“Perf”的版本间的差异

来自百问网嵌入式Linux wiki
 
第7行: 第7行:
  
 
==Yocto下安装==
 
==Yocto下安装==
 +
<syntaxhighlight lang="bash">
 +
RRECOMMENDS_${PN} = "\
 +
  ${PERF} \
 +
  trace-cmd \
 +
  blktrace \
 +
  ${PROFILE_TOOLS_X} \
 +
  ${PROFILE_TOOLS_SYSTEMD} \
 +
  "
 +
...
 +
PERF = "perf"
 +
</syntaxhighlight>
  
 +
==安卓系统下使用==
 +
:simpleperf<ref name="android">https://source.android.com/devices/tech/debug/eval_perf</ref>与perf等效,并且默认情况下已安装(/system/xbin/simpleperf),可以与所有适用于Android™的软件包一起使用
  
 +
:它支持较少的选项:
 +
<syntaxhighlight lang="bash">
 +
Board $> simpleperf --help
 +
Usage: simpleperf [common options] subcommand [args_for_subcommand]
 +
common options:
 +
    -h/--help    Print this help information.
 +
    --log <severity> Set the minimum severity of logging. Possible severities
 +
                      include verbose, debug, warning, info, error, fatal.
 +
                      Default is info.
 +
    --version    Print version of simpleperf.
 +
subcommands:
 +
    debug-unwind        Debug/test offline unwinding.
 +
    dump                dump perf record file
 +
    help                print help information for simpleperf
 +
    kmem                collect kernel memory allocation information
 +
    list                list available event types
 +
    record              record sampling info in perf.data
 +
    report              report sampling information in perf.data
 +
    report-sample      report raw sample information in perf.data
 +
    stat                gather performance counter information
  
 +
</syntaxhighlight>
 
=开始使用=
 
=开始使用=
 +
Board $> which perf
 +
/usr/bin/perf
 +
 
== Perf命令==
 
== Perf命令==
perf report (Linux kernel documentation[10]): breaks down the events by process, function, etc.
+
<syntaxhighlight lang="bash">
 +
usage: perf [--version] [--help] [OPTIONS] COMMAND [ARGS]
 +
 
 +
The most commonly used perf commands are:
 +
  annotate        Reads perf.data (created by perf record) and displays annotated code
 +
  archive        Creates archive with object files with build-ids found in perf.data file
 +
  bench          General framework for benchmark suites
 +
  buildid-cache  Manages build-id cache.
 +
  buildid-list    Lists the buildids in a perf.data file
 +
  c2c            Shared Data C2C/HITM Analyzer.
 +
  config          Gets and sets variables in a configuration file.
 +
  data            Data file related processing
 +
  diff            Reads perf.data files and displays the differential profile
 +
  evlist          Lists the event names in a perf.data file
 +
  ftrace          simple wrapper for kernel's ftrace functionality
 +
  inject          Filters to augment the events stream with additional information
 +
  kallsyms        Searches running kernel for symbols
 +
  kmem            Tool to trace/measure kernel memory properties
 +
  kvm            Tool to trace/measure kvm guest os
 +
  list            Lists all symbolic event types
 +
  lock            Analyzes lock events
 +
  mem            Profiles memory accesses
 +
  record          Runs a command and records its profile into perf.data
 +
  report         Reads perf.data (created by perf record) and displays the profile
 +
  sched          Tool to trace/measure scheduler properties (latencies)
 +
  script          Reads perf.data (created by perf record) and displays trace output
 +
  stat            Runs a command and gathers performance counter statistics
 +
  test            Runs sanity tests.
 +
  timechart      Tool to visualize total system behavior during a workload
 +
  top            System profiling tool.
 +
  probe          Defines new dynamic tracepoints
 +
 
 +
See 'perf COMMAND -h' for more information on a specific command.
 +
</syntaxhighlight>
 +
 
 +
*perf top(Linux内核文档<ref>https://github.com/STMicroelectronics/linux/blob/v4.19-stm32mp/tools/perf/Documentation/perf-top.txt</ref>):通过计算循环事件数来提供CPU负载;默认顺序是每个符号的采样数降序:
 +
<syntaxhighlight lang="bash">
 +
Board $> perf top
 +
40.62%  [kernel]                          [k] v7_dma_inv_range
 +
18.65%  [kernel]                          [k] _raw_spin_unlock_irqrestore
 +
17.01%  [kernel]                          [k] arch_cpu_idle
 +
  8.27%  [kernel]                          [k] v7_dma_clean_range
 +
  5.00%  [kernel]                          [k] rcu_idle_exit
 +
  1.70%  [kernel]                          [k] cpu_startup_entry
 +
  0.52%  [kernel]                          [k] trace_graph_return
 +
  0.48%  [kernel]                          [k] finish_task_switch
 +
  0.48%  libc-2.18.so                      [.] memcpy
 +
  0.47%  [kernel]                          [k] trace_graph_entry
 +
 
