134 lines · plain
1perf-mem(1)2===========3 4NAME5----6perf-mem - Profile memory accesses7 8SYNOPSIS9--------10[verse]11'perf mem' [<options>] (record [<command>] | report)12 13DESCRIPTION14-----------15"perf mem record" runs a command and gathers memory operation data16from it, into perf.data. Perf record options are accepted and are passed through.17 18"perf mem report" displays the result. It invokes perf report with the19right set of options to display a memory access profile. By default, loads20and stores are sampled. Use the -t option to limit to loads or stores.21 22Note that on Intel systems the memory latency reported is the use-latency,23not the pure load (or store latency). Use latency includes any pipeline24queuing delays in addition to the memory subsystem latency.25 26On Arm64 this uses SPE to sample load and store operations, therefore hardware27and kernel support is required. See linkperf:perf-arm-spe[1] for a setup guide.28Due to the statistical nature of SPE sampling, not every memory operation will29be sampled.30 31COMMON OPTIONS32--------------33-f::34--force::35 Don't do ownership validation36 37-t::38--type=<type>::39 Select the memory operation type: load or store (default: load,store)40 41-v::42--verbose::43 Be more verbose (show counter open errors, etc)44 45-p::46--phys-data::47 Record/Report sample physical addresses48 49--data-page-size::50 Record/Report sample data address page size51 52RECORD OPTIONS53--------------54<command>...::55 Any command you can specify in a shell.56 57-e::58--event <event>::59 Event selector. Use 'perf mem record -e list' to list available events.60 61-K::62--all-kernel::63 Configure all used events to run in kernel space.64 65-U::66--all-user::67 Configure all used events to run in user space.68 69--ldlat <n>::70 Specify desired latency for loads event. Supported on Intel and Arm6471 processors only. Ignored on other archs.72 73REPORT OPTIONS74--------------75-i::76--input=<file>::77 Input file name.78 79-C::80--cpu=<cpu>::81 Monitor only on the list of CPUs provided. Multiple CPUs can be provided as a82 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -83 like 0-2. Default is to monitor all CPUS.84 85-D::86--dump-raw-samples::87 Dump the raw decoded samples on the screen in a format that is easy to parse with88 one sample per line.89 90-s::91--sort=<key>::92 Group result by given key(s) - multiple keys can be specified93 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.96 97 - symbol_daddr: name of data symbol being executed on at the time of sample98 - symbol_iaddr: name of code symbol being executed on at the time of sample99 - dso_daddr: name of library or module containing the data being executed100 on at the time of the sample101 - locked: whether the bus was locked at the time of the sample102 - tlb: type of tlb access for the data at the time of the sample103 - mem: type of memory access for the data at the time of the sample104 - snoop: type of snoop (if any) for the data at the time of the sample105 - dcacheline: the cacheline the data address is on at the time of the sample106 - phys_daddr: physical address of data being executed on at the time of sample107 - data_page_size: the data page size of data being executed on at the time of sample108 - blocked: reason of blocked load access for the data at the time of the sample109 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.112 113-T::114--type-profile::115 Show data-type profile result instead of code symbols. This requires116 the debug information and it will change the default sort keys to:117 mem, snoop, tlb, type.118 119-U::120--hide-unresolved::121 Only display entries resolved to a symbol.122 123-x::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.127 128In addition, for report all perf report options are valid, and for record129all perf record options.130 131SEE ALSO132--------133linkperf:perf-record[1], linkperf:perf-report[1], linkperf:perf-arm-spe[1]134