6 perf-mem - Profile memory accesses
11 'perf mem' [<options>] (record [<command>] | report)
15 "perf mem record" runs a command and gathers memory operation data
16 from it, into perf.data. Perf record options are accepted and are passed through.
18 "perf mem report" displays the result. It invokes perf report with the
19 right set of options to display a memory access profile. By default, loads
20 and stores are sampled. Use the -t option to limit to loads or stores.
22 Note that on Intel systems the memory latency reported is the use-latency,
23 not the pure load (or store latency). Use latency includes any pipeline
24 queuing delays in addition to the memory subsystem latency.
26 On Arm64 this uses SPE to sample load and store operations, therefore hardware
27 and kernel support is required. See linkperf:perf-arm-spe[1] for a setup guide.
28 Due to the statistical nature of SPE sampling, not every memory operation will
35 Don't do ownership validation
39 Select the memory operation type: load or store (default: load,store)
43 Be more verbose (show counter open errors, etc)
47 Record/Report sample physical addresses
50 Record/Report sample data address page size
55 Any command you can specify in a shell.
59 Event selector. Use 'perf mem record -e list' to list available events.
63 Configure all used events to run in kernel space.
67 Configure all used events to run in user space.
70 Specify desired latency for loads event. Supported on Intel and Arm64
71 processors only. Ignored on other archs.
81 Monitor only on the list of CPUs provided. Multiple CPUs can be provided as a
82 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -
83 like 0-2. Default is to monitor all CPUS.
87 Dump the raw decoded samples on the screen in a format that is easy to parse with
92 Group result by given key(s) - multiple keys can be specified
93 in CSV format. The keys are specific to memory samples are:
94 symbol_daddr, symbol_iaddr, dso_daddr, locked, tlb, mem, snoop,
95 dcacheline, phys_daddr, data_page_size, blocked.
97 - symbol_daddr: name of data symbol being executed on at the time of sample
98 - symbol_iaddr: name of code symbol being executed on at the time of sample
99 - dso_daddr: name of library or module containing the data being executed
100 on at the time of the sample
101 - locked: whether the bus was locked at the time of the sample
102 - tlb: type of tlb access for the data at the time of the sample
103 - mem: type of memory access for the data at the time of the sample
104 - snoop: type of snoop (if any) for the data at the time of the sample
105 - dcacheline: the cacheline the data address is on at the time of the sample
106 - phys_daddr: physical address of data being executed on at the time of sample
107 - data_page_size: the data page size of data being executed on at the time of sample
108 - blocked: reason of blocked load access for the data at the time of the sample
110 And the default sort keys are changed to local_weight, mem, sym, dso,
111 symbol_daddr, dso_daddr, snoop, tlb, locked, blocked, local_ins_lat.
115 Show data-type profile result instead of code symbols. This requires
116 the debug information and it will change the default sort keys to:
117 mem, snoop, tlb, type.
121 Only display entries resolved to a symbol.
124 --field-separator=<separator>::
125 Specify the field separator used when dump raw samples (-D option). By default,
126 The separator is the space character.
128 In addition, for report all perf report options are valid, and for record
129 all perf record options.
133 linkperf:perf-record[1], linkperf:perf-report[1], linkperf:perf-arm-spe[1]