linux/arch/x86/include/asm/perf_event_p4.h

722 lines
20 KiB
C
Raw Normal View History

perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
/*
* Netburst Perfomance Events (P4, old Xeon)
*/
#ifndef PERF_EVENT_P4_H
#define PERF_EVENT_P4_H
#include <linux/cpu.h>
#include <linux/bitops.h>
/*
* NetBurst has perfomance MSRs shared between
* threads if HT is turned on, ie for both logical
* processors (mem: in turn in Atom with HT support
* perf-MSRs are not shared and every thread has its
* own perf-MSRs set)
*/
#define ARCH_P4_TOTAL_ESCR (46)
#define ARCH_P4_RESERVED_ESCR (2) /* IQ_ESCR(0,1) not always present */
#define ARCH_P4_MAX_ESCR (ARCH_P4_TOTAL_ESCR - ARCH_P4_RESERVED_ESCR)
#define ARCH_P4_MAX_CCCR (18)
#define ARCH_P4_MAX_COUNTER (ARCH_P4_MAX_CCCR / 2)
#define P4_EVNTSEL_EVENT_MASK 0x7e000000U
#define P4_EVNTSEL_EVENT_SHIFT 25
#define P4_EVNTSEL_EVENTMASK_MASK 0x01fffe00U
#define P4_EVNTSEL_EVENTMASK_SHIFT 9
#define P4_EVNTSEL_TAG_MASK 0x000001e0U
#define P4_EVNTSEL_TAG_SHIFT 5
#define P4_EVNTSEL_TAG_ENABLE 0x00000010U
#define P4_EVNTSEL_T0_OS 0x00000008U
#define P4_EVNTSEL_T0_USR 0x00000004U
#define P4_EVNTSEL_T1_OS 0x00000002U
#define P4_EVNTSEL_T1_USR 0x00000001U
/* Non HT mask */
#define P4_EVNTSEL_MASK \
(P4_EVNTSEL_EVENT_MASK | \
P4_EVNTSEL_EVENTMASK_MASK | \
P4_EVNTSEL_TAG_MASK | \
P4_EVNTSEL_TAG_ENABLE | \
P4_EVNTSEL_T0_OS | \
P4_EVNTSEL_T0_USR)
/* HT mask */
#define P4_EVNTSEL_MASK_HT \
(P4_EVNTSEL_MASK | \
P4_EVNTSEL_T1_OS | \
P4_EVNTSEL_T1_USR)
#define P4_CCCR_OVF 0x80000000U
#define P4_CCCR_CASCADE 0x40000000U
#define P4_CCCR_OVF_PMI_T0 0x04000000U
#define P4_CCCR_OVF_PMI_T1 0x08000000U
#define P4_CCCR_FORCE_OVF 0x02000000U
#define P4_CCCR_EDGE 0x01000000U
#define P4_CCCR_THRESHOLD_MASK 0x00f00000U
#define P4_CCCR_THRESHOLD_SHIFT 20
#define P4_CCCR_THRESHOLD(v) ((v) << P4_CCCR_THRESHOLD_SHIFT)
#define P4_CCCR_COMPLEMENT 0x00080000U
#define P4_CCCR_COMPARE 0x00040000U
#define P4_CCCR_ESCR_SELECT_MASK 0x0000e000U
#define P4_CCCR_ESCR_SELECT_SHIFT 13
#define P4_CCCR_ENABLE 0x00001000U
#define P4_CCCR_THREAD_SINGLE 0x00010000U
#define P4_CCCR_THREAD_BOTH 0x00020000U
#define P4_CCCR_THREAD_ANY 0x00030000U
#define P4_CCCR_RESERVED 0x00000fffU
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
/* Non HT mask */
#define P4_CCCR_MASK \
(P4_CCCR_OVF | \
P4_CCCR_CASCADE | \
P4_CCCR_OVF_PMI_T0 | \
P4_CCCR_FORCE_OVF | \
P4_CCCR_EDGE | \
P4_CCCR_THRESHOLD_MASK | \
P4_CCCR_COMPLEMENT | \
P4_CCCR_COMPARE | \
P4_CCCR_ESCR_SELECT_MASK | \
P4_CCCR_ENABLE)
/* HT mask */
#define P4_CCCR_MASK_HT \
(P4_CCCR_MASK | \
P4_CCCR_THREAD_ANY)
/*
* format is 32 bit: ee ss aa aa
* where
* ee - 8 bit event
* ss - 8 bit selector
* aa aa - 16 bits reserved for tags/attributes
*/
#define P4_EVENT_PACK(event, selector) (((event) << 24) | ((selector) << 16))
#define P4_EVENT_UNPACK_EVENT(packed) (((packed) >> 24) & 0xff)
#define P4_EVENT_UNPACK_SELECTOR(packed) (((packed) >> 16) & 0xff)