 +
</syntaxhighlight>
 +
:意味着CPU在功能v7_dma_inv_range中花费了40%的时间,在_raw_spin_unlock_irqrestore中花费了18.65%的时间。
 +
:有关更多信息和示例,请访问perf.wiki.kernel.org<ref>https://perf.wiki.kernel.org/index.php/Tutorial#Live_analysis_with_perf_top</ref>。
 +
:也可以按指定的排序显示结果:
 +
<syntaxhighlight lang="bash">
 +
sage: perf top [<options>]
 +
  -s, --sort <key[,key2...]>
 +
              sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline, ... Please refer to the main page for the complete list.
 +
</syntaxhighlight>
 +
 
 +
*perf stat(linux kernel documentation<ref>https://github.com/STMicroelectronics/linux/blob/v4.19-stm32mp/tools/perf/Documentation/perf-stat.txt</ref>): 获取事件计数
 +
<syntaxhighlight lang="bash">
 +
Board $> perf stat hello_world_example
 +
User space example: hello world from STMicroelectronics
 +
10 9 8 7 6 5 4 3 2 1 0
 +
User space example: goodbye from STMicroelectronics
 +
</syntaxhighlight>
 +
<syntaxhighlight lang="bash">
 +
Performance counter stats for 'hello_world_example':
 +
 
 +
        4.328249      task-clock (msec)        #    0.000 CPUs utilized         
 +
              11      context-switches          #    0.003 M/sec                 
 +
                0      cpu-migrations            #    0.000 K/sec                 
 +
              38      page-faults              #    0.009 M/sec                 
 +
          2710036      cycles                    #    0.626 GHz                   
 +
          640856      instructions              #    0.24  insn per cycle       
 +
            75644      branches                  #  17.477 M/sec                 
 +
            21764      branch-misses            #  28.77% of all branches       
 +
 
 +
    11.109859338 seconds time elapsed
 +
</syntaxhighlight>
 +
:有关更多信息和示例,请参见perf.wiki.kernel.org<ref>https://perf.wiki.kernel.org/index.php/Tutorial#Counting_with_perf_stat</ref>
 +
 
