Cleanup Intel PMU initialization code
Subsystems (2)
cover Cleanup Intel PMU initialization code
From: Kan Liang <[email protected]>
The patch set is to clean up the Intel PMU initialization code. There is no functional change.
Kan Liang (6):
perf/x86/intel: Use the common uarch name for the shared functions perf/x86/intel: Factor out the initialization code for SPR perf/x86/intel: Factor out the initialization code for ADL e-core perf/x86/intel: Apply the common initialization code for ADL perf/x86/intel: Cleanup the hybrid CPU type perf/x86/intel: Add common intel_pmu_init_hybrid
arch/x86/events/core.c | 6 +- arch/x86/events/intel/core.c | 397 ++++++++++++++++++----------------- arch/x86/events/intel/ds.c | 4 +- arch/x86/events/perf_event.h | 37 ++-- 4 files changed, 226 insertions(+), 218 deletions(-)
Signature · 1 line
2.35.1
PATCH 1/6 perf/x86/intel: Use the common uarch name for the shared functions
From: Kan Liang <[email protected]>
From PMU's perspective, the SPR/GNR server has a similar uarch to the ADL/MTL client p-core. Many functions are shared. However, the shared function name uses the abbreviation of the server product code name, rather than the common uarch code name.
Rename these internal shared functions by the common uarch name.
--- arch/x86/events/intel/core.c | 64 ++++++++++++++++++------------------ arch/x86/events/intel/ds.c | 2 +- arch/x86/events/perf_event.h | 2 +- 3 files changed, 34 insertions(+), 34 deletions(-)
arch/x86/events/intel/core.c +32 -32
@@ -299,7 +299,7 @@ static struct extra_reg intel_icl_extra_regs[] __read_mostly = {EVENT_EXTRA_END }; -static struct extra_reg intel_spr_extra_regs[]__read_mostly = { +static struct extra_reg intel_glc_extra_regs[]__read_mostly = {INTEL_UEVENT_EXTRA_REG (0x012a ,MSR_OFFCORE_RSP_0 ,0x3fffffffffull ,RSP_0 ),INTEL_UEVENT_EXTRA_REG (0x012b ,MSR_OFFCORE_RSP_1 ,0x3fffffffffull ,RSP_1 ),INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG (0x01cd ),
@@ -309,7 +309,7 @@ static struct extra_reg intel_spr_extra_regs[] __read_mostly = {EVENT_EXTRA_END }; -static struct event_constraint intel_spr_event_constraints[] = { +static struct event_constraint intel_glc_event_constraints[] = {FIXED_EVENT_CONSTRAINT (0x00c0 ,0 ),/* INST_RETIRED.ANY */ FIXED_EVENT_CONSTRAINT (0x0100 ,0 ),/* INST_RETIRED.PREC_DIST */ FIXED_EVENT_CONSTRAINT (0x003c ,1 ),/* CPU_CLK_UNHALTED.CORE */
@@ -349,7 +349,7 @@ static struct event_constraint intel_spr_event_constraints[] = {EVENT_CONSTRAINT_END }; -static struct extra_reg intel_gnr_extra_regs[]__read_mostly = { +static struct extra_reg intel_rwc_extra_regs[]__read_mostly = {INTEL_UEVENT_EXTRA_REG (0x012a ,MSR_OFFCORE_RSP_0 ,0x3fffffffffull ,RSP_0 ),INTEL_UEVENT_EXTRA_REG (0x012b ,MSR_OFFCORE_RSP_1 ,0x3fffffffffull ,RSP_1 ),INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG (0x01cd ),
@@ -473,7 +473,7 @@ static u64 intel_pmu_event_map(int hw_event)return intel_perfmon_event_map[hw_event]; } -static __initconst const u64 spr_hw_cache_event_ids +static __initconst const u64 glc_hw_cache_event_ids [PERF_COUNT_HW_CACHE_MAX ] [PERF_COUNT_HW_CACHE_OP_MAX ] [PERF_COUNT_HW_CACHE_RESULT_MAX] =
@@ -552,7 +552,7 @@ static __initconst const u64 spr_hw_cache_event_ids }, }; -static __initconst const u64 spr_hw_cache_extra_regs +static __initconst const u64 glc_hw_cache_extra_regs [PERF_COUNT_HW_CACHE_MAX ] [PERF_COUNT_HW_CACHE_OP_MAX ] [PERF_COUNT_HW_CACHE_RESULT_MAX] =
@@ -4266,7 +4266,7 @@ icl_get_event_constraints(struct cpu_hw_events *cpuc, int idx, }static struct event_constraint * -spr_get_event_constraints (struct cpu_hw_events *cpuc,int idx, +glc_get_event_constraints (struct cpu_hw_events *cpuc,int idx,struct perf_event *event) {struct event_constraint *c;
@@ -4355,7 +4355,7 @@ adl_get_event_constraints(struct cpu_hw_events *cpuc, int idx,struct x86_hybrid_pmu *pmu =hybrid_pmu (event->pmu);if (pmu->cpu_type == hybrid_big) -return spr_get_event_constraints (cpuc, idx, event); +return glc_get_event_constraints (cpuc, idx, event);else if (pmu->cpu_type == hybrid_small)return tnt_get_event_constraints (cpuc, idx, event);
@@ -4402,7 +4402,7 @@ rwc_get_event_constraints(struct cpu_hw_events *cpuc, int idx, {struct event_constraint *c; - c =spr_get_event_constraints (cpuc, idx, event); + c =glc_get_event_constraints (cpuc, idx, event);/* The Retire Latency is not supported by the fixed counter 0. */ if (event->attr.precise_ip &&
@@ -4483,7 +4483,7 @@ static void nhm_limit_period(struct perf_event *event, s64 *left) *left =max (*left,32LL ); } -static void spr_limit_period (struct perf_event *event,s64 *left) +static void glc_limit_period (struct perf_event *event,s64 *left) {if (event->attr.precise_ip ==3 ) *left =max (*left,128LL );
@@ -5330,14 +5330,14 @@ static struct attribute *icl_tsx_events_attrs[] = {EVENT_ATTR_STR (mem-stores, mem_st_spr,"event=0xcd,umask=0x2" );EVENT_ATTR_STR (mem-loads-aux, mem_ld_aux,"event=0x03,umask=0x82" ); -static struct attribute *spr_events_attrs[] = { +static struct attribute *glc_events_attrs[] = {EVENT_PTR (mem_ld_hsw),EVENT_PTR (mem_st_spr),EVENT_PTR (mem_ld_aux),NULL , }; -static struct attribute *spr_td_events_attrs[] = { +static struct attribute *glc_td_events_attrs[] = {EVENT_PTR (slots),EVENT_PTR (td_retiring),EVENT_PTR (td_bad_spec),
@@ -5350,7 +5350,7 @@ static struct attribute *spr_td_events_attrs[] = { NULL, }; -static struct attribute *spr_tsx_events_attrs[] = { +static struct attribute *glc_tsx_events_attrs[] = {EVENT_PTR (tx_start),EVENT_PTR (tx_abort),EVENT_PTR (tx_commit),
@@ -6208,7 +6208,7 @@ __init int intel_pmu_init(void)intel_pmu_pebs_data_source_grt (); x86_pmu.pebs_latency_data = adl_latency_data_small; x86_pmu.get_event_constraints = tnt_get_event_constraints; - x86_pmu.limit_period = spr_limit_period; + x86_pmu.limit_period = glc_limit_period; td_attr = tnt_events_attrs; mem_attr = grt_mem_attrs; extra_attr = nhm_format_attr;
@@ -6239,7 +6239,7 @@ __init int intel_pmu_init(void)intel_pmu_pebs_data_source_cmt (); x86_pmu.pebs_latency_data = mtl_latency_data_small; x86_pmu.get_event_constraints = cmt_get_event_constraints; - x86_pmu.limit_period = spr_limit_period; + x86_pmu.limit_period = glc_limit_period; td_attr = cmt_events_attrs; mem_attr = grt_mem_attrs; extra_attr = cmt_format_attr;
@@ -6556,20 +6556,20 @@ __init int intel_pmu_init(void)case INTEL_FAM6_SAPPHIRERAPIDS_X :case INTEL_FAM6_EMERALDRAPIDS_X : x86_pmu.flags |=PMU_FL_MEM_LOADS_AUX ; - x86_pmu.extra_regs = intel_spr_extra_regs; + x86_pmu.extra_regs = intel_glc_extra_regs; fallthrough;case INTEL_FAM6_GRANITERAPIDS_X :case INTEL_FAM6_GRANITERAPIDS_D : pmem = true; x86_pmu.late_ack = true; -memcpy (hw_cache_event_ids, spr_hw_cache_event_ids,sizeof (hw_cache_event_ids)); -memcpy (hw_cache_extra_regs, spr_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); +memcpy (hw_cache_event_ids, glc_hw_cache_event_ids,sizeof (hw_cache_event_ids)); +memcpy (hw_cache_extra_regs, glc_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); - x86_pmu.event_constraints = intel_spr_event_constraints; - x86_pmu.pebs_constraints = intel_spr_pebs_event_constraints; + x86_pmu.event_constraints = intel_glc_event_constraints; + x86_pmu.pebs_constraints = intel_glc_pebs_event_constraints;if (!x86_pmu.extra_regs) - x86_pmu.extra_regs = intel_gnr_extra_regs; - x86_pmu.limit_period = spr_limit_period; + x86_pmu.extra_regs = intel_rwc_extra_regs; + x86_pmu.limit_period = glc_limit_period; x86_pmu.pebs_ept =1 ; x86_pmu.pebs_aliases =NULL ; x86_pmu.pebs_prec_dist = true;
@@ -6579,13 +6579,13 @@ __init int intel_pmu_init(void) x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; x86_pmu.hw_config = hsw_hw_config; - x86_pmu.get_event_constraints = spr_get_event_constraints; + x86_pmu.get_event_constraints = glc_get_event_constraints; extra_attr =boot_cpu_has (X86_FEATURE_RTM ) ? hsw_format_attr : nhm_format_attr; extra_skl_attr = skl_format_attr; - mem_attr = spr_events_attrs; - td_attr = spr_td_events_attrs; - tsx_attr = spr_tsx_events_attrs; + mem_attr = glc_events_attrs; + td_attr = glc_td_events_attrs; + tsx_attr = glc_tsx_events_attrs; x86_pmu.rtm_abort_event = X86_CONFIG(.event=0xc9, .umask=0x04); x86_pmu.lbr_pt_coexist = true;intel_pmu_pebs_data_source_skl (pmem);
@@ -6635,7 +6635,7 @@ __init int intel_pmu_init(void) x86_pmu.filter = intel_pmu_filter; x86_pmu.get_event_constraints = adl_get_event_constraints; x86_pmu.hw_config = adl_hw_config; - x86_pmu.limit_period = spr_limit_period; + x86_pmu.limit_period = glc_limit_period; x86_pmu.get_hybrid_cpu_type = adl_get_hybrid_cpu_type;/* * The rtm_abort_event is used to check whether to enable GPRs
@@ -6684,11 +6684,11 @@ __init int intel_pmu_init(void) pmu->intel_cap.perf_metrics =1 ; pmu->intel_cap.pebs_output_pt_available =0 ; -memcpy (pmu->hw_cache_event_ids, spr_hw_cache_event_ids,sizeof (pmu->hw_cache_event_ids)); -memcpy (pmu->hw_cache_extra_regs, spr_hw_cache_extra_regs,sizeof (pmu->hw_cache_extra_regs)); - pmu->event_constraints = intel_spr_event_constraints; - pmu->pebs_constraints = intel_spr_pebs_event_constraints; - pmu->extra_regs = intel_spr_extra_regs; +memcpy (pmu->hw_cache_event_ids, glc_hw_cache_event_ids,sizeof (pmu->hw_cache_event_ids)); +memcpy (pmu->hw_cache_extra_regs, glc_hw_cache_extra_regs,sizeof (pmu->hw_cache_extra_regs)); + pmu->event_constraints = intel_glc_event_constraints; + pmu->pebs_constraints = intel_glc_pebs_event_constraints; + pmu->extra_regs = intel_glc_extra_regs;/* Initialize Atom core specific PerfMon capabilities.*/ pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_ATOM_IDX ];
@@ -6712,7 +6712,7 @@ __init int intel_pmu_init(void) pmu->pebs_constraints = intel_grt_pebs_event_constraints; pmu->extra_regs = intel_grt_extra_regs;if (is_mtl (boot_cpu_data.x86_model)) { - x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ].extra_regs = intel_gnr_extra_regs; + x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ].extra_regs = intel_rwc_extra_regs; x86_pmu.pebs_latency_data = mtl_latency_data_small; extra_attr =boot_cpu_has (X86_FEATURE_RTM ) ? mtl_hybrid_extra_attr_rtm : mtl_hybrid_extra_attr;
arch/x86/events/intel/ds.c +1 -1
@@ -1058,7 +1058,7 @@ struct event_constraint intel_icl_pebs_event_constraints[] = {EVENT_CONSTRAINT_END }; -struct event_constraint intel_spr_pebs_event_constraints[] = { +struct event_constraint intel_glc_pebs_event_constraints[] = {INTEL_FLAGS_UEVENT_CONSTRAINT (0x100 ,0x100000000ULL ),/* INST_RETIRED.PREC_DIST */ INTEL_FLAGS_UEVENT_CONSTRAINT (0x0400 ,0x800000000ULL ),
arch/x86/events/perf_event.h +1 -1
@@ -1521,7 +1521,7 @@ extern struct event_constraint intel_skl_pebs_event_constraints[];extern struct event_constraint intel_icl_pebs_event_constraints[]; -extern struct event_constraint intel_spr_pebs_event_constraints[]; +extern struct event_constraint intel_glc_pebs_event_constraints[];struct event_constraint *intel_pebs_constraints (struct perf_event *event);
Signature · 1 line
2.35.1
[tip: perf/core] perf/x86/intel: Use the common uarch name for the shared functions
The following commit has been merged into the perf/core branch of tip:
The following commit has been merged into the perf/core branch of tip:
Commit-ID: d4b5694c75d4eba8238d541a55da0c67e876213e Gitweb: https://git.kernel.org/tip/d4b5694c75d4eba8238d541a55da0c67e876213e Author: Kan Liang <[email protected]> AuthorDate: Tue, 29 Aug 2023 05:58:01 -07:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 29 Aug 2023 20:59:22 +02:00
perf/x86/intel: Use the common uarch name for the shared functions
Quoted 1 line
From PMU's perspective, the SPR/GNR server has a similar uarch to the
ADL/MTL client p-core. Many functions are shared. However, the shared function name uses the abbreviation of the server product code name, rather than the common uarch code name.