#define P4_EVENT_PACK_ATTR(attr) ((attr))
#define P4_EVENT_UNPACK_ATTR(packed) ((packed) & 0xffff)
#define P4_MAKE_EVENT_ATTR(class, name, bit) class##_##name = (1 << bit)
#define P4_EVENT_ATTR(class, name) class##_##name
#define P4_EVENT_ATTR_STR(class, name) __stringify(class##_##name)
/*
* config field is 64bit width and consists of
* HT << 63 | ESCR << 32 | CCCR
* where HT is HyperThreading bit (since ESCR
* has it reserved we may use it for own purpose)
*
* note that this is NOT the addresses of respective
* ESCR and CCCR but rather an only packed value should
* be unpacked and written to a proper addresses
*
* the base idea is to pack as much info as
* possible
*/
#define p4_config_pack_escr(v) (((u64)(v)) << 32)
#define p4_config_pack_cccr(v) (((u64)(v)) & 0xffffffffULL)
#define p4_config_unpack_escr(v) (((u64)(v)) >> 32)
#define p4_config_unpack_cccr(v) (((u64)(v)) & 0xfffff000ULL)
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
#define p4_config_unpack_emask(v) \
({ \
u32 t = p4_config_unpack_escr((v)); \
t &= P4_EVNTSEL_EVENTMASK_MASK; \
t >>= P4_EVNTSEL_EVENTMASK_SHIFT; \
t; \
})
#define p4_config_unpack_key(v) (((u64)(v)) & P4_CCCR_RESERVED)
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
#define P4_CONFIG_HT_SHIFT 63
#define P4_CONFIG_HT (1ULL << P4_CONFIG_HT_SHIFT)
static inline u32 p4_config_unpack_opcode(u64 config)
{
u32 e, s;
/*
* we don't care about HT presence here since
* event opcode doesn't depend on it
*/
e = (p4_config_unpack_escr(config) & P4_EVNTSEL_EVENT_MASK) >> P4_EVNTSEL_EVENT_SHIFT;
s = (p4_config_unpack_cccr(config) & P4_CCCR_ESCR_SELECT_MASK) >> P4_CCCR_ESCR_SELECT_SHIFT;
return P4_EVENT_PACK(e, s);
}
static inline bool p4_is_event_cascaded(u64 config)
{
u32 cccr = p4_config_unpack_cccr(config);
return !!(cccr & P4_CCCR_CASCADE);
}
static inline int p4_ht_config_thread(u64 config)
{
return !!(config & P4_CONFIG_HT);
}
static inline u64 p4_set_ht_bit(u64 config)
{
return config | P4_CONFIG_HT;
}
static inline u64 p4_clear_ht_bit(u64 config)
{
return config & ~P4_CONFIG_HT;
}
static inline int p4_ht_active(void)
{
#ifdef CONFIG_SMP
return smp_num_siblings > 1;
#endif
return 0;
}
static inline int p4_ht_thread(int cpu)
{
#ifdef CONFIG_SMP
if (smp_num_siblings == 2)
return cpu != cpumask_first(__get_cpu_var(cpu_sibling_map));
#endif
return 0;
}
static inline int p4_should_swap_ts(u64 config, int cpu)
{
return p4_ht_config_thread(config) ^ p4_ht_thread(cpu);
}
static inline u32 p4_default_cccr_conf(int cpu)
{
/*
* Note that P4_CCCR_THREAD_ANY is "required" on
* non-HT machines (on HT machines we count TS events
* regardless the state of second logical processor
*/
u32 cccr = P4_CCCR_THREAD_ANY;
if (!p4_ht_thread(cpu))
cccr |= P4_CCCR_OVF_PMI_T0;
else
cccr |= P4_CCCR_OVF_PMI_T1;
return cccr;
}
static inline u32 p4_default_escr_conf(int cpu, int exclude_os, int exclude_usr)
{
u32 escr = 0;
if (!p4_ht_thread(cpu)) {
if (!exclude_os)
escr |= P4_EVNTSEL_T0_OS;
