profiler-md 0.6.0 → 0.8.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/cli/index.mjs +12953 -4258
- package/dist/index.d.mts +1182 -435
- package/dist/index.js +2 -2
- package/docs/api.md +138 -0
- package/docs/formats/callgrind.md +25 -0
- package/docs/formats/collapsed.md +5 -3
- package/docs/formats/ghc-eventlog.md +20 -0
- package/docs/formats/ghc-json-profile.md +23 -0
- package/docs/formats/hprof.md +21 -0
- package/docs/formats/jfr.md +2 -2
- package/docs/formats/jsc-heap-snapshot.md +10 -4
- package/docs/formats/memray.md +30 -0
- package/docs/formats/perf.md +74 -0
- package/docs/formats/pprof.md +4 -2
- package/docs/formats/speedscope.md +3 -2
- package/docs/formats/systing.md +17 -15
- package/docs/formats/v8-cpu-profile.md +4 -3
- package/docs/formats/v8-heap-profile.md +8 -5
- package/docs/formats/v8-heap-snapshot.md +8 -3
- package/docs/formats/webkit-timeline-recording.md +6 -0
- package/docs/languages/c.md +119 -32
- package/docs/languages/csharp.md +5 -3
- package/docs/languages/elixir.md +19 -11
- package/docs/languages/fortran.md +141 -0
- package/docs/languages/go.md +33 -10
- package/docs/languages/haskell.md +134 -0
- package/docs/languages/java.md +81 -26
- package/docs/languages/javascript.md +43 -25
- package/docs/languages/julia.md +21 -18
- package/docs/languages/php.md +19 -0
- package/docs/languages/python.md +105 -14
- package/docs/languages/ruby.md +15 -16
- package/docs/languages/rust.md +11 -10
- package/docs/languages/swift.md +135 -0
- package/docs/languages/zig.md +155 -0
- package/package.json +42 -42
- package/readme.md +302 -259
- package/skills/profile-optimize/SKILL.md +4 -2
- package/examples/output/c.gperftools.cpu.base.pprof.md +0 -324
- package/examples/output/c.gperftools.cpu.current.pprof.md +0 -271
- package/examples/output/c.gperftools.cpu.diff.pprof.md +0 -98
- package/examples/output/c.gperftools.heap.base.pprof.md +0 -362
- package/examples/output/c.gperftools.heap.current.pprof.md +0 -362
- package/examples/output/c.gperftools.heap.diff.pprof.md +0 -28
- package/examples/output/c.systing.cpu.base.systing.md +0 -824
- package/examples/output/c.systing.cpu.current.systing.md +0 -825
- package/examples/output/c.systing.cpu.diff.systing.md +0 -238
- package/examples/output/cpp.gperftools.cpu.base.pprof.md +0 -557
- package/examples/output/cpp.gperftools.cpu.current.pprof.md +0 -573
- package/examples/output/cpp.gperftools.cpu.diff.pprof.md +0 -122
- package/examples/output/cpp.gperftools.heap.base.pprof.md +0 -67
- package/examples/output/cpp.gperftools.heap.current.pprof.md +0 -67
- package/examples/output/cpp.gperftools.heap.diff.pprof.md +0 -19
- package/examples/output/csharp.dotnet-trace.base.speedscope.json.md +0 -411
- package/examples/output/csharp.dotnet-trace.current.speedscope.json.md +0 -412
- package/examples/output/csharp.dotnet-trace.diff.speedscope.json.md +0 -122
- package/examples/output/elixir.eflambe.cpu.base.collapsed.md +0 -399
- package/examples/output/elixir.eflambe.cpu.current.collapsed.md +0 -398
- package/examples/output/elixir.eflambe.cpu.diff.collapsed.md +0 -92
- package/examples/output/erlang.eflambe.cpu.base.collapsed.md +0 -147
- package/examples/output/erlang.eflambe.cpu.current.collapsed.md +0 -147
- package/examples/output/erlang.eflambe.cpu.diff.collapsed.md +0 -18
- package/examples/output/fsharp.dotnet-trace.base.speedscope.json.md +0 -324
- package/examples/output/fsharp.dotnet-trace.current.speedscope.json.md +0 -314
- package/examples/output/fsharp.dotnet-trace.diff.speedscope.json.md +0 -119
- package/examples/output/go.pprof.block.base.pprof.md +0 -86
- package/examples/output/go.pprof.block.current.pprof.md +0 -83
- package/examples/output/go.pprof.block.diff.pprof.md +0 -42
