| title | date | tags |
|---|---|---|
Linux 性能调优 |
2023-02-02 11:08:04 -0800 |
$ perf
usage: perf [--version] [--help] [OPTIONS] COMMAND [ARGS]
The most commonly used perf commands are:
annotate Read perf.data (created by perf record) and display annotated code
archive Create archive with object files with build-ids found in perf.data file
bench General framework for benchmark suites
buildid-cache Manage build-id cache.
buildid-list List the buildids in a perf.data file
c2c Shared Data C2C/HITM Analyzer.
config Get and set variables in a configuration file.
data Data file related processing
diff Read perf.data files and display the differential profile
evlist List the event names in a perf.data file
ftrace simple wrapper for kernel's ftrace functionality
inject Filter 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 List all symbolic event types
lock Analyze lock events
mem Profile memory accesses
record Run a command and record its profile into perf.data
report Read perf.data (created by perf record) and display the profile
sched Tool to trace/measure scheduler properties (latencies)
script Read perf.data (created by perf record) and display trace output
stat Run a command and gather performance counter statistics
test Runs sanity tests.
timechart Tool to visualize total system behavior during a workload
top System profiling tool.
version display the version of perf binary
probe Define new dynamic tracepoints
trace strace inspired tool
perf list: 产看 perf的 events
perf top
perf record -a -g -p [pid]/perf report -g
`perf stat -p [pid]`` 开始了, 跑程序, 然后 ctrl+c
59,665.07 msec task-clock # 25.719 CPUs utilized
29,686 context-switches # 0.498 K/sec
4,043 cpu-migrations # 0.068 K/sec
904,315 page-faults # 0.015 M/sec
182,466,072,426 cycles # 3.058 GHz (83.06%)
48,393,560,647 stalled-cycles-frontend # 26.52% frontend cycles idle (83.27%)
41,766,561,738 stalled-cycles-backend # 22.89% backend cycles idle (83.35%)
127,710,645,250 instructions # 0.70 insn per cycle
# 0.38 stalled cycles per insn (83.58%)
24,547,226,083 branches # 411.417 M/sec (83.73%)
693,011,495 branch-misses # 2.82% of all branches (83.00%)
2.319854477 seconds time elapsedperf stat -e L1-dcache-load-misses -e L1-dcache-load -p
perf stat -e L1-icache-load-misses -e L1-icache-load -p [pid]
Performance counter stats for process id 'xxx':
22,036,023 L1-icache-load-misses # 1.03% of all L1-icache hits
2,132,515,373 L1-icache-load
2.590468195 seconds time elapsed#perf record -e cpu-clock -g -p $pid -- sleep $sec
perf record -F 99 -e cpu-clock -g -p $pid --call-graph dwarf -- sleep $sec
perf script > out.perf
/code/FlameGraph/stackcollapse-perf.pl out.perf > out.folded
/code/FlameGraph/flamegraph.pl out.folded > $file_name.svgnumactl --cpubind=0 绑定核数