 +
*perf list (Linux kernel documentation<ref>https://github.com/STMicroelectronics/linux/blob/v4.19-stm32mp/tools/perf/Documentation/perf-list.txt</ref>): 支持的符号事件类型.
 +
<div style="height:30em; overflow:auto; border: 1px solid #088">
 +
<syntaxhighlight lang="bash">
 +
Board $> perf list
 +
branch-instructions OR branches                    [Hardware event]
 +
branch-misses                                      [Hardware event]
 +
bus-cycles                                        [Hardware event]
 +
cache-misses                                      [Hardware event]
 +
cache-references                                  [Hardware event]
 +
cpu-cycles OR cycles                              [Hardware event]
 +
instructions                                      [Hardware event]
 +
alignment-faults                                  [Software event]
 +
bpf-output                                        [Software event]
 +
context-switches OR cs                            [Software event]
 +
cpu-clock                                          [Software event]
 +
cpu-migrations OR migrations                      [Software event]
 +
dummy                                              [Software event]
 +
emulation-faults                                  [Software event]
 +
major-faults                                      [Software event]
 +
minor-faults                                      [Software event]
 +
page-faults OR faults                              [Software event]
 +
task-clock                                        [Software event]
 +
L1-dcache-load-misses                              [Hardware cache event]
 +
L1-dcache-loads                                    [Hardware cache event]
 +
L1-dcache-store-misses                            [Hardware cache event]
 +
L1-dcache-stores                                  [Hardware cache event]
 +
L1-icache-load-misses                              [Hardware cache event]
 +
L1-icache-loads                                    [Hardware cache event]
 +
LLC-load-misses                                    [Hardware cache event]
 +
LLC-loads                                          [Hardware cache event]
 +
LLC-store-misses                                  [Hardware cache event]
 +
LLC-stores                                        [Hardware cache event]
 +
branch-load-misses                                [Hardware cache event]
 +
branch-loads                                      [Hardware cache event]
 +
dTLB-load-misses                                  [Hardware cache event]
 +
dTLB-store-misses                                  [Hardware cache event]
 +
iTLB-load-misses                                  [Hardware cache event]
 +
armv7_cortex_a7/br_immed_retired/                  [Kernel PMU event]
 +
armv7_cortex_a7/br_mis_pred/                      [Kernel PMU event]
 +
armv7_cortex_a7/br_pred/                          [Kernel PMU event]
 +
armv7_cortex_a7/br_return_retired/                [Kernel PMU event]
 +
armv7_cortex_a7/bus_access/                        [Kernel PMU event]
 +
armv7_cortex_a7/bus_cycles/                        [Kernel PMU event]
 +
armv7_cortex_a7/cid_write_retired/                [Kernel PMU event]
 +
armv7_cortex_a7/cpu_cycles/                        [Kernel PMU event]
 +
armv7_cortex_a7/exc_return/                        [Kernel PMU event]
 +
armv7_cortex_a7/exc_taken/                        [Kernel PMU event]
 +
armv7_cortex_a7/inst_retired/                      [Kernel PMU event]
 +
armv7_cortex_a7/inst_spec/                        [Kernel PMU event]
 +
armv7_cortex_a7/l1d_cache/                        [Kernel PMU event]
 +
armv7_cortex_a7/l1d_cache_refill/                  [Kernel PMU event]
 +
armv7_cortex_a7/l1d_cache_wb/                      [Kernel PMU event]
 +
armv7_cortex_a7/l1d_tlb_refill/                    [Kernel PMU event]
 +
armv7_cortex_a7/l1i_cache/                        [Kernel PMU event]
 +
armv7_cortex_a7/l1i_cache_refill/                  [Kernel PMU event]
 +
armv7_cortex_a7/l1i_tlb_refill/                    [Kernel PMU event]
 +
armv7_cortex_a7/l2d_cache/                        [Kernel PMU event]
 +
armv7_cortex_a7/l2d_cache_refill/                  [Kernel PMU event]
 +
armv7_cortex_a7/l2d_cache_wb/                      [Kernel PMU event]
 +
armv7_cortex_a7/ld_retired/                        [Kernel PMU event]
 +
armv7_cortex_a7/mem_access/                        [Kernel PMU event]
 +
armv7_cortex_a7/memory_error/                      [Kernel PMU event]
 +
armv7_cortex_a7/pc_write_retired/                  [Kernel PMU event]
 +
armv7_cortex_a7/st_retired/                        [Kernel PMU event]
 +
armv7_cortex_a7/sw_incr/                          [Kernel PMU event]
 +
armv7_cortex_a7/ttbr_write_retired/                [Kernel PMU event]
 +
armv7_cortex_a7/unaligned_ldst_retired/            [Kernel PMU event]
 +
rNNN                                              [Raw hardware event descriptor]
 +
cpu/t1=v1[,t2=v2,t3 ...]/modifier                  [Raw hardware event descriptor]
 +
mem:<addr>[/len][:access]                          [Hardware breakpoint]
 +
alarmtimer:alarmtimer_cancel                      [Tracepoint event]
 +
alarmtimer:alarmtimer_fired                        [Tracepoint event]
 +
alarmtimer:alarmtimer_start                        [Tracepoint event]
 +
alarmtimer:alarmtimer_suspend                      [Tracepoint event]
 +
asoc:snd_soc_bias_level_done                      [Tracepoint event]
 +
asoc:snd_soc_bias_level_start                      [Tracepoint event]
 +
asoc:snd_soc_dapm_connected                        [Tracepoint event]
 +
asoc:snd_soc_dapm_done                            [Tracepoint event]
 +
asoc:snd_soc_dapm_path                            [Tracepoint event]
 +
asoc:snd_soc_dapm_start                            [Tracepoint event]
 +
asoc:snd_soc_dapm_walk_done                        [Tracepoint event]
 +
asoc:snd_soc_dapm_widget_event_done                [Tracepoint event]
 +
asoc:snd_soc_dapm_widget_event_start              [Tracepoint event]
 +
...
 +
xhci-hcd:xhci_inc_enq                              [Tracepoint event]
 +
xhci-hcd:xhci_queue_trb                            [Tracepoint event]
 +
xhci-hcd:xhci_ring_alloc                          [Tracepoint event]
 +
xhci-hcd:xhci_ring_expansion                      [Tracepoint event]
 +
xhci-hcd:xhci_ring_free                            [Tracepoint event]
 +
xhci-hcd:xhci_setup_addressable_virt_device        [Tracepoint event]
 +
xhci-hcd:xhci_setup_device                        [Tracepoint event]
 +
xhci-hcd:xhci_setup_device_slot                    [Tracepoint event]
 +
xhci-hcd:xhci_stop_device                          [Tracepoint event]
 +
xhci-hcd:xhci_urb_dequeue                          [Tracepoint event]
 +
xhci-hcd:xhci_urb_enqueue                          [Tracepoint event]
 +
xhci-hcd:xhci_urb_giveback                        [Tracepoint event]
 +
</syntaxhighlight>
 +
</div>
 +
 
