Skip to content

L
linux
Commit 376d8b58 authored by Ian Rogers's avatar Ian Rogers Committed by Arnaldo Carvalho de Melo
Browse files
Options

perf vendor events: Update Intel jaketown 

Update to v21, the metrics are based on TMA 4.4 full.

Use script at:
https://github.com/intel/event-converter-for-linux-perf/blob/master/download_and_gen.py

to download and generate the latest events and metrics. Manually copy
the jaketown files into perf and update mapfile.csv.

Tested on a non-jaketown with 'perf test':
 10: PMU events                                                      :
 10.1: PMU event table sanity                                        : Ok
 10.2: PMU event map aliases                                         : Ok
 10.3: Parsing of PMU event table metrics                            : Ok
 10.4: Parsing of PMU event table metrics with fake PMUs             : Ok
Signed-off-by: default avatarIan Rogers <irogers@google.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexandre Torgue <alexandre.torgue@foss.st.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: James Clark <james.clark@arm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John Garry <john.garry@huawei.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Kshipra Bopardikar <kshipra.bopardikar@intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Maxime Coquelin <mcoquelin.stm32@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Xing Zhengjun <zhengjun.xing@linux.intel.com>
Link: http://lore.kernel.org/lkml/20220727220832.2865794-17-irogers@google.comSigned-off-by: default avatarArnaldo Carvalho de Melo <acme@redhat.com>
parent 62201368
Hide whitespace changes
Inline Side-by-side
Showing with 4910 additions and 273 deletions
tools/perf/pmu-events/arch/x86/jaketown/cache.json
View file @ 376d8b58
...@@ -1263,4 +1263,4 @@ ...@@ -1263,4 +1263,4 @@
"SampleAfterValue": "100003", "SampleAfterValue": "100003",
"UMask": "0x10" "UMask": "0x10"
} }
] ]
\ No newline at end of file
tools/perf/pmu-events/arch/x86/jaketown/floating-point.json
View file @ 376d8b58
...@@ -135,4 +135,4 @@ ...@@ -135,4 +135,4 @@
"SampleAfterValue": "2000003", "SampleAfterValue": "2000003",
"UMask": "0x1" "UMask": "0x1"
} }
] ]
\ No newline at end of file
tools/perf/pmu-events/arch/x86/jaketown/frontend.json
View file @ 376d8b58
...@@ -311,4 +311,4 @@ ...@@ -311,4 +311,4 @@
"SampleAfterValue": "2000003", "SampleAfterValue": "2000003",
"UMask": "0x1" "UMask": "0x1"
} }
] ]
\ No newline at end of file
tools/perf/pmu-events/arch/x86/jaketown/jkt-metrics.json
View file @ 376d8b58
...@@ -124,7 +124,7 @@ ...@@ -124,7 +124,7 @@
"MetricName": "FLOPc_SMT" "MetricName": "FLOPc_SMT"
}, },
{ {
"BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is at least 1 uop executed)", "BriefDescription": "Instruction-Level-Parallelism (average number of uops executed when there is execution) per-core",
"MetricExpr": "UOPS_DISPATCHED.THREAD / (( cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@)", "MetricExpr": "UOPS_DISPATCHED.THREAD / (( cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@ / 2 ) if #SMT_on else cpu@UOPS_DISPATCHED.CORE\\,cmask\\=1@)",
"MetricGroup": "Backend;Cor;Pipeline;PortsUtil", "MetricGroup": "Backend;Cor;Pipeline;PortsUtil",
"MetricName": "ILP" "MetricName": "ILP"
...@@ -141,6 +141,12 @@ ...@@ -141,6 +141,12 @@
"MetricGroup": "Summary;TmaL1", "MetricGroup": "Summary;TmaL1",
"MetricName": "Instructions" "MetricName": "Instructions"
}, },
{
"BriefDescription": "Average number of Uops retired in cycles where at least one uop has retired.",
"MetricExpr": "UOPS_RETIRED.RETIRE_SLOTS / cpu@UOPS_RETIRED.RETIRE_SLOTS\\,cmask\\=1@",
"MetricGroup": "Pipeline;Ret",
"MetricName": "Retire"
},
{ {
"BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)", "BriefDescription": "Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache)",
"MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )", "MetricExpr": "IDQ.DSB_UOPS / (( IDQ.DSB_UOPS + LSD.UOPS + IDQ.MITE_UOPS + IDQ.MS_UOPS ) )",
...@@ -163,7 +169,8 @@ ...@@ -163,7 +169,8 @@
"BriefDescription": "Giga Floating Point Operations Per Second", "BriefDescription": "Giga Floating Point Operations Per Second",
"MetricExpr": "( ( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / 1000000000 ) / duration_time", "MetricExpr": "( ( 1 * ( FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE + FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE ) + 2 * FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE + 4 * ( FP_COMP_OPS_EXE.SSE_PACKED_SINGLE + SIMD_FP_256.PACKED_DOUBLE ) + 8 * SIMD_FP_256.PACKED_SINGLE ) / 1000000000 ) / duration_time",
"MetricGroup": "Cor;Flops;HPC", "MetricGroup": "Cor;Flops;HPC",
"MetricName": "GFLOPs" "MetricName": "GFLOPs",
"PublicDescription": "Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width and AMX engine."
}, },
{ {
"BriefDescription": "Average Frequency Utilization relative nominal frequency", "BriefDescription": "Average Frequency Utilization relative nominal frequency",
......
tools/perf/pmu-events/arch/x86/jaketown/memory.json
View file @ 376d8b58
...@@ -419,4 +419,4 @@ ...@@ -419,4 +419,4 @@
"SampleAfterValue": "100003", "SampleAfterValue": "100003",
"UMask": "0x1" "UMask": "0x1"
} }
] ]
\ No newline at end of file
tools/perf/pmu-events/arch/x86/jaketown/other.json
View file @ 376d8b58
...@@ -55,4 +55,4 @@ ...@@ -55,4 +55,4 @@
"SampleAfterValue": "2000003", "SampleAfterValue": "2000003",
"UMask": "0x1" "UMask": "0x1"
} }
] ]
\ No newline at end of file
tools/perf/pmu-events/arch/x86/jaketown/pipeline.json
View file @ 376d8b58
...@@ -410,8 +410,8 @@ ...@@ -410,8 +410,8 @@
}, },
{ {
"BriefDescription": "Reference cycles when the core is not in halt state.", "BriefDescription": "Reference cycles when the core is not in halt state.",
"Counter": "Fixed counter 3", "Counter": "Fixed counter 2",
"CounterHTOff": "Fixed counter 3", "CounterHTOff": "Fixed counter 2",
"EventName": "CPU_CLK_UNHALTED.REF_TSC", "EventName": "CPU_CLK_UNHALTED.REF_TSC",
"PublicDescription": "This event counts the number of reference cycles when the core is not in a halt state. The core enters the halt state when it is running the HLT instruction or the MWAIT instruction. This event is not affected by core frequency changes (for example, P states, TM2 transitions) but has the same incrementing frequency as the time stamp counter. This event can approximate elapsed time while the core was not in a halt state. This event has a constant ratio with the CPU_CLK_UNHALTED.REF_XCLK event. It is counted on a dedicated fixed counter, leaving the four (eight when Hyperthreading is disabled) programmable counters available for other events.", "PublicDescription": "This event counts the number of reference cycles when the core is not in a halt state. The core enters the halt state when it is running the HLT instruction or the MWAIT instruction. This event is not affected by core frequency changes (for example, P states, TM2 transitions) but has the same incrementing frequency as the time stamp counter. This event can approximate elapsed time while the core was not in a halt state. This event has a constant ratio with the CPU_CLK_UNHALTED.REF_XCLK event. It is counted on a dedicated fixed counter, leaving the four (eight when Hyperthreading is disabled) programmable counters available for other events.",
"SampleAfterValue": "2000003", "SampleAfterValue": "2000003",
...@@ -439,8 +439,8 @@ ...@@ -439,8 +439,8 @@
}, },
{ {
"BriefDescription": "Core cycles when the thread is not in halt state.", "BriefDescription": "Core cycles when the thread is not in halt state.",
"Counter": "Fixed counter 2", "Counter": "Fixed counter 1",
"CounterHTOff": "Fixed counter 2", "CounterHTOff": "Fixed counter 1",
"EventName": "CPU_CLK_UNHALTED.THREAD", "EventName": "CPU_CLK_UNHALTED.THREAD",
"PublicDescription": "This event counts the number of core cycles while the thread is not in a halt state. The thread enters the halt state when it is running the HLT instruction. This event is a component in many key event ratios. The core frequency may change from time to time due to transitions associated with Enhanced Intel SpeedStep Technology or TM2. For this reason this event may have a changing ratio with regards to time. When the core frequency is constant, this event can approximate elapsed time while the core was not in the halt state. It is counted on a dedicated fixed counter, leaving the four (eight when Hyperthreading is disabled) programmable counters available for other events.", "PublicDescription": "This event counts the number of core cycles while the thread is not in a halt state. The thread enters the halt state when it is running the HLT instruction. This event is a component in many key event ratios. The core frequency may change from time to time due to transitions associated with Enhanced Intel SpeedStep Technology or TM2. For this reason this event may have a changing ratio with regards to time. When the core frequency is constant, this event can approximate elapsed time while the core was not in the halt state. It is counted on a dedicated fixed counter, leaving the four (eight when Hyperthreading is disabled) programmable counters available for other events.",
"SampleAfterValue": "2000003", "SampleAfterValue": "2000003",
...@@ -542,8 +542,8 @@ ...@@ -542,8 +542,8 @@
}, },
{ {
"BriefDescription": "Instructions retired from execution.", "BriefDescription": "Instructions retired from execution.",
"Counter": "Fixed counter 1", "Counter": "Fixed counter 0",
"CounterHTOff": "Fixed counter 1", "CounterHTOff": "Fixed counter 0",
"EventName": "INST_RETIRED.ANY", "EventName": "INST_RETIRED.ANY",
"PublicDescription": "This event counts the number of instructions retired from execution. For instructions that consist of multiple micro-ops, this event counts the retirement of the last micro-op of the instruction. Counting continues during hardware interrupts, traps, and inside interrupt handlers.", "PublicDescription": "This event counts the number of instructions retired from execution. For instructions that consist of multiple micro-ops, this event counts the retirement of the last micro-op of the instruction. Counting continues during hardware interrupts, traps, and inside interrupt handlers.",
"SampleAfterValue": "2000003", "SampleAfterValue": "2000003",
...@@ -599,7 +599,7 @@ ...@@ -599,7 +599,7 @@
"UMask": "0x3" "UMask": "0x3"
}, },
{ {
"BriefDescription": "Number of occurences waiting for the checkpoints in Resource Allocation Table (RAT) to be recovered after Nuke due to all other cases except JEClear (e.g. whenever a ucode assist is needed like SSE exception, memory disambiguation, etc...).", "BriefDescription": "Number of occurrences waiting for the checkpoints in Resource Allocation Table (RAT) to be recovered after Nuke due to all other cases except JEClear (e.g. whenever a ucode assist is needed like SSE exception, memory disambiguation, etc...).",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"CounterHTOff": "0,1,2,3,4,5,6,7", "CounterHTOff": "0,1,2,3,4,5,6,7",
"CounterMask": "1", "CounterMask": "1",
...@@ -1199,4 +1199,4 @@ ...@@ -1199,4 +1199,4 @@
"SampleAfterValue": "2000003", "SampleAfterValue": "2000003",
"UMask": "0x1" "UMask": "0x1"
} }
] ]
\ No newline at end of file
tools/perf/pmu-events/arch/x86/jaketown/uncore-cache.json
View file @ 376d8b58
This source diff could not be displayed because it is too large. You can view the blob instead.
tools/perf/pmu-events/arch/x86/jaketown/uncore-interconnect.json
View file @ 376d8b58
[ [
{ {
"BriefDescription": "QPI clock ticks. Used to get percentages of QPI cycles events", "BriefDescription": "Number of qfclks",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x14", "EventCode": "0x14",
"EventName": "UNC_Q_CLOCKTICKS", "EventName": "UNC_Q_CLOCKTICKS",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of clocks in the QPI LL. This clock runs at 1/8th the 'GT/s' speed of the QPI link. For example, a 8GT/s link will have qfclk or 1GHz. JKT does not support dynamic link speeds, so this frequency is fixed.",
"Unit": "QPI LL" "Unit": "QPI LL"
}, },
{ {
"BriefDescription": "Cycles where receiving QPI link is in half-width mode", "BriefDescription": "Count of CTO Events",
"Counter": "0,1,2,3",
"EventCode": "0x38",
"EventName": "UNC_Q_CTO_COUNT",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of CTO (cluster trigger outs) events that were asserted across the two slots. If both slots trigger in a given cycle, the event will increment by 2. You can use edge detect to count the number of cases when both events triggered.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Direct 2 Core Spawning; Spawn Failure - Egress Credits",
"Counter": "0,1,2,3",
"EventCode": "0x13",
"EventName": "UNC_Q_DIRECT2CORE.FAILURE_CREDITS",
"PerPkg": "1",
"PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "Direct 2 Core Spawning; Spawn Failure - Egress and RBT",
"Counter": "0,1,2,3",
"EventCode": "0x13",
"EventName": "UNC_Q_DIRECT2CORE.FAILURE_CREDITS_RBT",
"PerPkg": "1",
"PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.",
"UMask": "0x8",
"Unit": "QPI LL"
},
{
"BriefDescription": "Direct 2 Core Spawning; Spawn Failure - RBT Not Set",
"Counter": "0,1,2,3",
"EventCode": "0x13",
"EventName": "UNC_Q_DIRECT2CORE.FAILURE_RBT",
"PerPkg": "1",
"PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.",
"UMask": "0x4",
"Unit": "QPI LL"
},
{
"BriefDescription": "Direct 2 Core Spawning; Spawn Success",
"Counter": "0,1,2,3",
"EventCode": "0x13",
"EventName": "UNC_Q_DIRECT2CORE.SUCCESS",
"PerPkg": "1",
"PublicDescription": "Counts the number of DRS packets that we attempted to do direct2core on. There are 4 mutually exlusive filters. Filter [0] can be used to get successful spawns, while [1:3] provide the different failure cases. Note that this does not count packets that are not candidates for Direct2Core. The only candidates for Direct2Core are DRS packets destined for Cbos.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Cycles in L1",
"Counter": "0,1,2,3",
"EventCode": "0x12",
"EventName": "UNC_Q_L1_POWER_CYCLES",
"PerPkg": "1",
"PublicDescription": "Number of QPI qfclk cycles spent in L1 power mode. L1 is a mode that totally shuts down a QPI link. Use edge detect to count the number of instances when the QPI link entered L1. Link power states are per link and per direction, so for example the Tx direction could be in one state while Rx was in another. Because L1 totally shuts down the link, it takes a good amount of time to exit this mode.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Cycles in L0p",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x10", "EventCode": "0x10",
"EventName": "UNC_Q_RxL0P_POWER_CYCLES", "EventName": "UNC_Q_RxL0P_POWER_CYCLES",
"MetricExpr": "(UNC_Q_RxL0P_POWER_CYCLES / UNC_Q_CLOCKTICKS) * 100.",
"MetricName": "rxl0p_power_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of QPI qfclk cycles spent in L0p power mode. L0p is a mode where we disable 1/2 of the QPI lanes, decreasing our bandwidth in order to save power. It increases snoop and data transfer latencies and decreases overall bandwidth. This mode can be very useful in NUMA optimized workloads that largely only utilize QPI for snoops and their responses. Use edge detect to count the number of instances when the QPI link entered L0p. Link power states are per link and per direction, so for example the Tx direction could be in one state while Rx was in another.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Cycles in L0",
"Counter": "0,1,2,3",
"EventCode": "0xf",
"EventName": "UNC_Q_RxL0_POWER_CYCLES",
"PerPkg": "1",
"PublicDescription": "Number of QPI qfclk cycles spent in L0 power mode in the Link Layer. L0 is the default mode which provides the highest performance with the most power. Use edge detect to count the number of instances that the link entered L0. Link power states are per link and per direction, so for example the Tx direction could be in one state while Rx was in another. The phy layer sometimes leaves L0 for training, which will not be captured by this event.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Rx Flit Buffer Bypassed",
"Counter": "0,1,2,3",
"EventCode": "0x9",
"EventName": "UNC_Q_RxL_BYPASSED",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an incoming flit was able to bypass the flit buffer and pass directly across the BGF and into the Egress. This is a latency optimization, and should generally be the common case. If this value is less than the number of flits transfered, it implies that there was queueing getting onto the ring, and thus the transactions saw higher latency.",
"Unit": "QPI LL"
},
{
"BriefDescription": "CRC Errors Detected; LinkInit",
"Counter": "0,1,2,3",
"EventCode": "0x3",
"EventName": "UNC_Q_RxL_CRC_ERRORS.LINK_INIT",
"PerPkg": "1",
"PublicDescription": "Number of CRC errors detected in the QPI Agent. Each QPI flit incorporates 8 bits of CRC for error detection. This counts the number of flits where the CRC was able to detect an error. After an error has been detected, the QPI agent will send a request to the transmitting socket to resend the flit (as well as any flits that came after it).",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "CRC Errors Detected; Normal Operations",
"Counter": "0,1,2,3",
"EventCode": "0x3",
"EventName": "UNC_Q_RxL_CRC_ERRORS.NORMAL_OP",
"PerPkg": "1",
"PublicDescription": "Number of CRC errors detected in the QPI Agent. Each QPI flit incorporates 8 bits of CRC for error detection. This counts the number of flits where the CRC was able to detect an error. After an error has been detected, the QPI agent will send a request to the transmitting socket to resend the flit (as well as any flits that came after it).",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "VN0 Credit Consumed; DRS",
"Counter": "0,1,2,3",
"EventCode": "0x1e",
"EventName": "UNC_Q_RxL_CREDITS_CONSUMED_VN0.DRS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an RxQ VN0 credit was consumed (i.e. message uses a VN0 credit for the Rx Buffer). This includes packets that went through the RxQ and those that were bypasssed.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "VN0 Credit Consumed; HOM",
"Counter": "0,1,2,3",
"EventCode": "0x1e",
"EventName": "UNC_Q_RxL_CREDITS_CONSUMED_VN0.HOM",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an RxQ VN0 credit was consumed (i.e. message uses a VN0 credit for the Rx Buffer). This includes packets that went through the RxQ and those that were bypasssed.",
"UMask": "0x8",
"Unit": "QPI LL"
},
{
"BriefDescription": "VN0 Credit Consumed; NCB",
"Counter": "0,1,2,3",
"EventCode": "0x1e",
"EventName": "UNC_Q_RxL_CREDITS_CONSUMED_VN0.NCB",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an RxQ VN0 credit was consumed (i.e. message uses a VN0 credit for the Rx Buffer). This includes packets that went through the RxQ and those that were bypasssed.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "VN0 Credit Consumed; NCS",
"Counter": "0,1,2,3",
"EventCode": "0x1e",
"EventName": "UNC_Q_RxL_CREDITS_CONSUMED_VN0.NCS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an RxQ VN0 credit was consumed (i.e. message uses a VN0 credit for the Rx Buffer). This includes packets that went through the RxQ and those that were bypasssed.",
"UMask": "0x4",
"Unit": "QPI LL"
},
{
"BriefDescription": "VN0 Credit Consumed; NDR",
"Counter": "0,1,2,3",
"EventCode": "0x1e",
"EventName": "UNC_Q_RxL_CREDITS_CONSUMED_VN0.NDR",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an RxQ VN0 credit was consumed (i.e. message uses a VN0 credit for the Rx Buffer). This includes packets that went through the RxQ and those that were bypasssed.",
"UMask": "0x20",
"Unit": "QPI LL"
},
{
"BriefDescription": "VN0 Credit Consumed; SNP",
"Counter": "0,1,2,3",
"EventCode": "0x1e",
"EventName": "UNC_Q_RxL_CREDITS_CONSUMED_VN0.SNP",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an RxQ VN0 credit was consumed (i.e. message uses a VN0 credit for the Rx Buffer). This includes packets that went through the RxQ and those that were bypasssed.",