if (!exclude_usr)
escr |= P4_EVNTSEL_T0_USR;
} else {
if (!exclude_os)
escr |= P4_EVNTSEL_T1_OS;
if (!exclude_usr)
escr |= P4_EVNTSEL_T1_USR;
}
return escr;
}
/*
* Comments below the event represent ESCR restriction
* for this event and counter index per ESCR
*
* MSR_P4_IQ_ESCR0 and MSR_P4_IQ_ESCR1 are available only on early
* processor builds (family 0FH, models 01H-02H). These MSRs
* are not available on later versions, so that we don't use
* them completely
*
* Also note that CCCR1 do not have P4_CCCR_ENABLE bit properly
* working so that we should not use this CCCR and respective
* counter as result
*/
#define P4_TC_DELIVER_MODE P4_EVENT_PACK(0x01, 0x01)
/*
* MSR_P4_TC_ESCR0: 4, 5
* MSR_P4_TC_ESCR1: 6, 7
*/
#define P4_BPU_FETCH_REQUEST P4_EVENT_PACK(0x03, 0x00)
/*
* MSR_P4_BPU_ESCR0: 0, 1
* MSR_P4_BPU_ESCR1: 2, 3
*/
#define P4_ITLB_REFERENCE P4_EVENT_PACK(0x18, 0x03)
/*
* MSR_P4_ITLB_ESCR0: 0, 1
* MSR_P4_ITLB_ESCR1: 2, 3
*/
#define P4_MEMORY_CANCEL P4_EVENT_PACK(0x02, 0x05)
/*
* MSR_P4_DAC_ESCR0: 8, 9
* MSR_P4_DAC_ESCR1: 10, 11
*/
#define P4_MEMORY_COMPLETE P4_EVENT_PACK(0x08, 0x02)
/*
* MSR_P4_SAAT_ESCR0: 8, 9
* MSR_P4_SAAT_ESCR1: 10, 11
*/
#define P4_LOAD_PORT_REPLAY P4_EVENT_PACK(0x04, 0x02)
/*
* MSR_P4_SAAT_ESCR0: 8, 9
* MSR_P4_SAAT_ESCR1: 10, 11
*/
#define P4_STORE_PORT_REPLAY P4_EVENT_PACK(0x05, 0x02)
/*
* MSR_P4_SAAT_ESCR0: 8, 9
* MSR_P4_SAAT_ESCR1: 10, 11
*/
#define P4_MOB_LOAD_REPLAY P4_EVENT_PACK(0x03, 0x02)
/*
* MSR_P4_MOB_ESCR0: 0, 1
* MSR_P4_MOB_ESCR1: 2, 3
*/
#define P4_PAGE_WALK_TYPE P4_EVENT_PACK(0x01, 0x04)
/*
* MSR_P4_PMH_ESCR0: 0, 1
* MSR_P4_PMH_ESCR1: 2, 3
*/
#define P4_BSQ_CACHE_REFERENCE P4_EVENT_PACK(0x0c, 0x07)
/*
* MSR_P4_BSU_ESCR0: 0, 1
* MSR_P4_BSU_ESCR1: 2, 3
*/
#define P4_IOQ_ALLOCATION P4_EVENT_PACK(0x03, 0x06)
/*
* MSR_P4_FSB_ESCR0: 0, 1
* MSR_P4_FSB_ESCR1: 2, 3
*/
#define P4_IOQ_ACTIVE_ENTRIES P4_EVENT_PACK(0x1a, 0x06)
/*
* MSR_P4_FSB_ESCR1: 2, 3
*/
#define P4_FSB_DATA_ACTIVITY P4_EVENT_PACK(0x17, 0x06)
/*
* MSR_P4_FSB_ESCR0: 0, 1
* MSR_P4_FSB_ESCR1: 2, 3
*/
#define P4_BSQ_ALLOCATION P4_EVENT_PACK(0x05, 0x07)
/*
* MSR_P4_BSU_ESCR0: 0, 1
*/
#define P4_BSQ_ACTIVE_ENTRIES P4_EVENT_PACK(0x06, 0x07)
/*
* NOTE: no ESCR name in docs, it's guessed
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
* MSR_P4_BSU_ESCR1: 2, 3
*/
#define P4_SSE_INPUT_ASSIST P4_EVENT_PACK(0x34, 0x01)
/*
* MSR_P4_FIRM_ESCR0: 8, 9
* MSR_P4_FIRM_ESCR1: 10, 11
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
*/
#define P4_PACKED_SP_UOP P4_EVENT_PACK(0x08, 0x01)
/*
* MSR_P4_FIRM_ESCR0: 8, 9
* MSR_P4_FIRM_ESCR1: 10, 11
*/
#define P4_PACKED_DP_UOP P4_EVENT_PACK(0x0c, 0x01)
/*
* MSR_P4_FIRM_ESCR0: 8, 9
* MSR_P4_FIRM_ESCR1: 10, 11
*/
#define P4_SCALAR_SP_UOP P4_EVENT_PACK(0x0a, 0x01)
/*
* MSR_P4_FIRM_ESCR0: 8, 9
* MSR_P4_FIRM_ESCR1: 10, 11
*/
#define P4_SCALAR_DP_UOP P4_EVENT_PACK(0x0e, 0x01)
/*
* MSR_P4_FIRM_ESCR0: 8, 9
* MSR_P4_FIRM_ESCR1: 10, 11
*/