Rename these internal shared functions by the common uarch name.
--- arch/x86/events/intel/core.c | 64 +++++++++++++++++------------------ arch/x86/events/intel/ds.c | 2 +- arch/x86/events/perf_event.h | 2 +- 3 files changed, 34 insertions(+), 34 deletions(-)
arch/x86/events/intel/core.c +32 -32
@@ -299,7 +299,7 @@ static struct extra_reg intel_icl_extra_regs[] __read_mostly = {EVENT_EXTRA_END }; -static struct extra_reg intel_spr_extra_regs[]__read_mostly = { +static struct extra_reg intel_glc_extra_regs[]__read_mostly = {INTEL_UEVENT_EXTRA_REG (0x012a ,MSR_OFFCORE_RSP_0 ,0x3fffffffffull ,RSP_0 ),INTEL_UEVENT_EXTRA_REG (0x012b ,MSR_OFFCORE_RSP_1 ,0x3fffffffffull ,RSP_1 ),INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG (0x01cd ),
@@ -309,7 +309,7 @@ static struct extra_reg intel_spr_extra_regs[] __read_mostly = {EVENT_EXTRA_END }; -static struct event_constraint intel_spr_event_constraints[] = { +static struct event_constraint intel_glc_event_constraints[] = {FIXED_EVENT_CONSTRAINT (0x00c0 ,0 ),/* INST_RETIRED.ANY */ FIXED_EVENT_CONSTRAINT (0x0100 ,0 ),/* INST_RETIRED.PREC_DIST */ FIXED_EVENT_CONSTRAINT (0x003c ,1 ),/* CPU_CLK_UNHALTED.CORE */
@@ -349,7 +349,7 @@ static struct event_constraint intel_spr_event_constraints[] = {EVENT_CONSTRAINT_END }; -static struct extra_reg intel_gnr_extra_regs[]__read_mostly = { +static struct extra_reg intel_rwc_extra_regs[]__read_mostly = {INTEL_UEVENT_EXTRA_REG (0x012a ,MSR_OFFCORE_RSP_0 ,0x3fffffffffull ,RSP_0 ),INTEL_UEVENT_EXTRA_REG (0x012b ,MSR_OFFCORE_RSP_1 ,0x3fffffffffull ,RSP_1 ),INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG (0x01cd ),
@@ -473,7 +473,7 @@ static u64 intel_pmu_event_map(int hw_event)return intel_perfmon_event_map[hw_event]; } -static __initconst const u64 spr_hw_cache_event_ids +static __initconst const u64 glc_hw_cache_event_ids [PERF_COUNT_HW_CACHE_MAX ] [PERF_COUNT_HW_CACHE_OP_MAX ] [PERF_COUNT_HW_CACHE_RESULT_MAX] =
@@ -552,7 +552,7 @@ static __initconst const u64 spr_hw_cache_event_ids }, }; -static __initconst const u64 spr_hw_cache_extra_regs +static __initconst const u64 glc_hw_cache_extra_regs [PERF_COUNT_HW_CACHE_MAX ] [PERF_COUNT_HW_CACHE_OP_MAX ] [PERF_COUNT_HW_CACHE_RESULT_MAX] =
@@ -4273,7 +4273,7 @@ icl_get_event_constraints(struct cpu_hw_events *cpuc, int idx, }static struct event_constraint * -spr_get_event_constraints (struct cpu_hw_events *cpuc,int idx, +glc_get_event_constraints (struct cpu_hw_events *cpuc,int idx,struct perf_event *event) {struct event_constraint *c;
@@ -4362,7 +4362,7 @@ adl_get_event_constraints(struct cpu_hw_events *cpuc, int idx,struct x86_hybrid_pmu *pmu =hybrid_pmu (event->pmu);if (pmu->cpu_type == hybrid_big) -return spr_get_event_constraints (cpuc, idx, event); +return glc_get_event_constraints (cpuc, idx, event);else if (pmu->cpu_type == hybrid_small)return tnt_get_event_constraints (cpuc, idx, event);
@@ -4409,7 +4409,7 @@ rwc_get_event_constraints(struct cpu_hw_events *cpuc, int idx, {struct event_constraint *c; - c =spr_get_event_constraints (cpuc, idx, event); + c =glc_get_event_constraints (cpuc, idx, event);/* The Retire Latency is not supported by the fixed counter 0. */ if (event->attr.precise_ip &&
@@ -4490,7 +4490,7 @@ static void nhm_limit_period(struct perf_event *event, s64 *left) *left =max (*left,32LL ); } -static void spr_limit_period (struct perf_event *event,s64 *left) +static void glc_limit_period (struct perf_event *event,s64 *left) {if (event->attr.precise_ip ==3 ) *left =max (*left,128LL );
@@ -5337,14 +5337,14 @@ static struct attribute *icl_tsx_events_attrs[] = {EVENT_ATTR_STR (mem-stores, mem_st_spr,"event=0xcd,umask=0x2" );EVENT_ATTR_STR (mem-loads-aux, mem_ld_aux,"event=0x03,umask=0x82" ); -static struct attribute *spr_events_attrs[] = { +static struct attribute *glc_events_attrs[] = {EVENT_PTR (mem_ld_hsw),EVENT_PTR (mem_st_spr),EVENT_PTR (mem_ld_aux),NULL , }; -static struct attribute *spr_td_events_attrs[] = { +static struct attribute *glc_td_events_attrs[] = {EVENT_PTR (slots),EVENT_PTR (td_retiring),EVENT_PTR (td_bad_spec),
@@ -5357,7 +5357,7 @@ static struct attribute *spr_td_events_attrs[] = { NULL, }; -static struct attribute *spr_tsx_events_attrs[] = { +static struct attribute *glc_tsx_events_attrs[] = {EVENT_PTR (tx_start),EVENT_PTR (tx_abort),EVENT_PTR (tx_commit),
@@ -6215,7 +6215,7 @@ __init int intel_pmu_init(void)intel_pmu_pebs_data_source_grt (); x86_pmu.pebs_latency_data = adl_latency_data_small; x86_pmu.get_event_constraints = tnt_get_event_constraints; - x86_pmu.limit_period = spr_limit_period; + x86_pmu.limit_period = glc_limit_period; td_attr = tnt_events_attrs; mem_attr = grt_mem_attrs; extra_attr = nhm_format_attr;
@@ -6246,7 +6246,7 @@ __init int intel_pmu_init(void)intel_pmu_pebs_data_source_cmt (); x86_pmu.pebs_latency_data = mtl_latency_data_small; x86_pmu.get_event_constraints = cmt_get_event_constraints; - x86_pmu.limit_period = spr_limit_period; + x86_pmu.limit_period = glc_limit_period; td_attr = cmt_events_attrs; mem_attr = grt_mem_attrs; extra_attr = cmt_format_attr;
@@ -6563,20 +6563,20 @@ __init int intel_pmu_init(void)case INTEL_FAM6_SAPPHIRERAPIDS_X :case INTEL_FAM6_EMERALDRAPIDS_X : x86_pmu.flags |=PMU_FL_MEM_LOADS_AUX ; - x86_pmu.extra_regs = intel_spr_extra_regs; + x86_pmu.extra_regs = intel_glc_extra_regs; fallthrough;case INTEL_FAM6_GRANITERAPIDS_X :case INTEL_FAM6_GRANITERAPIDS_D : pmem = true; x86_pmu.late_ack = true; -memcpy (hw_cache_event_ids, spr_hw_cache_event_ids,sizeof (hw_cache_event_ids)); -memcpy (hw_cache_extra_regs, spr_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); +memcpy (hw_cache_event_ids, glc_hw_cache_event_ids,sizeof (hw_cache_event_ids)); +memcpy (hw_cache_extra_regs, glc_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); - x86_pmu.event_constraints = intel_spr_event_constraints; - x86_pmu.pebs_constraints = intel_spr_pebs_event_constraints; + x86_pmu.event_constraints = intel_glc_event_constraints; + x86_pmu.pebs_constraints = intel_glc_pebs_event_constraints;if (!x86_pmu.extra_regs) - x86_pmu.extra_regs = intel_gnr_extra_regs; - x86_pmu.limit_period = spr_limit_period; + x86_pmu.extra_regs = intel_rwc_extra_regs; + x86_pmu.limit_period = glc_limit_period; x86_pmu.pebs_ept =1 ; x86_pmu.pebs_aliases =NULL ; x86_pmu.pebs_prec_dist = true;
@@ -6586,13 +6586,13 @@ __init int intel_pmu_init(void) x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; x86_pmu.hw_config = hsw_hw_config; - x86_pmu.get_event_constraints = spr_get_event_constraints; + x86_pmu.get_event_constraints = glc_get_event_constraints; extra_attr =boot_cpu_has (X86_FEATURE_RTM ) ? hsw_format_attr : nhm_format_attr; extra_skl_attr = skl_format_attr; - mem_attr = spr_events_attrs; - td_attr = spr_td_events_attrs; - tsx_attr = spr_tsx_events_attrs; + mem_attr = glc_events_attrs; + td_attr = glc_td_events_attrs; + tsx_attr = glc_tsx_events_attrs; x86_pmu.rtm_abort_event = X86_CONFIG(.event=0xc9, .umask=0x04); x86_pmu.lbr_pt_coexist = true;intel_pmu_pebs_data_source_skl (pmem);
@@ -6642,7 +6642,7 @@ __init int intel_pmu_init(void) x86_pmu.filter = intel_pmu_filter; x86_pmu.get_event_constraints = adl_get_event_constraints; x86_pmu.hw_config = adl_hw_config; - x86_pmu.limit_period = spr_limit_period; + x86_pmu.limit_period = glc_limit_period; x86_pmu.get_hybrid_cpu_type = adl_get_hybrid_cpu_type;/* * The rtm_abort_event is used to check whether to enable GPRs
@@ -6691,11 +6691,11 @@ __init int intel_pmu_init(void) pmu->intel_cap.perf_metrics =1 ; pmu->intel_cap.pebs_output_pt_available =0 ; -memcpy (pmu->hw_cache_event_ids, spr_hw_cache_event_ids,sizeof (pmu->hw_cache_event_ids)); -memcpy (pmu->hw_cache_extra_regs, spr_hw_cache_extra_regs,sizeof (pmu->hw_cache_extra_regs)); - pmu->event_constraints = intel_spr_event_constraints; - pmu->pebs_constraints = intel_spr_pebs_event_constraints; - pmu->extra_regs = intel_spr_extra_regs; +memcpy (pmu->hw_cache_event_ids, glc_hw_cache_event_ids,sizeof (pmu->hw_cache_event_ids)); +memcpy (pmu->hw_cache_extra_regs, glc_hw_cache_extra_regs,sizeof (pmu->hw_cache_extra_regs)); + pmu->event_constraints = intel_glc_event_constraints; + pmu->pebs_constraints = intel_glc_pebs_event_constraints; + pmu->extra_regs = intel_glc_extra_regs;/* Initialize Atom core specific PerfMon capabilities.*/ pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_ATOM_IDX ];
