake I see:
``` $ perf stat -M tma_slow_pause -a sleep 0.1 Performance counter stats for 'system wide': 105,210,913 TOPDOWN.SLOTS # 0.1 % tma_slow_pause (72.65%) 6,701,129 topdown-retiring (72.65%) 52,359,712 topdown-fe-bound (72.65%) 32,904,532 topdown-be-bound (72.65%) 14,117,814 topdown-bad-spec (72.65%) 6,602,391 RESOURCE_STALLS.SCOREBOARD (76.17%) 4,220,773 cpu/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (76.73%) 421,812 EXE_ACTIVITY.BOUND_ON_STORES (76.69%) 5,164,088 EXE_ACTIVITY.1_PORTS_UTIL (76.70%) 299,681 cpu/INT_MISC.RECOVERY_CYCLES,cmask=1,edge/ (76.69%) 245 MISC_RETIRED.PAUSE_INST (76.67%) 58,403,687 CPU_CLK_UNHALTED.THREAD (76.72%) 25,297,841 CYCLE_ACTIVITY.STALLS_MEM_ANY (76.67%) 3,788,772 EXE_ACTIVITY.2_PORTS_UTIL (62.69%) 20,973,875 CYCLE_ACTIVITY.STALLS_TOTAL (62.16%) 68,053 ARITH.DIVIDER_ACTIVE (62.18%) 0.102624327 seconds time elapsed ```Quoted 7 lines
so 10 generic counters which would never fit and the weak group is broken - the difference in the metric explaining why I've not been seeing the issue. I think I need to add alderlake/sapphirerapids constraints here:
https://github.com/captain5050/perfmon/blob/main/scripts/create_perf_json.py#L1382
Ideally we'd automate the constraint generation (or the PMU driver would help us out by failing to open the weak group).Yes, an automation will be great. The
NO_GROUP_EVENTS_NMIcan be set for a group which has CPU_CLK_UNHALTED.THREAD and the number of core events (expect topdown) == the max number of GP counters + 1.Thanks,
KanQuoted 14 lines
Thanks,
IanQuoted 8 lines
Thanks,
KanQuoted 4 lines
Fwiw, if we
switch to the buddy watchdog mechanism then we'll no longer need to disable the NMI watchdog:
https://lore.kernel.org/lkml/20230421155255.1.I6bf789d21d0c3d75d382e7e51a804a7a51315f2c@changeid/
Arnaldo Carvalho de Melo
maintainerEm Thu, Jan 04, 2024 at 05:56:22AM -0800, Ian Rogers escreveu:
Quoted 4 lines · Arnaldo Carvalho de Melo
Quoted 3 lines
Em Wed, Jan 03, 2024 at 11:42:56PM -0800, Ian Rogers escreveu:
Quoted 2 lines
Fix that the core PMU is being specified for 2 uncore events. Specify a PMU for the alderlake
UNCORE_FREQmetric.
<SNIP>
Quoted 12 lines
Quoted 12 lines
101: perf all metricgroups test : Ok 102: perf all metrics test : FAILED! 107: perf metrics value validation : Ok102 is now failing due to some other problem:
root@number:~# perf test -v 102 102: perf all metrics test : --- start --- test child forked, pid 2701034 Testing tma_core_bound Testing tma_info_core_ilp
<SNIP>
Quoted 35 lines
Quoted 35 lines
Testing
tma_memory_fence
Metric 'tma_memory_fence' not printed in:
# Running 'internals/synthesize' benchmark:
Computing performance of single threaded perf event synthesis by synthesizing events on the perf process itself:
Average synthesis took: 49.458 usec (+- 0.033 usec) Average num. events: 47.000 (+- 0.000)
Average time per event 1.052 usec
Average data synthesis took: 53.268 usec (+- 0.027 usec) Average num. events: 244.000 (+- 0.000)
Average time per event 0.218 usecPerformance counter stats for 'perf bench internals synthesize':
<not counted> cpu_core/TOPDOWN.SLOTS/ (0.00%) <not counted> cpu_core/topdown-retiring/ (0.00%) <not counted> cpu_core/topdown-mem-bound/ (0.00%) <not counted> cpu_core/topdown-bad-spec/ (0.00%) <not counted> cpu_core/topdown-fe-bound/ (0.00%) <not counted> cpu_core/topdown-be-bound/ (0.00%) <not counted> cpu_core/RESOURCE_STALLS.SCOREBOARD/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.1_PORTS_UTIL/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.BOUND_ON_LOADS/ (0.00%) <not counted> cpu_core/MISC2_RETIRED.LFENCE/ (0.00%) <not counted> cpu_core/CYCLE_ACTIVITY.STALLS_TOTAL/ (0.00%) <not counted> cpu_core/CPU_CLK_UNHALTED.THREAD/ (0.00%) <not counted> cpu_core/ARITH.DIV_ACTIVE/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.2_PORTS_UTIL,umask=0xc/ (0.00%) <not counted> cpu_core/EXE_ACTIVITY.3_PORTS_UTIL,umask=0x80/ (0.00%)1.177929044 seconds time elapsed 0.434552000 seconds user 0.736874000 seconds sysTesting
tma_port_1
<SNIP>
Quoted 6 lines
Quoted 4 lines
test child finished with -1
---- end ----
perf all metrics test: FAILED!