#define P4_64BIT_MMX_UOP P4_EVENT_PACK(0x02, 0x01)
/*
* MSR_P4_FIRM_ESCR0: 8, 9
* MSR_P4_FIRM_ESCR1: 10, 11
*/
#define P4_128BIT_MMX_UOP P4_EVENT_PACK(0x1a, 0x01)
/*
* MSR_P4_FIRM_ESCR0: 8, 9
* MSR_P4_FIRM_ESCR1: 10, 11
*/
#define P4_X87_FP_UOP P4_EVENT_PACK(0x04, 0x01)
/*
* MSR_P4_FIRM_ESCR0: 8, 9
* MSR_P4_FIRM_ESCR1: 10, 11
*/
#define P4_TC_MISC P4_EVENT_PACK(0x06, 0x01)
/*
* MSR_P4_TC_ESCR0: 4, 5
* MSR_P4_TC_ESCR1: 6, 7
*/
#define P4_GLOBAL_POWER_EVENTS P4_EVENT_PACK(0x13, 0x06)
/*
* MSR_P4_FSB_ESCR0: 0, 1
* MSR_P4_FSB_ESCR1: 2, 3
*/
#define P4_TC_MS_XFER P4_EVENT_PACK(0x05, 0x00)
/*
* MSR_P4_MS_ESCR0: 4, 5
* MSR_P4_MS_ESCR1: 6, 7
*/
#define P4_UOP_QUEUE_WRITES P4_EVENT_PACK(0x09, 0x00)
/*
* MSR_P4_MS_ESCR0: 4, 5
* MSR_P4_MS_ESCR1: 6, 7
*/
#define P4_RETIRED_MISPRED_BRANCH_TYPE P4_EVENT_PACK(0x05, 0x02)
/*
* MSR_P4_TBPU_ESCR0: 4, 5
* MSR_P4_TBPU_ESCR1: 6, 7
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
*/
#define P4_RETIRED_BRANCH_TYPE P4_EVENT_PACK(0x04, 0x02)
/*
* MSR_P4_TBPU_ESCR0: 4, 5
* MSR_P4_TBPU_ESCR1: 6, 7
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
*/
#define P4_RESOURCE_STALL P4_EVENT_PACK(0x01, 0x01)
/*
* MSR_P4_ALF_ESCR0: 12, 13, 16
* MSR_P4_ALF_ESCR1: 14, 15, 17
*/
#define P4_WC_BUFFER P4_EVENT_PACK(0x05, 0x05)
/*
* MSR_P4_DAC_ESCR0: 8, 9
* MSR_P4_DAC_ESCR1: 10, 11
*/
#define P4_B2B_CYCLES P4_EVENT_PACK(0x16, 0x03)
/*
* MSR_P4_FSB_ESCR0: 0, 1
* MSR_P4_FSB_ESCR1: 2, 3
*/
#define P4_BNR P4_EVENT_PACK(0x08, 0x03)
/*
* MSR_P4_FSB_ESCR0: 0, 1
* MSR_P4_FSB_ESCR1: 2, 3
*/
#define P4_SNOOP P4_EVENT_PACK(0x06, 0x03)
/*
* MSR_P4_FSB_ESCR0: 0, 1
* MSR_P4_FSB_ESCR1: 2, 3
*/
#define P4_RESPONSE P4_EVENT_PACK(0x04, 0x03)
/*
* MSR_P4_FSB_ESCR0: 0, 1
* MSR_P4_FSB_ESCR1: 2, 3
*/
#define P4_FRONT_END_EVENT P4_EVENT_PACK(0x08, 0x05)
/*
* MSR_P4_CRU_ESCR2: 12, 13, 16
* MSR_P4_CRU_ESCR3: 14, 15, 17
*/
#define P4_EXECUTION_EVENT P4_EVENT_PACK(0x0c, 0x05)
/*
* MSR_P4_CRU_ESCR2: 12, 13, 16
* MSR_P4_CRU_ESCR3: 14, 15, 17
*/
#define P4_REPLAY_EVENT P4_EVENT_PACK(0x09, 0x05)
/*
* MSR_P4_CRU_ESCR2: 12, 13, 16
* MSR_P4_CRU_ESCR3: 14, 15, 17
*/
#define P4_INSTR_RETIRED P4_EVENT_PACK(0x02, 0x04)
/*
* MSR_P4_CRU_ESCR0: 12, 13, 16
* MSR_P4_CRU_ESCR1: 14, 15, 17
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
*/
#define P4_UOPS_RETIRED P4_EVENT_PACK(0x01, 0x04)
/*
* MSR_P4_CRU_ESCR0: 12, 13, 16
* MSR_P4_CRU_ESCR1: 14, 15, 17
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
*/
#define P4_UOP_TYPE P4_EVENT_PACK(0x02, 0x02)
/*
* MSR_P4_RAT_ESCR0: 12, 13, 16
* MSR_P4_RAT_ESCR1: 14, 15, 17
*/
#define P4_BRANCH_RETIRED P4_EVENT_PACK(0x06, 0x05)
/*
* MSR_P4_CRU_ESCR2: 12, 13, 16
* MSR_P4_CRU_ESCR3: 14, 15, 17
*/
#define P4_MISPRED_BRANCH_RETIRED P4_EVENT_PACK(0x03, 0x04)
/*
* MSR_P4_CRU_ESCR0: 12, 13, 16
* MSR_P4_CRU_ESCR1: 14, 15, 17
*/
#define P4_X87_ASSIST P4_EVENT_PACK(0x03, 0x05)
/*
* MSR_P4_CRU_ESCR2: 12, 13, 16
* MSR_P4_CRU_ESCR3: 14, 15, 17
*/
#define P4_MACHINE_CLEAR P4_EVENT_PACK(0x02, 0x05)