@@ -6719,7 +6719,7 @@ __init int intel_pmu_init(void) pmu->pebs_constraints = intel_grt_pebs_event_constraints; pmu->extra_regs = intel_grt_extra_regs;if (is_mtl (boot_cpu_data.x86_model)) { - x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ].extra_regs = intel_gnr_extra_regs; + x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ].extra_regs = intel_rwc_extra_regs; x86_pmu.pebs_latency_data = mtl_latency_data_small; extra_attr =boot_cpu_has (X86_FEATURE_RTM ) ? mtl_hybrid_extra_attr_rtm : mtl_hybrid_extra_attr;
arch/x86/events/intel/ds.c +1 -1
@@ -1058,7 +1058,7 @@ struct event_constraint intel_icl_pebs_event_constraints[] = {EVENT_CONSTRAINT_END }; -struct event_constraint intel_spr_pebs_event_constraints[] = { +struct event_constraint intel_glc_pebs_event_constraints[] = {INTEL_FLAGS_UEVENT_CONSTRAINT (0x100 ,0x100000000ULL ),/* INST_RETIRED.PREC_DIST */ INTEL_FLAGS_UEVENT_CONSTRAINT (0x0400 ,0x800000000ULL ),
arch/x86/events/perf_event.h +1 -1
@@ -1521,7 +1521,7 @@ extern struct event_constraint intel_skl_pebs_event_constraints[];extern struct event_constraint intel_icl_pebs_event_constraints[]; -extern struct event_constraint intel_spr_pebs_event_constraints[]; +extern struct event_constraint intel_glc_pebs_event_constraints[];struct event_constraint *intel_pebs_constraints (struct perf_event *event);
PATCH 2/6 perf/x86/intel: Factor out the initialization code for SPR
In reply to [email protected]From: Kan Liang <[email protected]>
The SPR and ADL p-core have a similar uarch. Most of the initialization code can be shared.
Factor out intel_pmu_init_glc() for the common initialization code. The common part of the ADL p-core will be replaced by the later patch.
--- arch/x86/events/intel/core.c | 49 +++++++++++++++++++----------------- 1 file changed, 26 insertions(+), 23 deletions(-)
arch/x86/events/intel/core.c +26 -23
@@ -5909,6 +5909,30 @@ static __always_inline bool is_mtl(u8 x86_model) (x86_model ==INTEL_FAM6_METEORLAKE_L ); } +static __always_inline void intel_pmu_init_glc(struct pmu *pmu) +{ + x86_pmu.late_ack = true; + x86_pmu.limit_period = glc_limit_period; + x86_pmu.pebs_aliases =NULL ; + x86_pmu.pebs_prec_dist = true; + x86_pmu.pebs_block = true; + x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; + x86_pmu.flags |=PMU_FL_NO_HT_SHARING ; + x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; + x86_pmu.rtm_abort_event = X86_CONFIG(.event=0xc9, .umask=0x04); + x86_pmu.lbr_pt_coexist = true; + x86_pmu.num_topdown_events =8 ; +static_call_update (intel_pmu_update_topdown_event, + &icl_update_topdown_event); +static_call_update (intel_pmu_set_topdown_event_period, + &icl_set_topdown_event_period); + +memcpy (hybrid_var (pmu, hw_cache_event_ids), glc_hw_cache_event_ids,sizeof (hw_cache_event_ids)); +memcpy (hybrid_var (pmu, hw_cache_extra_regs), glc_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); +hybrid (pmu, event_constraints) = intel_glc_event_constraints; +hybrid (pmu, pebs_constraints) = intel_glc_pebs_event_constraints; +} +__init int intel_pmu_init (void ) {struct attribute **extra_skl_attr = &empty_attrs;
@@ -6560,24 +6584,10 @@ __init int intel_pmu_init(void) fallthrough;case INTEL_FAM6_GRANITERAPIDS_X :case INTEL_FAM6_GRANITERAPIDS_D : - pmem = true; - x86_pmu.late_ack = true; -memcpy (hw_cache_event_ids, glc_hw_cache_event_ids,sizeof (hw_cache_event_ids)); -memcpy (hw_cache_extra_regs, glc_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); - - x86_pmu.event_constraints = intel_glc_event_constraints; - x86_pmu.pebs_constraints = intel_glc_pebs_event_constraints; +intel_pmu_init_glc (NULL );if (!x86_pmu.extra_regs) x86_pmu.extra_regs = intel_rwc_extra_regs; - x86_pmu.limit_period = glc_limit_period; x86_pmu.pebs_ept =1 ; - x86_pmu.pebs_aliases =NULL ; - x86_pmu.pebs_prec_dist = true; - x86_pmu.pebs_block = true; - x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; - x86_pmu.flags |=PMU_FL_NO_HT_SHARING ; - x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; - x86_pmu.hw_config = hsw_hw_config; x86_pmu.get_event_constraints = glc_get_event_constraints; extra_attr = boot_cpu_has(X86_FEATURE_RTM) ?
@@ -6586,14 +6596,7 @@ __init int intel_pmu_init(void) mem_attr = glc_events_attrs; td_attr = glc_td_events_attrs; tsx_attr = glc_tsx_events_attrs; - x86_pmu.rtm_abort_event = X86_CONFIG(.event=0xc9, .umask=0x04); - x86_pmu.lbr_pt_coexist = true; -intel_pmu_pebs_data_source_skl (pmem); - x86_pmu.num_topdown_events =8 ; -static_call_update (intel_pmu_update_topdown_event, - &icl_update_topdown_event); -static_call_update (intel_pmu_set_topdown_event_period, - &icl_set_topdown_event_period); +intel_pmu_pebs_data_source_skl (true);pr_cont ("Sapphire Rapids events, " ); name ="sapphire_rapids" ;break ;
Signature · 1 line
2.35.1
[tip: perf/core] perf/x86/intel: Factor out the initialization code for SPR
The following commit has been merged into the perf/core branch of tip:
The following commit has been merged into the perf/core branch of tip:
Commit-ID: 0ba0c03528e918a8f6b5aa63d502fdc6a9d80fc7 Gitweb: https://git.kernel.org/tip/0ba0c03528e918a8f6b5aa63d502fdc6a9d80fc7 Author: Kan Liang <[email protected]> AuthorDate: Tue, 29 Aug 2023 05:58:02 -07:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 29 Aug 2023 20:59:22 +02:00
perf/x86/intel: Factor out the initialization code for SPR
The SPR and ADL p-core have a similar uarch. Most of the initialization code can be shared.
Factor out intel_pmu_init_glc() for the common initialization code. The common part of the ADL p-core will be replaced by the later patch.
--- arch/x86/events/intel/core.c | 49 ++++++++++++++++++----------------- 1 file changed, 26 insertions(+), 23 deletions(-)
arch/x86/events/intel/core.c +26 -23
@@ -5916,6 +5916,30 @@ static __always_inline bool is_mtl(u8 x86_model) (x86_model ==INTEL_FAM6_METEORLAKE_L ); } +static __always_inline void intel_pmu_init_glc(struct pmu *pmu) +{ + x86_pmu.late_ack = true; + x86_pmu.limit_period = glc_limit_period; + x86_pmu.pebs_aliases =NULL ; + x86_pmu.pebs_prec_dist = true; + x86_pmu.pebs_block = true; + x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; + x86_pmu.flags |=PMU_FL_NO_HT_SHARING ; + x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; + x86_pmu.rtm_abort_event = X86_CONFIG(.event=0xc9, .umask=0x04); + x86_pmu.lbr_pt_coexist = true; + x86_pmu.num_topdown_events =8 ; +static_call_update (intel_pmu_update_topdown_event, + &icl_update_topdown_event); +static_call_update (intel_pmu_set_topdown_event_period, + &icl_set_topdown_event_period); + +memcpy (hybrid_var (pmu, hw_cache_event_ids), glc_hw_cache_event_ids,sizeof (hw_cache_event_ids)); +memcpy (hybrid_var (pmu, hw_cache_extra_regs), glc_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); +hybrid (pmu, event_constraints) = intel_glc_event_constraints; +hybrid (pmu, pebs_constraints) = intel_glc_pebs_event_constraints; +} +__init int intel_pmu_init (void ) {struct attribute **extra_skl_attr = &empty_attrs;
@@ -6567,24 +6591,10 @@ __init int intel_pmu_init(void) fallthrough;case INTEL_FAM6_GRANITERAPIDS_X :case INTEL_FAM6_GRANITERAPIDS_D : - pmem = true; - x86_pmu.late_ack = true; -memcpy (hw_cache_event_ids, glc_hw_cache_event_ids,sizeof (hw_cache_event_ids)); -memcpy (hw_cache_extra_regs, glc_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); - - x86_pmu.event_constraints = intel_glc_event_constraints; - x86_pmu.pebs_constraints = intel_glc_pebs_event_constraints; +intel_pmu_init_glc (NULL );if (!x86_pmu.extra_regs) x86_pmu.extra_regs = intel_rwc_extra_regs; - x86_pmu.limit_period = glc_limit_period; x86_pmu.pebs_ept =1 ; - x86_pmu.pebs_aliases =NULL ; - x86_pmu.pebs_prec_dist = true; - x86_pmu.pebs_block = true; - x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; - x86_pmu.flags |=PMU_FL_NO_HT_SHARING ; - x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; - x86_pmu.hw_config = hsw_hw_config; x86_pmu.get_event_constraints = glc_get_event_constraints; extra_attr = boot_cpu_has(X86_FEATURE_RTM) ?
@@ -6593,14 +6603,7 @@ __init int intel_pmu_init(void) mem_attr = glc_events_attrs; td_attr = glc_td_events_attrs; tsx_attr = glc_tsx_events_attrs; - x86_pmu.rtm_abort_event = X86_CONFIG(.event=0xc9, .umask=0x04); - x86_pmu.lbr_pt_coexist = true; -intel_pmu_pebs_data_source_skl (pmem); - x86_pmu.num_topdown_events =8 ; -static_call_update (intel_pmu_update_topdown_event, - &icl_update_topdown_event); -static_call_update (intel_pmu_set_topdown_event_period, - &icl_set_topdown_event_period); +intel_pmu_pebs_data_source_skl (true);pr_cont ("Sapphire Rapids events, " ); name ="sapphire_rapids" ;break ;
PATCH 3/6 perf/x86/intel: Factor out the initialization code for ADL e-core
In reply to [email protected]From: Kan Liang <[email protected]>
From PMU's perspective, the ADL e-core and newer SRF/GRR have a similar uarch. Most of the initialization code can be shared.