"UMask": "0x10",
"Unit": "QPI LL"
},
{
"BriefDescription": "VNA Credit Consumed",
"Counter": "0,1,2,3",
"EventCode": "0x1d",
"EventName": "UNC_Q_RxL_CREDITS_CONSUMED_VNA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an RxQ VNA credit was consumed (i.e. message uses a VNA credit for the Rx Buffer). This includes packets that went through the RxQ and those that were bypasssed.",
"Unit": "QPI LL"
},
{
"BriefDescription": "RxQ Cycles Not Empty",
"Counter": "0,1,2,3",
"EventCode": "0xa",
"EventName": "UNC_Q_RxL_CYCLES_NE",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the QPI RxQ was not empty. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Occupancy Accumulator event to calculate the average occupancy.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 0; Data Tx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_RxL_FLITS_G0.DATA",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 0; Idle and Null Flits",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_RxL_FLITS_G0.IDLE",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 0; Non-Data protocol Tx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_RxL_FLITS_G0.NON_DATA",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.",
"UMask": "0x4",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 1; DRS Flits (both Header and Data)",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_RxL_FLITS_G1.DRS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x18",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 1; DRS Data Flits",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_RxL_FLITS_G1.DRS_DATA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x8",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 1; DRS Header Flits",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_RxL_FLITS_G1.DRS_NONDATA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x10",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 1; HOM Flits",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_RxL_FLITS_G1.HOM",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x6",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 1; HOM Non-Request Flits",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_RxL_FLITS_G1.HOM_NONREQ",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x4",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 1; HOM Request Flits",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_RxL_FLITS_G1.HOM_REQ",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 1; SNP Flits",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_RxL_FLITS_G1.SNP",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 2; Non-Coherent Rx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x3",
"EventName": "UNC_Q_RxL_FLITS_G2.NCB",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0xc",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 2; Non-Coherent data Rx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x3",
"EventName": "UNC_Q_RxL_FLITS_G2.NCB_DATA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x4",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 2; Non-Coherent non-data Rx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x3",
"EventName": "UNC_Q_RxL_FLITS_G2.NCB_NONDATA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x8",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 2; Non-Coherent standard Rx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x3",
"EventName": "UNC_Q_RxL_FLITS_G2.NCS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x10",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 2; Non-Data Response Rx Flits - AD",
"Counter": "0,1,2,3",
"EventCode": "0x3",
"EventName": "UNC_Q_RxL_FLITS_G2.NDR_AD",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Received - Group 2; Non-Data Response Rx Flits - AK",
"Counter": "0,1,2,3",
"EventCode": "0x3",
"EventName": "UNC_Q_RxL_FLITS_G2.NDR_AK",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits received from the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "Rx Flit Buffer Allocations",
"Counter": "0,1,2,3",
"EventCode": "0x8",
"EventName": "UNC_Q_RxL_INSERTS",
"PerPkg": "1",
"PublicDescription": "Number of allocations into the QPI Rx Flit Buffer. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Occupancy event in order to calculate the average flit buffer lifetime.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Rx Flit Buffer Allocations - DRS",
"Counter": "0,1,2,3",
"EventCode": "0x9",
"EventName": "UNC_Q_RxL_INSERTS_DRS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of allocations into the QPI Rx Flit Buffer. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Occupancy event in order to calculate the average flit buffer lifetime. This monitors only DRS flits.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Rx Flit Buffer Allocations - HOM",
"Counter": "0,1,2,3",
"EventCode": "0xc",
"EventName": "UNC_Q_RxL_INSERTS_HOM",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of allocations into the QPI Rx Flit Buffer. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Occupancy event in order to calculate the average flit buffer lifetime. This monitors only HOM flits.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Rx Flit Buffer Allocations - NCB",
"Counter": "0,1,2,3",
"EventCode": "0xa",
"EventName": "UNC_Q_RxL_INSERTS_NCB",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of allocations into the QPI Rx Flit Buffer. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Occupancy event in order to calculate the average flit buffer lifetime. This monitors only NCB flits.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Rx Flit Buffer Allocations - NCS",
"Counter": "0,1,2,3",
"EventCode": "0xb",
"EventName": "UNC_Q_RxL_INSERTS_NCS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of allocations into the QPI Rx Flit Buffer. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Occupancy event in order to calculate the average flit buffer lifetime. This monitors only NCS flits.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Rx Flit Buffer Allocations - NDR",
"Counter": "0,1,2,3",
"EventCode": "0xe",
"EventName": "UNC_Q_RxL_INSERTS_NDR",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of allocations into the QPI Rx Flit Buffer. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Occupancy event in order to calculate the average flit buffer lifetime. This monitors only NDR flits.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Rx Flit Buffer Allocations - SNP",
"Counter": "0,1,2,3",
"EventCode": "0xd",
"EventName": "UNC_Q_RxL_INSERTS_SNP",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of allocations into the QPI Rx Flit Buffer. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Occupancy event in order to calculate the average flit buffer lifetime. This monitors only SNP flits.",
"Unit": "QPI LL"
},
{
"BriefDescription": "RxQ Occupancy - All Packets",
"Counter": "0,1,2,3",
"EventCode": "0xb",
"EventName": "UNC_Q_RxL_OCCUPANCY",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of elements in the QPI RxQ in each cycle. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Not Empty event to calculate average occupancy, or with the Flit Buffer Allocations event to track average lifetime.",
"Unit": "QPI LL"
},
{
"BriefDescription": "RxQ Occupancy - DRS",
"Counter": "0,1,2,3",
"EventCode": "0x15",
"EventName": "UNC_Q_RxL_OCCUPANCY_DRS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of elements in the QPI RxQ in each cycle. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Not Empty event to calculate average occupancy, or with the Flit Buffer Allocations event to track average lifetime. This monitors DRS flits only.",
"Unit": "QPI LL"
},
{
"BriefDescription": "RxQ Occupancy - HOM",
"Counter": "0,1,2,3",
"EventCode": "0x18",
"EventName": "UNC_Q_RxL_OCCUPANCY_HOM",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of elements in the QPI RxQ in each cycle. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Not Empty event to calculate average occupancy, or with the Flit Buffer Allocations event to track average lifetime. This monitors HOM flits only.",
"Unit": "QPI LL"
},
{
"BriefDescription": "RxQ Occupancy - NCB",
"Counter": "0,1,2,3",
"EventCode": "0x16",
"EventName": "UNC_Q_RxL_OCCUPANCY_NCB",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of elements in the QPI RxQ in each cycle. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Not Empty event to calculate average occupancy, or with the Flit Buffer Allocations event to track average lifetime. This monitors NCB flits only.",
"Unit": "QPI LL"
},
{
"BriefDescription": "RxQ Occupancy - NCS",
"Counter": "0,1,2,3",
"EventCode": "0x17",
"EventName": "UNC_Q_RxL_OCCUPANCY_NCS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of elements in the QPI RxQ in each cycle. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Not Empty event to calculate average occupancy, or with the Flit Buffer Allocations event to track average lifetime. This monitors NCS flits only.",
"Unit": "QPI LL"
},
{
"BriefDescription": "RxQ Occupancy - NDR",
"Counter": "0,1,2,3",
"EventCode": "0x1a",
"EventName": "UNC_Q_RxL_OCCUPANCY_NDR",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of elements in the QPI RxQ in each cycle. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Not Empty event to calculate average occupancy, or with the Flit Buffer Allocations event to track average lifetime. This monitors NDR flits only.",
"Unit": "QPI LL"
},
{
"BriefDescription": "RxQ Occupancy - SNP",
"Counter": "0,1,2,3",
"EventCode": "0x19",
"EventName": "UNC_Q_RxL_OCCUPANCY_SNP",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of elements in the QPI RxQ in each cycle. Generally, when data is transmitted across QPI, it will bypass the RxQ and pass directly to the ring interface. If things back up getting transmitted onto the ring, however, it may need to allocate into this buffer, thus increasing the latency. This event can be used in conjunction with the Flit Buffer Not Empty event to calculate average occupancy, or with the Flit Buffer Allocations event to track average lifetime. This monitors SNP flits only.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Stalls Sending to R3QPI; BGF Stall - HOM",
"Counter": "0,1,2,3",
"EventCode": "0x35",
"EventName": "UNC_Q_RxL_STALLS.BGF_DRS",
"PerPkg": "1",
"PublicDescription": "Number of stalls trying to send to R3QPI.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Stalls Sending to R3QPI; BGF Stall - DRS",
"Counter": "0,1,2,3",
"EventCode": "0x35",
"EventName": "UNC_Q_RxL_STALLS.BGF_HOM",
"PerPkg": "1",
"PublicDescription": "Number of stalls trying to send to R3QPI.",
"UMask": "0x8",
"Unit": "QPI LL"
},
{
"BriefDescription": "Stalls Sending to R3QPI; BGF Stall - SNP",
"Counter": "0,1,2,3",
"EventCode": "0x35",
"EventName": "UNC_Q_RxL_STALLS.BGF_NCB",
"PerPkg": "1",
"PublicDescription": "Number of stalls trying to send to R3QPI.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "Stalls Sending to R3QPI; BGF Stall - NDR",
"Counter": "0,1,2,3",
"EventCode": "0x35",
"EventName": "UNC_Q_RxL_STALLS.BGF_NCS",
"PerPkg": "1",
"PublicDescription": "Number of stalls trying to send to R3QPI.",
"UMask": "0x4",
"Unit": "QPI LL"
},
{
"BriefDescription": "Stalls Sending to R3QPI; BGF Stall - NCS",
"Counter": "0,1,2,3",
"EventCode": "0x35",
"EventName": "UNC_Q_RxL_STALLS.BGF_NDR",
"PerPkg": "1",
"PublicDescription": "Number of stalls trying to send to R3QPI.",
"UMask": "0x20",
"Unit": "QPI LL"
},
{
"BriefDescription": "Stalls Sending to R3QPI; BGF Stall - NCB",
"Counter": "0,1,2,3",
"EventCode": "0x35",
"EventName": "UNC_Q_RxL_STALLS.BGF_SNP",
"PerPkg": "1",
"PublicDescription": "Number of stalls trying to send to R3QPI.",
"UMask": "0x10",
"Unit": "QPI LL"
},
{
"BriefDescription": "Stalls Sending to R3QPI; Egress Credits",
"Counter": "0,1,2,3",
"EventCode": "0x35",
"EventName": "UNC_Q_RxL_STALLS.EGRESS_CREDITS",
"PerPkg": "1",
"PublicDescription": "Number of stalls trying to send to R3QPI.",
"UMask": "0x40",
"Unit": "QPI LL" "Unit": "QPI LL"
}, },
{ {
"BriefDescription": "Cycles where transmitting QPI link is in half-width mode", "BriefDescription": "Stalls Sending to R3QPI; GV",
"Counter": "0,1,2,3",
"EventCode": "0x35",
"EventName": "UNC_Q_RxL_STALLS.GV",
"PerPkg": "1",
"PublicDescription": "Number of stalls trying to send to R3QPI.",
"UMask": "0x80",
"Unit": "QPI LL"
},
{
"BriefDescription": "Cycles in L0p",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xd", "EventCode": "0xd",
"EventName": "UNC_Q_TxL0P_POWER_CYCLES", "EventName": "UNC_Q_TxL0P_POWER_CYCLES",
"MetricExpr": "(UNC_Q_TxL0P_POWER_CYCLES / UNC_Q_CLOCKTICKS) * 100.",
"MetricName": "txl0p_power_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of QPI qfclk cycles spent in L0p power mode. L0p is a mode where we disable 1/2 of the QPI lanes, decreasing our bandwidth in order to save power. It increases snoop and data transfer latencies and decreases overall bandwidth. This mode can be very useful in NUMA optimized workloads that largely only utilize QPI for snoops and their responses. Use edge detect to count the number of instances when the QPI link entered L0p. Link power states are per link and per direction, so for example the Tx direction could be in one state while Rx was in another.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Cycles in L0",
"Counter": "0,1,2,3",
"EventCode": "0xc",
"EventName": "UNC_Q_TxL0_POWER_CYCLES",
"PerPkg": "1",
"PublicDescription": "Number of QPI qfclk cycles spent in L0 power mode in the Link Layer. L0 is the default mode which provides the highest performance with the most power. Use edge detect to count the number of instances that the link entered L0. Link power states are per link and per direction, so for example the Tx direction could be in one state while Rx was in another. The phy layer sometimes leaves L0 for training, which will not be captured by this event.",
"Unit": "QPI LL" "Unit": "QPI LL"
}, },
{ {
"BriefDescription": "Number of data flits transmitted ", "BriefDescription": "Tx Flit Buffer Bypassed",
"Counter": "0,1,2,3",
"EventCode": "0x5",
"EventName": "UNC_Q_TxL_BYPASSED",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that an incoming flit was able to bypass the Tx flit buffer and pass directly out the QPI Link. Generally, when data is transmitted across QPI, it will bypass the TxQ and pass directly to the link. However, the TxQ will be used with L0p and when LLR occurs, increasing latency to transfer out to the link.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Cycles Stalled with no LLR Credits; LLR is almost full",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_TxL_CRC_NO_CREDITS.ALMOST_FULL",
"PerPkg": "1",
"PublicDescription": "Number of cycles when the Tx side ran out of Link Layer Retry credits, causing the Tx to stall.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "Cycles Stalled with no LLR Credits; LLR is full",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_Q_TxL_CRC_NO_CREDITS.FULL",
"PerPkg": "1",
"PublicDescription": "Number of cycles when the Tx side ran out of Link Layer Retry credits, causing the Tx to stall.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Tx Flit Buffer Cycles not Empty",
"Counter": "0,1,2,3",
"EventCode": "0x6",
"EventName": "UNC_Q_TxL_CYCLES_NE",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the TxQ is not empty. Generally, when data is transmitted across QPI, it will bypass the TxQ and pass directly to the link. However, the TxQ will be used with L0p and when LLR occurs, increasing latency to transfer out to the link.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 0; Data Tx Flits",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G0.DATA", "EventName": "UNC_Q_TxL_FLITS_G0.DATA",
"PerPkg": "1", "PerPkg": "1",
"ScaleUnit": "8Bytes", "PublicDescription": "Counts the number of flits transmitted across the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.",
"UMask": "0x2", "UMask": "0x2",
"Unit": "QPI LL" "Unit": "QPI LL"
}, },
{ {
"BriefDescription": "Number of non data (control) flits transmitted ", "BriefDescription": "Flits Transferred - Group 0; Idle and Null Flits",
"Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G0.IDLE",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits transmitted across the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 0; Non-Data protocol Tx Flits",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G0.NON_DATA", "EventName": "UNC_Q_TxL_FLITS_G0.NON_DATA",
"PerPkg": "1", "PerPkg": "1",
"ScaleUnit": "8Bytes", "PublicDescription": "Counts the number of flits transmitted across the QPI Link. It includes filters for Idle, protocol, and Data Flits. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time (for L0) or 4B instead of 8B for L0p.",
"UMask": "0x4",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 1; DRS Flits (both Header and Data)",
"Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G1.DRS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x18",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 1; DRS Data Flits",
"Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G1.DRS_DATA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x8",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 1; DRS Header Flits",
"Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G1.DRS_NONDATA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x10",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 1; HOM Flits",
"Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G1.HOM",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x6",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 1; HOM Non-Request Flits",
"Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G1.HOM_NONREQ",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x4", "UMask": "0x4",
"Unit": "QPI LL" "Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 1; HOM Request Flits",
"Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G1.HOM_REQ",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 1; SNP Flits",
"Counter": "0,1,2,3",
"EventName": "UNC_Q_TxL_FLITS_G1.SNP",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for SNP, HOM, and DRS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 2; Non-Coherent Bypass Tx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_TxL_FLITS_G2.NCB",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0xc",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 2; Non-Coherent data Tx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_TxL_FLITS_G2.NCB_DATA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x4",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 2; Non-Coherent non-data Tx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_TxL_FLITS_G2.NCB_NONDATA",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x8",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 2; Non-Coherent standard Tx Flits",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_TxL_FLITS_G2.NCS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x10",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 2; Non-Data Response Tx Flits - AD",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_TxL_FLITS_G2.NDR_AD",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x1",
"Unit": "QPI LL"
},
{
"BriefDescription": "Flits Transferred - Group 2; Non-Data Response Tx Flits - AK",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_Q_TxL_FLITS_G2.NDR_AK",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of flits trasmitted across the QPI Link. This is one of three 'groups' that allow us to track flits. It includes filters for NDR, NCB, and NCS message classes. Each 'flit' is made up of 80 bits of information (in addition to some ECC data). In full-width (L0) mode, flits are made up of four 'fits', each of which contains 20 bits of data (along with some additional ECC data). In half-width (L0p) mode, the fits are only 10 bits, and therefore it takes twice as many fits to transmit a flit. When one talks about QPI 'speed' (for example, 8.0 GT/s), the 'transfers' here refer to 'fits'. Therefore, in L0, the system will transfer 1 'flit' at the rate of 1/4th the QPI speed. One can calculate the bandwidth of the link by taking: flits*80b/time. Note that this is not the same as 'data' bandwidth. For example, when we are transfering a 64B cacheline across QPI, we will break it into 9 flits -- 1 with header information and 8 with 64 bits of actual 'data' and an additional 16 bits of other information. To calculate 'data' bandwidth, one should therefore do: data flits * 8B / time.",