/*
* MSR_P4_CRU_ESCR2: 12, 13, 16
* MSR_P4_CRU_ESCR3: 14, 15, 17
*/
#define P4_INSTR_COMPLETED P4_EVENT_PACK(0x07, 0x04)
/*
* MSR_P4_CRU_ESCR0: 12, 13, 16
* MSR_P4_CRU_ESCR1: 14, 15, 17
*/
/*
* a caller should use P4_EVENT_ATTR helper to
* pick the attribute needed, for example
*
* P4_EVENT_ATTR(P4_TC_DELIVER_MODE, DD)
*/
enum P4_EVENTS_ATTR {
P4_MAKE_EVENT_ATTR(P4_TC_DELIVER_MODE, DD, 0),
P4_MAKE_EVENT_ATTR(P4_TC_DELIVER_MODE, DB, 1),
P4_MAKE_EVENT_ATTR(P4_TC_DELIVER_MODE, DI, 2),
P4_MAKE_EVENT_ATTR(P4_TC_DELIVER_MODE, BD, 3),
P4_MAKE_EVENT_ATTR(P4_TC_DELIVER_MODE, BB, 4),
P4_MAKE_EVENT_ATTR(P4_TC_DELIVER_MODE, BI, 5),
P4_MAKE_EVENT_ATTR(P4_TC_DELIVER_MODE, ID, 6),
P4_MAKE_EVENT_ATTR(P4_BPU_FETCH_REQUEST, TCMISS, 0),
P4_MAKE_EVENT_ATTR(P4_ITLB_REFERENCE, HIT, 0),
P4_MAKE_EVENT_ATTR(P4_ITLB_REFERENCE, MISS, 1),
P4_MAKE_EVENT_ATTR(P4_ITLB_REFERENCE, HIT_UK, 2),
P4_MAKE_EVENT_ATTR(P4_MEMORY_CANCEL, ST_RB_FULL, 2),
P4_MAKE_EVENT_ATTR(P4_MEMORY_CANCEL, 64K_CONF, 3),
P4_MAKE_EVENT_ATTR(P4_MEMORY_COMPLETE, LSC, 0),
P4_MAKE_EVENT_ATTR(P4_MEMORY_COMPLETE, SSC, 1),
P4_MAKE_EVENT_ATTR(P4_LOAD_PORT_REPLAY, SPLIT_LD, 1),
P4_MAKE_EVENT_ATTR(P4_STORE_PORT_REPLAY, SPLIT_ST, 1),
P4_MAKE_EVENT_ATTR(P4_MOB_LOAD_REPLAY, NO_STA, 1),
P4_MAKE_EVENT_ATTR(P4_MOB_LOAD_REPLAY, NO_STD, 3),
P4_MAKE_EVENT_ATTR(P4_MOB_LOAD_REPLAY, PARTIAL_DATA, 4),
P4_MAKE_EVENT_ATTR(P4_MOB_LOAD_REPLAY, UNALGN_ADDR, 5),
P4_MAKE_EVENT_ATTR(P4_PAGE_WALK_TYPE, DTMISS, 0),
P4_MAKE_EVENT_ATTR(P4_PAGE_WALK_TYPE, ITMISS, 1),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, RD_2ndL_HITS, 0),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, RD_2ndL_HITE, 1),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, RD_2ndL_HITM, 2),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, RD_3rdL_HITS, 3),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, RD_3rdL_HITE, 4),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, RD_3rdL_HITM, 5),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, RD_2ndL_MISS, 8),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, RD_3rdL_MISS, 9),
P4_MAKE_EVENT_ATTR(P4_BSQ_CACHE_REFERENCE, WR_2ndL_MISS, 10),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, DEFAULT, 0),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, ALL_READ, 5),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, ALL_WRITE, 6),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, MEM_UC, 7),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, MEM_WC, 8),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, MEM_WT, 9),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, MEM_WP, 10),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, MEM_WB, 11),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, OWN, 13),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, OTHER, 14),