Factor out intel_pmu_init_grt() for the common initialization code. The common part of the ADL e-core will be replaced by the later patch.
--- arch/x86/events/intel/core.c | 58 +++++++++++++----------------------- 1 file changed, 21 insertions(+), 37 deletions(-)
arch/x86/events/intel/core.c +21 -37
@@ -5933,6 +5933,25 @@ static __always_inline void intel_pmu_init_glc(struct pmu *pmu)hybrid (pmu, pebs_constraints) = intel_glc_pebs_event_constraints; } +static __always_inline void intel_pmu_init_grt(struct pmu *pmu) +{ + x86_pmu.mid_ack = true; + x86_pmu.limit_period = glc_limit_period; + x86_pmu.pebs_aliases =NULL ; + x86_pmu.pebs_prec_dist = true; + x86_pmu.pebs_block = true; + x86_pmu.lbr_pt_coexist = true; + x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; + x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; + +memcpy (hybrid_var (pmu, hw_cache_event_ids), glp_hw_cache_event_ids,sizeof (hw_cache_event_ids)); +memcpy (hybrid_var (pmu, hw_cache_extra_regs), tnt_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); +hybrid_var (pmu, hw_cache_event_ids)[C (ITLB )][C (OP_READ )][C (RESULT_ACCESS )] =-1 ; +hybrid (pmu, event_constraints) = intel_slm_event_constraints; +hybrid (pmu, pebs_constraints) = intel_grt_pebs_event_constraints; +hybrid (pmu, extra_regs) = intel_grt_extra_regs; +} +__init int intel_pmu_init (void ) {struct attribute **extra_skl_attr = &empty_attrs;
@@ -6211,28 +6230,10 @@ __init int intel_pmu_init(void)break ;case INTEL_FAM6_ATOM_GRACEMONT : - x86_pmu.mid_ack = true; -memcpy (hw_cache_event_ids, glp_hw_cache_event_ids, -sizeof (hw_cache_event_ids)); -memcpy (hw_cache_extra_regs, tnt_hw_cache_extra_regs, -sizeof (hw_cache_extra_regs)); - hw_cache_event_ids[C (ITLB )][C (OP_READ )][C (RESULT_ACCESS )] =-1 ; - - x86_pmu.event_constraints = intel_slm_event_constraints; - x86_pmu.pebs_constraints = intel_grt_pebs_event_constraints; - x86_pmu.extra_regs = intel_grt_extra_regs; - - x86_pmu.pebs_aliases =NULL ; - x86_pmu.pebs_prec_dist = true; - x86_pmu.pebs_block = true; - x86_pmu.lbr_pt_coexist = true; - x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; - x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; - +intel_pmu_init_grt (NULL );intel_pmu_pebs_data_source_grt (); x86_pmu.pebs_latency_data = adl_latency_data_small; x86_pmu.get_event_constraints = tnt_get_event_constraints; - x86_pmu.limit_period = glc_limit_period; td_attr = tnt_events_attrs; mem_attr = grt_mem_attrs; extra_attr = nhm_format_attr;
@@ -6242,28 +6243,11 @@ __init int intel_pmu_init(void)case INTEL_FAM6_ATOM_CRESTMONT :case INTEL_FAM6_ATOM_CRESTMONT_X : - x86_pmu.mid_ack = true; -memcpy (hw_cache_event_ids, glp_hw_cache_event_ids, -sizeof (hw_cache_event_ids)); -memcpy (hw_cache_extra_regs, tnt_hw_cache_extra_regs, -sizeof (hw_cache_extra_regs)); - hw_cache_event_ids[C (ITLB )][C (OP_READ )][C (RESULT_ACCESS )] =-1 ; - - x86_pmu.event_constraints = intel_slm_event_constraints; - x86_pmu.pebs_constraints = intel_grt_pebs_event_constraints; +intel_pmu_init_grt (NULL ); x86_pmu.extra_regs = intel_cmt_extra_regs; - - x86_pmu.pebs_aliases =NULL ; - x86_pmu.pebs_prec_dist = true; - x86_pmu.lbr_pt_coexist = true; - x86_pmu.pebs_block = true; - x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; - x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; -intel_pmu_pebs_data_source_cmt (); x86_pmu.pebs_latency_data = mtl_latency_data_small; x86_pmu.get_event_constraints = cmt_get_event_constraints; - x86_pmu.limit_period = glc_limit_period; td_attr = cmt_events_attrs; mem_attr = grt_mem_attrs; extra_attr = cmt_format_attr;
Signature · 1 line
2.35.1
[tip: perf/core] perf/x86/intel: Factor out the initialization code for ADL e-core
The following commit has been merged into the perf/core branch of tip:
The following commit has been merged into the perf/core branch of tip:
Commit-ID: d87d221f854b62f5e8026505497d33404ef6050c Gitweb: https://git.kernel.org/tip/d87d221f854b62f5e8026505497d33404ef6050c Author: Kan Liang <[email protected]> AuthorDate: Tue, 29 Aug 2023 05:58:03 -07:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 29 Aug 2023 20:59:22 +02:00
perf/x86/intel: Factor out the initialization code for ADL e-core
Quoted 1 line
From PMU's perspective, the ADL e-core and newer SRF/GRR have a similar
uarch. Most of the initialization code can be shared.
Factor out intel_pmu_init_grt() for the common initialization code. The common part of the ADL e-core will be replaced by the later patch.
--- arch/x86/events/intel/core.c | 58 ++++++++++++----------------------- 1 file changed, 21 insertions(+), 37 deletions(-)
arch/x86/events/intel/core.c +21 -37
@@ -5940,6 +5940,25 @@ static __always_inline void intel_pmu_init_glc(struct pmu *pmu)hybrid (pmu, pebs_constraints) = intel_glc_pebs_event_constraints; } +static __always_inline void intel_pmu_init_grt(struct pmu *pmu) +{ + x86_pmu.mid_ack = true; + x86_pmu.limit_period = glc_limit_period; + x86_pmu.pebs_aliases =NULL ; + x86_pmu.pebs_prec_dist = true; + x86_pmu.pebs_block = true; + x86_pmu.lbr_pt_coexist = true; + x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; + x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; + +memcpy (hybrid_var (pmu, hw_cache_event_ids), glp_hw_cache_event_ids,sizeof (hw_cache_event_ids)); +memcpy (hybrid_var (pmu, hw_cache_extra_regs), tnt_hw_cache_extra_regs,sizeof (hw_cache_extra_regs)); +hybrid_var (pmu, hw_cache_event_ids)[C (ITLB )][C (OP_READ )][C (RESULT_ACCESS )] =-1 ; +hybrid (pmu, event_constraints) = intel_slm_event_constraints; +hybrid (pmu, pebs_constraints) = intel_grt_pebs_event_constraints; +hybrid (pmu, extra_regs) = intel_grt_extra_regs; +} +__init int intel_pmu_init (void ) {struct attribute **extra_skl_attr = &empty_attrs;
@@ -6218,28 +6237,10 @@ __init int intel_pmu_init(void)break ;case INTEL_FAM6_ATOM_GRACEMONT : - x86_pmu.mid_ack = true; -memcpy (hw_cache_event_ids, glp_hw_cache_event_ids, -sizeof (hw_cache_event_ids)); -memcpy (hw_cache_extra_regs, tnt_hw_cache_extra_regs, -sizeof (hw_cache_extra_regs)); - hw_cache_event_ids[C (ITLB )][C (OP_READ )][C (RESULT_ACCESS )] =-1 ; - - x86_pmu.event_constraints = intel_slm_event_constraints; - x86_pmu.pebs_constraints = intel_grt_pebs_event_constraints; - x86_pmu.extra_regs = intel_grt_extra_regs; - - x86_pmu.pebs_aliases =NULL ; - x86_pmu.pebs_prec_dist = true; - x86_pmu.pebs_block = true; - x86_pmu.lbr_pt_coexist = true; - x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; - x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; - +intel_pmu_init_grt (NULL );intel_pmu_pebs_data_source_grt (); x86_pmu.pebs_latency_data = adl_latency_data_small; x86_pmu.get_event_constraints = tnt_get_event_constraints; - x86_pmu.limit_period = glc_limit_period; td_attr = tnt_events_attrs; mem_attr = grt_mem_attrs; extra_attr = nhm_format_attr;
@@ -6249,28 +6250,11 @@ __init int intel_pmu_init(void)case INTEL_FAM6_ATOM_CRESTMONT :case INTEL_FAM6_ATOM_CRESTMONT_X : - x86_pmu.mid_ack = true; -memcpy (hw_cache_event_ids, glp_hw_cache_event_ids, -sizeof (hw_cache_event_ids)); -memcpy (hw_cache_extra_regs, tnt_hw_cache_extra_regs, -sizeof (hw_cache_extra_regs)); - hw_cache_event_ids[C (ITLB )][C (OP_READ )][C (RESULT_ACCESS )] =-1 ; - - x86_pmu.event_constraints = intel_slm_event_constraints; - x86_pmu.pebs_constraints = intel_grt_pebs_event_constraints; +intel_pmu_init_grt (NULL ); x86_pmu.extra_regs = intel_cmt_extra_regs; - - x86_pmu.pebs_aliases =NULL ; - x86_pmu.pebs_prec_dist = true; - x86_pmu.lbr_pt_coexist = true; - x86_pmu.pebs_block = true; - x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; - x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; -intel_pmu_pebs_data_source_cmt (); x86_pmu.pebs_latency_data = mtl_latency_data_small; x86_pmu.get_event_constraints = cmt_get_event_constraints; - x86_pmu.limit_period = glc_limit_period; td_attr = cmt_events_attrs; mem_attr = grt_mem_attrs; extra_attr = cmt_format_attr;
PATCH 4/6 perf/x86/intel: Apply the common initialization code for ADL
In reply to [email protected]From: Kan Liang <[email protected]>
Use the intel_pmu_init_glc() and intel_pmu_init_grt() to replace the duplicate code for ADL.
The current code already checks the PERF_X86_EVENT_TOPDOWN flag before invoking the Topdown metrics functions. (The PERF_X86_EVENT_TOPDOWN flag is to indicate the Topdown metric feature, which is only available for the p-core.) Drop the unnecessary adl_set_topdown_event_period() and adl_update_topdown_event().