 +
*perf record (Linux kernel documentation<ref>https://github.com/STMicroelectronics/linux/blob/v4.19-stm32mp/tools/perf/Documentation/perf-record.txt</ref>): 记录事件以供以后报告
 +
<syntaxhighlight lang="bash">
 +
 
 +
Board $> perf record hello_world_example
 +
 
 +
User space example: hello world from STMicroelectronics
 +
10 9 8 7 6 5 4 3 2 1 0
 +
User space example: goodbye from STMicroelectronics
 +
[ perf record: Woken up 1 time to write data ]
 +
[ perf record: Captured and wrote 0.004 MB perf.data (28 samples) ]
 +
</syntaxhighlight>
 +
 
 +
:这可以过滤事件(由perf list命令提供)。有关更多信息,选项和示例,请参见perf.wiki.kernel.org<ref>https://perf.wiki.kernel.org/index.php/Tutorial#Sampling_with_perf_record</ref>。
 +
:默认情况下,事件记录在perf.data文件中。如果要指定另一个输出文件名,则必须添加-o,--output <file>选项。
 +
 
 +
*perf report (Linux kernel documentation<ref>https://github.com/STMicroelectronics/linux/blob/v4.19-stm32mp/tools/perf/Documentation/perf-record.txt</ref>): 按进程,功能等细分事件。.
 +
<syntaxhighlight lang="bash">
 +
Example after previous command "perf record hello_world_example"
 +
Board $> perf report
 +
Samples: 28  of event 'cycles:ppp', Event count (approx.):2737925               
 +
Overhead  Command          Shared Object Symbol                 
 +
  12.66%  hello_world_exa  ld-2.26.so        [.] _dl_relocate_object
 +
  11.71%  hello_world_exa  [kernel.kallsyms]  [k] filemap_map_pages
 +
  10.65%  hello_world_exa  [kernel.kallsyms]  [k] n_tty_write
 +
  6.43%  hello_world_exa  [kernel.kallsyms]  [k] percpu_counter_add_batch
 +
  6.43%  hello_world_exa  ld-2.26.so        [.] sbrk
 +
  6.24%  hello_world_exa  [kernel.kallsyms]  [k] cpu_v7_set_pte_ext
 +
  5.56%  hello_world_exa  [kernel.kallsyms]  [k] alloc_set_pte
 +
  5.56%  hello_world_exa  libc-2.26.so      [.] __sbrk
 +
  5.37%  hello_world_exa  [kernel.kallsyms]  [k] __vma_link_file
 +
  5.32%  hello_world_exa  [kernel.kallsyms]  [k] __fput
 +
  5.32%  hello_world_exa  [kernel.kallsyms]  [k] ldsem_up_read
 +
  5.32%  hello_world_exa  [kernel.kallsyms]  [k] unmap_page_range
 +
  5.32%  hello_world_exa  libc-2.26.so      [.] printf
 +
  5.24%  hello_world_exa  [kernel.kallsyms]  [k] _raw_spin_unlock_irqrestore
 +
  2.23%  hello_world_exa  [kernel.kallsyms]  [k] perf_event_mmap
 +
  0.48%  hello_world_exa  [kernel.kallsyms]  [k] perf_output_begin
 +
  0.13%  perf            [kernel.kallsyms]  [k] perf_event_exec
 +
</syntaxhighlight>
 +
 
 +
默认情况下,报告文件perf.data被读取为输入文件。如果要指定另一个输入文件名,则必须添加-i,--input <file> option。
 +
有关更多信息和示例,请参见perf.wiki.kernel.org<ref>https://perf.wiki.kernel.org/index.php/Tutorial#Sample_analysis_with_perf_report</ref>。
 +
*perf bench (Linux kernel documentation<ref>https://github.com/STMicroelectronics/linux/blob/v4.19-stm32mp/tools/perf/Documentation/perf-record.txt</ref>): 运行不同的内核微基准测试:
 +
<syntaxhighlight lang="bash">
 +
# List of all available benchmark collections:
 +
 
 +
        sched: Scheduler and IPC benchmarks
 +
          mem: Memory access benchmarks
 +
        futex: Futex stressing benchmarks
 +
          all: All benchmarks
 +
</syntaxhighlight>
 +
 