P4_MAKE_EVENT_ATTR(P4_IOQ_ALLOCATION, PREFETCH, 15),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, DEFAULT, 0),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, ALL_READ, 5),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, ALL_WRITE, 6),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, MEM_UC, 7),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, MEM_WC, 8),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, MEM_WT, 9),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, MEM_WP, 10),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, MEM_WB, 11),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, OWN, 13),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, OTHER, 14),
P4_MAKE_EVENT_ATTR(P4_IOQ_ACTIVE_ENTRIES, PREFETCH, 15),
P4_MAKE_EVENT_ATTR(P4_FSB_DATA_ACTIVITY, DRDY_DRV, 0),
P4_MAKE_EVENT_ATTR(P4_FSB_DATA_ACTIVITY, DRDY_OWN, 1),
P4_MAKE_EVENT_ATTR(P4_FSB_DATA_ACTIVITY, DRDY_OTHER, 2),
P4_MAKE_EVENT_ATTR(P4_FSB_DATA_ACTIVITY, DBSY_DRV, 3),
P4_MAKE_EVENT_ATTR(P4_FSB_DATA_ACTIVITY, DBSY_OWN, 4),
P4_MAKE_EVENT_ATTR(P4_FSB_DATA_ACTIVITY, DBSY_OTHER, 5),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_TYPE0, 0),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_TYPE1, 1),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_LEN0, 2),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_LEN1, 3),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_IO_TYPE, 5),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_LOCK_TYPE, 6),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_CACHE_TYPE, 7),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_SPLIT_TYPE, 8),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_DEM_TYPE, 9),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, REQ_ORD_TYPE, 10),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, MEM_TYPE0, 11),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, MEM_TYPE1, 12),
P4_MAKE_EVENT_ATTR(P4_BSQ_ALLOCATION, MEM_TYPE2, 13),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_TYPE0, 0),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_TYPE1, 1),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_LEN0, 2),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_LEN1, 3),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_IO_TYPE, 5),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_LOCK_TYPE, 6),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_CACHE_TYPE, 7),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_SPLIT_TYPE, 8),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_DEM_TYPE, 9),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, REQ_ORD_TYPE, 10),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, MEM_TYPE0, 11),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, MEM_TYPE1, 12),
P4_MAKE_EVENT_ATTR(P4_BSQ_ACTIVE_ENTRIES, MEM_TYPE2, 13),
P4_MAKE_EVENT_ATTR(P4_SSE_INPUT_ASSIST, ALL, 15),
P4_MAKE_EVENT_ATTR(P4_PACKED_SP_UOP, ALL, 15),
P4_MAKE_EVENT_ATTR(P4_PACKED_DP_UOP, ALL, 15),
P4_MAKE_EVENT_ATTR(P4_SCALAR_SP_UOP, ALL, 15),
P4_MAKE_EVENT_ATTR(P4_SCALAR_DP_UOP, ALL, 15),