--- arch/x86/events/intel/core.c | 53 ++---------------------------------- 1 file changed, 2 insertions(+), 51 deletions(-)
arch/x86/events/intel/core.c +2 -51
@@ -2556,16 +2556,6 @@ static int icl_set_topdown_event_period(struct perf_event *event)return 0 ; } -static int adl_set_topdown_event_period (struct perf_event *event) -{ -struct x86_hybrid_pmu *pmu =hybrid_pmu (event->pmu); - -if (pmu->cpu_type != hybrid_big) -return 0 ; - -return icl_set_topdown_event_period (event); -} -DEFINE_STATIC_CALL (intel_pmu_set_topdown_event_period, x86_perf_event_set_period);static inline u64 icl_get_metrics_event_value (u64 metric,u64 slots,int idx)
@@ -2708,16 +2698,6 @@ static u64 icl_update_topdown_event(struct perf_event *event) x86_pmu.num_topdown_events - 1); } -static u64 adl_update_topdown_event (struct perf_event *event) -{ -struct x86_hybrid_pmu *pmu =hybrid_pmu (event->pmu); - -if (pmu->cpu_type != hybrid_big) -return 0 ; - -return icl_update_topdown_event (event); -} -DEFINE_STATIC_CALL (intel_pmu_update_topdown_event, x86_perf_event_update);static void intel_pmu_read_topdown_event (struct perf_event *event)
@@ -6605,32 +6585,11 @@ __init int intel_pmu_init(void)static_branch_enable (&perf_is_hybrid); x86_pmu.num_hybrid_pmus =X86_HYBRID_NUM_PMUS ; - x86_pmu.pebs_aliases =NULL ; - x86_pmu.pebs_prec_dist = true; - x86_pmu.pebs_block = true; - x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; - x86_pmu.flags |=PMU_FL_NO_HT_SHARING ; - x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; - x86_pmu.lbr_pt_coexist = true; x86_pmu.pebs_latency_data = adl_latency_data_small; - x86_pmu.num_topdown_events =8 ; -static_call_update (intel_pmu_update_topdown_event, - &adl_update_topdown_event); -static_call_update (intel_pmu_set_topdown_event_period, - &adl_set_topdown_event_period); - x86_pmu.filter = intel_pmu_filter; x86_pmu.get_event_constraints = adl_get_event_constraints; x86_pmu.hw_config = adl_hw_config; - x86_pmu.limit_period = glc_limit_period; x86_pmu.get_hybrid_cpu_type = adl_get_hybrid_cpu_type; -/* -* The rtm_abort_event is used to check whether to enable GPRs -* for the RTM abort event. Atom doesn't have the RTM abort -* event. There is no harmful to set it in the common -* x86_pmu.rtm_abort_event. -*/ - x86_pmu.rtm_abort_event = X86_CONFIG(.event=0xc9, .umask=0x04); td_attr = adl_hybrid_events_attrs; mem_attr = adl_hybrid_mem_attrs;
@@ -6642,6 +6601,7 @@ __init int intel_pmu_init(void) pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ]; pmu->name ="cpu_core" ; pmu->cpu_type = hybrid_big; +intel_pmu_init_glc (&pmu->pmu); pmu->late_ack = true;if (cpu_feature_enabled (X86_FEATURE_HYBRID_CPU )) { pmu->num_counters = x86_pmu.num_counters +2 ;
@@ -6671,16 +6631,13 @@ __init int intel_pmu_init(void) pmu->intel_cap.perf_metrics =1 ; pmu->intel_cap.pebs_output_pt_available =0 ; -memcpy (pmu->hw_cache_event_ids, glc_hw_cache_event_ids,sizeof (pmu->hw_cache_event_ids)); -memcpy (pmu->hw_cache_extra_regs, glc_hw_cache_extra_regs,sizeof (pmu->hw_cache_extra_regs)); - pmu->event_constraints = intel_glc_event_constraints; - pmu->pebs_constraints = intel_glc_pebs_event_constraints; pmu->extra_regs = intel_glc_extra_regs;/* Initialize Atom core specific PerfMon capabilities.*/ pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_ATOM_IDX ]; pmu->name ="cpu_atom" ; pmu->cpu_type = hybrid_small; +intel_pmu_init_grt (&pmu->pmu); pmu->mid_ack = true; pmu->num_counters = x86_pmu.num_counters; pmu->num_counters_fixed = x86_pmu.num_counters_fixed;
@@ -6692,12 +6649,6 @@ __init int intel_pmu_init(void) pmu->intel_cap.perf_metrics =0 ; pmu->intel_cap.pebs_output_pt_available =1 ; -memcpy (pmu->hw_cache_event_ids, glp_hw_cache_event_ids,sizeof (pmu->hw_cache_event_ids)); -memcpy (pmu->hw_cache_extra_regs, tnt_hw_cache_extra_regs,sizeof (pmu->hw_cache_extra_regs)); - pmu->hw_cache_event_ids[C (ITLB )][C (OP_READ )][C (RESULT_ACCESS )] =-1 ; - pmu->event_constraints = intel_slm_event_constraints; - pmu->pebs_constraints = intel_grt_pebs_event_constraints; - pmu->extra_regs = intel_grt_extra_regs;if (is_mtl (boot_cpu_data.x86_model)) { x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ].extra_regs = intel_rwc_extra_regs; x86_pmu.pebs_latency_data = mtl_latency_data_small;
Signature · 1 line
2.35.1
[tip: perf/core] perf/x86/intel: Apply the common initialization code for ADL
The following commit has been merged into the perf/core branch of tip:
The following commit has been merged into the perf/core branch of tip:
Commit-ID: 299a5fc8e783eed705015e83e381912dbbf3eabc Gitweb: https://git.kernel.org/tip/299a5fc8e783eed705015e83e381912dbbf3eabc Author: Kan Liang <[email protected]> AuthorDate: Tue, 29 Aug 2023 05:58:04 -07:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 29 Aug 2023 20:59:23 +02:00
perf/x86/intel: Apply the common initialization code for ADL
Use the intel_pmu_init_glc() and intel_pmu_init_grt() to replace the duplicate code for ADL.
The current code already checks the PERF_X86_EVENT_TOPDOWN flag before invoking the Topdown metrics functions. (The PERF_X86_EVENT_TOPDOWN flag is to indicate the Topdown metric feature, which is only available for the p-core.) Drop the unnecessary adl_set_topdown_event_period() and adl_update_topdown_event().
--- arch/x86/events/intel/core.c | 53 +---------------------------------- 1 file changed, 2 insertions(+), 51 deletions(-)
arch/x86/events/intel/core.c +2 -51
@@ -2556,16 +2556,6 @@ static int icl_set_topdown_event_period(struct perf_event *event)return 0 ; } -static int adl_set_topdown_event_period (struct perf_event *event) -{ -struct x86_hybrid_pmu *pmu =hybrid_pmu (event->pmu); - -if (pmu->cpu_type != hybrid_big) -return 0 ; - -return icl_set_topdown_event_period (event); -} -DEFINE_STATIC_CALL (intel_pmu_set_topdown_event_period, x86_perf_event_set_period);static inline u64 icl_get_metrics_event_value (u64 metric,u64 slots,int idx)
@@ -2708,16 +2698,6 @@ static u64 icl_update_topdown_event(struct perf_event *event) x86_pmu.num_topdown_events - 1); } -static u64 adl_update_topdown_event (struct perf_event *event) -{ -struct x86_hybrid_pmu *pmu =hybrid_pmu (event->pmu); - -if (pmu->cpu_type != hybrid_big) -return 0 ; - -return icl_update_topdown_event (event); -} -DEFINE_STATIC_CALL (intel_pmu_update_topdown_event, x86_perf_event_update);static void intel_pmu_read_topdown_event (struct perf_event *event)
@@ -6612,32 +6592,11 @@ __init int intel_pmu_init(void)static_branch_enable (&perf_is_hybrid); x86_pmu.num_hybrid_pmus =X86_HYBRID_NUM_PMUS ; - x86_pmu.pebs_aliases =NULL ; - x86_pmu.pebs_prec_dist = true; - x86_pmu.pebs_block = true; - x86_pmu.flags |=PMU_FL_HAS_RSP_1 ; - x86_pmu.flags |=PMU_FL_NO_HT_SHARING ; - x86_pmu.flags |=PMU_FL_INSTR_LATENCY ; - x86_pmu.lbr_pt_coexist = true; x86_pmu.pebs_latency_data = adl_latency_data_small; - x86_pmu.num_topdown_events =8 ; -static_call_update (intel_pmu_update_topdown_event, - &adl_update_topdown_event); -static_call_update (intel_pmu_set_topdown_event_period, - &adl_set_topdown_event_period); - x86_pmu.filter = intel_pmu_filter; x86_pmu.get_event_constraints = adl_get_event_constraints; x86_pmu.hw_config = adl_hw_config; - x86_pmu.limit_period = glc_limit_period; x86_pmu.get_hybrid_cpu_type = adl_get_hybrid_cpu_type; -/* -* The rtm_abort_event is used to check whether to enable GPRs -* for the RTM abort event. Atom doesn't have the RTM abort -* event. There is no harmful to set it in the common -* x86_pmu.rtm_abort_event. -*/ - x86_pmu.rtm_abort_event = X86_CONFIG(.event=0xc9, .umask=0x04); td_attr = adl_hybrid_events_attrs; mem_attr = adl_hybrid_mem_attrs;
@@ -6649,6 +6608,7 @@ __init int intel_pmu_init(void) pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ]; pmu->name ="cpu_core" ; pmu->cpu_type = hybrid_big; +intel_pmu_init_glc (&pmu->pmu); pmu->late_ack = true;if (cpu_feature_enabled (X86_FEATURE_HYBRID_CPU )) { pmu->num_counters = x86_pmu.num_counters +2 ;
@@ -6678,16 +6638,13 @@ __init int intel_pmu_init(void) pmu->intel_cap.perf_metrics =1 ; pmu->intel_cap.pebs_output_pt_available =0 ; -memcpy (pmu->hw_cache_event_ids, glc_hw_cache_event_ids,sizeof (pmu->hw_cache_event_ids)); -memcpy (pmu->hw_cache_extra_regs, glc_hw_cache_extra_regs,sizeof (pmu->hw_cache_extra_regs)); - pmu->event_constraints = intel_glc_event_constraints; - pmu->pebs_constraints = intel_glc_pebs_event_constraints; pmu->extra_regs = intel_glc_extra_regs;/* Initialize Atom core specific PerfMon capabilities.*/ pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_ATOM_IDX ]; pmu->name ="cpu_atom" ; pmu->cpu_type = hybrid_small; +intel_pmu_init_grt (&pmu->pmu); pmu->mid_ack = true; pmu->num_counters = x86_pmu.num_counters; pmu->num_counters_fixed = x86_pmu.num_counters_fixed;
@@ -6699,12 +6656,6 @@ __init int intel_pmu_init(void) pmu->intel_cap.perf_metrics =0 ; pmu->intel_cap.pebs_output_pt_available =1 ; -memcpy (pmu->hw_cache_event_ids, glp_hw_cache_event_ids,sizeof (pmu->hw_cache_event_ids)); -memcpy (pmu->hw_cache_extra_regs, tnt_hw_cache_extra_regs,sizeof (pmu->hw_cache_extra_regs)); - pmu->hw_cache_event_ids[C (ITLB )][C (OP_READ )][C (RESULT_ACCESS )] =-1 ; - pmu->event_constraints = intel_slm_event_constraints; - pmu->pebs_constraints = intel_grt_pebs_event_constraints; - pmu->extra_regs = intel_grt_extra_regs;if (is_mtl (boot_cpu_data.x86_model)) { x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ].extra_regs = intel_rwc_extra_regs; x86_pmu.pebs_latency_data = mtl_latency_data_small;
PATCH 5/6 perf/x86/intel: Cleanup the hybrid CPU type
In reply to [email protected]From: Kan Liang <[email protected]>
There is a fairly long list of grievances about the current code. The main beefs:
1. hybrid_big_small assumes that the *HARDWARE* (CPUID) provided core types are a bitmap. They are not. If Intel happened to make a core type of 0xff, hilarity would ensue.
2. adl_get_hybrid_cpu_type() utterly inscrutable. There are precisely zero comments and zero changelog about what it is attempting to do.
According to Kan, the adl_get_hybrid_cpu_type() is there because some Alder Lake (ADL) CPUs can do some silly things. Some ADL models are *supposed* to be hybrid CPUs with big and little cores, but there are some SKUs that only have big cores. CPUID(0x1a) on those CPUs does not say that the CPUs are big cores. It apparently just returns 0x0. It confuses perf because it expects to see either 0x40 (Core) or 0x20 (Atom).
The perf workaround for this is to watch for a CPU core saying it is type 0x0. If that happens on an Alder Lake, it calls x86_pmu.get_hybrid_cpu_type() and just assumes that the core is a Core (0x40) CPU.
To fix up the mess, separate out the CPU types and the 'pmu' types. This allows 'hybrid_pmu_type' bitmaps without worrying that some future CPU type will set multiple bits.
Since the types are now separate, add a function to glue them back together again. Actual comment on the situation in the glue function (find_hybrid_pmu_for_cpu()).
Also, give ->get_hybrid_cpu_type() a real return type and make it clear that it is overriding the *CPU* type, not the PMU type.
Rename cpu_type to pmu_type in the struct x86_hybrid_pmu to reflect the change.