 +
<syntaxhighlight lang="bash">
 +
Example of getting memcpy benchmark for 100MB:
 +
Board $> perf bench mem memcpy --size 100MB
 +
# Running 'mem/memcpy' benchmark:
 +
# function 'default' (Default memcpy() provided by glibc)
 +
# Copying 100MB bytes ...
 +
 
 +
      1.426138 GB/sec
 +
</syntaxhighlight>
 +
:有关更多信息和示例,请参见perf.wiki.kernel.org<ref>https://perf.wiki.kernel.org/index.php/Tutorial#Benchmarking_with_perf_bench</ref>
 +
 
  
 
=进阶使用=
 
=进阶使用=
 +
作为Flame Graphs [14 ]的一部分,可以可视化来自perf的痕迹
 +
:作为Flame Graphs<ref>http://www.brendangregg.com/flamegraphs.html</ref>的一部分, 用来可视化perf的trace数据分析.
 +
 +
[[File:Perf_002.png]]
 +
:火焰图是使用火焰图工具套件生成<ref>https://github.com/brendangregg/FlameGraph</ref>
 +
 +
*在主机端安装Flame Graph工具套件。
 +
<syntaxhighlight lang="bash">
 +
PC $> cd <your_local_path>
 +
PC $> git clone https://github.com/brendangregg/FlameGraph.git
 +
PC $> cd FlameGraph
 +
</syntaxhighlight>
 +
*从perf tool生成火焰图
 +
:生成性能记录时,必须添加-g选项。
 +
 +
生成火焰图并查看示例:
 +
<syntaxhighlight lang="bash">
 +
- 在板上执行perf record命令
 +
Board $> perf record -a -g top
 +
Board $> perf script > perf_top.out
 +
 +
- 复制perf_top.out到主机PC中(即在FlameGraph目录中)
 +
- 使用stackcollapse-perf.pl脚本在主机PC端执行火焰图生成命令。
 +
PC $> ./stackcollapse-perf.pl perf_top.out > out.top_folded
 +
 +
- 使用flamegraph.pl渲染SVG(可缩放矢量图形)文件。
 +
PC $> ./flamegraph.pl out.top_folded > top.svg
 +
 +
- 例如使用网络浏览器查看SVG
 +
PC $> firefox top.svg
 +
 +
</syntaxhighlight>
 +
  
 
=参考=
 
=参考=
 
<references />
 
<references />
 
[[Category:Linux_Operating_System]][[Category:Linux_monitoring_tools]][[Category:Perf]]
 
[[Category:Linux_Operating_System]][[Category:Linux_monitoring_tools]][[Category:Perf]]

2019年12月12日 (四) 16:00的最新版本

简介

perf[1]是Linux用户空间工具,它可以获取系统性能数据

如何安装

手动编译安装

BuildRoot下安装

Yocto下安装

RRECOMMENDS_${PN} = "\
   ${PERF} \
   trace-cmd \
   blktrace \
   ${PROFILE_TOOLS_X} \
   ${PROFILE_TOOLS_SYSTEMD} \
   "
...
PERF = "perf"

安卓系统下使用

simpleperf[2]与perf等效,并且默认情况下已安装(/system/xbin/simpleperf),可以与所有适用于Android™的软件包一起使用
它支持较少的选项:
Board $> simpleperf --help
Usage: simpleperf [common options] subcommand [args_for_subcommand]
common options:
    -h/--help     Print this help information.
    --log <severity> Set the minimum severity of logging. Possible severities
                       include verbose, debug, warning, info, error, fatal.
                       Default is info.
    --version     Print version of simpleperf.
subcommands:
    debug-unwind        Debug/test offline unwinding.
    dump                dump perf record file
    help                print help information for simpleperf
    kmem                collect kernel memory allocation information
    list                list available event types
    record              record sampling info in perf.data
    report              report sampling information in perf.data
    report-sample       report raw sample information in perf.data
    stat                gather performance counter information

开始使用

Board $> which perf
/usr/bin/perf

Perf命令

usage: perf [--version] [--help] [OPTIONS] COMMAND [ARGS]