P4_MAKE_EVENT_ATTR(P4_64BIT_MMX_UOP, ALL, 15),
P4_MAKE_EVENT_ATTR(P4_128BIT_MMX_UOP, ALL, 15),
P4_MAKE_EVENT_ATTR(P4_X87_FP_UOP, ALL, 15),
P4_MAKE_EVENT_ATTR(P4_TC_MISC, FLUSH, 4),
P4_MAKE_EVENT_ATTR(P4_GLOBAL_POWER_EVENTS, RUNNING, 0),
P4_MAKE_EVENT_ATTR(P4_TC_MS_XFER, CISC, 0),
P4_MAKE_EVENT_ATTR(P4_UOP_QUEUE_WRITES, FROM_TC_BUILD, 0),
P4_MAKE_EVENT_ATTR(P4_UOP_QUEUE_WRITES, FROM_TC_DELIVER, 1),
P4_MAKE_EVENT_ATTR(P4_UOP_QUEUE_WRITES, FROM_ROM, 2),
P4_MAKE_EVENT_ATTR(P4_RETIRED_MISPRED_BRANCH_TYPE, CONDITIONAL, 1),
P4_MAKE_EVENT_ATTR(P4_RETIRED_MISPRED_BRANCH_TYPE, CALL, 2),
P4_MAKE_EVENT_ATTR(P4_RETIRED_MISPRED_BRANCH_TYPE, RETURN, 3),
P4_MAKE_EVENT_ATTR(P4_RETIRED_MISPRED_BRANCH_TYPE, INDIRECT, 4),
P4_MAKE_EVENT_ATTR(P4_RETIRED_BRANCH_TYPE, CONDITIONAL, 1),
P4_MAKE_EVENT_ATTR(P4_RETIRED_BRANCH_TYPE, CALL, 2),
P4_MAKE_EVENT_ATTR(P4_RETIRED_BRANCH_TYPE, RETURN, 3),
P4_MAKE_EVENT_ATTR(P4_RETIRED_BRANCH_TYPE, INDIRECT, 4),
P4_MAKE_EVENT_ATTR(P4_RESOURCE_STALL, SBFULL, 5),
P4_MAKE_EVENT_ATTR(P4_WC_BUFFER, WCB_EVICTS, 0),
P4_MAKE_EVENT_ATTR(P4_WC_BUFFER, WCB_FULL_EVICTS, 1),
P4_MAKE_EVENT_ATTR(P4_FRONT_END_EVENT, NBOGUS, 0),
P4_MAKE_EVENT_ATTR(P4_FRONT_END_EVENT, BOGUS, 1),
P4_MAKE_EVENT_ATTR(P4_EXECUTION_EVENT, NBOGUS0, 0),
P4_MAKE_EVENT_ATTR(P4_EXECUTION_EVENT, NBOGUS1, 1),
P4_MAKE_EVENT_ATTR(P4_EXECUTION_EVENT, NBOGUS2, 2),
P4_MAKE_EVENT_ATTR(P4_EXECUTION_EVENT, NBOGUS3, 3),
P4_MAKE_EVENT_ATTR(P4_EXECUTION_EVENT, BOGUS0, 4),
P4_MAKE_EVENT_ATTR(P4_EXECUTION_EVENT, BOGUS1, 5),
P4_MAKE_EVENT_ATTR(P4_EXECUTION_EVENT, BOGUS2, 6),
P4_MAKE_EVENT_ATTR(P4_EXECUTION_EVENT, BOGUS3, 7),
P4_MAKE_EVENT_ATTR(P4_REPLAY_EVENT, NBOGUS, 0),
P4_MAKE_EVENT_ATTR(P4_REPLAY_EVENT, BOGUS, 1),
P4_MAKE_EVENT_ATTR(P4_INSTR_RETIRED, NBOGUSNTAG, 0),
P4_MAKE_EVENT_ATTR(P4_INSTR_RETIRED, NBOGUSTAG, 1),
P4_MAKE_EVENT_ATTR(P4_INSTR_RETIRED, BOGUSNTAG, 2),
P4_MAKE_EVENT_ATTR(P4_INSTR_RETIRED, BOGUSTAG, 3),
P4_MAKE_EVENT_ATTR(P4_UOPS_RETIRED, NBOGUS, 0),
P4_MAKE_EVENT_ATTR(P4_UOPS_RETIRED, BOGUS, 1),
P4_MAKE_EVENT_ATTR(P4_UOP_TYPE, TAGLOADS, 1),
P4_MAKE_EVENT_ATTR(P4_UOP_TYPE, TAGSTORES, 2),
P4_MAKE_EVENT_ATTR(P4_BRANCH_RETIRED, MMNP, 0),
P4_MAKE_EVENT_ATTR(P4_BRANCH_RETIRED, MMNM, 1),
P4_MAKE_EVENT_ATTR(P4_BRANCH_RETIRED, MMTP, 2),
P4_MAKE_EVENT_ATTR(P4_BRANCH_RETIRED, MMTM, 3),
P4_MAKE_EVENT_ATTR(P4_MISPRED_BRANCH_RETIRED, NBOGUS, 0),
P4_MAKE_EVENT_ATTR(P4_X87_ASSIST, FPSU, 0),
P4_MAKE_EVENT_ATTR(P4_X87_ASSIST, FPSO, 1),
P4_MAKE_EVENT_ATTR(P4_X87_ASSIST, POAO, 2),
P4_MAKE_EVENT_ATTR(P4_X87_ASSIST, POAU, 3),
P4_MAKE_EVENT_ATTR(P4_X87_ASSIST, PREA, 4),
P4_MAKE_EVENT_ATTR(P4_MACHINE_CLEAR, CLEAR, 0),
P4_MAKE_EVENT_ATTR(P4_MACHINE_CLEAR, MOCLEAR, 1),
P4_MAKE_EVENT_ATTR(P4_MACHINE_CLEAR, SMCLEAR, 2),
P4_MAKE_EVENT_ATTR(P4_INSTR_COMPLETED, NBOGUS, 0),
P4_MAKE_EVENT_ATTR(P4_INSTR_COMPLETED, BOGUS, 1),
};
enum {
KEY_P4_L1D_OP_READ_RESULT_MISS = PERF_COUNT_HW_MAX,
KEY_P4_LL_OP_READ_RESULT_MISS,