--- arch/x86/events/core.c | 6 ++-- arch/x86/events/intel/core.c | 69 ++++++++++++++++++++++++------------ arch/x86/events/intel/ds.c | 2 +- arch/x86/events/perf_event.h | 35 ++++++++++-------- 4 files changed, 72 insertions(+), 40 deletions(-)
arch/x86/events/core.c +3 -3
@@ -1887,9 +1887,9 @@ ssize_t events_hybrid_sysfs_show(struct device *dev, str = pmu_attr->event_str;for (i =0 ; i < x86_pmu.num_hybrid_pmus; i++) { -if (!(x86_pmu.hybrid_pmu[i].cpu_type & pmu_attr->pmu_type)) +if (!(x86_pmu.hybrid_pmu[i].pmu_type & pmu_attr->pmu_type))continue ; -if (x86_pmu.hybrid_pmu[i].cpu_type & pmu->cpu_type) { +if (x86_pmu.hybrid_pmu[i].pmu_type & pmu->pmu_type) { next_str =strchr (str,';' );if (next_str)return snprintf (page, next_str - str +1 ,"%s" , str);
@@ -2169,7 +2169,7 @@ static int __init init_hw_perf_events(void) hybrid_pmu->pmu.capabilities |=PERF_PMU_CAP_EXTENDED_HW_TYPE ; err =perf_pmu_register (&hybrid_pmu->pmu, hybrid_pmu->name, - (hybrid_pmu->cpu_type == hybrid_big) ?PERF_TYPE_RAW :-1 ); + (hybrid_pmu->pmu_type == hybrid_big) ?PERF_TYPE_RAW :-1 );if (err)break ; }
arch/x86/events/intel/core.c +47 -22
140 lines not on this page · raw message
arch/x86/events/intel/ds.c +1 -1
@@ -261,7 +261,7 @@ static u64 __adl_latency_data_small(struct perf_event *event, u64 status, {u64 val; -WARN_ON_ONCE (hybrid_pmu (event->pmu)->cpu_type == hybrid_big); +WARN_ON_ONCE (hybrid_pmu (event->pmu)->pmu_type == hybrid_big); dse &=PERF_PEBS_DATA_SOURCE_MASK ; val =hybrid_var (event->pmu, pebs_data_source)[dse];
arch/x86/events/perf_event.h +21 -14
@@ -652,10 +652,29 @@ enum {#define PERF_PEBS_DATA_SOURCE_MAX 0x10#define PERF_PEBS_DATA_SOURCE_MASK (PERF_PEBS_DATA_SOURCE_MAX - 1) +enum hybrid_cpu_type { +HYBRID_INTEL_NONE , +HYBRID_INTEL_ATOM =0x20 , +HYBRID_INTEL_CORE =0x40 , +}; + +enum hybrid_pmu_type { + not_hybrid, + hybrid_small =BIT (0 ), + hybrid_big =BIT (1 ), + + hybrid_big_small = hybrid_big | hybrid_small,/* only used for matching */ +}; + +#define X86_HYBRID_PMU_ATOM_IDX 0 +#define X86_HYBRID_PMU_CORE_IDX 1 + +#define X86_HYBRID_NUM_PMUS 2 +struct x86_hybrid_pmu {struct pmu pmu;const char *name; -u8 cpu_type; +enum hybrid_pmu_type pmu_type;cpumask_t supported_cpus;union perf_capabilities intel_cap;u64 intel_ctrl;
@@ -721,18 +740,6 @@ extern struct static_key_false perf_is_hybrid; __Fp; \ }) -enum hybrid_pmu_type { - hybrid_big =0x40 , - hybrid_small =0x20 , - - hybrid_big_small = hybrid_big | hybrid_small, -}; - -#define X86_HYBRID_PMU_ATOM_IDX 0 -#define X86_HYBRID_PMU_CORE_IDX 1 - -#define X86_HYBRID_NUM_PMUS 2 -/* * struct x86_pmu - generic x86 pmu */
@@ -940,7 +947,7 @@ struct x86_pmu {*/ int num_hybrid_pmus;struct x86_hybrid_pmu *hybrid_pmu; -u8 (*get_hybrid_cpu_type) (void); +enum hybrid_cpu_type (*get_hybrid_cpu_type) (void ); };struct x86_perf_task_context_opt {
Signature · 1 line
2.35.1
[tip: perf/core] perf/x86/intel: Clean up the hybrid CPU type handling code
The following commit has been merged into the perf/core branch of tip:
The following commit has been merged into the perf/core branch of tip:
Commit-ID: b0560bfd4b70277a4936c82e50e940aa253c95bf Gitweb: https://git.kernel.org/tip/b0560bfd4b70277a4936c82e50e940aa253c95bf Author: Kan Liang <[email protected]> AuthorDate: Tue, 29 Aug 2023 05:58:05 -07:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 29 Aug 2023 20:59:23 +02:00
perf/x86/intel: Clean up the hybrid CPU type handling code
There is a fairly long list of grievances about the current code. The main beefs:
1. hybrid_big_small assumes that the *HARDWARE* (CPUID) provided core types are a bitmap. They are not. If Intel happened to make a core type of 0xff, hilarity would ensue.
2. adl_get_hybrid_cpu_type() utterly inscrutable. There are precisely zero comments and zero changelog about what it is attempting to do.
According to Kan, the adl_get_hybrid_cpu_type() is there because some Alder Lake (ADL) CPUs can do some silly things. Some ADL models are *supposed* to be hybrid CPUs with big and little cores, but there are some SKUs that only have big cores. CPUID(0x1a) on those CPUs does not say that the CPUs are big cores. It apparently just returns 0x0. It confuses perf because it expects to see either 0x40 (Core) or 0x20 (Atom).
The perf workaround for this is to watch for a CPU core saying it is type 0x0. If that happens on an Alder Lake, it calls x86_pmu.get_hybrid_cpu_type() and just assumes that the core is a Core (0x40) CPU.
To fix up the mess, separate out the CPU types and the 'pmu' types. This allows 'hybrid_pmu_type' bitmaps without worrying that some future CPU type will set multiple bits.
Since the types are now separate, add a function to glue them back together again. Actual comment on the situation in the glue function (find_hybrid_pmu_for_cpu()).
Also, give ->get_hybrid_cpu_type() a real return type and make it clear that it is overriding the *CPU* type, not the PMU type.
Rename cpu_type to pmu_type in the struct x86_hybrid_pmu to reflect the change.
--- arch/x86/events/core.c | 6 +-- arch/x86/events/intel/core.c | 69 +++++++++++++++++++++++------------ arch/x86/events/intel/ds.c | 2 +- arch/x86/events/perf_event.h | 35 ++++++++++-------- 4 files changed, 72 insertions(+), 40 deletions(-)
arch/x86/events/core.c +3 -3
@@ -1887,9 +1887,9 @@ ssize_t events_hybrid_sysfs_show(struct device *dev, str = pmu_attr->event_str;for (i =0 ; i < x86_pmu.num_hybrid_pmus; i++) { -if (!(x86_pmu.hybrid_pmu[i].cpu_type & pmu_attr->pmu_type)) +if (!(x86_pmu.hybrid_pmu[i].pmu_type & pmu_attr->pmu_type))continue ; -if (x86_pmu.hybrid_pmu[i].cpu_type & pmu->cpu_type) { +if (x86_pmu.hybrid_pmu[i].pmu_type & pmu->pmu_type) { next_str =strchr (str,';' );if (next_str)return snprintf (page, next_str - str +1 ,"%s" , str);
@@ -2169,7 +2169,7 @@ static int __init init_hw_perf_events(void) hybrid_pmu->pmu.capabilities |=PERF_PMU_CAP_EXTENDED_HW_TYPE ; err =perf_pmu_register (&hybrid_pmu->pmu, hybrid_pmu->name, - (hybrid_pmu->cpu_type == hybrid_big) ?PERF_TYPE_RAW :-1 ); + (hybrid_pmu->pmu_type == hybrid_big) ?PERF_TYPE_RAW :-1 );if (err)break ; }
arch/x86/events/intel/core.c +47 -22
140 lines not on this page · raw message
arch/x86/events/intel/ds.c +1 -1
@@ -261,7 +261,7 @@ static u64 __adl_latency_data_small(struct perf_event *event, u64 status, {u64 val; -WARN_ON_ONCE (hybrid_pmu (event->pmu)->cpu_type == hybrid_big); +WARN_ON_ONCE (hybrid_pmu (event->pmu)->pmu_type == hybrid_big); dse &=PERF_PEBS_DATA_SOURCE_MASK ; val =hybrid_var (event->pmu, pebs_data_source)[dse];
arch/x86/events/perf_event.h +21 -14
@@ -652,10 +652,29 @@ enum {#define PERF_PEBS_DATA_SOURCE_MAX 0x10#define PERF_PEBS_DATA_SOURCE_MASK (PERF_PEBS_DATA_SOURCE_MAX - 1) +enum hybrid_cpu_type { +HYBRID_INTEL_NONE , +HYBRID_INTEL_ATOM =0x20 , +HYBRID_INTEL_CORE =0x40 , +}; + +enum hybrid_pmu_type { + not_hybrid, + hybrid_small =BIT (0 ), + hybrid_big =BIT (1 ), + + hybrid_big_small = hybrid_big | hybrid_small,/* only used for matching */ +}; + +#define X86_HYBRID_PMU_ATOM_IDX 0 +#define X86_HYBRID_PMU_CORE_IDX 1 + +#define X86_HYBRID_NUM_PMUS 2 +struct x86_hybrid_pmu {struct pmu pmu;const char *name; -u8 cpu_type; +enum hybrid_pmu_type pmu_type;cpumask_t supported_cpus;union perf_capabilities intel_cap;u64 intel_ctrl;
@@ -721,18 +740,6 @@ extern struct static_key_false perf_is_hybrid; __Fp; \ }) -enum hybrid_pmu_type { - hybrid_big =0x40 , - hybrid_small =0x20 , - - hybrid_big_small = hybrid_big | hybrid_small, -}; - -#define X86_HYBRID_PMU_ATOM_IDX 0 -#define X86_HYBRID_PMU_CORE_IDX 1 - -#define X86_HYBRID_NUM_PMUS 2 -/* * struct x86_pmu - generic x86 pmu */
@@ -940,7 +947,7 @@ struct x86_pmu {*/ int num_hybrid_pmus;struct x86_hybrid_pmu *hybrid_pmu; -u8 (*get_hybrid_cpu_type) (void); +enum hybrid_cpu_type (*get_hybrid_cpu_type) (void ); };struct x86_perf_task_context_opt {
PATCH 6/6 perf/x86/intel: Add common intel_pmu_init_hybrid
In reply to [email protected]From: Kan Liang <[email protected]>
The current hybrid initialization codes aren't well organized and are hard to read.
Factor out intel_pmu_init_hybrid() to do a common setup for each hybrid PMU. The PMU-specific capability will be updated later via either hard code (ADL) or CPUID hybrid enumeration (MTL).
Splitting the ADL and MTL initialization codes, since they have different uarches. The hard code PMU capabilities are not required for MTL either. They can be enumerated by the new leaf 0x23 and IA32_PERF_CAPABILITIES MSR.
The hybrid enumeration of the IA32_PERF_CAPABILITIES MSR is broken on MTL. Using the default value.
