0001 perf-mem(1)
0002 ===========
0003
0004 NAME
0005 ----
0006 perf-mem - Profile memory accesses
0007
0008 SYNOPSIS
0009 --------
0010 [verse]
0011 'perf mem' [<options>] (record [<command>] | report)
0012
0013 DESCRIPTION
0014 -----------
0015 "perf mem record" runs a command and gathers memory operation data
0016 from it, into perf.data. Perf record options are accepted and are passed through.
0017
0018 "perf mem report" displays the result. It invokes perf report with the
0019 right set of options to display a memory access profile. By default, loads
0020 and stores are sampled. Use the -t option to limit to loads or stores.
0021
0022 Note that on Intel systems the memory latency reported is the use-latency,
0023 not the pure load (or store latency). Use latency includes any pipeline
0024 queueing delays in addition to the memory subsystem latency.
0025
0026 OPTIONS
0027 -------
0028 <command>...::
0029 Any command you can specify in a shell.
0030
0031 -i::
0032 --input=<file>::
0033 Input file name.
0034
0035 -f::
0036 --force::
0037 Don't do ownership validation
0038
0039 -t::
0040 --type=<type>::
0041 Select the memory operation type: load or store (default: load,store)
0042
0043 -D::
0044 --dump-raw-samples::
0045 Dump the raw decoded samples on the screen in a format that is easy to parse with
0046 one sample per line.
0047
0048 -x::
0049 --field-separator=<separator>::
0050 Specify the field separator used when dump raw samples (-D option). By default,
0051 The separator is the space character.
0052
0053 -C::
0054 --cpu=<cpu>::
0055 Monitor only on the list of CPUs provided. Multiple CPUs can be provided as a
0056 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. Default
0057 is to monitor all CPUS.
0058 -U::
0059 --hide-unresolved::
0060 Only display entries resolved to a symbol.
0061
0062 -p::
0063 --phys-data::
0064 Record/Report sample physical addresses
0065
0066 --data-page-size::
0067 Record/Report sample data address page size
0068
0069 RECORD OPTIONS
0070 --------------
0071 -e::
0072 --event <event>::
0073 Event selector. Use 'perf mem record -e list' to list available events.
0074
0075 -K::
0076 --all-kernel::
0077 Configure all used events to run in kernel space.
0078
0079 -U::
0080 --all-user::
0081 Configure all used events to run in user space.
0082
0083 -v::
0084 --verbose::
0085 Be more verbose (show counter open errors, etc)
0086
0087 --ldlat <n>::
0088 Specify desired latency for loads event. (x86 only)
0089
0090 In addition, for report all perf report options are valid, and for record
0091 all perf record options.
0092
0093 SEE ALSO
0094 --------
0095 linkperf:perf-record[1], linkperf:perf-report[1]