KEY_P4_DTLB_OP_READ_RESULT_MISS,
KEY_P4_DTLB_OP_WRITE_RESULT_MISS,
KEY_P4_ITLB_OP_READ_RESULT_ACCESS,
KEY_P4_ITLB_OP_READ_RESULT_MISS,
KEY_P4_UOP_TYPE,
};
perf, x86: Implement initial P4 PMU driver The netburst PMU is way different from the "architectural perfomance monitoring" specification that current CPUs use. P4 uses a tuple of ESCR+CCCR+COUNTER MSR registers to handle perfomance monitoring events. A few implementational details: 1) We need a separate x86_pmu::hw_config helper in struct x86_pmu since register bit-fields are quite different from P6, Core and later cpu series. 2) For the same reason is a x86_pmu::schedule_events helper introduced. 3) hw_perf_event::config consists of packed ESCR+CCCR values. It's allowed since in reality both registers only use a half of their size. Of course before making a real write into a particular MSR we need to unpack the value and extend it to a proper size. 4) The tuple of packed ESCR+CCCR in hw_perf_event::config doesn't describe the memory address of ESCR MSR register so that we need to keep a mapping between these tuples used and available ESCR (various P4 events may use same ESCRs but not simultaneously), for this sake every active event has a per-cpu map of hw_perf_event::idx <--> ESCR addresses. 5) Since hw_perf_event::idx is an offset to counter/control register we need to lift X86_PMC_MAX_GENERIC up, otherwise kernel strips it down to 8 registers and event armed may never be turned off (ie the bit in active_mask is set but the loop never reaches this index to check), thanks to Peter Zijlstra Restrictions: - No cascaded counters support (do we ever need them?) - No dependent events support (so PERF_COUNT_HW_INSTRUCTIONS doesn't work for now) - There are events with same counters which can't work simultaneously (need to use intersected ones due to broken counter 1) - No PERF_COUNT_HW_CACHE_ events yet Todo: - Implement dependent events - Need proper hashing for event opcodes (no linear search, good for debugging stage but not in real loads) - Some events counted during a clock cycle -- need to set threshold for them and count every clock cycle just to get summary statistics (ie to behave the same way as other PMUs do) - Need to swicth to use event_constraints - To support RAW events we need to encode a global list of P4 events into p4_templates - Cache events need to be added Event support status matrix: Event status ----------------------------- cycles works cache-references works cache-misses works branch-misses works bus-cycles partially (does not work on 64bit cpu with HT enabled) instruction doesnt work (needs dependent event [mop tagging]) branches doesnt work Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> Signed-off-by: Lin Ming <ming.m.lin@intel.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <20100311165439.GB5129@lenovo> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-11 16:54:39 +00:00
#endif /* PERF_EVENT_P4_H */