"UMask": "0x2",
"Unit": "QPI LL"
},
{
"BriefDescription": "Tx Flit Buffer Allocations",
"Counter": "0,1,2,3",
"EventCode": "0x4",
"EventName": "UNC_Q_TxL_INSERTS",
"PerPkg": "1",
"PublicDescription": "Number of allocations into the QPI Tx Flit Buffer. Generally, when data is transmitted across QPI, it will bypass the TxQ and pass directly to the link. However, the TxQ will be used with L0p and when LLR occurs, increasing latency to transfer out to the link. This event can be used in conjunction with the Flit Buffer Occupancy event in order to calculate the average flit buffer lifetime.",
"Unit": "QPI LL"
},
{
"BriefDescription": "Tx Flit Buffer Occupancy",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_Q_TxL_OCCUPANCY",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of flits in the TxQ. Generally, when data is transmitted across QPI, it will bypass the TxQ and pass directly to the link. However, the TxQ will be used with L0p and when LLR occurs, increasing latency to transfer out to the link. This can be used with the cycles not empty event to track average occupancy, or the allocations event to track average lifetime in the TxQ.",
"Unit": "QPI LL"
},
{
"BriefDescription": "VNA Credits Returned",
"Counter": "0,1,2,3",
"EventCode": "0x1c",
"EventName": "UNC_Q_VNA_CREDIT_RETURNS",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of VNA credits returned.",
"Unit": "QPI LL"
},
{
"BriefDescription": "VNA Credits Pending Return - Occupancy",
"Counter": "0,1,2,3",
"EventCode": "0x1b",
"EventName": "UNC_Q_VNA_CREDIT_RETURN_OCCUPANCY",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of VNA credits in the Rx side that are waitng to be returned back across the link.",
"Unit": "QPI LL"
} }
] ]
tools/perf/pmu-events/arch/x86/jaketown/uncore-memory.json
View file @ 376d8b58
[ [
{ {
"BriefDescription": "Memory page activates", "BriefDescription": "DRAM Activate Count",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x1", "EventCode": "0x1",
"EventName": "UNC_M_ACT_COUNT", "EventName": "UNC_M_ACT_COUNT",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of DRAM Activate commands sent on this channel. Activate commands are issued to open up a page on the DRAM devices so that it can be read or written to with a CAS. One can calculate the number of Page Misses by subtracting the number of Page Miss precharges from the number of Activates.",
"Unit": "iMC" "Unit": "iMC"
}, },
{ {
"BriefDescription": "read requests to memory controller. Derived from unc_m_cas_count.rd", "BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; All DRAM WR_CAS (w/ and w/out auto-pre)",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x4", "EventCode": "0x4",
"EventName": "LLC_MISSES.MEM_READ", "EventName": "UNC_M_CAS_COUNT.ALL",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "DRAM RD_CAS and WR_CAS Commands",
"UMask": "0xf",
"Unit": "iMC"
},
{
"BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; All DRAM Reads (RD_CAS + Underfills)",
"Counter": "0,1,2,3",
"EventCode": "0x4",
"EventName": "UNC_M_CAS_COUNT.RD",
"PerPkg": "1",
"PublicDescription": "DRAM RD_CAS and WR_CAS Commands",
"UMask": "0x3", "UMask": "0x3",
"Unit": "iMC" "Unit": "iMC"
}, },
{ {
"BriefDescription": "write requests to memory controller. Derived from unc_m_cas_count.wr", "BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; All DRAM RD_CAS (w/ and w/out auto-pre)",
"Counter": "0,1,2,3",
"EventCode": "0x4",
"EventName": "UNC_M_CAS_COUNT.RD_REG",
"PerPkg": "1",
"PublicDescription": "DRAM RD_CAS and WR_CAS Commands",
"UMask": "0x1",
"Unit": "iMC"
},
{
"BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; Underfill Read Issued",
"Counter": "0,1,2,3",
"EventCode": "0x4",
"EventName": "UNC_M_CAS_COUNT.RD_UNDERFILL",
"PerPkg": "1",
"PublicDescription": "DRAM RD_CAS and WR_CAS Commands",
"UMask": "0x2",
"Unit": "iMC"
},
{
"BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; All DRAM WR_CAS (both Modes)",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x4", "EventCode": "0x4",
"EventName": "LLC_MISSES.MEM_WRITE", "EventName": "UNC_M_CAS_COUNT.WR",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "DRAM RD_CAS and WR_CAS Commands",
"UMask": "0xc", "UMask": "0xc",
"Unit": "iMC" "Unit": "iMC"
}, },
{ {
"BriefDescription": "Memory controller clock ticks. Used to get percentages of memory controller cycles events", "BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; DRAM WR_CAS (w/ and w/out auto-pre) in Read Major Mode",
"Counter": "0,1,2,3",
"EventCode": "0x4",
"EventName": "UNC_M_CAS_COUNT.WR_RMM",
"PerPkg": "1",
"PublicDescription": "DRAM RD_CAS and WR_CAS Commands",
"UMask": "0x8",
"Unit": "iMC"
},
{
"BriefDescription": "DRAM RD_CAS and WR_CAS Commands.; DRAM WR_CAS (w/ and w/out auto-pre) in Write Major Mode",
"Counter": "0,1,2,3",
"EventCode": "0x4",
"EventName": "UNC_M_CAS_COUNT.WR_WMM",
"PerPkg": "1",
"PublicDescription": "DRAM RD_CAS and WR_CAS Commands",
"UMask": "0x4",
"Unit": "iMC"
},
{
"BriefDescription": "uclks",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventName": "UNC_M_CLOCKTICKS", "EventName": "UNC_M_CLOCKTICKS",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Uncore Fixed Counter - uclks",
"Unit": "iMC"
},
{
"BriefDescription": "DRAM Precharge All Commands",
"Counter": "0,1,2,3",
"EventCode": "0x6",
"EventName": "UNC_M_DRAM_PRE_ALL",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that the precharge all command was sent.",
"Unit": "iMC"
},
{
"BriefDescription": "Number of DRAM Refreshes Issued",
"Counter": "0,1,2,3",
"EventCode": "0x5",
"EventName": "UNC_M_DRAM_REFRESH.HIGH",
"PerPkg": "1",
"PublicDescription": "Counts the number of refreshes issued.",
"UMask": "0x4",
"Unit": "iMC"
},
{
"BriefDescription": "Number of DRAM Refreshes Issued",
"Counter": "0,1,2,3",
"EventCode": "0x5",
"EventName": "UNC_M_DRAM_REFRESH.PANIC",
"PerPkg": "1",
"PublicDescription": "Counts the number of refreshes issued.",
"UMask": "0x2",
"Unit": "iMC"
},
{
"BriefDescription": "ECC Correctable Errors",
"Counter": "0,1,2,3",
"EventCode": "0x9",
"EventName": "UNC_M_ECC_CORRECTABLE_ERRORS",
"PerPkg": "1",
"PublicDescription": "Counts the number of ECC errors detected and corrected by the iMC on this channel. This counter is only useful with ECC DRAM devices. This count will increment one time for each correction regardless of the number of bits corrected. The iMC can correct up to 4 bit errors in independent channel mode and 8 bit erros in lockstep mode.",
"Unit": "iMC"
},
{
"BriefDescription": "Cycles in a Major Mode; Isoch Major Mode",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_M_MAJOR_MODES.ISOCH",
"PerPkg": "1",
"PublicDescription": "Counts the total number of cycles spent in a major mode (selected by a filter) on the given channel. Major modea are channel-wide, and not a per-rank (or dimm or bank) mode.",
"UMask": "0x8",
"Unit": "iMC" "Unit": "iMC"
}, },
{ {
"BriefDescription": "Cycles where DRAM ranks are in power down (CKE) mode", "BriefDescription": "Cycles in a Major Mode; Partial Major Mode",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_M_MAJOR_MODES.PARTIAL",
"PerPkg": "1",
"PublicDescription": "Counts the total number of cycles spent in a major mode (selected by a filter) on the given channel. Major modea are channel-wide, and not a per-rank (or dimm or bank) mode.",
"UMask": "0x4",
"Unit": "iMC"
},
{
"BriefDescription": "Cycles in a Major Mode; Read Major Mode",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_M_MAJOR_MODES.READ",
"PerPkg": "1",
"PublicDescription": "Counts the total number of cycles spent in a major mode (selected by a filter) on the given channel. Major modea are channel-wide, and not a per-rank (or dimm or bank) mode.",
"UMask": "0x1",
"Unit": "iMC"
},
{
"BriefDescription": "Cycles in a Major Mode; Write Major Mode",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_M_MAJOR_MODES.WRITE",
"PerPkg": "1",
"PublicDescription": "Counts the total number of cycles spent in a major mode (selected by a filter) on the given channel. Major modea are channel-wide, and not a per-rank (or dimm or bank) mode.",
"UMask": "0x2",
"Unit": "iMC"
},
{
"BriefDescription": "Channel DLLOFF Cycles",
"Counter": "0,1,2,3",
"EventCode": "0x84",
"EventName": "UNC_M_POWER_CHANNEL_DLLOFF",
"PerPkg": "1",
"PublicDescription": "Number of cycles when all the ranks in the channel are in CKE Slow (DLLOFF) mode.",
"Unit": "iMC"
},
{
"BriefDescription": "Channel PPD Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x85", "EventCode": "0x85",
"EventName": "UNC_M_POWER_CHANNEL_PPD", "EventName": "UNC_M_POWER_CHANNEL_PPD",
"MetricExpr": "(UNC_M_POWER_CHANNEL_PPD / UNC_M_CLOCKTICKS) * 100.",
"MetricName": "power_channel_ppd %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles when all the ranks in the channel are in PPD mode. If IBT=off is enabled, then this can be used to count those cycles. If it is not enabled, then this can count the number of cycles when that could have been taken advantage of.",
"Unit": "iMC" "Unit": "iMC"
}, },
{ {
"BriefDescription": "Cycles all ranks are in critical thermal throttle", "BriefDescription": "CKE_ON_CYCLES by Rank; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x83",
"EventName": "UNC_M_POWER_CKE_CYCLES.RANK0",
"PerPkg": "1",
"PublicDescription": "Number of cycles spent in CKE ON mode. The filter allows you to select a rank to monitor. If multiple ranks are in CKE ON mode at one time, the counter will ONLY increment by one rather than doing accumulation. Multiple counters will need to be used to track multiple ranks simultaneously. There is no distinction between the different CKE modes (APD, PPDS, PPDF). This can be determined based on the system programming. These events should commonly be used with Invert to get the number of cycles in power saving mode. Edge Detect is also useful here. Make sure that you do NOT use Invert with Edge Detect (this just confuses the system and is not necessary).",
"UMask": "0x1",
"Unit": "iMC"
},
{
"BriefDescription": "CKE_ON_CYCLES by Rank; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x83",
"EventName": "UNC_M_POWER_CKE_CYCLES.RANK1",
"PerPkg": "1",
"PublicDescription": "Number of cycles spent in CKE ON mode. The filter allows you to select a rank to monitor. If multiple ranks are in CKE ON mode at one time, the counter will ONLY increment by one rather than doing accumulation. Multiple counters will need to be used to track multiple ranks simultaneously. There is no distinction between the different CKE modes (APD, PPDS, PPDF). This can be determined based on the system programming. These events should commonly be used with Invert to get the number of cycles in power saving mode. Edge Detect is also useful here. Make sure that you do NOT use Invert with Edge Detect (this just confuses the system and is not necessary).",
"UMask": "0x2",
"Unit": "iMC"
},
{
"BriefDescription": "CKE_ON_CYCLES by Rank; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x83",
"EventName": "UNC_M_POWER_CKE_CYCLES.RANK2",
"PerPkg": "1",
"PublicDescription": "Number of cycles spent in CKE ON mode. The filter allows you to select a rank to monitor. If multiple ranks are in CKE ON mode at one time, the counter will ONLY increment by one rather than doing accumulation. Multiple counters will need to be used to track multiple ranks simultaneously. There is no distinction between the different CKE modes (APD, PPDS, PPDF). This can be determined based on the system programming. These events should commonly be used with Invert to get the number of cycles in power saving mode. Edge Detect is also useful here. Make sure that you do NOT use Invert with Edge Detect (this just confuses the system and is not necessary).",
"UMask": "0x4",
"Unit": "iMC"
},
{
"BriefDescription": "CKE_ON_CYCLES by Rank; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x83",
"EventName": "UNC_M_POWER_CKE_CYCLES.RANK3",
"PerPkg": "1",
"PublicDescription": "Number of cycles spent in CKE ON mode. The filter allows you to select a rank to monitor. If multiple ranks are in CKE ON mode at one time, the counter will ONLY increment by one rather than doing accumulation. Multiple counters will need to be used to track multiple ranks simultaneously. There is no distinction between the different CKE modes (APD, PPDS, PPDF). This can be determined based on the system programming. These events should commonly be used with Invert to get the number of cycles in power saving mode. Edge Detect is also useful here. Make sure that you do NOT use Invert with Edge Detect (this just confuses the system and is not necessary).",
"UMask": "0x8",
"Unit": "iMC"
},
{
"BriefDescription": "CKE_ON_CYCLES by Rank; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x83",
"EventName": "UNC_M_POWER_CKE_CYCLES.RANK4",
"PerPkg": "1",
"PublicDescription": "Number of cycles spent in CKE ON mode. The filter allows you to select a rank to monitor. If multiple ranks are in CKE ON mode at one time, the counter will ONLY increment by one rather than doing accumulation. Multiple counters will need to be used to track multiple ranks simultaneously. There is no distinction between the different CKE modes (APD, PPDS, PPDF). This can be determined based on the system programming. These events should commonly be used with Invert to get the number of cycles in power saving mode. Edge Detect is also useful here. Make sure that you do NOT use Invert with Edge Detect (this just confuses the system and is not necessary).",
"UMask": "0x10",
"Unit": "iMC"
},
{
"BriefDescription": "CKE_ON_CYCLES by Rank; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x83",
"EventName": "UNC_M_POWER_CKE_CYCLES.RANK5",
"PerPkg": "1",
"PublicDescription": "Number of cycles spent in CKE ON mode. The filter allows you to select a rank to monitor. If multiple ranks are in CKE ON mode at one time, the counter will ONLY increment by one rather than doing accumulation. Multiple counters will need to be used to track multiple ranks simultaneously. There is no distinction between the different CKE modes (APD, PPDS, PPDF). This can be determined based on the system programming. These events should commonly be used with Invert to get the number of cycles in power saving mode. Edge Detect is also useful here. Make sure that you do NOT use Invert with Edge Detect (this just confuses the system and is not necessary).",
"UMask": "0x20",
"Unit": "iMC"
},
{
"BriefDescription": "CKE_ON_CYCLES by Rank; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x83",
"EventName": "UNC_M_POWER_CKE_CYCLES.RANK6",
"PerPkg": "1",
"PublicDescription": "Number of cycles spent in CKE ON mode. The filter allows you to select a rank to monitor. If multiple ranks are in CKE ON mode at one time, the counter will ONLY increment by one rather than doing accumulation. Multiple counters will need to be used to track multiple ranks simultaneously. There is no distinction between the different CKE modes (APD, PPDS, PPDF). This can be determined based on the system programming. These events should commonly be used with Invert to get the number of cycles in power saving mode. Edge Detect is also useful here. Make sure that you do NOT use Invert with Edge Detect (this just confuses the system and is not necessary).",
"UMask": "0x40",
"Unit": "iMC"
},
{
"BriefDescription": "CKE_ON_CYCLES by Rank; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x83",
"EventName": "UNC_M_POWER_CKE_CYCLES.RANK7",
"PerPkg": "1",
"PublicDescription": "Number of cycles spent in CKE ON mode. The filter allows you to select a rank to monitor. If multiple ranks are in CKE ON mode at one time, the counter will ONLY increment by one rather than doing accumulation. Multiple counters will need to be used to track multiple ranks simultaneously. There is no distinction between the different CKE modes (APD, PPDS, PPDF). This can be determined based on the system programming. These events should commonly be used with Invert to get the number of cycles in power saving mode. Edge Detect is also useful here. Make sure that you do NOT use Invert with Edge Detect (this just confuses the system and is not necessary).",
"UMask": "0x80",
"Unit": "iMC"
},
{
"BriefDescription": "Critical Throttle Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x86", "EventCode": "0x86",
"EventName": "UNC_M_POWER_CRITICAL_THROTTLE_CYCLES", "EventName": "UNC_M_POWER_CRITICAL_THROTTLE_CYCLES",
"MetricExpr": "(UNC_M_POWER_CRITICAL_THROTTLE_CYCLES / UNC_M_CLOCKTICKS) * 100.",
"MetricName": "power_critical_throttle_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the iMC is in critical thermal throttling. When this happens, all traffic is blocked. This should be rare unless something bad is going on in the platform. There is no filtering by rank for this event.",
"Unit": "iMC" "Unit": "iMC"
}, },
{ {
"BriefDescription": "Cycles Memory is in self refresh power mode", "BriefDescription": "Clock-Enabled Self-Refresh",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x43", "EventCode": "0x43",
"EventName": "UNC_M_POWER_SELF_REFRESH", "EventName": "UNC_M_POWER_SELF_REFRESH",
"MetricExpr": "(UNC_M_POWER_SELF_REFRESH / UNC_M_CLOCKTICKS) * 100.",
"MetricName": "power_self_refresh %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the iMC is in self-refresh and the iMC still has a clock. This happens in some package C-states. For example, the PCU may ask the iMC to enter self-refresh even though some of the cores are still processing. One use of this is for Monroe technology. Self-refresh is required during package C3 and C6, but there is no clock in the iMC at this time, so it is not possible to count these cases.",
"Unit": "iMC"
},
{
"BriefDescription": "Throttle Cycles for Rank 0; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x41",
"EventName": "UNC_M_POWER_THROTTLE_CYCLES.RANK0",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles while the iMC is being throttled by either thermal constraints or by the PCU throttling. It is not possible to distinguish between the two. This can be filtered by rank. If multiple ranks are selected and are being throttled at the same time, the counter will only increment by 1.",
"UMask": "0x1",
"Unit": "iMC"
},
{
"BriefDescription": "Throttle Cycles for Rank 0; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x41",
"EventName": "UNC_M_POWER_THROTTLE_CYCLES.RANK1",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles while the iMC is being throttled by either thermal constraints or by the PCU throttling. It is not possible to distinguish between the two. This can be filtered by rank. If multiple ranks are selected and are being throttled at the same time, the counter will only increment by 1.",
"UMask": "0x2",
"Unit": "iMC"
},
{
"BriefDescription": "Throttle Cycles for Rank 0; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x41",
"EventName": "UNC_M_POWER_THROTTLE_CYCLES.RANK2",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles while the iMC is being throttled by either thermal constraints or by the PCU throttling. It is not possible to distinguish between the two. This can be filtered by rank. If multiple ranks are selected and are being throttled at the same time, the counter will only increment by 1.",
"UMask": "0x4",
"Unit": "iMC"
},
{
"BriefDescription": "Throttle Cycles for Rank 0; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x41",
"EventName": "UNC_M_POWER_THROTTLE_CYCLES.RANK3",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles while the iMC is being throttled by either thermal constraints or by the PCU throttling. It is not possible to distinguish between the two. This can be filtered by rank. If multiple ranks are selected and are being throttled at the same time, the counter will only increment by 1.",
"UMask": "0x8",
"Unit": "iMC"
},
{
"BriefDescription": "Throttle Cycles for Rank 0; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x41",
"EventName": "UNC_M_POWER_THROTTLE_CYCLES.RANK4",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles while the iMC is being throttled by either thermal constraints or by the PCU throttling. It is not possible to distinguish between the two. This can be filtered by rank. If multiple ranks are selected and are being throttled at the same time, the counter will only increment by 1.",
"UMask": "0x10",
"Unit": "iMC"
},
{
"BriefDescription": "Throttle Cycles for Rank 0; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x41",
"EventName": "UNC_M_POWER_THROTTLE_CYCLES.RANK5",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles while the iMC is being throttled by either thermal constraints or by the PCU throttling. It is not possible to distinguish between the two. This can be filtered by rank. If multiple ranks are selected and are being throttled at the same time, the counter will only increment by 1.",
"UMask": "0x20",
"Unit": "iMC"
},
{
"BriefDescription": "Throttle Cycles for Rank 0; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x41",
"EventName": "UNC_M_POWER_THROTTLE_CYCLES.RANK6",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles while the iMC is being throttled by either thermal constraints or by the PCU throttling. It is not possible to distinguish between the two. This can be filtered by rank. If multiple ranks are selected and are being throttled at the same time, the counter will only increment by 1.",
"UMask": "0x40",
"Unit": "iMC"
},
{
"BriefDescription": "Throttle Cycles for Rank 0; DIMM ID",
"Counter": "0,1,2,3",
"EventCode": "0x41",
"EventName": "UNC_M_POWER_THROTTLE_CYCLES.RANK7",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles while the iMC is being throttled by either thermal constraints or by the PCU throttling. It is not possible to distinguish between the two. This can be filtered by rank. If multiple ranks are selected and are being throttled at the same time, the counter will only increment by 1.",
"UMask": "0x80",
"Unit": "iMC"
},
{
"BriefDescription": "Read Preemption Count; Read over Read Preemption",
"Counter": "0,1,2,3",
"EventCode": "0x8",
"EventName": "UNC_M_PREEMPTION.RD_PREEMPT_RD",
"PerPkg": "1",
"PublicDescription": "Counts the number of times a read in the iMC preempts another read or write. Generally reads to an open page are issued ahead of requests to closed pages. This improves the page hit rate of the system. However, high priority requests can cause pages of active requests to be closed in order to get them out. This will reduce the latency of the high-priority request at the expense of lower bandwidth and increased overall average latency.",
"UMask": "0x1",
"Unit": "iMC"
},
{
"BriefDescription": "Read Preemption Count; Read over Write Preemption",
"Counter": "0,1,2,3",
"EventCode": "0x8",
"EventName": "UNC_M_PREEMPTION.RD_PREEMPT_WR",
"PerPkg": "1",
"PublicDescription": "Counts the number of times a read in the iMC preempts another read or write. Generally reads to an open page are issued ahead of requests to closed pages. This improves the page hit rate of the system. However, high priority requests can cause pages of active requests to be closed in order to get them out. This will reduce the latency of the high-priority request at the expense of lower bandwidth and increased overall average latency.",
"UMask": "0x2",
"Unit": "iMC"
},
{
"BriefDescription": "DRAM Precharge commands.; Precharge due to timer expiration",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_M_PRE_COUNT.PAGE_CLOSE",
"PerPkg": "1",
"PublicDescription": "Counts the number of DRAM Precharge commands sent on this channel.",
"UMask": "0x2",
"Unit": "iMC" "Unit": "iMC"
}, },
{ {
"BriefDescription": "Memory page conflicts", "BriefDescription": "DRAM Precharge commands.; Precharges due to page miss",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x2", "EventCode": "0x2",
"EventName": "UNC_M_PRE_COUNT.PAGE_MISS", "EventName": "UNC_M_PRE_COUNT.PAGE_MISS",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of DRAM Precharge commands sent on this channel.",
"UMask": "0x1", "UMask": "0x1",
"Unit": "iMC" "Unit": "iMC"
}, },
{ {
"BriefDescription": "Occupancy counter for memory read queue", "BriefDescription": "Read Pending Queue Full Cycles",
"Counter": "0,1,2,3",
"EventCode": "0x12",
"EventName": "UNC_M_RPQ_CYCLES_FULL",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the Read Pending Queue is full. When the RPQ is full, the HA will not be able to issue any additional read requests into the iMC. This count should be similar count in the HA which tracks the number of cycles that the HA has no RPQ credits, just somewhat smaller to account for the credit return overhead. We generally do not expect to see RPQ become full except for potentially during Write Major Mode or while running with slow DRAM. This event only tracks non-ISOC queue entries.",
"Unit": "iMC"
},
{
"BriefDescription": "Read Pending Queue Not Empty",
"Counter": "0,1,2,3",
"EventCode": "0x11",
"EventName": "UNC_M_RPQ_CYCLES_NE",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the Read Pending Queue is not empty. This can then be used to calculate the average occupancy (in conjunction with the Read Pending Queue Occupancy count). The RPQ is used to schedule reads out to the memory controller and to track the requests. Requests allocate into the RPQ soon after they enter the memory controller, and need credits for an entry in this buffer before being sent from the HA to the iMC. They deallocate after the CAS command has been issued to memory. This filter is to be used in conjunction with the occupancy filter so that one can correctly track the average occupancies for schedulable entries and scheduled requests.",
"Unit": "iMC"
},
{
"BriefDescription": "Read Pending Queue Allocations",
"Counter": "0,1,2,3",
"EventCode": "0x10",
"EventName": "UNC_M_RPQ_INSERTS",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the Read Pending Queue. This queue is used to schedule reads out to the memory controller and to track the requests. Requests allocate into the RPQ soon after they enter the memory controller, and need credits for an entry in this buffer before being sent from the HA to the iMC. They deallocate after the CAS command has been issued to memory. This includes both ISOCH and non-ISOCH requests.",
"Unit": "iMC"
},
{
"BriefDescription": "Read Pending Queue Occupancy",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x80", "EventCode": "0x80",
"EventName": "UNC_M_RPQ_OCCUPANCY", "EventName": "UNC_M_RPQ_OCCUPANCY",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Accumulates the occupancies of the Read Pending Queue each cycle. This can then be used to calculate both the average occupancy (in conjunction with the number of cycles not empty) and the average latency (in conjunction with the number of allocations). The RPQ is used to schedule reads out to the memory controller and to track the requests. Requests allocate into the RPQ soon after they enter the memory controller, and need credits for an entry in this buffer before being sent from the HA to the iMC. They deallocate after the CAS command has been issued to memory.",
"Unit": "iMC"
},
{
"BriefDescription": "Write Pending Queue Full Cycles",
"Counter": "0,1,2,3",
"EventCode": "0x22",
"EventName": "UNC_M_WPQ_CYCLES_FULL",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the Write Pending Queue is full. When the WPQ is full, the HA will not be able to issue any additional read requests into the iMC. This count should be similar count in the HA which tracks the number of cycles that the HA has no WPQ credits, just somewhat smaller to account for the credit return overhead.",
"Unit": "iMC"
},
{
"BriefDescription": "Write Pending Queue Not Empty",
"Counter": "0,1,2,3",
"EventCode": "0x21",
"EventName": "UNC_M_WPQ_CYCLES_NE",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the Write Pending Queue is not empty. This can then be used to calculate the average queue occupancy (in conjunction with the WPQ Occupancy Accumulation count). The WPQ is used to schedule write out to the memory controller and to track the writes. Requests allocate into the WPQ soon after they enter the memory controller, and need credits for an entry in this buffer before being sent from the HA to the iMC. They deallocate after being issued to DRAM. Write requests themselves are able to complete (from the perspective of the rest of the system) as soon they have 'posted' to the iMC. This is not to be confused with actually performing the write to DRAM. Therefore, the average latency for this queue is actually not useful for deconstruction intermediate write latencies.",
"Unit": "iMC"
},
{
"BriefDescription": "Write Pending Queue Allocations",
"Counter": "0,1,2,3",
"EventCode": "0x20",
"EventName": "UNC_M_WPQ_INSERTS",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the Write Pending Queue. This can then be used to calculate the average queuing latency (in conjunction with the WPQ occupancy count). The WPQ is used to schedule write out to the memory controller and to track the writes. Requests allocate into the WPQ soon after they enter the memory controller, and need credits for an entry in this buffer before being sent from the HA to the iMC. They deallocate after being issued to DRAM. Write requests themselves are able to complete (from the perspective of the rest of the system) as soon they have 'posted' to the iMC.",
"Unit": "iMC"
},
{
"BriefDescription": "Write Pending Queue Occupancy",
"Counter": "0,1,2,3",
"EventCode": "0x81",
"EventName": "UNC_M_WPQ_OCCUPANCY",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancies of the Write Pending Queue each cycle. This can then be used to calculate both the average queue occupancy (in conjunction with the number of cycles not empty) and the average latency (in conjunction with the number of allocations). The WPQ is used to schedule write out to the memory controller and to track the writes. Requests allocate into the WPQ soon after they enter the memory controller, and need credits for an entry in this buffer before being sent from the HA to the iMC. They deallocate after being issued to DRAM. Write requests themselves are able to complete (from the perspective of the rest of the system) as soon they have 'posted' to the iMC. This is not to be confused with actually performing the write to DRAM. Therefore, the average latency for this queue is actually not useful for deconstruction intermediate write latencies. So, we provide filtering based on if the request has posted or not. By using the 'not posted' filter, we can track how long writes spent in the iMC before completions were sent to the HA. The 'posted' filter, on the other hand, provides information about how much queueing is actually happenning in the iMC for writes before they are actually issued to memory. High average occupancies will generally coincide with high write major mode counts.",
"Unit": "iMC"
},
{
"BriefDescription": "Write Pending Queue CAM Match",
"Counter": "0,1,2,3",
"EventCode": "0x23",
"EventName": "UNC_M_WPQ_READ_HIT",
"PerPkg": "1",
"PublicDescription": "Counts the number of times a request hits in the WPQ (write-pending queue). The iMC allows writes and reads to pass up other writes to different addresses. Before a read or a write is issued, it will first CAM the WPQ to see if there is a write pending to that address. When reads hit, they are able to directly pull their data from the WPQ instead of going to memory. Writes that hit will overwrite the existing data. Partial writes that hit will not need to do underfill reads and will simply update their relevant sections.",
"Unit": "iMC"
},
{
"BriefDescription": "Write Pending Queue CAM Match",
"Counter": "0,1,2,3",
"EventCode": "0x24",
"EventName": "UNC_M_WPQ_WRITE_HIT",
"PerPkg": "1",
"PublicDescription": "Counts the number of times a request hits in the WPQ (write-pending queue). The iMC allows writes and reads to pass up other writes to different addresses. Before a read or a write is issued, it will first CAM the WPQ to see if there is a write pending to that address. When reads hit, they are able to directly pull their data from the WPQ instead of going to memory. Writes that hit will overwrite the existing data. Partial writes that hit will not need to do underfill reads and will simply update their relevant sections.",
"Unit": "iMC" "Unit": "iMC"
} }
] ]
tools/perf/pmu-events/arch/x86/jaketown/uncore-other.json 0 → 100644
View file @ 376d8b58
[
{
"BriefDescription": "Address Match (Conflict) Count; Conflict Merges",
"Counter": "0,1",
"EventCode": "0x17",
"EventName": "UNC_I_ADDRESS_MATCH.MERGE_COUNT",
"PerPkg": "1",
"PublicDescription": "Counts the number of times when an inbound write (from a device to memory or another device) had an address match with another request in the write cache.",
"UMask": "0x2",
"Unit": "IRP"
},
{
"BriefDescription": "Address Match (Conflict) Count; Conflict Stalls",
"Counter": "0,1",
"EventCode": "0x17",
"EventName": "UNC_I_ADDRESS_MATCH.STALL_COUNT",
"PerPkg": "1",
"PublicDescription": "Counts the number of times when an inbound write (from a device to memory or another device) had an address match with another request in the write cache.",
"UMask": "0x1",
"Unit": "IRP"
},
{
"BriefDescription": "Write Ack Pending Occupancy; Any Source",
"Counter": "0,1",
"EventCode": "0x14",
"EventName": "UNC_I_CACHE_ACK_PENDING_OCCUPANCY.ANY",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of writes that have acquired ownership but have not yet returned their data to the uncore. These writes are generally queued up in the switch trying to get to the head of their queues so that they can post their data. The queue occuapancy increments when the ACK is received, and decrements when either the data is returned OR a tickle is received and ownership is released. Note that a single tickle can result in multiple decrements.",
"UMask": "0x1",
"Unit": "IRP"
},
{
"BriefDescription": "Write Ack Pending Occupancy; Select Source",
"Counter": "0,1",
"EventCode": "0x14",
"EventName": "UNC_I_CACHE_ACK_PENDING_OCCUPANCY.SOURCE",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of writes that have acquired ownership but have not yet returned their data to the uncore. These writes are generally queued up in the switch trying to get to the head of their queues so that they can post their data. The queue occuapancy increments when the ACK is received, and decrements when either the data is returned OR a tickle is received and ownership is released. Note that a single tickle can result in multiple decrements.",
"UMask": "0x2",
"Unit": "IRP"
},
{
"BriefDescription": "Outstanding Write Ownership Occupancy; Any Source",
"Counter": "0,1",
"EventCode": "0x13",
"EventName": "UNC_I_CACHE_OWN_OCCUPANCY.ANY",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of writes (and write prefetches) that are outstanding in the uncore trying to acquire ownership in each cycle. This can be used with the write transaction count to calculate the average write latency in the uncore. The occupancy increments when a write request is issued, and decrements when the data is returned.",
"UMask": "0x1",
"Unit": "IRP"
},
{
"BriefDescription": "Outstanding Write Ownership Occupancy; Select Source",
"Counter": "0,1",
"EventCode": "0x13",
"EventName": "UNC_I_CACHE_OWN_OCCUPANCY.SOURCE",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of writes (and write prefetches) that are outstanding in the uncore trying to acquire ownership in each cycle. This can be used with the write transaction count to calculate the average write latency in the uncore. The occupancy increments when a write request is issued, and decrements when the data is returned.",
"UMask": "0x2",
"Unit": "IRP"
},
{
"BriefDescription": "Outstanding Read Occupancy; Any Source",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_I_CACHE_READ_OCCUPANCY.ANY",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of reads that are outstanding in the uncore in each cycle. This can be used with the read transaction count to calculate the average read latency in the uncore. The occupancy increments when a read request is issued, and decrements when the data is returned.",
"UMask": "0x1",
"Unit": "IRP"
},
{
"BriefDescription": "Outstanding Read Occupancy; Select Source",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_I_CACHE_READ_OCCUPANCY.SOURCE",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of reads that are outstanding in the uncore in each cycle. This can be used with the read transaction count to calculate the average read latency in the uncore. The occupancy increments when a read request is issued, and decrements when the data is returned.",
"UMask": "0x2",
"Unit": "IRP"
},
{
"BriefDescription": "Total Write Cache Occupancy; Any Source",
"Counter": "0,1",
"EventCode": "0x12",
"EventName": "UNC_I_CACHE_TOTAL_OCCUPANCY.ANY",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of reads and writes that are outstanding in the uncore in each cycle. This is effectively the sum of the READ_OCCUPANCY and WRITE_OCCUPANCY events.",
"UMask": "0x1",
"Unit": "IRP"
},
{
"BriefDescription": "Total Write Cache Occupancy; Select Source",
"Counter": "0,1",
"EventCode": "0x12",
"EventName": "UNC_I_CACHE_TOTAL_OCCUPANCY.SOURCE",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of reads and writes that are outstanding in the uncore in each cycle. This is effectively the sum of the READ_OCCUPANCY and WRITE_OCCUPANCY events.",
"UMask": "0x2",
"Unit": "IRP"
},
{
"BriefDescription": "Outstanding Write Occupancy; Any Source",
"Counter": "0,1",
"EventCode": "0x11",
"EventName": "UNC_I_CACHE_WRITE_OCCUPANCY.ANY",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of writes (and write prefetches) that are outstanding in the uncore in each cycle. This can be used with the transaction count event to calculate the average latency in the uncore. The occupancy increments when the ownership fetch/prefetch is issued, and decrements the data is returned to the uncore.",
"UMask": "0x1",
"Unit": "IRP"
},
{
"BriefDescription": "Outstanding Write Occupancy; Select Source",
"Counter": "0,1",
"EventCode": "0x11",
"EventName": "UNC_I_CACHE_WRITE_OCCUPANCY.SOURCE",
"PerPkg": "1",
"PublicDescription": "Accumulates the number of writes (and write prefetches) that are outstanding in the uncore in each cycle. This can be used with the transaction count event to calculate the average latency in the uncore. The occupancy increments when the ownership fetch/prefetch is issued, and decrements the data is returned to the uncore.",
"UMask": "0x2",
"Unit": "IRP"
},
{
"BriefDescription": "Clocks in the IRP",
"Counter": "0,1",
"EventName": "UNC_I_CLOCKTICKS",
"PerPkg": "1",
"PublicDescription": "Number of clocks in the IRP.",
"Unit": "IRP"
},
{
"Counter": "0,1",
"EventCode": "0xB",
"EventName": "UNC_I_RxR_AK_CYCLES_FULL",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the AK Ingress is full. This queue is where the IRP receives responses from R2PCIe (the ring).",
"Unit": "IRP"
},
{
"BriefDescription": "AK Ingress Occupancy",
"Counter": "0,1",
"EventCode": "0xA",
"EventName": "UNC_I_RxR_AK_INSERTS",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the AK Ingress. This queue is where the IRP receives responses from R2PCIe (the ring).",
"Unit": "IRP"
},
{
"Counter": "0,1",
"EventCode": "0xC",
"EventName": "UNC_I_RxR_AK_OCCUPANCY",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of the AK Ingress in each cycles. This queue is where the IRP receives responses from R2PCIe (the ring).",
"Unit": "IRP"
},
{
"Counter": "0,1",
"EventCode": "0x4",
"EventName": "UNC_I_RxR_BL_DRS_CYCLES_FULL",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the BL Ingress is full. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"BriefDescription": "BL Ingress Occupancy - DRS",
"Counter": "0,1",
"EventCode": "0x1",
"EventName": "UNC_I_RxR_BL_DRS_INSERTS",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the BL Ingress. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"Counter": "0,1",
"EventCode": "0x7",
"EventName": "UNC_I_RxR_BL_DRS_OCCUPANCY",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of the BL Ingress in each cycles. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"Counter": "0,1",
"EventCode": "0x5",
"EventName": "UNC_I_RxR_BL_NCB_CYCLES_FULL",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the BL Ingress is full. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"BriefDescription": "BL Ingress Occupancy - NCB",
"Counter": "0,1",
"EventCode": "0x2",
"EventName": "UNC_I_RxR_BL_NCB_INSERTS",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the BL Ingress. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"Counter": "0,1",
"EventCode": "0x8",
"EventName": "UNC_I_RxR_BL_NCB_OCCUPANCY",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of the BL Ingress in each cycles. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"Counter": "0,1",
"EventCode": "0x6",
"EventName": "UNC_I_RxR_BL_NCS_CYCLES_FULL",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the BL Ingress is full. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"BriefDescription": "BL Ingress Occupancy - NCS",
"Counter": "0,1",
"EventCode": "0x3",
"EventName": "UNC_I_RxR_BL_NCS_INSERTS",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the BL Ingress. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"Counter": "0,1",
"EventCode": "0x9",
"EventName": "UNC_I_RxR_BL_NCS_OCCUPANCY",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of the BL Ingress in each cycles. This queue is where the IRP receives data from R2PCIe (the ring). It is used for data returns from read requets as well as outbound MMIO writes.",
"Unit": "IRP"
},
{
"BriefDescription": "Tickle Count; Ownership Lost",
"Counter": "0,1",
"EventCode": "0x16",
"EventName": "UNC_I_TICKLES.LOST_OWNERSHIP",
"PerPkg": "1",
"PublicDescription": "Counts the number of tickles that are received. This is for both explicit (from Cbo) and implicit (internal conflict) tickles.",
"UMask": "0x1",
"Unit": "IRP"
},
{
"BriefDescription": "Tickle Count; Data Returned",
"Counter": "0,1",
"EventCode": "0x16",
"EventName": "UNC_I_TICKLES.TOP_OF_QUEUE",
"PerPkg": "1",
"PublicDescription": "Counts the number of tickles that are received. This is for both explicit (from Cbo) and implicit (internal conflict) tickles.",
"UMask": "0x2",
"Unit": "IRP"
},
{
"BriefDescription": "Inbound Transaction Count; Select Source",
"Counter": "0,1",
"EventCode": "0x15",
"EventName": "UNC_I_TRANSACTIONS.ORDERINGQ",
"Filter": "IRPFilter[4:0]",
"PerPkg": "1",
"PublicDescription": "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.",
"UMask": "0x8",
"Unit": "IRP"
},
{
"BriefDescription": "Inbound Transaction Count; Read Prefetches",
"Counter": "0,1",
"EventCode": "0x15",
"EventName": "UNC_I_TRANSACTIONS.PD_PREFETCHES",
"PerPkg": "1",
"PublicDescription": "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.",
"UMask": "0x4",
"Unit": "IRP"
},
{
"BriefDescription": "Inbound Transaction Count; Reads",
"Counter": "0,1",
"EventCode": "0x15",
"EventName": "UNC_I_TRANSACTIONS.READS",
"PerPkg": "1",
"PublicDescription": "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.",
"UMask": "0x1",
"Unit": "IRP"
},
{
"BriefDescription": "Inbound Transaction Count; Writes",
"Counter": "0,1",
"EventCode": "0x15",
"EventName": "UNC_I_TRANSACTIONS.WRITES",
"PerPkg": "1",
"PublicDescription": "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.",
"UMask": "0x2",
"Unit": "IRP"
},
{
"BriefDescription": "No AD Egress Credit Stalls",
"Counter": "0,1",
"EventCode": "0x18",
"EventName": "UNC_I_TxR_AD_STALL_CREDIT_CYCLES",
"PerPkg": "1",
"PublicDescription": "Counts the number times when it is not possible to issue a request to the R2PCIe because there are no AD Egress Credits available.",
"Unit": "IRP"
},
{
"BriefDescription": "No BL Egress Credit Stalls",
"Counter": "0,1",
"EventCode": "0x19",
"EventName": "UNC_I_TxR_BL_STALL_CREDIT_CYCLES",
"PerPkg": "1",
"PublicDescription": "Counts the number times when it is not possible to issue data to the R2PCIe because there are no BL Egress Credits available.",
"Unit": "IRP"
},
{
"BriefDescription": "Outbound Read Requests",
"Counter": "0,1",
"EventCode": "0xE",
"EventName": "UNC_I_TxR_DATA_INSERTS_NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of requests issued to the switch (towards the devices).",
"Unit": "IRP"
},
{
"BriefDescription": "Outbound Read Requests",
"Counter": "0,1",
"EventCode": "0xF",
"EventName": "UNC_I_TxR_DATA_INSERTS_NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of requests issued to the switch (towards the devices).",
"Unit": "IRP"
},
{
"BriefDescription": "Outbound Request Queue Occupancy",
"Counter": "0,1",
"EventCode": "0xD",
"EventName": "UNC_I_TxR_REQUEST_OCCUPANCY",
"PerPkg": "1",
"PublicDescription": "Accumultes the number of outstanding outbound requests from the IRP to the switch (towards the devices). This can be used in conjuection with the allocations event in order to calculate average latency of outbound requests.",
"Unit": "IRP"
},
{
"BriefDescription": "Write Ordering Stalls",
"Counter": "0,1",
"EventCode": "0x1A",
"EventName": "UNC_I_WRITE_ORDERING_STALL_CYCLES",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when there are pending write ACK's in the switch but the switch->IRP pipeline is not utilized.",
"Unit": "IRP"
},
{
"BriefDescription": "Number of uclks in domain",
"Counter": "0,1,2,3",
"EventCode": "0x1",
"EventName": "UNC_R2_CLOCKTICKS",
"PerPkg": "1",
"PublicDescription": "Counts the number of uclks in the R2PCIe uclk domain. This could be slightly different than the count in the Ubox because of enable/freeze delays. However, because the R2PCIe is close to the Ubox, they generally should not diverge by more than a handful of cycles.",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Credit Acquired; DRS",
"Counter": "0,1",
"EventCode": "0x33",
"EventName": "UNC_R2_IIO_CREDITS_ACQUIRED.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of credits that are acquired in the R2PCIe agent for sending transactions into the IIO on either NCB or NCS are in use. Transactions from the BL ring going into the IIO Agent must first acquire a credit. These credits are for either the NCB or NCS message classes. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x8",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Credit Acquired; NCB",
"Counter": "0,1",
"EventCode": "0x33",
"EventName": "UNC_R2_IIO_CREDITS_ACQUIRED.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of credits that are acquired in the R2PCIe agent for sending transactions into the IIO on either NCB or NCS are in use. Transactions from the BL ring going into the IIO Agent must first acquire a credit. These credits are for either the NCB or NCS message classes. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x10",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Credit Acquired; NCS",
"Counter": "0,1",
"EventCode": "0x33",
"EventName": "UNC_R2_IIO_CREDITS_ACQUIRED.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of credits that are acquired in the R2PCIe agent for sending transactions into the IIO on either NCB or NCS are in use. Transactions from the BL ring going into the IIO Agent must first acquire a credit. These credits are for either the NCB or NCS message classes. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x20",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Failed to Acquire a Credit; DRS",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R2_IIO_CREDITS_REJECT.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that a request pending in the BL Ingress attempted to acquire either a NCB or NCS credit to transmit into the IIO, but was rejected because no credits were available. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x8",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Failed to Acquire a Credit; NCB",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R2_IIO_CREDITS_REJECT.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that a request pending in the BL Ingress attempted to acquire either a NCB or NCS credit to transmit into the IIO, but was rejected because no credits were available. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x10",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Failed to Acquire a Credit; NCS",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R2_IIO_CREDITS_REJECT.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that a request pending in the BL Ingress attempted to acquire either a NCB or NCS credit to transmit into the IIO, but was rejected because no credits were available. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x20",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Credits in Use; DRS",
"Counter": "0,1",
"EventCode": "0x32",
"EventName": "UNC_R2_IIO_CREDITS_USED.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when one or more credits in the R2PCIe agent for sending transactions into the IIO on either NCB or NCS are in use. Transactions from the BL ring going into the IIO Agent must first acquire a credit. These credits are for either the NCB or NCS message classes. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x8",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Credits in Use; NCB",
"Counter": "0,1",
"EventCode": "0x32",
"EventName": "UNC_R2_IIO_CREDITS_USED.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when one or more credits in the R2PCIe agent for sending transactions into the IIO on either NCB or NCS are in use. Transactions from the BL ring going into the IIO Agent must first acquire a credit. These credits are for either the NCB or NCS message classes. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x10",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2PCIe IIO Credits in Use; NCS",
"Counter": "0,1",
"EventCode": "0x32",
"EventName": "UNC_R2_IIO_CREDITS_USED.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when one or more credits in the R2PCIe agent for sending transactions into the IIO on either NCB or NCS are in use. Transactions from the BL ring going into the IIO Agent must first acquire a credit. These credits are for either the NCB or NCS message classes. NCB, or non-coherent bypass messages are used to transmit data without coherency (and are common). NCS is used for reads to PCIe (and should be used sparingly).",
"UMask": "0x20",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 AD Ring in Use; Counterclockwise and Even",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_R2_RING_AD_USED.CCW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AD ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x4",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 AD Ring in Use; Counterclockwise and Odd",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_R2_RING_AD_USED.CCW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AD ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x8",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 AD Ring in Use; Clockwise and Even",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_R2_RING_AD_USED.CW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AD ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x1",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 AD Ring in Use; Clockwise and Odd",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_R2_RING_AD_USED.CW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AD ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x2",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 AK Ring in Use; Counterclockwise and Even",
"Counter": "0,1,2,3",
"EventCode": "0x8",
"EventName": "UNC_R2_RING_AK_USED.CCW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AK ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x4",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 AK Ring in Use; Counterclockwise and Odd",
"Counter": "0,1,2,3",
"EventCode": "0x8",
"EventName": "UNC_R2_RING_AK_USED.CCW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AK ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x8",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 AK Ring in Use; Clockwise and Even",
"Counter": "0,1,2,3",
"EventCode": "0x8",
"EventName": "UNC_R2_RING_AK_USED.CW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AK ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x1",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 AK Ring in Use; Clockwise and Odd",
"Counter": "0,1,2,3",
"EventCode": "0x8",
"EventName": "UNC_R2_RING_AK_USED.CW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AK ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x2",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 BL Ring in Use; Counterclockwise and Even",
"Counter": "0,1,2,3",
"EventCode": "0x9",
"EventName": "UNC_R2_RING_BL_USED.CCW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the BL ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x4",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 BL Ring in Use; Counterclockwise and Odd",
"Counter": "0,1,2,3",
"EventCode": "0x9",
"EventName": "UNC_R2_RING_BL_USED.CCW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the BL ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x8",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 BL Ring in Use; Clockwise and Even",
"Counter": "0,1,2,3",
"EventCode": "0x9",
"EventName": "UNC_R2_RING_BL_USED.CW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the BL ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x1",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 BL Ring in Use; Clockwise and Odd",
"Counter": "0,1,2,3",
"EventCode": "0x9",
"EventName": "UNC_R2_RING_BL_USED.CW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the BL ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x2",
"Unit": "R2PCIe"
},
{
"BriefDescription": "R2 IV Ring in Use; Any",
"Counter": "0,1,2,3",
"EventCode": "0xa",
"EventName": "UNC_R2_RING_IV_USED.ANY",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the IV ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sunk into the ring stop. The IV ring is unidirectional. Whether UP or DN is used is dependent on the system programming. Thereofore, one should generally set both the UP and DN bits for a given polarity (or both) at a given time.",
"UMask": "0xf",
"Unit": "R2PCIe"
},
{
"BriefDescription": "AK Ingress Bounced",
"EventCode": "0x12",
"EventName": "UNC_R2_RxR_AK_BOUNCES",
"PerPkg": "1",
"PublicDescription": "Counts the number of times when a request destined for the AK ingress bounced.",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Ingress Cycles Not Empty; DRS",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R2_RxR_CYCLES_NE.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Ingress is not empty. This tracks one of the three rings that are used by the R2PCIe agent. This can be used in conjunction with the R2PCIe Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x8",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Ingress Cycles Not Empty; NCB",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R2_RxR_CYCLES_NE.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Ingress is not empty. This tracks one of the three rings that are used by the R2PCIe agent. This can be used in conjunction with the R2PCIe Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x10",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Ingress Cycles Not Empty; NCS",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R2_RxR_CYCLES_NE.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Ingress is not empty. This tracks one of the three rings that are used by the R2PCIe agent. This can be used in conjunction with the R2PCIe Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x20",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress Cycles Full; AD",
"EventCode": "0x25",
"EventName": "UNC_R2_TxR_CYCLES_FULL.AD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Egress buffer is full.",
"UMask": "0x1",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress Cycles Full; AK",
"EventCode": "0x25",
"EventName": "UNC_R2_TxR_CYCLES_FULL.AK",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Egress buffer is full.",
"UMask": "0x2",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress Cycles Full; BL",
"EventCode": "0x25",
"EventName": "UNC_R2_TxR_CYCLES_FULL.BL",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Egress buffer is full.",
"UMask": "0x4",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress Cycles Not Empty; AD",
"EventCode": "0x23",
"EventName": "UNC_R2_TxR_CYCLES_NE.AD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Egress is not empty. This tracks one of the three rings that are used by the R2PCIe agent. This can be used in conjunction with the R2PCIe Egress Occupancy Accumulator event in order to calculate average queue occupancy. Only a single Egress queue can be tracked at any given time. It is not possible to filter based on direction or polarity.",
"UMask": "0x1",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress Cycles Not Empty; AK",
"EventCode": "0x23",
"EventName": "UNC_R2_TxR_CYCLES_NE.AK",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Egress is not empty. This tracks one of the three rings that are used by the R2PCIe agent. This can be used in conjunction with the R2PCIe Egress Occupancy Accumulator event in order to calculate average queue occupancy. Only a single Egress queue can be tracked at any given time. It is not possible to filter based on direction or polarity.",
"UMask": "0x2",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress Cycles Not Empty; BL",
"EventCode": "0x23",
"EventName": "UNC_R2_TxR_CYCLES_NE.BL",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the R2PCIe Egress is not empty. This tracks one of the three rings that are used by the R2PCIe agent. This can be used in conjunction with the R2PCIe Egress Occupancy Accumulator event in order to calculate average queue occupancy. Only a single Egress queue can be tracked at any given time. It is not possible to filter based on direction or polarity.",
"UMask": "0x4",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress NACK; AD",
"Counter": "0,1",
"EventCode": "0x26",
"EventName": "UNC_R2_TxR_NACKS.AD",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that the Egress received a NACK from the ring and could not issue a transaction.",
"UMask": "0x1",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress NACK; AK",
"Counter": "0,1",
"EventCode": "0x26",
"EventName": "UNC_R2_TxR_NACKS.AK",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that the Egress received a NACK from the ring and could not issue a transaction.",
"UMask": "0x2",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Egress NACK; BL",
"Counter": "0,1",
"EventCode": "0x26",
"EventName": "UNC_R2_TxR_NACKS.BL",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that the Egress received a NACK from the ring and could not issue a transaction.",
"UMask": "0x4",
"Unit": "R2PCIe"
},
{
"BriefDescription": "Number of uclks in domain",
"Counter": "0,1,2",
"EventCode": "0x1",
"EventName": "UNC_R3_CLOCKTICKS",
"PerPkg": "1",
"PublicDescription": "Counts the number of uclks in the QPI uclk domain. This could be slightly different than the count in the Ubox because of enable/freeze delays. However, because the QPI Agent is close to the Ubox, they generally should not diverge by more than a handful of cycles.",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit Acquired",
"Counter": "0,1",
"EventCode": "0x20",
"EventName": "UNC_R3_IIO_CREDITS_ACQUIRED.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of times the NCS/NCB/DRS credit is acquried in the QPI for sending messages on BL to the IIO. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit Acquired",
"Counter": "0,1",
"EventCode": "0x20",
"EventName": "UNC_R3_IIO_CREDITS_ACQUIRED.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of times the NCS/NCB/DRS credit is acquried in the QPI for sending messages on BL to the IIO. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit Acquired",
"Counter": "0,1",
"EventCode": "0x20",
"EventName": "UNC_R3_IIO_CREDITS_ACQUIRED.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of times the NCS/NCB/DRS credit is acquried in the QPI for sending messages on BL to the IIO. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit Rejected",
"Counter": "0,1",
"EventCode": "0x21",
"EventName": "UNC_R3_IIO_CREDITS_REJECT.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that a request attempted to acquire an NCS/NCB/DRS credit in the QPI for sending messages on BL to the IIO but was rejected because no credit was available. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit Rejected",
"Counter": "0,1",
"EventCode": "0x21",
"EventName": "UNC_R3_IIO_CREDITS_REJECT.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that a request attempted to acquire an NCS/NCB/DRS credit in the QPI for sending messages on BL to the IIO but was rejected because no credit was available. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit Rejected",
"Counter": "0,1",
"EventCode": "0x21",
"EventName": "UNC_R3_IIO_CREDITS_REJECT.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of times that a request attempted to acquire an NCS/NCB/DRS credit in the QPI for sending messages on BL to the IIO but was rejected because no credit was available. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit In Use",
"Counter": "0,1",
"EventCode": "0x22",
"EventName": "UNC_R3_IIO_CREDITS_USED.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the NCS/NCB/DRS credit is in use in the QPI for sending messages on BL to the IIO. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit In Use",
"Counter": "0,1",
"EventCode": "0x22",
"EventName": "UNC_R3_IIO_CREDITS_USED.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the NCS/NCB/DRS credit is in use in the QPI for sending messages on BL to the IIO. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "to IIO BL Credit In Use",
"Counter": "0,1",
"EventCode": "0x22",
"EventName": "UNC_R3_IIO_CREDITS_USED.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the NCS/NCB/DRS credit is in use in the QPI for sending messages on BL to the IIO. There is one credit for each of these three message classes (three credits total). NCS is used for reads to PCIe space, NCB is used for transfering data without coherency, and DRS is used for transfering data with coherency (cachable PCI transactions). This event can only track one message class at a time.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 AD Ring in Use; Counterclockwise and Even",
"Counter": "0,1,2",
"EventCode": "0x7",
"EventName": "UNC_R3_RING_AD_USED.CCW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AD ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 AD Ring in Use; Counterclockwise and Odd",
"Counter": "0,1,2",
"EventCode": "0x7",
"EventName": "UNC_R3_RING_AD_USED.CCW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AD ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 AD Ring in Use; Clockwise and Even",
"Counter": "0,1,2",
"EventCode": "0x7",
"EventName": "UNC_R3_RING_AD_USED.CW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AD ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 AD Ring in Use; Clockwise and Odd",
"Counter": "0,1,2",
"EventCode": "0x7",
"EventName": "UNC_R3_RING_AD_USED.CW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AD ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 AK Ring in Use; Counterclockwise and Even",
"Counter": "0,1,2",
"EventCode": "0x8",
"EventName": "UNC_R3_RING_AK_USED.CCW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AK ring is being used at this ring stop. This includes when packets are passing by and when packets are being sent, but does not include when packets are being sunk into the ring stop.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 AK Ring in Use; Counterclockwise and Odd",
"Counter": "0,1,2",
"EventCode": "0x8",
"EventName": "UNC_R3_RING_AK_USED.CCW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AK ring is being used at this ring stop. This includes when packets are passing by and when packets are being sent, but does not include when packets are being sunk into the ring stop.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 AK Ring in Use; Clockwise and Even",
"Counter": "0,1,2",
"EventCode": "0x8",
"EventName": "UNC_R3_RING_AK_USED.CW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AK ring is being used at this ring stop. This includes when packets are passing by and when packets are being sent, but does not include when packets are being sunk into the ring stop.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 AK Ring in Use; Clockwise and Odd",
"Counter": "0,1,2",
"EventCode": "0x8",
"EventName": "UNC_R3_RING_AK_USED.CW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the AK ring is being used at this ring stop. This includes when packets are passing by and when packets are being sent, but does not include when packets are being sunk into the ring stop.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 BL Ring in Use; Counterclockwise and Even",
"Counter": "0,1,2",
"EventCode": "0x9",
"EventName": "UNC_R3_RING_BL_USED.CCW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the BL ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 BL Ring in Use; Counterclockwise and Odd",
"Counter": "0,1,2",
"EventCode": "0x9",
"EventName": "UNC_R3_RING_BL_USED.CCW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the BL ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 BL Ring in Use; Clockwise and Even",
"Counter": "0,1,2",
"EventCode": "0x9",
"EventName": "UNC_R3_RING_BL_USED.CW_EVEN",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the BL ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 BL Ring in Use; Clockwise and Odd",
"Counter": "0,1,2",
"EventCode": "0x9",
"EventName": "UNC_R3_RING_BL_USED.CW_ODD",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the BL ring is being used at this ring stop. This includes when packets are passing by and when packets are being sunk, but does not include when packets are being sent from the ring stop.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "R3 IV Ring in Use; Any",
"Counter": "0,1,2",
"EventCode": "0xa",
"EventName": "UNC_R3_RING_IV_USED.ANY",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the IV ring is being used at this ring stop. This includes when packets are passing by and when packets are being sent, but does not include when packets are being sunk into the ring stop. The IV ring is unidirectional. Whether UP or DN is used is dependent on the system programming. Thereofore, one should generally set both the UP and DN bits for a given polarity (or both) at a given time.",
"UMask": "0xf",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Bypassed",
"Counter": "0,1",
"EventCode": "0x12",
"EventName": "UNC_R3_RxR_BYPASSED.AD",
"PerPkg": "1",
"PublicDescription": "Counts the number of times when the Ingress was bypassed and an incoming transaction was bypassed directly across the BGF and into the qfclk domain.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Cycles Not Empty; DRS",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R3_RxR_CYCLES_NE.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the QPI Ingress is not empty. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Cycles Not Empty; HOM",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R3_RxR_CYCLES_NE.HOM",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the QPI Ingress is not empty. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Cycles Not Empty; NCB",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R3_RxR_CYCLES_NE.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the QPI Ingress is not empty. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Cycles Not Empty; NCS",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R3_RxR_CYCLES_NE.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the QPI Ingress is not empty. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Cycles Not Empty; NDR",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R3_RxR_CYCLES_NE.NDR",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the QPI Ingress is not empty. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Cycles Not Empty; SNP",
"Counter": "0,1",
"EventCode": "0x10",
"EventName": "UNC_R3_RxR_CYCLES_NE.SNP",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the QPI Ingress is not empty. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue occupancy. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Allocations; DRS",
"Counter": "0,1",
"EventCode": "0x11",
"EventName": "UNC_R3_RxR_INSERTS.DRS",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the QPI Ingress. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue latency. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Allocations; HOM",
"Counter": "0,1",
"EventCode": "0x11",
"EventName": "UNC_R3_RxR_INSERTS.HOM",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the QPI Ingress. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue latency. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Allocations; NCB",
"Counter": "0,1",
"EventCode": "0x11",
"EventName": "UNC_R3_RxR_INSERTS.NCB",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the QPI Ingress. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue latency. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Allocations; NCS",
"Counter": "0,1",
"EventCode": "0x11",
"EventName": "UNC_R3_RxR_INSERTS.NCS",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the QPI Ingress. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue latency. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Allocations; NDR",
"Counter": "0,1",
"EventCode": "0x11",
"EventName": "UNC_R3_RxR_INSERTS.NDR",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the QPI Ingress. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue latency. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Allocations; SNP",
"Counter": "0,1",
"EventCode": "0x11",
"EventName": "UNC_R3_RxR_INSERTS.SNP",
"PerPkg": "1",
"PublicDescription": "Counts the number of allocations into the QPI Ingress. This tracks one of the three rings that are used by the QPI agent. This can be used in conjunction with the QPI Ingress Occupancy Accumulator event in order to calculate average queue latency. Multiple ingress buffers can be tracked at a given time using multiple counters.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Occupancy Accumulator; DRS",
"EventCode": "0x13",
"EventName": "UNC_R3_RxR_OCCUPANCY.DRS",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of a given QPI Ingress queue in each cycles. This tracks one of the three ring Ingress buffers. This can be used with the QPI Ingress Not Empty event to calculate average occupancy or the QPI Ingress Allocations event in order to calculate average queuing latency.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Occupancy Accumulator; HOM",
"EventCode": "0x13",
"EventName": "UNC_R3_RxR_OCCUPANCY.HOM",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of a given QPI Ingress queue in each cycles. This tracks one of the three ring Ingress buffers. This can be used with the QPI Ingress Not Empty event to calculate average occupancy or the QPI Ingress Allocations event in order to calculate average queuing latency.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Occupancy Accumulator; NCB",
"EventCode": "0x13",
"EventName": "UNC_R3_RxR_OCCUPANCY.NCB",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of a given QPI Ingress queue in each cycles. This tracks one of the three ring Ingress buffers. This can be used with the QPI Ingress Not Empty event to calculate average occupancy or the QPI Ingress Allocations event in order to calculate average queuing latency.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Occupancy Accumulator; NCS",
"EventCode": "0x13",
"EventName": "UNC_R3_RxR_OCCUPANCY.NCS",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of a given QPI Ingress queue in each cycles. This tracks one of the three ring Ingress buffers. This can be used with the QPI Ingress Not Empty event to calculate average occupancy or the QPI Ingress Allocations event in order to calculate average queuing latency.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Occupancy Accumulator; NDR",
"EventCode": "0x13",
"EventName": "UNC_R3_RxR_OCCUPANCY.NDR",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of a given QPI Ingress queue in each cycles. This tracks one of the three ring Ingress buffers. This can be used with the QPI Ingress Not Empty event to calculate average occupancy or the QPI Ingress Allocations event in order to calculate average queuing latency.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "Ingress Occupancy Accumulator; SNP",
"EventCode": "0x13",
"EventName": "UNC_R3_RxR_OCCUPANCY.SNP",
"PerPkg": "1",
"PublicDescription": "Accumulates the occupancy of a given QPI Ingress queue in each cycles. This tracks one of the three ring Ingress buffers. This can be used with the QPI Ingress Not Empty event to calculate average occupancy or the QPI Ingress Allocations event in order to calculate average queuing latency.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Acquisition Failed on DRS; DRS Message Class",
"Counter": "0,1",
"EventCode": "0x37",
"EventName": "UNC_R3_VN0_CREDITS_REJECT.DRS",
"PerPkg": "1",
"PublicDescription": "Number of times a request failed to acquire a DRS VN0 credit. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This therefore counts the number of times when a request failed to acquire either a VNA or VN0 credit and is delayed. This should generally be a rare situation.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Acquisition Failed on DRS; HOM Message Class",
"Counter": "0,1",
"EventCode": "0x37",
"EventName": "UNC_R3_VN0_CREDITS_REJECT.HOM",
"PerPkg": "1",
"PublicDescription": "Number of times a request failed to acquire a DRS VN0 credit. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This therefore counts the number of times when a request failed to acquire either a VNA or VN0 credit and is delayed. This should generally be a rare situation.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Acquisition Failed on DRS; NCB Message Class",
"Counter": "0,1",
"EventCode": "0x37",
"EventName": "UNC_R3_VN0_CREDITS_REJECT.NCB",
"PerPkg": "1",
"PublicDescription": "Number of times a request failed to acquire a DRS VN0 credit. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This therefore counts the number of times when a request failed to acquire either a VNA or VN0 credit and is delayed. This should generally be a rare situation.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Acquisition Failed on DRS; NCS Message Class",
"Counter": "0,1",
"EventCode": "0x37",
"EventName": "UNC_R3_VN0_CREDITS_REJECT.NCS",
"PerPkg": "1",
"PublicDescription": "Number of times a request failed to acquire a DRS VN0 credit. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This therefore counts the number of times when a request failed to acquire either a VNA or VN0 credit and is delayed. This should generally be a rare situation.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Acquisition Failed on DRS; NDR Message Class",
"Counter": "0,1",
"EventCode": "0x37",
"EventName": "UNC_R3_VN0_CREDITS_REJECT.NDR",
"PerPkg": "1",
"PublicDescription": "Number of times a request failed to acquire a DRS VN0 credit. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This therefore counts the number of times when a request failed to acquire either a VNA or VN0 credit and is delayed. This should generally be a rare situation.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Acquisition Failed on DRS; SNP Message Class",
"Counter": "0,1",
"EventCode": "0x37",
"EventName": "UNC_R3_VN0_CREDITS_REJECT.SNP",
"PerPkg": "1",
"PublicDescription": "Number of times a request failed to acquire a DRS VN0 credit. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This therefore counts the number of times when a request failed to acquire either a VNA or VN0 credit and is delayed. This should generally be a rare situation.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Used; DRS Message Class",
"Counter": "0,1",
"EventCode": "0x36",
"EventName": "UNC_R3_VN0_CREDITS_USED.DRS",
"PerPkg": "1",
"PublicDescription": "Number of times a VN0 credit was used on the DRS message channel. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This counts the number of times a VN0 credit was used. Note that a single VN0 credit holds access to potentially multiple flit buffers. For example, a transfer that uses VNA could use 9 flit buffers and in that case uses 9 credits. A transfer on VN0 will only count a single credit even though it may use multiple buffers.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Used; HOM Message Class",
"Counter": "0,1",
"EventCode": "0x36",
"EventName": "UNC_R3_VN0_CREDITS_USED.HOM",
"PerPkg": "1",
"PublicDescription": "Number of times a VN0 credit was used on the DRS message channel. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This counts the number of times a VN0 credit was used. Note that a single VN0 credit holds access to potentially multiple flit buffers. For example, a transfer that uses VNA could use 9 flit buffers and in that case uses 9 credits. A transfer on VN0 will only count a single credit even though it may use multiple buffers.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Used; NCB Message Class",
"Counter": "0,1",
"EventCode": "0x36",
"EventName": "UNC_R3_VN0_CREDITS_USED.NCB",
"PerPkg": "1",
"PublicDescription": "Number of times a VN0 credit was used on the DRS message channel. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This counts the number of times a VN0 credit was used. Note that a single VN0 credit holds access to potentially multiple flit buffers. For example, a transfer that uses VNA could use 9 flit buffers and in that case uses 9 credits. A transfer on VN0 will only count a single credit even though it may use multiple buffers.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Used; NCS Message Class",
"Counter": "0,1",
"EventCode": "0x36",
"EventName": "UNC_R3_VN0_CREDITS_USED.NCS",
"PerPkg": "1",
"PublicDescription": "Number of times a VN0 credit was used on the DRS message channel. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This counts the number of times a VN0 credit was used. Note that a single VN0 credit holds access to potentially multiple flit buffers. For example, a transfer that uses VNA could use 9 flit buffers and in that case uses 9 credits. A transfer on VN0 will only count a single credit even though it may use multiple buffers.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Used; NDR Message Class",
"Counter": "0,1",
"EventCode": "0x36",
"EventName": "UNC_R3_VN0_CREDITS_USED.NDR",
"PerPkg": "1",
"PublicDescription": "Number of times a VN0 credit was used on the DRS message channel. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This counts the number of times a VN0 credit was used. Note that a single VN0 credit holds access to potentially multiple flit buffers. For example, a transfer that uses VNA could use 9 flit buffers and in that case uses 9 credits. A transfer on VN0 will only count a single credit even though it may use multiple buffers.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "VN0 Credit Used; SNP Message Class",
"Counter": "0,1",
"EventCode": "0x36",
"EventName": "UNC_R3_VN0_CREDITS_USED.SNP",
"PerPkg": "1",
"PublicDescription": "Number of times a VN0 credit was used on the DRS message channel. In order for a request to be transferred across QPI, it must be guaranteed to have a flit buffer on the remote socket to sink into. There are two credit pools, VNA and VN0. VNA is a shared pool used to achieve high performance. The VN0 pool has reserved entries for each message class and is used to prevent deadlock. Requests first attempt to acquire a VNA credit, and then fall back to VN0 if they fail. This counts the number of times a VN0 credit was used. Note that a single VN0 credit holds access to potentially multiple flit buffers. For example, a transfer that uses VNA could use 9 flit buffers and in that case uses 9 credits. A transfer on VN0 will only count a single credit even though it may use multiple buffers.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "VNA credit Acquisitions",
"Counter": "0,1",
"EventCode": "0x33",
"EventName": "UNC_R3_VNA_CREDITS_ACQUIRED",
"PerPkg": "1",
"PublicDescription": "Number of QPI VNA Credit acquisitions. This event can be used in conjunction with the VNA In-Use Accumulator to calculate the average lifetime of a credit holder. VNA credits are used by all message classes in order to communicate across QPI. If a packet is unable to acquire credits, it will then attempt to use credts from the VN0 pool. Note that a single packet may require multiple flit buffers (i.e. when data is being transfered). Therefore, this event will increment by the number of credits acquired in each cycle. Filtering based on message class is not provided. One can count the number of packets transfered in a given message class using an qfclk event.",
"Unit": "R3QPI"
},
{
"BriefDescription": "VNA Credit Reject; DRS Message Class",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R3_VNA_CREDITS_REJECT.DRS",
"PerPkg": "1",
"PublicDescription": "Number of attempted VNA credit acquisitions that were rejected because the VNA credit pool was full (or almost full). It is possible to filter this event by message class. Some packets use more than one flit buffer, and therefore must acquire multiple credits. Therefore, one could get a reject even if the VNA credits were not fully used up. The VNA pool is generally used to provide the bulk of the QPI bandwidth (as opposed to the VN0 pool which is used to guarantee forward progress). VNA credits can run out if the flit buffer on the receiving side starts to queue up substantially. This can happen if the rest of the uncore is unable to drain the requests fast enough.",
"UMask": "0x8",
"Unit": "R3QPI"
},
{
"BriefDescription": "VNA Credit Reject; HOM Message Class",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R3_VNA_CREDITS_REJECT.HOM",
"PerPkg": "1",
"PublicDescription": "Number of attempted VNA credit acquisitions that were rejected because the VNA credit pool was full (or almost full). It is possible to filter this event by message class. Some packets use more than one flit buffer, and therefore must acquire multiple credits. Therefore, one could get a reject even if the VNA credits were not fully used up. The VNA pool is generally used to provide the bulk of the QPI bandwidth (as opposed to the VN0 pool which is used to guarantee forward progress). VNA credits can run out if the flit buffer on the receiving side starts to queue up substantially. This can happen if the rest of the uncore is unable to drain the requests fast enough.",
"UMask": "0x1",
"Unit": "R3QPI"
},
{
"BriefDescription": "VNA Credit Reject; NCB Message Class",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R3_VNA_CREDITS_REJECT.NCB",
"PerPkg": "1",
"PublicDescription": "Number of attempted VNA credit acquisitions that were rejected because the VNA credit pool was full (or almost full). It is possible to filter this event by message class. Some packets use more than one flit buffer, and therefore must acquire multiple credits. Therefore, one could get a reject even if the VNA credits were not fully used up. The VNA pool is generally used to provide the bulk of the QPI bandwidth (as opposed to the VN0 pool which is used to guarantee forward progress). VNA credits can run out if the flit buffer on the receiving side starts to queue up substantially. This can happen if the rest of the uncore is unable to drain the requests fast enough.",
"UMask": "0x10",
"Unit": "R3QPI"
},
{
"BriefDescription": "VNA Credit Reject; NCS Message Class",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R3_VNA_CREDITS_REJECT.NCS",
"PerPkg": "1",
"PublicDescription": "Number of attempted VNA credit acquisitions that were rejected because the VNA credit pool was full (or almost full). It is possible to filter this event by message class. Some packets use more than one flit buffer, and therefore must acquire multiple credits. Therefore, one could get a reject even if the VNA credits were not fully used up. The VNA pool is generally used to provide the bulk of the QPI bandwidth (as opposed to the VN0 pool which is used to guarantee forward progress). VNA credits can run out if the flit buffer on the receiving side starts to queue up substantially. This can happen if the rest of the uncore is unable to drain the requests fast enough.",
"UMask": "0x20",
"Unit": "R3QPI"
},
{
"BriefDescription": "VNA Credit Reject; NDR Message Class",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R3_VNA_CREDITS_REJECT.NDR",
"PerPkg": "1",
"PublicDescription": "Number of attempted VNA credit acquisitions that were rejected because the VNA credit pool was full (or almost full). It is possible to filter this event by message class. Some packets use more than one flit buffer, and therefore must acquire multiple credits. Therefore, one could get a reject even if the VNA credits were not fully used up. The VNA pool is generally used to provide the bulk of the QPI bandwidth (as opposed to the VN0 pool which is used to guarantee forward progress). VNA credits can run out if the flit buffer on the receiving side starts to queue up substantially. This can happen if the rest of the uncore is unable to drain the requests fast enough.",
"UMask": "0x4",
"Unit": "R3QPI"
},
{
"BriefDescription": "VNA Credit Reject; SNP Message Class",
"Counter": "0,1",
"EventCode": "0x34",
"EventName": "UNC_R3_VNA_CREDITS_REJECT.SNP",
"PerPkg": "1",
"PublicDescription": "Number of attempted VNA credit acquisitions that were rejected because the VNA credit pool was full (or almost full). It is possible to filter this event by message class. Some packets use more than one flit buffer, and therefore must acquire multiple credits. Therefore, one could get a reject even if the VNA credits were not fully used up. The VNA pool is generally used to provide the bulk of the QPI bandwidth (as opposed to the VN0 pool which is used to guarantee forward progress). VNA credits can run out if the flit buffer on the receiving side starts to queue up substantially. This can happen if the rest of the uncore is unable to drain the requests fast enough.",
"UMask": "0x2",
"Unit": "R3QPI"
},
{
"BriefDescription": "Cycles with no VNA credits available",
"Counter": "0,1",
"EventCode": "0x31",
"EventName": "UNC_R3_VNA_CREDIT_CYCLES_OUT",
"PerPkg": "1",
"PublicDescription": "Number of QPI uclk cycles when the transmitted has no VNA credits available and therefore cannot send any requests on this channel. Note that this does not mean that no flits can be transmitted, as those holding VN0 credits will still (potentially) be able to transmit. Generally it is the goal of the uncore that VNA credits should not run out, as this can substantially throttle back useful QPI bandwidth.",
"Unit": "R3QPI"
},
{
"BriefDescription": "Cycles with 1 or more VNA credits in use",
"Counter": "0,1",
"EventCode": "0x32",
"EventName": "UNC_R3_VNA_CREDIT_CYCLES_USED",
"PerPkg": "1",
"PublicDescription": "Number of QPI uclk cycles with one or more VNA credits in use. This event can be used in conjunction with the VNA In-Use Accumulator to calculate the average number of used VNA credits.",
"Unit": "R3QPI"
},
{
"BriefDescription": "VLW Received",
"Counter": "0,1",
"EventCode": "0x42",
"EventName": "UNC_U_EVENT_MSG.DOORBELL_RCVD",
"PerPkg": "1",
"PublicDescription": "Virtual Logical Wire (legacy) message were received from Uncore. Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x8",
"Unit": "UBOX"
},
{
"BriefDescription": "VLW Received",
"Counter": "0,1",
"EventCode": "0x42",
"EventName": "UNC_U_EVENT_MSG.INT_PRIO",
"PerPkg": "1",
"PublicDescription": "Virtual Logical Wire (legacy) message were received from Uncore. Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x10",
"Unit": "UBOX"
},
{
"BriefDescription": "VLW Received",
"Counter": "0,1",
"EventCode": "0x42",
"EventName": "UNC_U_EVENT_MSG.IPI_RCVD",
"PerPkg": "1",
"PublicDescription": "Virtual Logical Wire (legacy) message were received from Uncore. Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x4",
"Unit": "UBOX"
},
{
"BriefDescription": "VLW Received",
"Counter": "0,1",
"EventCode": "0x42",
"EventName": "UNC_U_EVENT_MSG.MSI_RCVD",
"PerPkg": "1",
"PublicDescription": "Virtual Logical Wire (legacy) message were received from Uncore. Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x2",
"Unit": "UBOX"
},
{
"BriefDescription": "VLW Received",
"Counter": "0,1",
"EventCode": "0x42",
"EventName": "UNC_U_EVENT_MSG.VLW_RCVD",
"PerPkg": "1",
"PublicDescription": "Virtual Logical Wire (legacy) message were received from Uncore. Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x1",
"Unit": "UBOX"
},
{
"BriefDescription": "Filter Match",
"Counter": "0,1",
"EventCode": "0x41",
"EventName": "UNC_U_FILTER_MATCH.DISABLE",
"PerPkg": "1",
"PublicDescription": "Filter match per thread (w/ or w/o Filter Enable). Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x2",
"Unit": "UBOX"
},
{
"BriefDescription": "Filter Match",
"Counter": "0,1",
"EventCode": "0x41",
"EventName": "UNC_U_FILTER_MATCH.ENABLE",
"Filter": "UBoxFilter[3:0]",
"PerPkg": "1",
"PublicDescription": "Filter match per thread (w/ or w/o Filter Enable). Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x1",
"Unit": "UBOX"
},
{
"BriefDescription": "Filter Match",
"Counter": "0,1",
"EventCode": "0x41",
"EventName": "UNC_U_FILTER_MATCH.U2C_DISABLE",
"PerPkg": "1",
"PublicDescription": "Filter match per thread (w/ or w/o Filter Enable). Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x8",
"Unit": "UBOX"
},
{
"BriefDescription": "Filter Match",
"Counter": "0,1",
"EventCode": "0x41",
"EventName": "UNC_U_FILTER_MATCH.U2C_ENABLE",
"Filter": "UBoxFilter[3:0]",
"PerPkg": "1",
"PublicDescription": "Filter match per thread (w/ or w/o Filter Enable). Specify the thread to filter on using NCUPMONCTRLGLCTR.ThreadID.",
"UMask": "0x4",
"Unit": "UBOX"
},
{
"BriefDescription": "IDI Lock/SplitLock Cycles",
"Counter": "0,1",
"EventCode": "0x44",
"EventName": "UNC_U_LOCK_CYCLES",
"PerPkg": "1",
"PublicDescription": "Number of times an IDI Lock/SplitLock sequence was started",
"Unit": "UBOX"
},
{
"BriefDescription": "MsgCh Requests by Size; 4B Requests",
"Counter": "0,1",
"EventCode": "0x47",
"EventName": "UNC_U_MSG_CHNL_SIZE_COUNT.4B",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of transactions on the message channel filtered by request size. This includes both reads and writes.",
"UMask": "0x1",
"Unit": "UBOX"
},
{
"BriefDescription": "MsgCh Requests by Size; 8B Requests",
"Counter": "0,1",
"EventCode": "0x47",
"EventName": "UNC_U_MSG_CHNL_SIZE_COUNT.8B",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Number of transactions on the message channel filtered by request size. This includes both reads and writes.",
"UMask": "0x2",
"Unit": "UBOX"
},
{
"BriefDescription": "Cycles PHOLD Assert to Ack; ACK to Deassert",
"Counter": "0,1",
"EventCode": "0x45",
"EventName": "UNC_U_PHOLD_CYCLES.ACK_TO_DEASSERT",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "PHOLD cycles. Filter from source CoreID.",
"UMask": "0x2",
"Unit": "UBOX"
},
{
"BriefDescription": "Cycles PHOLD Assert to Ack; Assert to ACK",
"Counter": "0,1",
"EventCode": "0x45",
"EventName": "UNC_U_PHOLD_CYCLES.ASSERT_TO_ACK",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "PHOLD cycles. Filter from source CoreID.",
"UMask": "0x1",
"Unit": "UBOX"
},
{
"BriefDescription": "RACU Request",
"Counter": "0,1",
"EventCode": "0x46",
"EventName": "UNC_U_RACU_REQUESTS.COUNT",
"ExtSel": "1",
"PerPkg": "1",
"UMask": "0x1",
"Unit": "UBOX"
},
{
"BriefDescription": "Monitor Sent to T0; Correctable Machine Check",
"Counter": "0,1",
"EventCode": "0x43",
"EventName": "UNC_U_U2C_EVENTS.CMC",
"PerPkg": "1",
"PublicDescription": "Events coming from Uncore can be sent to one or all cores",
"UMask": "0x10",
"Unit": "UBOX"
},
{
"BriefDescription": "Monitor Sent to T0; Livelock",
"Counter": "0,1",
"EventCode": "0x43",
"EventName": "UNC_U_U2C_EVENTS.LIVELOCK",
"PerPkg": "1",
"PublicDescription": "Events coming from Uncore can be sent to one or all cores",
"UMask": "0x4",
"Unit": "UBOX"
},
{
"BriefDescription": "Monitor Sent to T0; LTError",
"Counter": "0,1",
"EventCode": "0x43",
"EventName": "UNC_U_U2C_EVENTS.LTERROR",
"PerPkg": "1",
"PublicDescription": "Events coming from Uncore can be sent to one or all cores",
"UMask": "0x8",
"Unit": "UBOX"
},
{
"BriefDescription": "Monitor Sent to T0; Monitor T0",
"Counter": "0,1",
"EventCode": "0x43",
"EventName": "UNC_U_U2C_EVENTS.MONITOR_T0",
"PerPkg": "1",
"PublicDescription": "Events coming from Uncore can be sent to one or all cores",
"UMask": "0x1",
"Unit": "UBOX"
},
{
"BriefDescription": "Monitor Sent to T0; Monitor T1",
"Counter": "0,1",
"EventCode": "0x43",
"EventName": "UNC_U_U2C_EVENTS.MONITOR_T1",
"PerPkg": "1",
"PublicDescription": "Events coming from Uncore can be sent to one or all cores",
"UMask": "0x2",
"Unit": "UBOX"
},
{
"BriefDescription": "Monitor Sent to T0; Other",
"Counter": "0,1",
"EventCode": "0x43",
"EventName": "UNC_U_U2C_EVENTS.OTHER",
"PerPkg": "1",
"PublicDescription": "Events coming from Uncore can be sent to one or all cores",
"UMask": "0x80",
"Unit": "UBOX"
},
{
"BriefDescription": "Monitor Sent to T0; Trap",
"Counter": "0,1",
"EventCode": "0x43",
"EventName": "UNC_U_U2C_EVENTS.TRAP",
"PerPkg": "1",
"PublicDescription": "Events coming from Uncore can be sent to one or all cores",
"UMask": "0x40",
"Unit": "UBOX"
},
{
"BriefDescription": "Monitor Sent to T0; Uncorrectable Machine Check",
"Counter": "0,1",
"EventCode": "0x43",
"EventName": "UNC_U_U2C_EVENTS.UMC",
"PerPkg": "1",
"PublicDescription": "Events coming from Uncore can be sent to one or all cores",
"UMask": "0x20",
"Unit": "UBOX"
}
]
tools/perf/pmu-events/arch/x86/jaketown/uncore-power.json
View file @ 376d8b58
[ [
{ {
"BriefDescription": "PCU clock ticks. Use to get percentages of PCU cycles events", "BriefDescription": "pclk Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventName": "UNC_P_CLOCKTICKS", "EventName": "UNC_P_CLOCKTICKS",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "The PCU runs off a fixed 800 MHz clock. This event counts the number of pclk cycles measured while the counter was enabled. The pclk, like the Memory Controller's dclk, counts at a constant rate making it a good measure of actual wall time.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to the frequency that is configured in the filter. (filter_band0=XXX with XXX in 100Mhz units). One can also use inversion (filter_inv=1) to track cycles when we were less than the configured frequency", "BriefDescription": "Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xb", "EventCode": "0x3",
"EventName": "UNC_P_FREQ_BAND0_CYCLES", "EventName": "UNC_P_CORE0_TRANSITION_CYCLES",
"MetricExpr": "(UNC_P_FREQ_BAND0_CYCLES / UNC_P_CLOCKTICKS) * 100.", "ExtSel": "1",
"MetricName": "freq_band0_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions. There is one event per core.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to the frequency that is configured in the filter. (filter_band1=XXX with XXX in 100Mhz units). One can also use inversion (filter_inv=1) to track cycles when we were less than the configured frequency", "BriefDescription": "Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xc", "EventCode": "0x4",
"EventName": "UNC_P_FREQ_BAND1_CYCLES", "EventName": "UNC_P_CORE1_TRANSITION_CYCLES",
"MetricExpr": "(UNC_P_FREQ_BAND1_CYCLES / UNC_P_CLOCKTICKS) * 100.", "ExtSel": "1",
"MetricName": "freq_band1_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions. There is one event per core.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to the frequency that is configured in the filter. (filter_band2=XXX with XXX in 100Mhz units). One can also use inversion (filter_inv=1) to track cycles when we were less than the configured frequency", "BriefDescription": "Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xd", "EventCode": "0x5",
"EventName": "UNC_P_FREQ_BAND2_CYCLES", "EventName": "UNC_P_CORE2_TRANSITION_CYCLES",
"MetricExpr": "(UNC_P_FREQ_BAND2_CYCLES / UNC_P_CLOCKTICKS) * 100.", "ExtSel": "1",
"MetricName": "freq_band2_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions. There is one event per core.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to the frequency that is configured in the filter. (filter_band3=XXX, with XXX in 100Mhz units). One can also use inversion (filter_inv=1) to track cycles when we were less than the configured frequency", "BriefDescription": "Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xe", "EventCode": "0x6",
"EventName": "UNC_P_FREQ_BAND3_CYCLES", "EventName": "UNC_P_CORE3_TRANSITION_CYCLES",
"MetricExpr": "(UNC_P_FREQ_BAND3_CYCLES / UNC_P_CLOCKTICKS) * 100.", "ExtSel": "1",
"MetricName": "freq_band3_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions. There is one event per core.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of times that the uncore transitioned a frequency greater than or equal to the frequency that is configured in the filter. (filter_band0=XXX with XXX in 100Mhz units). One can also use inversion (filter_inv=1) to track cycles when we were less than the configured frequency. Derived from unc_p_freq_band0_cycles", "BriefDescription": "Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xb", "EventCode": "0x7",
"EventName": "UNC_P_FREQ_BAND0_TRANSITIONS", "EventName": "UNC_P_CORE4_TRANSITION_CYCLES",
"Filter": "edge=1", "ExtSel": "1",
"MetricExpr": "(UNC_P_FREQ_BAND0_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_band0_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions. There is one event per core.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of times that the uncore transistioned to a frequency greater than or equal to the frequency that is configured in the filter. (filter_band1=XXX with XXX in 100Mhz units). One can also use inversion (filter_inv=1) to track cycles when we were less than the configured frequency. Derived from unc_p_freq_band1_cycles", "BriefDescription": "Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xc", "EventCode": "0x8",
"EventName": "UNC_P_FREQ_BAND1_TRANSITIONS", "EventName": "UNC_P_CORE5_TRANSITION_CYCLES",
"Filter": "edge=1", "ExtSel": "1",
"MetricExpr": "(UNC_P_FREQ_BAND1_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_band1_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions. There is one event per core.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore transitioned to a frequency greater than or equal to the frequency that is configured in the filter. (filter_band2=XXX with XXX in 100Mhz units). One can also use inversion (filter_inv=1) to track cycles when we were less than the configured frequency. Derived from unc_p_freq_band2_cycles", "BriefDescription": "Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xd", "EventCode": "0x9",
"EventName": "UNC_P_FREQ_BAND2_TRANSITIONS", "EventName": "UNC_P_CORE6_TRANSITION_CYCLES",
"Filter": "edge=1", "ExtSel": "1",
"MetricExpr": "(UNC_P_FREQ_BAND2_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_band2_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions. There is one event per core.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore transitioned to a frequency greater than or equal to the frequency that is configured in the filter. (filter_band3=XXX, with XXX in 100Mhz units). One can also use inversion (filter_inv=1) to track cycles when we were less than the configured frequency. Derived from unc_p_freq_band3_cycles", "BriefDescription": "Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xe", "EventCode": "0xa",
"EventName": "UNC_P_FREQ_BAND3_TRANSITIONS", "EventName": "UNC_P_CORE7_TRANSITION_CYCLES",
"Filter": "edge=1", "ExtSel": "1",
"MetricExpr": "(UNC_P_FREQ_BAND3_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_band3_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions. There is one event per core.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "This is an occupancy event that tracks the number of cores that are in C0. It can be used by itself to get the average number of cores in C0, with threshholding to generate histograms, or with other PCU events and occupancy triggering to capture other details", "BriefDescription": "Core C State Demotions",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x80", "EventCode": "0x1e",
"EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C0", "EventName": "UNC_P_DEMOTIONS_CORE0",
"Filter": "occ_sel=1", "Filter": "PCUFilter[7:0]",
"MetricExpr": "(UNC_P_POWER_STATE_OCCUPANCY.CORES_C0 / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "power_state_occupancy.cores_c0 %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of times when a configurable cores had a C-state demotion",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "This is an occupancy event that tracks the number of cores that are in C3. It can be used by itself to get the average number of cores in C0, with threshholding to generate histograms, or with other PCU events and occupancy triggering to capture other details", "BriefDescription": "Core C State Demotions",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x80", "EventCode": "0x1f",
"EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C3", "EventName": "UNC_P_DEMOTIONS_CORE1",
"Filter": "occ_sel=2", "Filter": "PCUFilter[7:0]",
"MetricExpr": "(UNC_P_POWER_STATE_OCCUPANCY.CORES_C3 / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "power_state_occupancy.cores_c3 %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of times when a configurable cores had a C-state demotion",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "This is an occupancy event that tracks the number of cores that are in C6. It can be used by itself to get the average number of cores in C0, with threshholding to generate histograms, or with other PCU events ", "BriefDescription": "Core C State Demotions",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x80", "EventCode": "0x20",
"EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C6", "EventName": "UNC_P_DEMOTIONS_CORE2",
"Filter": "occ_sel=3",
"MetricExpr": "(UNC_P_POWER_STATE_OCCUPANCY.CORES_C6 / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "power_state_occupancy.cores_c6 %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of times when a configurable cores had a C-state demotion",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that we are in external PROCHOT mode. This mode is triggered when a sensor off the die determines that something off-die (like DRAM) is too hot and must throttle to avoid damaging the chip", "BriefDescription": "Core C State Demotions",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xa", "EventCode": "0x21",
"EventName": "UNC_P_PROCHOT_EXTERNAL_CYCLES", "EventName": "UNC_P_DEMOTIONS_CORE3",
"MetricExpr": "(UNC_P_PROCHOT_EXTERNAL_CYCLES / UNC_P_CLOCKTICKS) * 100.", "Filter": "PCUFilter[7:0]",
"MetricName": "prochot_external_cycles %", "PerPkg": "1",
"PublicDescription": "Counts the number of times when a configurable cores had a C-state demotion",
"Unit": "PCU"
},
{
"BriefDescription": "Core C State Demotions",
"Counter": "0,1,2,3",
"EventCode": "0x22",
"EventName": "UNC_P_DEMOTIONS_CORE4",
"Filter": "PCUFilter[7:0]",
"PerPkg": "1",
"PublicDescription": "Counts the number of times when a configurable cores had a C-state demotion",
"Unit": "PCU"
},
{
"BriefDescription": "Core C State Demotions",
"Counter": "0,1,2,3",
"EventCode": "0x23",
"EventName": "UNC_P_DEMOTIONS_CORE5",
"Filter": "PCUFilter[7:0]",
"PerPkg": "1",
"PublicDescription": "Counts the number of times when a configurable cores had a C-state demotion",
"Unit": "PCU"
},
{
"BriefDescription": "Core C State Demotions",
"Counter": "0,1,2,3",
"EventCode": "0x24",
"EventName": "UNC_P_DEMOTIONS_CORE6",
"Filter": "PCUFilter[7:0]",
"PerPkg": "1",
"PublicDescription": "Counts the number of times when a configurable cores had a C-state demotion",
"Unit": "PCU"
},
{
"BriefDescription": "Core C State Demotions",
"Counter": "0,1,2,3",
"EventCode": "0x25",
"EventName": "UNC_P_DEMOTIONS_CORE7",
"Filter": "PCUFilter[7:0]",
"PerPkg": "1",
"PublicDescription": "Counts the number of times when a configurable cores had a C-state demotion",
"Unit": "PCU"
},
{
"BriefDescription": "Frequency Residency",
"Counter": "0,1,2,3",
"EventCode": "0xb",
"EventName": "UNC_P_FREQ_BAND0_CYCLES",
"Filter": "PCUFilter[7:0]",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to the frequency that is configured in the filter. One can use all four counters with this event, so it is possible to track up to 4 configurable bands. One can use edge detect in conjunction with this event to track the number of times that we transitioned into a frequency greater than or equal to the configurable frequency. One can also use inversion to track cycles when we were less than the configured frequency.",
"Unit": "PCU"
},
{
"BriefDescription": "Frequency Residency",
"Counter": "0,1,2,3",
"EventCode": "0xc",
"EventName": "UNC_P_FREQ_BAND1_CYCLES",
"Filter": "PCUFilter[15:8]",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to the frequency that is configured in the filter. One can use all four counters with this event, so it is possible to track up to 4 configurable bands. One can use edge detect in conjunction with this event to track the number of times that we transitioned into a frequency greater than or equal to the configurable frequency. One can also use inversion to track cycles when we were less than the configured frequency.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles when temperature is the upper limit on frequency", "BriefDescription": "Frequency Residency",
"Counter": "0,1,2,3",
"EventCode": "0xd",
"EventName": "UNC_P_FREQ_BAND2_CYCLES",
"Filter": "PCUFilter[23:16]",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to the frequency that is configured in the filter. One can use all four counters with this event, so it is possible to track up to 4 configurable bands. One can use edge detect in conjunction with this event to track the number of times that we transitioned into a frequency greater than or equal to the configurable frequency. One can also use inversion to track cycles when we were less than the configured frequency.",
"Unit": "PCU"
},
{
"BriefDescription": "Frequency Residency",
"Counter": "0,1,2,3",
"EventCode": "0xe",
"EventName": "UNC_P_FREQ_BAND3_CYCLES",
"Filter": "PCUFilter[31:24]",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to the frequency that is configured in the filter. One can use all four counters with this event, so it is possible to track up to 4 configurable bands. One can use edge detect in conjunction with this event to track the number of times that we transitioned into a frequency greater than or equal to the configurable frequency. One can also use inversion to track cycles when we were less than the configured frequency.",
"Unit": "PCU"
},
{
"BriefDescription": "Current Strongest Upper Limit Cycles",
"Counter": "0,1,2,3",
"EventCode": "0x7",
"EventName": "UNC_P_FREQ_MAX_CURRENT_CYCLES",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when current is the upper limit on frequency.",
"Unit": "PCU"
},
{
"BriefDescription": "Thermal Strongest Upper Limit Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x4", "EventCode": "0x4",
"EventName": "UNC_P_FREQ_MAX_LIMIT_THERMAL_CYCLES", "EventName": "UNC_P_FREQ_MAX_LIMIT_THERMAL_CYCLES",
"MetricExpr": "(UNC_P_FREQ_MAX_LIMIT_THERMAL_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_max_limit_thermal_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when thermal conditions are the upper limit on frequency. This is related to the THERMAL_THROTTLE CYCLES_ABOVE_TEMP event, which always counts cycles when we are above the thermal temperature. This event (STRONGEST_UPPER_LIMIT) is sampled at the output of the algorithm that determines the actual frequency, while THERMAL_THROTTLE looks at the input.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles when the OS is the upper limit on frequency", "BriefDescription": "OS Strongest Upper Limit Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x6", "EventCode": "0x6",
"EventName": "UNC_P_FREQ_MAX_OS_CYCLES", "EventName": "UNC_P_FREQ_MAX_OS_CYCLES",
"MetricExpr": "(UNC_P_FREQ_MAX_OS_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_max_os_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the OS is the upper limit on frequency.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles when power is the upper limit on frequency", "BriefDescription": "Power Strongest Upper Limit Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x5", "EventCode": "0x5",
"EventName": "UNC_P_FREQ_MAX_POWER_CYCLES", "EventName": "UNC_P_FREQ_MAX_POWER_CYCLES",
"MetricExpr": "(UNC_P_FREQ_MAX_POWER_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_max_power_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when power is the upper limit on frequency.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles when current is the upper limit on frequency", "BriefDescription": "IO P Limit Strongest Lower Limit Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0x7", "EventCode": "0x1",
"EventName": "UNC_P_FREQ_MAX_CURRENT_CYCLES", "EventName": "UNC_P_FREQ_MIN_IO_P_CYCLES",
"MetricExpr": "(UNC_P_FREQ_MAX_CURRENT_CYCLES / UNC_P_CLOCKTICKS) * 100.", "ExtSel": "1",
"MetricName": "freq_max_current_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when IO P Limit is preventing us from dropping the frequency lower. This algorithm monitors the needs to the IO subsystem on both local and remote sockets and will maintain a frequency high enough to maintain good IO BW. This is necessary for when all the IA cores on a socket are idle but a user still would like to maintain high IO Bandwidth.",
"Unit": "PCU"
},
{
"BriefDescription": "Perf P Limit Strongest Lower Limit Cycles",
"Counter": "0,1,2,3",
"EventCode": "0x2",
"EventName": "UNC_P_FREQ_MIN_PERF_P_CYCLES",
"ExtSel": "1",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles when Perf P Limit is preventing us from dropping the frequency lower. Perf P Limit is an algorithm that takes input from remote sockets when determining if a socket should drop it's frequency down. This is largely to minimize increases in snoop and remote read latencies.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Cycles spent changing Frequency", "BriefDescription": "Cycles spent changing Frequency",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventName": "UNC_P_FREQ_TRANS_CYCLES", "EventName": "UNC_P_FREQ_TRANS_CYCLES",
"MetricExpr": "(UNC_P_FREQ_TRANS_CYCLES / UNC_P_CLOCKTICKS) * 100.", "ExtSel": "1",
"MetricName": "freq_trans_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the system is changing frequency. This can not be filtered by thread ID. One can also use it with the occupancy counter that monitors number of threads in C0 to estimate the performance impact that frequency transitions had on the system.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to 1.2Ghz. Derived from unc_p_freq_band0_cycles", "BriefDescription": "Memory Phase Shedding Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xb", "EventCode": "0x2f",
"EventName": "UNC_P_FREQ_GE_1200MHZ_CYCLES", "EventName": "UNC_P_MEMORY_PHASE_SHEDDING_CYCLES",
"Filter": "filter_band0=12",
"MetricExpr": "(UNC_P_FREQ_GE_1200MHZ_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_ge_1200mhz_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles that the PCU has triggered memory phase shedding. This is a mode that can be run in the iMC physicals that saves power at the expense of additional latency.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to 2Ghz. Derived from unc_p_freq_band1_cycles", "BriefDescription": "Number of cores in C0",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xc", "EventCode": "0x80",
"EventName": "UNC_P_FREQ_GE_2000MHZ_CYCLES", "EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C0",
"Filter": "filter_band1=20",
"MetricExpr": "(UNC_P_FREQ_GE_2000MHZ_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_ge_2000mhz_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "This is an occupancy event that tracks the number of cores that are in C0. It can be used by itself to get the average number of cores in C0, with threshholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to 3Ghz. Derived from unc_p_freq_band2_cycles", "BriefDescription": "Number of cores in C0",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xd", "EventCode": "0x80",
"EventName": "UNC_P_FREQ_GE_3000MHZ_CYCLES", "EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C3",
"Filter": "filter_band2=30",
"MetricExpr": "(UNC_P_FREQ_GE_3000MHZ_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_ge_3000mhz_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "This is an occupancy event that tracks the number of cores that are in C0. It can be used by itself to get the average number of cores in C0, with threshholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore was running at a frequency greater than or equal to 4Ghz. Derived from unc_p_freq_band3_cycles", "BriefDescription": "Number of cores in C0",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xe", "EventCode": "0x80",
"EventName": "UNC_P_FREQ_GE_4000MHZ_CYCLES", "EventName": "UNC_P_POWER_STATE_OCCUPANCY.CORES_C6",
"Filter": "filter_band3=40",
"MetricExpr": "(UNC_P_FREQ_GE_4000MHZ_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_ge_4000mhz_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "This is an occupancy event that tracks the number of cores that are in C0. It can be used by itself to get the average number of cores in C0, with threshholding to generate histograms, or with other PCU events and occupancy triggering to capture other details.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of times that the uncore transitioned to a frequency greater than or equal to 1.2Ghz. Derived from unc_p_freq_band0_cycles", "BriefDescription": "External Prochot",
"Counter": "0,1,2,3",
"EventCode": "0xa",
"EventName": "UNC_P_PROCHOT_EXTERNAL_CYCLES",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that we are in external PROCHOT mode. This mode is triggered when a sensor off the die determines that something off-die (like DRAM) is too hot and must throttle to avoid damaging the chip.",
"Unit": "PCU"
},
{
"BriefDescription": "Internal Prochot",
"Counter": "0,1,2,3",
"EventCode": "0x9",
"EventName": "UNC_P_PROCHOT_INTERNAL_CYCLES",
"PerPkg": "1",
"PublicDescription": "Counts the number of cycles that we are in Interal PROCHOT mode. This mode is triggered when a sensor on the die determines that we are too hot and must throttle to avoid damaging the chip.",
"Unit": "PCU"
},
{
"BriefDescription": "Total Core C State Transition Cycles",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xb", "EventCode": "0xb",
"EventName": "UNC_P_FREQ_GE_1200MHZ_TRANSITIONS", "EventName": "UNC_P_TOTAL_TRANSITION_CYCLES",
"Filter": "edge=1,filter_band0=12", "ExtSel": "1",
"MetricExpr": "(UNC_P_FREQ_GE_1200MHZ_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_ge_1200mhz_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Number of cycles spent performing core C state transitions across all cores.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of times that the uncore transitioned to a frequency greater than or equal to 2Ghz. Derived from unc_p_freq_band1_cycles", "BriefDescription": "Cycles Changing Voltage",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xc", "EventCode": "0x3",
"EventName": "UNC_P_FREQ_GE_2000MHZ_TRANSITIONS", "EventName": "UNC_P_VOLT_TRANS_CYCLES_CHANGE",
"Filter": "edge=1,filter_band1=20",
"MetricExpr": "(UNC_P_FREQ_GE_2000MHZ_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_ge_2000mhz_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the system is changing voltage. There is no filtering supported with this event. One can use it as a simple event, or use it conjunction with the occupancy events to monitor the number of cores or threads that were impacted by the transition. This event is calculated by or'ing together the increasing and decreasing events.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore transitioned to a frequency greater than or equal to 3Ghz. Derived from unc_p_freq_band2_cycles", "BriefDescription": "Cycles Decreasing Voltage",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xd", "EventCode": "0x2",
"EventName": "UNC_P_FREQ_GE_3000MHZ_TRANSITIONS", "EventName": "UNC_P_VOLT_TRANS_CYCLES_DECREASE",
"Filter": "edge=1,filter_band2=30",
"MetricExpr": "(UNC_P_FREQ_GE_3000MHZ_CYCLES / UNC_P_CLOCKTICKS) * 100.",
"MetricName": "freq_ge_3000mhz_cycles %",
"PerPkg": "1", "PerPkg": "1",
"PublicDescription": "Counts the number of cycles when the system is decreasing voltage. There is no filtering supported with this event. One can use it as a simple event, or use it conjunction with the occupancy events to monitor the number of cores or threads that were impacted by the transition.",
"Unit": "PCU" "Unit": "PCU"
}, },
{ {
"BriefDescription": "Counts the number of cycles that the uncore transitioned to a frequency greater than or equal to 4Ghz. Derived from unc_p_freq_band3_cycles", "BriefDescription": "Cycles Increasing Voltage",
"Counter": "0,1,2,3", "Counter": "0,1,2,3",
"EventCode": "0xe", "EventCode": "0x1",
"EventName": "UNC_P_FREQ_GE_4000MHZ_TRANSITIONS", "EventName": "UNC_P_VOLT_TRANS_CYCLES_INCREASE",
"Filter": "edge=1,filter_band3=40", "PerPkg": "1",
"MetricExpr": "(UNC_P_FREQ_GE_4000MHZ_CYCLES / UNC_P_CLOCKTICKS) * 100.", "PublicDescription": "Counts the number of cycles when the system is increasing voltage. There is no filtering supported with this event. One can use it as a simple event, or use it conjunction with the occupancy events to monitor the number of cores or threads that were impacted by the transition.",
"MetricName": "freq_ge_4000mhz_cycles %", "Unit": "PCU"
},
{
"BriefDescription": "VR Hot",
"Counter": "0,1,2,3",
"EventCode": "0x32",
"EventName": "UNC_P_VR_HOT_CYCLES",
"PerPkg": "1", "PerPkg": "1",
"Unit": "PCU" "Unit": "PCU"
} }
......
tools/perf/pmu-events/arch/x86/jaketown/virtual-memory.json
View file @ 376d8b58
...@@ -146,4 +146,4 @@ ...@@ -146,4 +146,4 @@
"SampleAfterValue": "100007", "SampleAfterValue": "100007",
"UMask": "0x20" "UMask": "0x20"
} }
] ]
\ No newline at end of file
tools/perf/pmu-events/arch/x86/mapfile.csv
View file @ 376d8b58
...@@ -14,7 +14,7 @@ GenuineIntel-6-(7D|7E|A7),v1.14,icelake,core ...@@ -14,7 +14,7 @@ GenuineIntel-6-(7D|7E|A7),v1.14,icelake,core
GenuineIntel-6-6[AC],v1.15,icelakex,core GenuineIntel-6-6[AC],v1.15,icelakex,core
GenuineIntel-6-3A,v22,ivybridge,core GenuineIntel-6-3A,v22,ivybridge,core
GenuineIntel-6-3E,v21,ivytown,core GenuineIntel-6-3E,v21,ivytown,core
GenuineIntel-6-2D,v20,jaketown,core GenuineIntel-6-2D,v21,jaketown,core
GenuineIntel-6-57,v9,knightslanding,core GenuineIntel-6-57,v9,knightslanding,core
GenuineIntel-6-85,v9,knightslanding,core GenuineIntel-6-85,v9,knightslanding,core
GenuineIntel-6-1E,v2,nehalemep,core GenuineIntel-6-1E,v2,nehalemep,core
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7