--- arch/x86/events/intel/core.c | 162 ++++++++++++++++++++++++----------- 1 file changed, 111 insertions(+), 51 deletions(-)
arch/x86/events/intel/core.c +111 -51
@@ -4591,6 +4591,16 @@ static void intel_pmu_check_num_counters(int *num_counters,int *num_counters_fixed, u64 *intel_ctrl, u64 fixed_mask); +static inline bool intel_pmu_broken_perf_cap (void ) +{ +/* The Perf Metric (Bit 15) is always cleared */ +if ((boot_cpu_data.x86_model ==INTEL_FAM6_METEORLAKE ) || + (boot_cpu_data.x86_model ==INTEL_FAM6_METEORLAKE_L )) +return true; + +return false; +} +static void update_pmu_cap (struct x86_hybrid_pmu *pmu) {unsigned int sub_bitmaps =cpuid_eax (ARCH_PERFMON_EXT_LEAF );
@@ -4603,7 +4613,27 @@ static void update_pmu_cap(struct x86_hybrid_pmu *pmu) pmu->num_counters_fixed =fls (ebx);intel_pmu_check_num_counters (&pmu->num_counters, &pmu->num_counters_fixed, &pmu->intel_ctrl, ebx); + pmu->max_pebs_events =min_t (unsigned,MAX_PEBS_EVENTS , pmu->num_counters); + pmu->unconstrained = (struct event_constraint ) +__EVENT_CONSTRAINT (0 , (1ULL << pmu->num_counters) -1 , +0 , pmu->num_counters,0 ,0 ); } + + +if (!intel_pmu_broken_perf_cap ()) { +/* Perf Metric (Bit 15) and PEBS via PT (Bit 16) are hybrid enumeration */ +rdmsrl (MSR_IA32_PERF_CAPABILITIES , pmu->intel_cap.capabilities); + } + +if (pmu->intel_cap.perf_metrics) + pmu->intel_ctrl |=1ULL <<GLOBAL_CTRL_EN_PERF_METRICS ; +else + pmu->intel_ctrl &= ~(1ULL <<GLOBAL_CTRL_EN_PERF_METRICS ); + +if (pmu->intel_cap.pebs_output_pt_available) + pmu->pmu.capabilities |=PERF_PMU_CAP_AUX_OUTPUT ; +else + pmu->pmu.capabilities |= ~PERF_PMU_CAP_AUX_OUTPUT ; }static struct x86_hybrid_pmu *find_hybrid_pmu_for_cpu (void )
@@ -5908,10 +5938,52 @@ static void intel_pmu_check_hybrid_pmus(u64 fixed_mask) } } -static __always_inline bool is_mtl(u8 x86_model) +static const struct {enum hybrid_pmu_type id;char *name; } intel_hybrid_pmu_type_map[]__initconst = { + { hybrid_small,"cpu_atom" }, + { hybrid_big,"cpu_core" }, +}; + +static __always_inline int intel_pmu_init_hybrid(enum hybrid_pmu_type pmus) { -return (x86_model ==INTEL_FAM6_METEORLAKE ) || - (x86_model ==INTEL_FAM6_METEORLAKE_L ); +unsigned long pmus_mask = pmus; +struct x86_hybrid_pmu *pmu; +int idx =0 , bit; + + x86_pmu.num_hybrid_pmus =hweight_long (pmus_mask); + x86_pmu.hybrid_pmu =kcalloc (x86_pmu.num_hybrid_pmus, +sizeof (struct x86_hybrid_pmu ), +GFP_KERNEL ); +if (!x86_pmu.hybrid_pmu) +return -ENOMEM ; + +static_branch_enable (&perf_is_hybrid); + x86_pmu.filter = intel_pmu_filter; + +for_each_set_bit (bit, &pmus_mask,ARRAY_SIZE (intel_hybrid_pmu_type_map)) { + pmu = &x86_pmu.hybrid_pmu[idx++]; + pmu->pmu_type = intel_hybrid_pmu_type_map[bit].id; + pmu->name = intel_hybrid_pmu_type_map[bit].name; + + pmu->num_counters = x86_pmu.num_counters; + pmu->num_counters_fixed = x86_pmu.num_counters_fixed; + pmu->max_pebs_events =min_t (unsigned,MAX_PEBS_EVENTS , pmu->num_counters); + pmu->unconstrained = (struct event_constraint ) +__EVENT_CONSTRAINT (0 , (1ULL << pmu->num_counters) -1 , +0 , pmu->num_counters,0 ,0 ); + + pmu->intel_cap.capabilities = x86_pmu.intel_cap.capabilities; +if (pmu->pmu_type & hybrid_small) { + pmu->intel_cap.perf_metrics =0 ; + pmu->intel_cap.pebs_output_pt_available =1 ; + pmu->mid_ack = true; + }else if (pmu->pmu_type & hybrid_big) { + pmu->intel_cap.perf_metrics =1 ; + pmu->intel_cap.pebs_output_pt_available =0 ; + pmu->late_ack = true; + } + } + +return 0 ; } static __always_inline void intel_pmu_init_glc(struct pmu *pmu)
@@ -6595,23 +6667,14 @@ __init int intel_pmu_init(void)case INTEL_FAM6_RAPTORLAKE :case INTEL_FAM6_RAPTORLAKE_P :case INTEL_FAM6_RAPTORLAKE_S : -case INTEL_FAM6_METEORLAKE : -case INTEL_FAM6_METEORLAKE_L :/* * Alder Lake has 2 types of CPU, core and atom. * * Initialize the common PerfMon capabilities here. */ - x86_pmu.hybrid_pmu =kcalloc (X86_HYBRID_NUM_PMUS , -sizeof (struct x86_hybrid_pmu ), -GFP_KERNEL ); -if (!x86_pmu.hybrid_pmu) -return -ENOMEM ; -static_branch_enable (&perf_is_hybrid); - x86_pmu.num_hybrid_pmus =X86_HYBRID_NUM_PMUS ; +intel_pmu_init_hybrid (hybrid_big_small); x86_pmu.pebs_latency_data = adl_latency_data_small; - x86_pmu.filter = intel_pmu_filter; x86_pmu.get_event_constraints = adl_get_event_constraints; x86_pmu.hw_config = adl_hw_config; x86_pmu.get_hybrid_cpu_type = adl_get_hybrid_cpu_type;
@@ -6624,10 +6687,7 @@ __init int intel_pmu_init(void)/* Initialize big core specific PerfMon capabilities.*/ pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ]; - pmu->name ="cpu_core" ; - pmu->pmu_type = hybrid_big;intel_pmu_init_glc (&pmu->pmu); - pmu->late_ack = true;if (cpu_feature_enabled (X86_FEATURE_HYBRID_CPU )) { pmu->num_counters = x86_pmu.num_counters +2 ; pmu->num_counters_fixed = x86_pmu.num_counters_fixed +1 ;
@@ -6652,45 +6712,45 @@ __init int intel_pmu_init(void) pmu->unconstrained = (struct event_constraint )__EVENT_CONSTRAINT (0 , (1ULL << pmu->num_counters) -1 ,0 , pmu->num_counters,0 ,0 ); - pmu->intel_cap.capabilities = x86_pmu.intel_cap.capabilities; - pmu->intel_cap.perf_metrics =1 ; - pmu->intel_cap.pebs_output_pt_available =0 ; - pmu->extra_regs = intel_glc_extra_regs;/* Initialize Atom core specific PerfMon capabilities.*/ pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_ATOM_IDX ]; - pmu->name ="cpu_atom" ; - pmu->pmu_type = hybrid_small;intel_pmu_init_grt (&pmu->pmu); - pmu->mid_ack = true; - pmu->num_counters = x86_pmu.num_counters; - pmu->num_counters_fixed = x86_pmu.num_counters_fixed; - pmu->max_pebs_events = x86_pmu.max_pebs_events; - pmu->unconstrained = (struct event_constraint ) -__EVENT_CONSTRAINT (0 , (1ULL << pmu->num_counters) -1 , -0 , pmu->num_counters,0 ,0 ); - pmu->intel_cap.capabilities = x86_pmu.intel_cap.capabilities; - pmu->intel_cap.perf_metrics =0 ; - pmu->intel_cap.pebs_output_pt_available =1 ; - -if (is_mtl (boot_cpu_data.x86_model)) { - x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ].extra_regs = intel_rwc_extra_regs; - x86_pmu.pebs_latency_data = mtl_latency_data_small; - extra_attr =boot_cpu_has (X86_FEATURE_RTM ) ? - mtl_hybrid_extra_attr_rtm : mtl_hybrid_extra_attr; - mem_attr = mtl_hybrid_mem_attrs; -intel_pmu_pebs_data_source_mtl (); - x86_pmu.get_event_constraints = mtl_get_event_constraints; - pmu->extra_regs = intel_cmt_extra_regs; -pr_cont ("Meteorlake Hybrid events, " ); - name ="meteorlake_hybrid" ; - }else { - x86_pmu.flags |=PMU_FL_MEM_LOADS_AUX ; -intel_pmu_pebs_data_source_adl (); -pr_cont ("Alderlake Hybrid events, " ); - name ="alderlake_hybrid" ; - } + + x86_pmu.flags |=PMU_FL_MEM_LOADS_AUX ; +intel_pmu_pebs_data_source_adl (); +pr_cont ("Alderlake Hybrid events, " ); + name ="alderlake_hybrid" ; +break ; + +case INTEL_FAM6_METEORLAKE : +case INTEL_FAM6_METEORLAKE_L : +intel_pmu_init_hybrid (hybrid_big_small); + + x86_pmu.pebs_latency_data = mtl_latency_data_small; + x86_pmu.get_event_constraints = mtl_get_event_constraints; + x86_pmu.hw_config = adl_hw_config; + + td_attr = adl_hybrid_events_attrs; + mem_attr = mtl_hybrid_mem_attrs; + tsx_attr = adl_hybrid_tsx_attrs; + extra_attr =boot_cpu_has (X86_FEATURE_RTM ) ? + mtl_hybrid_extra_attr_rtm : mtl_hybrid_extra_attr; + +/* Initialize big core specific PerfMon capabilities.*/ + pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ]; +intel_pmu_init_glc (&pmu->pmu); + pmu->extra_regs = intel_rwc_extra_regs; + +/* Initialize Atom core specific PerfMon capabilities.*/ + pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_ATOM_IDX ]; +intel_pmu_init_grt (&pmu->pmu); + pmu->extra_regs = intel_cmt_extra_regs; + +intel_pmu_pebs_data_source_mtl (); +pr_cont ("Meteorlake Hybrid events, " ); + name ="meteorlake_hybrid" ;break ;default :
@@ -6802,7 +6862,7 @@ __init int intel_pmu_init(void)if (!is_hybrid () && x86_pmu.intel_cap.perf_metrics) x86_pmu.intel_ctrl |=1ULL <<GLOBAL_CTRL_EN_PERF_METRICS ; -if (is_hybrid ()) +if (is_hybrid () && !boot_cpu_has (X86_FEATURE_ARCH_PERFMON_EXT ))intel_pmu_check_hybrid_pmus ((u64 )fixed_mask);if (x86_pmu.intel_cap.pebs_timing_info)
Signature · 1 line
2.35.1
[tip: perf/core] perf/x86/intel: Add common intel_pmu_init_hybrid()
The following commit has been merged into the perf/core branch of tip:
The following commit has been merged into the perf/core branch of tip:
Commit-ID: 97588df87b56e27fd2b5d928d61c7a53e38afbb0 Gitweb: https://git.kernel.org/tip/97588df87b56e27fd2b5d928d61c7a53e38afbb0 Author: Kan Liang <[email protected]> AuthorDate: Tue, 29 Aug 2023 05:58:06 -07:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 29 Aug 2023 20:59:23 +02:00
perf/x86/intel: Add common intel_pmu_init_hybrid()
The current hybrid initialization codes aren't well organized and are hard to read.
Factor out intel_pmu_init_hybrid() to do a common setup for each hybrid PMU. The PMU-specific capability will be updated later via either hard code (ADL) or CPUID hybrid enumeration (MTL).
Splitting the ADL and MTL initialization codes, since they have different uarches. The hard code PMU capabilities are not required for MTL either. They can be enumerated by the new leaf 0x23 and IA32_PERF_CAPABILITIES MSR.
The hybrid enumeration of the IA32_PERF_CAPABILITIES MSR is broken on MTL. Using the default value.