The most commonly used perf commands are:
  annotate        Reads perf.data (created by perf record) and displays annotated code
  archive         Creates archive with object files with build-ids found in perf.data file
  bench           General framework for benchmark suites
  buildid-cache   Manages build-id cache.
  buildid-list    Lists the buildids in a perf.data file
  c2c             Shared Data C2C/HITM Analyzer.
  config          Gets and sets variables in a configuration file.
  data            Data file related processing
  diff            Reads perf.data files and displays the differential profile
  evlist          Lists the event names in a perf.data file
  ftrace          simple wrapper for kernel's ftrace functionality
  inject          Filters to augment the events stream with additional information
  kallsyms        Searches running kernel for symbols
  kmem            Tool to trace/measure kernel memory properties
  kvm             Tool to trace/measure kvm guest os
  list            Lists all symbolic event types
  lock            Analyzes lock events
  mem             Profiles memory accesses
  record          Runs a command and records its profile into perf.data
  report          Reads perf.data (created by perf record) and displays the profile
  sched           Tool to trace/measure scheduler properties (latencies)
  script          Reads perf.data (created by perf record) and displays trace output
  stat            Runs a command and gathers performance counter statistics
  test            Runs sanity tests.
  timechart       Tool to visualize total system behavior during a workload
  top             System profiling tool.
  probe           Defines new dynamic tracepoints

See 'perf COMMAND -h' for more information on a specific command.
  • perf top(Linux内核文档[3]):通过计算循环事件数来提供CPU负载;默认顺序是每个符号的采样数降序:
Board $> perf top
 40.62%  [kernel]                           [k] v7_dma_inv_range
 18.65%  [kernel]                           [k] _raw_spin_unlock_irqrestore
 17.01%  [kernel]                           [k] arch_cpu_idle
  8.27%  [kernel]                           [k] v7_dma_clean_range
  5.00%  [kernel]                           [k] rcu_idle_exit
  1.70%  [kernel]                           [k] cpu_startup_entry
  0.52%  [kernel]                           [k] trace_graph_return
  0.48%  [kernel]                           [k] finish_task_switch
  0.48%  libc-2.18.so                       [.] memcpy
  0.47%  [kernel]                           [k] trace_graph_entry
意味着CPU在功能v7_dma_inv_range中花费了40%的时间,在_raw_spin_unlock_irqrestore中花费了18.65%的时间。
有关更多信息和示例,请访问perf.wiki.kernel.org[4]
也可以按指定的排序显示结果:
sage: perf top [<options>]
   -s, --sort <key[,key2...]>
              sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline, ... Please refer to the main page for the complete list.
  • perf stat(linux kernel documentation[5]): 获取事件计数
Board $> perf stat hello_world_example
User space example: hello world from STMicroelectronics
10 9 8 7 6 5 4 3 2 1 0 
User space example: goodbye from STMicroelectronics
Performance counter stats for 'hello_world_example':

         4.328249      task-clock (msec)         #    0.000 CPUs utilized          
               11      context-switches          #    0.003 M/sec                  
                0      cpu-migrations            #    0.000 K/sec                  
               38      page-faults               #    0.009 M/sec                  
          2710036      cycles                    #    0.626 GHz                    
           640856      instructions              #    0.24  insn per cycle         
            75644      branches                  #   17.477 M/sec                  
            21764      branch-misses             #   28.77% of all branches        

     11.109859338 seconds time elapsed
有关更多信息和示例,请参见perf.wiki.kernel.org[6]
  • perf list (Linux kernel documentation[7]): 支持的符号事件类型.
Board $> perf list
 branch-instructions OR branches                    [Hardware event]
 branch-misses                                      [Hardware event]
 bus-cycles                                         [Hardware event]
 cache-misses                                       [Hardware event]
 cache-references                                   [Hardware event]
 cpu-cycles OR cycles                               [Hardware event]
 instructions                                       [Hardware event]
 alignment-faults                                   [Software event]
 bpf-output                                         [Software event]
 context-switches OR cs                             [Software event]
 cpu-clock                                          [Software event]
 cpu-migrations OR migrations                       [Software event]
 dummy                                              [Software event]
 emulation-faults                                   [Software event]
 major-faults                                       [Software event]
 minor-faults                                       [Software event]
 page-faults OR faults                              [Software event]
 task-clock                                         [Software event]
 L1-dcache-load-misses                              [Hardware cache event]
 L1-dcache-loads                                    [Hardware cache event]
 L1-dcache-store-misses                             [Hardware cache event]
 L1-dcache-stores                                   [Hardware cache event]
 L1-icache-load-misses                              [Hardware cache event]
 L1-icache-loads                                    [Hardware cache event]
 LLC-load-misses                                    [Hardware cache event]
 LLC-loads                                          [Hardware cache event]
 LLC-store-misses                                   [Hardware cache event]
 LLC-stores                                         [Hardware cache event]
 branch-load-misses                                 [Hardware cache event]
 branch-loads                                       [Hardware cache event]
 dTLB-load-misses                                   [Hardware cache event]
 dTLB-store-misses                                  [Hardware cache event]
 iTLB-load-misses                                   [Hardware cache event]
 armv7_cortex_a7/br_immed_retired/                  [Kernel PMU event]
 armv7_cortex_a7/br_mis_pred/                       [Kernel PMU event]
 armv7_cortex_a7/br_pred/                           [Kernel PMU event]
 armv7_cortex_a7/br_return_retired/                 [Kernel PMU event]
 armv7_cortex_a7/bus_access/                        [Kernel PMU event]
 armv7_cortex_a7/bus_cycles/                        [Kernel PMU event]
 armv7_cortex_a7/cid_write_retired/                 [Kernel PMU event]
 armv7_cortex_a7/cpu_cycles/                        [Kernel PMU event]
 armv7_cortex_a7/exc_return/                        [Kernel PMU event]
 armv7_cortex_a7/exc_taken/                         [Kernel PMU event]
 armv7_cortex_a7/inst_retired/                      [Kernel PMU event]
 armv7_cortex_a7/inst_spec/                         [Kernel PMU event]
 armv7_cortex_a7/l1d_cache/                         [Kernel PMU event]
 armv7_cortex_a7/l1d_cache_refill/                  [Kernel PMU event]
 armv7_cortex_a7/l1d_cache_wb/                      [Kernel PMU event]
 armv7_cortex_a7/l1d_tlb_refill/                    [Kernel PMU event]
 armv7_cortex_a7/l1i_cache/                         [Kernel PMU event]
 armv7_cortex_a7/l1i_cache_refill/                  [Kernel PMU event]
 armv7_cortex_a7/l1i_tlb_refill/                    [Kernel PMU event]
 armv7_cortex_a7/l2d_cache/                         [Kernel PMU event]
 armv7_cortex_a7/l2d_cache_refill/                  [Kernel PMU event]
 armv7_cortex_a7/l2d_cache_wb/                      [Kernel PMU event]
 armv7_cortex_a7/ld_retired/                        [Kernel PMU event]
 armv7_cortex_a7/mem_access/                        [Kernel PMU event]
 armv7_cortex_a7/memory_error/                      [Kernel PMU event]
 armv7_cortex_a7/pc_write_retired/                  [Kernel PMU event]
 armv7_cortex_a7/st_retired/                        [Kernel PMU event]
 armv7_cortex_a7/sw_incr/                           [Kernel PMU event]
 armv7_cortex_a7/ttbr_write_retired/                [Kernel PMU event]
 armv7_cortex_a7/unaligned_ldst_retired/            [Kernel PMU event]
 rNNN                                               [Raw hardware event descriptor]
 cpu/t1=v1[,t2=v2,t3 ...]/modifier                  [Raw hardware event descriptor]
 mem:<addr>[/len][:access]                          [Hardware breakpoint]
 alarmtimer:alarmtimer_cancel                       [Tracepoint event]
 alarmtimer:alarmtimer_fired                        [Tracepoint event]
 alarmtimer:alarmtimer_start                        [Tracepoint event]
 alarmtimer:alarmtimer_suspend                      [Tracepoint event]
 asoc:snd_soc_bias_level_done                       [Tracepoint event]
 asoc:snd_soc_bias_level_start                      [Tracepoint event]
 asoc:snd_soc_dapm_connected                        [Tracepoint event]
 asoc:snd_soc_dapm_done                             [Tracepoint event]
 asoc:snd_soc_dapm_path                             [Tracepoint event]
 asoc:snd_soc_dapm_start                            [Tracepoint event]
 asoc:snd_soc_dapm_walk_done                        [Tracepoint event]
 asoc:snd_soc_dapm_widget_event_done                [Tracepoint event]
 asoc:snd_soc_dapm_widget_event_start               [Tracepoint event]
...
 xhci-hcd:xhci_inc_enq                              [Tracepoint event]
 xhci-hcd:xhci_queue_trb                            [Tracepoint event]
 xhci-hcd:xhci_ring_alloc                           [Tracepoint event]
 xhci-hcd:xhci_ring_expansion                       [Tracepoint event]
 xhci-hcd:xhci_ring_free                            [Tracepoint event]
 xhci-hcd:xhci_setup_addressable_virt_device        [Tracepoint event]
 xhci-hcd:xhci_setup_device                         [Tracepoint event]
 xhci-hcd:xhci_setup_device_slot                    [Tracepoint event]
 xhci-hcd:xhci_stop_device                          [Tracepoint event]
 xhci-hcd:xhci_urb_dequeue                          [Tracepoint event]
 xhci-hcd:xhci_urb_enqueue                          [Tracepoint event]
 xhci-hcd:xhci_urb_giveback                         [Tracepoint event]
  • perf record (Linux kernel documentation[8]): 记录事件以供以后报告
Board $> perf record hello_world_example 

User space example: hello world from STMicroelectronics
10 9 8 7 6 5 4 3 2 1 0 
User space example: goodbye from STMicroelectronics
[ perf record: Woken up 1 time to write data ]
[ perf record: Captured and wrote 0.004 MB perf.data (28 samples) ]
这可以过滤事件(由perf list命令提供)。有关更多信息,选项和示例,请参见perf.wiki.kernel.org[9]
默认情况下,事件记录在perf.data文件中。如果要指定另一个输出文件名,则必须添加-o,--output <file>选项。
  • perf report (Linux kernel documentation[10]): 按进程,功能等细分事件。.
Example after previous command "perf record hello_world_example"
Board $> perf report
Samples: 28  of event 'cycles:ppp', Event count (approx.):2737925                
Overhead  Command          Shared Object Symbol                   
  12.66%  hello_world_exa  ld-2.26.so         [.] _dl_relocate_object
  11.71%  hello_world_exa  [kernel.kallsyms]  [k] filemap_map_pages
  10.65%  hello_world_exa  [kernel.kallsyms]  [k] n_tty_write
   6.43%  hello_world_exa  [kernel.kallsyms]  [k] percpu_counter_add_batch
   6.43%  hello_world_exa  ld-2.26.so         [.] sbrk
   6.24%  hello_world_exa  [kernel.kallsyms]  [k] cpu_v7_set_pte_ext
   5.56%  hello_world_exa  [kernel.kallsyms]  [k] alloc_set_pte
   5.56%  hello_world_exa  libc-2.26.so       [.] __sbrk
   5.37%  hello_world_exa  [kernel.kallsyms]  [k] __vma_link_file
   5.32%  hello_world_exa  [kernel.kallsyms]  [k] __fput
   5.32%  hello_world_exa  [kernel.kallsyms]  [k] ldsem_up_read
   5.32%  hello_world_exa  [kernel.kallsyms]  [k] unmap_page_range
   5.32%  hello_world_exa  libc-2.26.so       [.] printf
   5.24%  hello_world_exa  [kernel.kallsyms]  [k] _raw_spin_unlock_irqrestore
   2.23%  hello_world_exa  [kernel.kallsyms]  [k] perf_event_mmap
   0.48%  hello_world_exa  [kernel.kallsyms]  [k] perf_output_begin
   0.13%  perf             [kernel.kallsyms]  [k] perf_event_exec

默认情况下,报告文件perf.data被读取为输入文件。如果要指定另一个输入文件名,则必须添加-i,--input <file> option。 有关更多信息和示例,请参见perf.wiki.kernel.org[11]

  • perf bench (Linux kernel documentation[12]): 运行不同的内核微基准测试:
# List of all available benchmark collections:

        sched: Scheduler and IPC benchmarks
          mem: Memory access benchmarks
        futex: Futex stressing benchmarks
          all: All benchmarks
Example of getting memcpy benchmark for 100MB:
Board $> perf bench mem memcpy --size 100MB
# Running 'mem/memcpy' benchmark:
# function 'default' (Default memcpy() provided by glibc)
# Copying 100MB bytes ...

       1.426138 GB/sec
有关更多信息和示例,请参见perf.wiki.kernel.org[13]


进阶使用

作为Flame Graphs [14 ]的一部分,可以可视化来自perf的痕迹

作为Flame Graphs[14]的一部分, 用来可视化perf的trace数据分析.

Perf 002.png

火焰图是使用火焰图工具套件生成[15]
  • 在主机端安装Flame Graph工具套件。
PC $> cd <your_local_path>
PC $> git clone https://github.com/brendangregg/FlameGraph.git
PC $> cd FlameGraph
  • 从perf tool生成火焰图
生成性能记录时,必须添加-g选项。

生成火焰图并查看示例:

- 在板上执行perf record命令 
Board $> perf record -a -g top
Board $> perf script > perf_top.out

- 复制perf_top.out到主机PC中(即在FlameGraph目录中)
- 使用stackcollapse-perf.pl脚本在主机PC端执行火焰图生成命令。
PC $> ./stackcollapse-perf.pl perf_top.out > out.top_folded

- 使用flamegraph.pl渲染SVG(可缩放矢量图形)文件。
PC $> ./flamegraph.pl out.top_folded > top.svg

- 例如使用网络浏览器查看SVG
PC $> firefox top.svg