1perf-list(1) 2============ 3 4NAME 5---- 6perf-list - List all symbolic event types 7 8SYNOPSIS 9-------- 10[verse] 11'perf list' [hw|sw|cache|tracepoint|pmu|event_glob] 12 13DESCRIPTION 14----------- 15This command displays the symbolic event types which can be selected in the 16various perf commands with the -e option. 17 18[[EVENT_MODIFIERS]] 19EVENT MODIFIERS 20--------------- 21 22Events can optionally have a modifier by appending a colon and one or 23more modifiers. Modifiers allow the user to restrict the events to be 24counted. The following modifiers exist: 25 26 u - user-space counting 27 k - kernel counting 28 h - hypervisor counting 29 G - guest counting (in KVM guests) 30 H - host counting (not in KVM guests) 31 p - precise level 32 S - read sample value (PERF_SAMPLE_READ) 33 D - pin the event to the PMU 34 35The 'p' modifier can be used for specifying how precise the instruction 36address should be. The 'p' modifier can be specified multiple times: 37 38 0 - SAMPLE_IP can have arbitrary skid 39 1 - SAMPLE_IP must have constant skid 40 2 - SAMPLE_IP requested to have 0 skid 41 3 - SAMPLE_IP must have 0 skid 42 43For Intel systems precise event sampling is implemented with PEBS 44which supports up to precise-level 2. 45 46On AMD systems it is implemented using IBS (up to precise-level 2). 47The precise modifier works with event types 0x76 (cpu-cycles, CPU 48clocks not halted) and 0xC1 (micro-ops retired). Both events map to 49IBS execution sampling (IBS op) with the IBS Op Counter Control bit 50(IbsOpCntCtl) set respectively (see AMD64 Architecture Programmer’s 51Manual Volume 2: System Programming, 13.3 Instruction-Based 52Sampling). Examples to use IBS: 53 54 perf record -a -e cpu-cycles:p ... # use ibs op counting cycles 55 perf record -a -e r076:p ... # same as -e cpu-cycles:p 56 perf record -a -e r0C1:p ... # use ibs op counting micro-ops 57 58RAW HARDWARE EVENT DESCRIPTOR 59----------------------------- 60Even when an event is not available in a symbolic form within perf right now, 61it can be encoded in a per processor specific way. 62 63For instance For x86 CPUs NNN represents the raw register encoding with the 64layout of IA32_PERFEVTSELx MSRs (see [Intel® 64 and IA-32 Architectures Software Developer's Manual Volume 3B: System Programming Guide] Figure 30-1 Layout 65of IA32_PERFEVTSELx MSRs) or AMD's PerfEvtSeln (see [AMD64 Architecture Programmer’s Manual Volume 2: System Programming], Page 344, 66Figure 13-7 Performance Event-Select Register (PerfEvtSeln)). 67 68Note: Only the following bit fields can be set in x86 counter 69registers: event, umask, edge, inv, cmask. Esp. guest/host only and 70OS/user mode flags must be setup using <<EVENT_MODIFIERS, EVENT 71MODIFIERS>>. 72 73Example: 74 75If the Intel docs for a QM720 Core i7 describe an event as: 76 77 Event Umask Event Mask 78 Num. Value Mnemonic Description Comment 79 80 A8H 01H LSD.UOPS Counts the number of micro-ops Use cmask=1 and 81 delivered by loop stream detector invert to count 82 cycles 83 84raw encoding of 0x1A8 can be used: 85 86 perf stat -e r1a8 -a sleep 1 87 perf record -e r1a8 ... 88 89You should refer to the processor specific documentation for getting these 90details. Some of them are referenced in the SEE ALSO section below. 91 92PARAMETERIZED EVENTS 93-------------------- 94 95Some pmu events listed by 'perf-list' will be displayed with '?' in them. For 96example: 97 98 hv_gpci/dtbp_ptitc,phys_processor_idx=?/ 99 100This means that when provided as an event, a value for '?' must 101also be supplied. For example: 102 103 perf stat -C 0 -e 'hv_gpci/dtbp_ptitc,phys_processor_idx=0x2/' ... 104 105OPTIONS 106------- 107 108Without options all known events will be listed. 109 110To limit the list use: 111 112. 'hw' or 'hardware' to list hardware events such as cache-misses, etc. 113 114. 'sw' or 'software' to list software events such as context switches, etc. 115 116. 'cache' or 'hwcache' to list hardware cache events such as L1-dcache-loads, etc. 117 118. 'tracepoint' to list all tracepoint events, alternatively use 119 'subsys_glob:event_glob' to filter by tracepoint subsystems such as sched, 120 block, etc. 121 122. 'pmu' to print the kernel supplied PMU events. 123 124. If none of the above is matched, it will apply the supplied glob to all 125 events, printing the ones that match. 126 127One or more types can be used at the same time, listing the events for the 128types specified. 129 130SEE ALSO 131-------- 132linkperf:perf-stat[1], linkperf:perf-top[1], 133linkperf:perf-record[1], 134http://www.intel.com/Assets/PDF/manual/253669.pdf[Intel® 64 and IA-32 Architectures Software Developer's Manual Volume 3B: System Programming Guide], 135http://support.amd.com/us/Processor_TechDocs/24593_APM_v2.pdf[AMD64 Architecture Programmer’s Manual Volume 2: System Programming] 136