--- arch/x86/events/intel/core.c | 162 +++++++++++++++++++++++----------- 1 file changed, 111 insertions(+), 51 deletions(-)
arch/x86/events/intel/core.c +111 -51
@@ -4598,6 +4598,16 @@ static void intel_pmu_check_num_counters(int *num_counters,int *num_counters_fixed, u64 *intel_ctrl, u64 fixed_mask); +static inline bool intel_pmu_broken_perf_cap (void ) +{ +/* The Perf Metric (Bit 15) is always cleared */ +if ((boot_cpu_data.x86_model ==INTEL_FAM6_METEORLAKE ) || + (boot_cpu_data.x86_model ==INTEL_FAM6_METEORLAKE_L )) +return true; + +return false; +} +static void update_pmu_cap (struct x86_hybrid_pmu *pmu) {unsigned int sub_bitmaps =cpuid_eax (ARCH_PERFMON_EXT_LEAF );
@@ -4610,7 +4620,27 @@ static void update_pmu_cap(struct x86_hybrid_pmu *pmu) pmu->num_counters_fixed =fls (ebx);intel_pmu_check_num_counters (&pmu->num_counters, &pmu->num_counters_fixed, &pmu->intel_ctrl, ebx); + pmu->max_pebs_events =min_t (unsigned,MAX_PEBS_EVENTS , pmu->num_counters); + pmu->unconstrained = (struct event_constraint ) +__EVENT_CONSTRAINT (0 , (1ULL << pmu->num_counters) -1 , +0 , pmu->num_counters,0 ,0 ); } + + +if (!intel_pmu_broken_perf_cap ()) { +/* Perf Metric (Bit 15) and PEBS via PT (Bit 16) are hybrid enumeration */ +rdmsrl (MSR_IA32_PERF_CAPABILITIES , pmu->intel_cap.capabilities); + } + +if (pmu->intel_cap.perf_metrics) + pmu->intel_ctrl |=1ULL <<GLOBAL_CTRL_EN_PERF_METRICS ; +else + pmu->intel_ctrl &= ~(1ULL <<GLOBAL_CTRL_EN_PERF_METRICS ); + +if (pmu->intel_cap.pebs_output_pt_available) + pmu->pmu.capabilities |=PERF_PMU_CAP_AUX_OUTPUT ; +else + pmu->pmu.capabilities |= ~PERF_PMU_CAP_AUX_OUTPUT ; }static struct x86_hybrid_pmu *find_hybrid_pmu_for_cpu (void )
@@ -5915,10 +5945,52 @@ static void intel_pmu_check_hybrid_pmus(u64 fixed_mask) } } -static __always_inline bool is_mtl(u8 x86_model) +static const struct {enum hybrid_pmu_type id;char *name; } intel_hybrid_pmu_type_map[]__initconst = { + { hybrid_small,"cpu_atom" }, + { hybrid_big,"cpu_core" }, +}; + +static __always_inline int intel_pmu_init_hybrid(enum hybrid_pmu_type pmus) { -return (x86_model ==INTEL_FAM6_METEORLAKE ) || - (x86_model ==INTEL_FAM6_METEORLAKE_L ); +unsigned long pmus_mask = pmus; +struct x86_hybrid_pmu *pmu; +int idx =0 , bit; + + x86_pmu.num_hybrid_pmus =hweight_long (pmus_mask); + x86_pmu.hybrid_pmu =kcalloc (x86_pmu.num_hybrid_pmus, +sizeof (struct x86_hybrid_pmu ), +GFP_KERNEL ); +if (!x86_pmu.hybrid_pmu) +return -ENOMEM ; + +static_branch_enable (&perf_is_hybrid); + x86_pmu.filter = intel_pmu_filter; + +for_each_set_bit (bit, &pmus_mask,ARRAY_SIZE (intel_hybrid_pmu_type_map)) { + pmu = &x86_pmu.hybrid_pmu[idx++]; + pmu->pmu_type = intel_hybrid_pmu_type_map[bit].id; + pmu->name = intel_hybrid_pmu_type_map[bit].name; + + pmu->num_counters = x86_pmu.num_counters; + pmu->num_counters_fixed = x86_pmu.num_counters_fixed; + pmu->max_pebs_events =min_t (unsigned,MAX_PEBS_EVENTS , pmu->num_counters); + pmu->unconstrained = (struct event_constraint ) +__EVENT_CONSTRAINT (0 , (1ULL << pmu->num_counters) -1 , +0 , pmu->num_counters,0 ,0 ); + + pmu->intel_cap.capabilities = x86_pmu.intel_cap.capabilities; +if (pmu->pmu_type & hybrid_small) { + pmu->intel_cap.perf_metrics =0 ; + pmu->intel_cap.pebs_output_pt_available =1 ; + pmu->mid_ack = true; + }else if (pmu->pmu_type & hybrid_big) { + pmu->intel_cap.perf_metrics =1 ; + pmu->intel_cap.pebs_output_pt_available =0 ; + pmu->late_ack = true; + } + } + +return 0 ; } static __always_inline void intel_pmu_init_glc(struct pmu *pmu)
@@ -6602,23 +6674,14 @@ __init int intel_pmu_init(void)case INTEL_FAM6_RAPTORLAKE :case INTEL_FAM6_RAPTORLAKE_P :case INTEL_FAM6_RAPTORLAKE_S : -case INTEL_FAM6_METEORLAKE : -case INTEL_FAM6_METEORLAKE_L :/* * Alder Lake has 2 types of CPU, core and atom. * * Initialize the common PerfMon capabilities here. */ - x86_pmu.hybrid_pmu =kcalloc (X86_HYBRID_NUM_PMUS , -sizeof (struct x86_hybrid_pmu ), -GFP_KERNEL ); -if (!x86_pmu.hybrid_pmu) -return -ENOMEM ; -static_branch_enable (&perf_is_hybrid); - x86_pmu.num_hybrid_pmus =X86_HYBRID_NUM_PMUS ; +intel_pmu_init_hybrid (hybrid_big_small); x86_pmu.pebs_latency_data = adl_latency_data_small; - x86_pmu.filter = intel_pmu_filter; x86_pmu.get_event_constraints = adl_get_event_constraints; x86_pmu.hw_config = adl_hw_config; x86_pmu.get_hybrid_cpu_type = adl_get_hybrid_cpu_type;
@@ -6631,10 +6694,7 @@ __init int intel_pmu_init(void)/* Initialize big core specific PerfMon capabilities.*/ pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ]; - pmu->name ="cpu_core" ; - pmu->pmu_type = hybrid_big;intel_pmu_init_glc (&pmu->pmu); - pmu->late_ack = true;if (cpu_feature_enabled (X86_FEATURE_HYBRID_CPU )) { pmu->num_counters = x86_pmu.num_counters +2 ; pmu->num_counters_fixed = x86_pmu.num_counters_fixed +1 ;
@@ -6659,45 +6719,45 @@ __init int intel_pmu_init(void) pmu->unconstrained = (struct event_constraint )__EVENT_CONSTRAINT (0 , (1ULL << pmu->num_counters) -1 ,0 , pmu->num_counters,0 ,0 ); - pmu->intel_cap.capabilities = x86_pmu.intel_cap.capabilities; - pmu->intel_cap.perf_metrics =1 ; - pmu->intel_cap.pebs_output_pt_available =0 ; - pmu->extra_regs = intel_glc_extra_regs;/* Initialize Atom core specific PerfMon capabilities.*/ pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_ATOM_IDX ]; - pmu->name ="cpu_atom" ; - pmu->pmu_type = hybrid_small;intel_pmu_init_grt (&pmu->pmu); - pmu->mid_ack = true; - pmu->num_counters = x86_pmu.num_counters; - pmu->num_counters_fixed = x86_pmu.num_counters_fixed; - pmu->max_pebs_events = x86_pmu.max_pebs_events; - pmu->unconstrained = (struct event_constraint ) -__EVENT_CONSTRAINT (0 , (1ULL << pmu->num_counters) -1 , -0 , pmu->num_counters,0 ,0 ); - pmu->intel_cap.capabilities = x86_pmu.intel_cap.capabilities; - pmu->intel_cap.perf_metrics =0 ; - pmu->intel_cap.pebs_output_pt_available =1 ; - -if (is_mtl (boot_cpu_data.x86_model)) { - x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ].extra_regs = intel_rwc_extra_regs; - x86_pmu.pebs_latency_data = mtl_latency_data_small; - extra_attr =boot_cpu_has (X86_FEATURE_RTM ) ? - mtl_hybrid_extra_attr_rtm : mtl_hybrid_extra_attr; - mem_attr = mtl_hybrid_mem_attrs; -intel_pmu_pebs_data_source_mtl (); - x86_pmu.get_event_constraints = mtl_get_event_constraints; - pmu->extra_regs = intel_cmt_extra_regs; -pr_cont ("Meteorlake Hybrid events, " ); - name ="meteorlake_hybrid" ; - }else { - x86_pmu.flags |=PMU_FL_MEM_LOADS_AUX ; -intel_pmu_pebs_data_source_adl (); -pr_cont ("Alderlake Hybrid events, " ); - name ="alderlake_hybrid" ; - } + + x86_pmu.flags |=PMU_FL_MEM_LOADS_AUX ; +intel_pmu_pebs_data_source_adl (); +pr_cont ("Alderlake Hybrid events, " ); + name ="alderlake_hybrid" ; +break ; + +case INTEL_FAM6_METEORLAKE : +case INTEL_FAM6_METEORLAKE_L : +intel_pmu_init_hybrid (hybrid_big_small); + + x86_pmu.pebs_latency_data = mtl_latency_data_small; + x86_pmu.get_event_constraints = mtl_get_event_constraints; + x86_pmu.hw_config = adl_hw_config; + + td_attr = adl_hybrid_events_attrs; + mem_attr = mtl_hybrid_mem_attrs; + tsx_attr = adl_hybrid_tsx_attrs; + extra_attr =boot_cpu_has (X86_FEATURE_RTM ) ? + mtl_hybrid_extra_attr_rtm : mtl_hybrid_extra_attr; + +/* Initialize big core specific PerfMon capabilities.*/ + pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_CORE_IDX ]; +intel_pmu_init_glc (&pmu->pmu); + pmu->extra_regs = intel_rwc_extra_regs; + +/* Initialize Atom core specific PerfMon capabilities.*/ + pmu = &x86_pmu.hybrid_pmu[X86_HYBRID_PMU_ATOM_IDX ]; +intel_pmu_init_grt (&pmu->pmu); + pmu->extra_regs = intel_cmt_extra_regs; + +intel_pmu_pebs_data_source_mtl (); +pr_cont ("Meteorlake Hybrid events, " ); + name ="meteorlake_hybrid" ;break ;default :
@@ -6809,7 +6869,7 @@ __init int intel_pmu_init(void)if (!is_hybrid () && x86_pmu.intel_cap.perf_metrics) x86_pmu.intel_ctrl |=1ULL <<GLOBAL_CTRL_EN_PERF_METRICS ; -if (is_hybrid ()) +if (is_hybrid () && !boot_cpu_has (X86_FEATURE_ARCH_PERFMON_EXT ))intel_pmu_check_hybrid_pmus ((u64 )fixed_mask);if (x86_pmu.intel_cap.pebs_timing_info)
Ingo Molnar
AppliedQuoted 13 lines · * [email protected] <[email protected]> wrote:
From: Kan Liang <[email protected]>
The patch set is to clean up the Intel PMU initialization code. There is no functional change.
Kan Liang (6):
perf/x86/intel: Use the common uarch name for the shared functions perf/x86/intel: Factor out the initialization code for SPR perf/x86/intel: Factor out the initialization code for ADL e-core perf/x86/intel: Apply the common initialization code for ADL perf/x86/intel: Cleanup the hybrid CPU type perf/x86/intel: Add commonintel_pmu_init_hybrid
Quoted 5 lines
arch/x86/events/core.c | 6 +- arch/x86/events/intel/core.c | 397 ++++++++++++++++++----------------- arch/x86/events/intel/ds.c | 4 +- arch/x86/events/perf_event.h | 37 ++-- 4 files changed, 226 insertions(+), 218 deletions(-)
Applied, thanks.
Ingo