perf性能瓶颈分析小试牛刀
文档
https://perf.wiki.kernel.org/index.php/Tutorial#Live_analysis_with_perf_top
内核配置
CONFIG_PERF_EVENTS=y
编译
make -C tools/perf
Auto-detecting system features:
... dwarf: [ OFF ]
... dwarf_getlocations: [ OFF ]
... glibc: [ on ]
... gtk2: [ OFF ]
... libaudit: [ OFF ]
... libbfd: [ on ]
... libelf: [ on ]
... libnuma: [ OFF ]
... numa_num_possible_cpus: [ OFF ]
... libperl: [ OFF ]
... libpython: [ OFF ]
... libslang: [ OFF ]
... libcrypto: [ OFF ]
... libunwind: [ on ]
... libdw-dwarf-unwind: [ OFF ]
... zlib: [ on ]
... lzma: [ OFF ]
... get_cpuid: [ OFF ]
... bpf: [ on ]
文件系统一定要添加libunwind,否则 perf 无法找到函数名,只能显示一些符号地址。
应用实例
lvgl 应用界面卡顿,简单的用 perf top 分析哪些函数消耗cpu资源较大,进一步优化应用程序
perf top
PerfTop: 630 irqs/sec kernel:90.8% exact: 0.0% [4000Hz cycles:ppp], (all, 4 CPUs)
-------------------------------------------------------------------------------
19.15% lvgl_launcher [.] lodepng_inflatev.isra.11
12.19% libc-2.23.so [.] memcpy
7.77% lvgl_launcher [.] fbdev_flush
6.41% [kernel] [k] cpuidle_enter_state
3.74% [kernel] [k] _raw_spin_unlock_irqrestore
3.73% lvgl_launcher [.] _lv_blend_map
3.49% lvgl_launcher [.] lodepng_zlib_decompress
3.46% lvgl_launcher [.] decodeGeneric
3.45% lvgl_launcher [.] unfilter
3.38% [kernel] [k] __slab_alloc.isra.19.constprop.23
2.77% lvgl_launcher [.] _lv_blend_fill
2.54% lvgl_launcher [.] lv_refr_vdb_flush
2.37% [kernel] [k] _raw_spin_unlock_irq
1.86% lvgl_launcher [.] lv_draw_label
1.85% lvgl_launcher [.] lv_draw_map
1.35% lvgl_launcher [.] decoder_open
1.28% lvgl_launcher [.] fread@plt
1.08% lvgl_launcher [.] lv_color_fill
1.00% [kernel] [k] __softirqentry_text_start
0.92% [kernel] [k] tick_nohz_idle_enter
perf top -p PID 单独分析lvgl_launcher进程
PerfTop: 3972 irqs/sec kernel: 3.8% exact: 0.0% [4000Hz cycles:ppp], (target_pid: 2766)
-------------------------------------------------------------------------------
33.50% lvgl_launcher [.] lodepng_inflatev.isra.11
12.89% libc-2.23.so [.] memcpy
6.92% lvgl_launcher [.] fbdev_flush
6.59% lvgl_launcher [.] _lv_blend_map
6.12% lvgl_launcher [.] lodepng_zlib_decompress
6.12% lvgl_launcher [.] unfilter
6.05% lvgl_launcher [.] decodeGeneric
3.15% lvgl_launcher [.] lv_draw_map
2.81% lvgl_launcher [.] _lv_blend_fill
2.33% lvgl_launcher [.] lv_refr_vdb_flush
2.32% lvgl_launcher [.] decoder_open
1.90% lvgl_launcher [.] lv_draw_label
1.30% lvgl_launcher [.] fread@plt
1.27% lvgl_launcher [.] lv_color_fill
0.68% lvgl_launcher [.] HuffmanTree_makeFromLengths2
0.68% [kernel] [k] __slab_alloc.isra.19.constprop.23
0.42% [kernel] [k] _raw_spin_unlock_irqrestore
0.28% [kernel] [k] __arch_copy_to_user
0.28% [kernel] [k] __softirqentry_text_start
0.19% [kernel] [k] clear_page
perf top -g -p PID 打印调用栈
99.92% 0.00% lvgl_launcher [.] _start
|
---_start
__libc_start_main
|
--99.91%--main
|
--99.82%--lv_task_handler
|
--99.64%--_lv_disp_refr_task
|
--99.62%--lv_refr_area_part
|
|--71.73%--lv_refr_obj_and_children
| lv_refr_obj.part.3
| |
| |--70.50%--lv_refr_obj.part.3
| | |
| | |--58.45%--lv_refr_obj.part.3
| | | |
| | | |--40.61%--lv_img_design
| | | | |
| | | | --40.53%--lv_draw_img
| | | | |
| | | | |--36.02%--_lv_img_cache_open
| | | | | |
| | | | | --35.86%--lv_img_decoder_open
| | | | | |
| | | | | --35.56%--decoder_open
| | | | | |
| | | | | --33.89%--lodepng_decode_memory
| | | | | |
| | | | | --33.75%--decodeGeneric
| | | | | |
| | | | | |--27.36%--lodepng_zlib_decompress
| | | | | | |
| | | | | | |--22.28%--lodepng_inflatev.isra.11
| | | | | | |
| | | | | | --1.85%--memcpy
| | | | | |
| | | | | --3.19%--postProcessScanlines
| | | | | unfilter
| | | | |
| | | | --4.49%--lv_draw_map
| | | | |
| | | | --2.75%--_lv_blend_map