root@number:~#Have a try disabling the NMI watchdog. Agreed that there is more to
Did the trick, added this to the cset log message:
--------------------------------------- 8< ---------------------------- Test 102 is failing for another reason, not being able to get as many counters as needed, Ian Rogers suggested disabling the NMI watchdog to have more counters available:
root@number:/home/acme# cat /proc/sys/kernel/nmi_watchdog 1 root@number:/home/acme# echo 0 > /proc/sys/kernel/nmi_watchdog root@number:/home/acme# perf test 102 102: perf all metrics test : Ok root@number:/home/acme# --------------------------------------- 8< ----------------------------
- Arnaldo
Quoted 5 lines
fix here but I think the PMU driver is in part to blame because manually breaking the weak group of events is a fix. Fwiw, if we switch to the buddy watchdog mechanism then we'll no longer need to disable the NMI watchdog:
https://lore.kernel.org/lkml/20230421155255.1.I6bf789d21d0c3d75d382e7e51a804a7a51315f2c@changeid/
Quoted 9 lines · Ian Rogers
Fix that the core PMU is being specified for 2 uncore events. Specify a PMU for the alderlake
UNCORE_FREQmetric.Conversion script updated in:
https://github.com/intel/perfmon/pull/126Reported-by Arnaldo Carvalho de Melo <[email protected]>Signed-off-by Ian Rogers <[email protected]>
Thanks Ian. For the series,
Thanks,
Kan
Quoted 58 lines
--- .../arch/x86/alderlake/adl-metrics.json | 15 ++++++++------- .../arch/x86/rocketlake/rkl-metrics.json | 2 +- 2 files changed, 9 insertions(+), 8 deletions(-)tools/perf/pmu-events/arch/x86/alderlake/adl-metrics.json +8
-7@@ -69,12 +69,6 @@ "MetricName": "C9_Pkg_Residency", "ScaleUnit": "100%" }, - { - "BriefDescription": "Uncore frequency per die [GHZ]", - "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", - "MetricGroup": "SoC", - "MetricName": "UNCORE_FREQ" - }, { "BriefDescription": "Percentage of cycles spent in System Management Interrupts.", "MetricExpr": "((msr@aperf@ - cycles) / msr@aperf@ if msr@smi@ > 0 else 0)",@@ -809,6 +803,13 @@ "ScaleUnit": "100%", "Unit": "cpu_atom" }, + { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ", + "Unit": "cpu_core" + }, { "BriefDescription": "This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.", "MetricExpr": "(cpu_core@UOPS_DISPATCHED.PORT_0@ + cpu_core@UOPS_DISPATCHED.PORT_1@ + cpu_core@UOPS_DISPATCHED.PORT_5_11@ + cpu_core@UOPS_DISPATCHED.PORT_6@) / (5 * tma_info_core_core_clks)",@@ -1838,7 +1839,7 @@ }, { "BriefDescription": "Average number of parallel data read requests to external memory", - "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu_core@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@", + "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@", "MetricGroup": "Mem;MemoryBW;SoC", "MetricName": "tma_info_system_mem_parallel_reads", "PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches",tools/perf/pmu-events/arch/x86/rocketlake/rkl-metrics.json +1
-1@@ -985,7 +985,7 @@ }, { "BriefDescription": "Average number of parallel data read requests to external memory", - "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / cpu@UNC_ARB_DAT_OCCUPANCY.RD\\,cmask\\=1@", + "MetricExpr": "UNC_ARB_DAT_OCCUPANCY.RD / UNC_ARB_DAT_OCCUPANCY.RD@cmask\\=1@", "MetricGroup": "Mem;MemoryBW;SoC", "MetricName": "tma_info_system_mem_parallel_reads", "PublicDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches"