perf vendor events intel: Fix broadwellde tma_info_system_dram_bw_use metric
Ian Rogers
reviewerPATCH v2 1/3 perf vendor events intel: Fix broadwellde tma_info_system_dram_bw_use metric
Broadwell-de has a consumer core and server uncore. The uncore_arb PMU isn't present and the broadwellx style cbox PMU should be used instead. Fix the tma_info_system_dram_bw_use metric to use the server metric rather than client.
The associated converter script fix is in:
https://github.com/intel/perfmon/pull/111
--- tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json +1 -1
@@ -652,7 +652,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1e6 / duration_time / 1e3", + "MetricExpr": "64 * (UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR) / 1e9 / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC;tma_issueBW", "MetricName": "tma_info_system_dram_bw_use", "PublicDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]. Related metrics: tma_fb_full, tma_mem_bandwidth, tma_sq_full"
Signature · 1 line
2.42.0.515.g380fc7ccd1-goog
Ian Rogers
reviewerPATCH v2 2/3 perf vendor events intel: Add broadwellde two metrics
Add tma_info_system_socket_clks and uncore_freq metrics that require a broadwellx style uncore event for UNC_CLOCK.
The associated converter script fix is in:
https://github.com/intel/perfmon/pull/112
--- .../arch/x86/broadwellde/bdwde-metrics.json | 12 ++++++++++++ 1 file changed, 12 insertions(+)
tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json +12 -0
@@ -48,6 +48,12 @@ "MetricName": "C7_Pkg_Residency", "ScaleUnit": "100%" }, + { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" + }, { "BriefDescription": "Percentage of cycles spent in System Management Interrupts.", "MetricExpr": "((msr@aperf@ - cycles) / msr@aperf@ if msr@smi@ > 0 else 0)",
@@ -690,6 +696,12 @@ "MetricGroup": "SMT", "MetricName": "tma_info_system_smt_2t_utilization" }, + { + "BriefDescription": "Socket actual clocks when any core is active on that socket", + "MetricExpr": "cbox_0@event\\=0x0@", + "MetricGroup": "SoC", + "MetricName": "tma_info_system_socket_clks" + }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", "MetricExpr": "tma_info_thread_clks / CPU_CLK_UNHALTED.REF_TSC",
Signature · 1 line
2.42.0.515.g380fc7ccd1-goog
Ian Rogers
reviewerPATCH v2 3/3 perf vendor events intel: Add tigerlake two metrics
In reply to Ian RogersAdd tma_info_system_socket_clks and uncore_freq metrics.
The associated converter script fix is in:
https://github.com/intel/perfmon/pull/112
--- .../pmu-events/arch/x86/tigerlake/tgl-metrics.json | 12 ++++++++++++ 1 file changed, 12 insertions(+)
tools/perf/pmu-events/arch/x86/tigerlake/tgl-metrics.json +12 -0
@@ -62,6 +62,12 @@ "MetricName": "C9_Pkg_Residency", "ScaleUnit": "100%" }, + { + "BriefDescription": "Uncore frequency per die [GHZ]", + "MetricExpr": "tma_info_system_socket_clks / #num_dies / duration_time / 1e9", + "MetricGroup": "SoC", + "MetricName": "UNCORE_FREQ" + }, { "BriefDescription": "Percentage of cycles spent in System Management Interrupts.", "MetricExpr": "((msr@aperf@ - cycles) / msr@aperf@ if msr@smi@ > 0 else 0)",
@@ -1026,6 +1032,12 @@ "MetricGroup": "SMT", "MetricName": "tma_info_system_smt_2t_utilization" }, + { + "BriefDescription": "Socket actual clocks when any core is active on that socket", + "MetricExpr": "UNC_CLOCK.SOCKET", + "MetricGroup": "SoC", + "MetricName": "tma_info_system_socket_clks" + }, { "BriefDescription": "Average Frequency Utilization relative nominal frequency", "MetricExpr": "tma_info_thread_clks / CPU_CLK_UNHALTED.REF_TSC",
Signature · 1 line
2.42.0.515.g380fc7ccd1-goog
Quoted 11 lines · Ian Rogers
Broadwell-de has a consumer core and server uncore. The
uncore_arbPMU isn't present and the broadwellx style cbox PMU should be used instead. Fix thetma_info_system_dram_bw_usemetric to use the server metric rather than client.The associated converter script fix is in:
https://github.com/intel/perfmon/pull/111Fixes 7d124303d620 ("perf vendor events intel: Update broadwell variant events/metrics")Signed-off-by Ian Rogers <[email protected]>---
Thanks Ian. The whole patch series looks good to me.
Thanks,
Kan
Quoted 16 lines
tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json +1
-1@@ -652,7 +652,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1e6 / duration_time / 1e3", + "MetricExpr": "64 * (UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR) / 1e9 / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC;tma_issueBW", "MetricName": "tma_info_system_dram_bw_use", "PublicDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]. Related metrics: tma_fb_full, tma_mem_bandwidth, tma_sq_full"
Ian Rogers
reviewerQuoted 12 lines · Kan
Quoted 8 lines · Ian Rogers
Broadwell-de has a consumer core and server uncore. The
uncore_arbPMU isn't present and the broadwellx style cbox PMU should be used instead. Fix thetma_info_system_dram_bw_usemetric to use the server metric rather than client.The associated converter script fix is in:
https://github.com/intel/perfmon/pull/111
Quoted 7 lines
Quoted 3 lines
Fixes 7d124303d620 ("perf vendor events intel: Update broadwell variant events/metrics")Signed-off-by Ian Rogers <[email protected]>---
Thanks Ian. The whole patch series looks good to me.
Reviewed-by Kan Liang <[email protected]>
Would be good to pick this up in perf-tools-next.
Thanks,
Ian
Quoted 19 lines
Thanks,
KanQuoted 16 lines
tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json +1
-1@@ -652,7 +652,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1e6 / duration_time / 1e3", + "MetricExpr": "64 * (UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR) / 1e9 / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC;tma_issueBW", "MetricName": "tma_info_system_dram_bw_use", "PublicDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]. Related metrics: tma_fb_full, tma_mem_bandwidth, tma_sq_full"
Ian Rogers
reviewerQuoted 18 lines · Ian Rogers
Quoted 16 lines · Kan
Quoted 11 lines · Ian Rogers
Broadwell-de has a consumer core and server uncore. The
uncore_arbPMU isn't present and the broadwellx style cbox PMU should be used instead. Fix thetma_info_system_dram_bw_usemetric to use the server metric rather than client.The associated converter script fix is in:
https://github.com/intel/perfmon/pull/111Fixes 7d124303d620 ("perf vendor events intel: Update broadwell variant events/metrics")Signed-off-by Ian Rogers <[email protected]>---
Quoted 8 lines
Quoted 3 lines
Thanks Ian. The whole patch series looks good to me.
Reviewed-by Kan Liang <[email protected]>Would be good to pick this up in perf-tools-next.
Thanks,
Ian
Ping.
Thanks,
Ian
Quoted 19 lines
Quoted 19 lines
Thanks,
KanQuoted 16 lines
tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json +1
-1@@ -652,7 +652,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1e6 / duration_time / 1e3", + "MetricExpr": "64 * (UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR) / 1e9 / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC;tma_issueBW", "MetricName": "tma_info_system_dram_bw_use", "PublicDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]. Related metrics: tma_fb_full, tma_mem_bandwidth, tma_sq_full"
Ian Rogers
reviewerQuoted 22 lines · Ian Rogers
Quoted 20 lines · Ian Rogers
Quoted 18 lines · Kan
Quoted 11 lines · Ian Rogers
Broadwell-de has a consumer core and server uncore. The
uncore_arbPMU isn't present and the broadwellx style cbox PMU should be used instead. Fix thetma_info_system_dram_bw_usemetric to use the server metric rather than client.The associated converter script fix is in:
https://github.com/intel/perfmon/pull/111Fixes 7d124303d620 ("perf vendor events intel: Update broadwell variant events/metrics")Signed-off-by Ian Rogers <[email protected]>---
Thanks Ian. The whole patch series looks good to me.
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Reviewed-by Kan Liang <[email protected]>Would be good to pick this up in perf-tools-next.
Thanks,
IanPing.
Thanks Namhyung for picking this up. There were two other patches in the v2 patch set that Kan reviewed:
https://lore.kernel.org/all/[email protected]/
that I don't see in perf-tools-next.
Ian
Quoted 22 lines
Thanks,
IanQuoted 19 lines
Quoted 19 lines
Thanks,
KanQuoted 16 lines
tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)tools/perf/pmu-events/arch/x86/broadwellde/bdwde-metrics.json +1
-1@@ -652,7 +652,7 @@ }, { "BriefDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]", - "MetricExpr": "64 * (arb@event\\=0x81\\,umask\\=0x1@ + arb@event\\=0x84\\,umask\\=0x1@) / 1e6 / duration_time / 1e3", + "MetricExpr": "64 * (UNC_M_CAS_COUNT.RD + UNC_M_CAS_COUNT.WR) / 1e9 / duration_time", "MetricGroup": "HPC;Mem;MemoryBW;SoC;tma_issueBW", "MetricName": "tma_info_system_dram_bw_use", "PublicDescription": "Average external Memory Bandwidth Use for reads and writes [GB / sec]. Related metrics: tma_fb_full, tma_mem_bandwidth, tma_sq_full"
Namhyung Kim
maintainerHi Ian,
Quoted 31 lines · Ian Rogers
Quoted 29 lines · Ian Rogers
Quoted 26 lines · Ian Rogers
Quoted 12 lines · Kan
Quoted 8 lines · Ian Rogers
Broadwell-de has a consumer core and server uncore. The
uncore_arbPMU isn't present and the broadwellx style cbox PMU should be used instead. Fix thetma_info_system_dram_bw_usemetric to use the server metric rather than client.The associated converter script fix is in:
https://github.com/intel/perfmon/pull/111Quoted 7 lines
Quoted 3 lines
Fixes 7d124303d620 ("perf vendor events intel: Update broadwell variant events/metrics")Signed-off-by Ian Rogers <[email protected]>---
Thanks Ian. The whole patch series looks good to me.
Reviewed-by Kan Liang <[email protected]>Would be good to pick this up in perf-tools-next.
Thanks,
Ian
Quoted 6 lines
Quoted 1 line
Ping.
Thanks Namhyung for picking this up. There were two other patches in the v2 patch set that Kan reviewed:
https://lore.kernel.org/all/[email protected]/
that I don't see in perf-tools-next.
Sorry about that. I'll take care of them too.
Thanks,
Namhyung
Quoted 35 lines · Namhyung Kim
Hi Ian,
Quoted 31 lines · Ian Rogers
Quoted 29 lines · Ian Rogers
Quoted 26 lines · Ian Rogers
Quoted 12 lines · Kan
Quoted 8 lines · Ian Rogers
Broadwell-de has a consumer core and server uncore. The
uncore_arbPMU isn't present and the broadwellx style cbox PMU should be used instead. Fix thetma_info_system_dram_bw_usemetric to use the server metric rather than client.The associated converter script fix is in:
https://github.com/intel/perfmon/pull/111Quoted 7 lines
Quoted 3 lines
Fixes 7d124303d620 ("perf vendor events intel: Update broadwell variant events/metrics")Signed-off-by Ian Rogers <[email protected]>---
Thanks Ian. The whole patch series looks good to me.
Reviewed-by Kan Liang <[email protected]>Would be good to pick this up in perf-tools-next.
Thanks,
Ian
Quoted 8 lines
Quoted 6 lines
Quoted 1 line
Ping.
Thanks Namhyung for picking this up. There were two other patches in the v2 patch set that Kan reviewed:
https://lore.kernel.org/all/[email protected]/
that I don't see in perf-tools-next.Sorry about that. I'll take care of them too.
Applied to perf-tools-next, thanks!