- package/examples/output/go.pprof.cpu.base.pprof.md +0 -81
- package/examples/output/go.pprof.cpu.current.pprof.md +0 -70
- package/examples/output/go.pprof.cpu.diff.pprof.md +0 -52
- package/examples/output/go.pprof.goroutine.base.pprof.md +0 -43
- package/examples/output/go.pprof.goroutine.current.pprof.md +0 -43
- package/examples/output/go.pprof.goroutine.diff.pprof.md +0 -13
- package/examples/output/go.pprof.heap-alloc.base.pprof.md +0 -154
- package/examples/output/go.pprof.heap-alloc.current.pprof.md +0 -146
- package/examples/output/go.pprof.heap-alloc.diff.pprof.md +0 -67
- package/examples/output/go.pprof.heap.base.pprof.md +0 -154
- package/examples/output/go.pprof.heap.current.pprof.md +0 -146
- package/examples/output/go.pprof.heap.diff.pprof.md +0 -67
- package/examples/output/go.pprof.mutex.base.pprof.md +0 -55
- package/examples/output/go.pprof.mutex.current.pprof.md +0 -41
- package/examples/output/go.pprof.mutex.diff.pprof.md +0 -31
- package/examples/output/go.pprof.threadcreate.base.pprof.md +0 -55
- package/examples/output/go.pprof.threadcreate.current.pprof.md +0 -55
- package/examples/output/go.pprof.threadcreate.diff.pprof.md +0 -19
- package/examples/output/java.async-profiler.all.base.jfr.md +0 -1071
- package/examples/output/java.async-profiler.all.current.jfr.md +0 -1007
- package/examples/output/java.async-profiler.all.diff.jfr.md +0 -268
- package/examples/output/java.async-profiler.alloc.base.jfr.md +0 -425
- package/examples/output/java.async-profiler.alloc.current.jfr.md +0 -413
- package/examples/output/java.async-profiler.alloc.diff.jfr.md +0 -99
- package/examples/output/java.async-profiler.cpu.base.collapsed.md +0 -367
- package/examples/output/java.async-profiler.cpu.base.jfr.md +0 -485
- package/examples/output/java.async-profiler.cpu.current.collapsed.md +0 -355
- package/examples/output/java.async-profiler.cpu.current.jfr.md +0 -488
- package/examples/output/java.async-profiler.cpu.diff.collapsed.md +0 -117
- package/examples/output/java.async-profiler.cpu.diff.jfr.md +0 -118
- package/examples/output/java.async-profiler.live.base.jfr.md +0 -697
- package/examples/output/java.async-profiler.live.current.jfr.md +0 -684
- package/examples/output/java.async-profiler.live.diff.jfr.md +0 -191
- package/examples/output/java.async-profiler.lock.base.jfr.md +0 -140
- package/examples/output/java.async-profiler.lock.current.jfr.md +0 -108
- package/examples/output/java.async-profiler.lock.diff.jfr.md +0 -44
- package/examples/output/java.async-profiler.nativemem.base.jfr.md +0 -222
- package/examples/output/java.async-profiler.nativemem.current.jfr.md +0 -222
- package/examples/output/java.async-profiler.nativemem.diff.jfr.md +0 -65
- package/examples/output/java.async-profiler.wall.base.jfr.md +0 -400
- package/examples/output/java.async-profiler.wall.current.jfr.md +0 -420
- package/examples/output/java.async-profiler.wall.diff.jfr.md +0 -94
- package/examples/output/java.jdk.all.base.jfr.md +0 -1024
- package/examples/output/java.jdk.all.current.jfr.md +0 -1010
- package/examples/output/java.jdk.all.diff.jfr.md +0 -311
- package/examples/output/java.jdk.alloc.base.jfr.md +0 -1024
- package/examples/output/java.jdk.alloc.current.jfr.md +0 -1046
- package/examples/output/java.jdk.alloc.diff.jfr.md +0 -312
- package/examples/output/java.jdk.cpu.base.jfr.md +0 -1023
- package/examples/output/java.jdk.cpu.current.jfr.md +0 -1039
- package/examples/output/java.jdk.cpu.diff.jfr.md +0 -285
- package/examples/output/java.jdk.live.base.jfr.md +0 -1029
- package/examples/output/java.jdk.live.current.jfr.md +0 -1035
- package/examples/output/java.jdk.live.diff.jfr.md +0 -289
- package/examples/output/java.jdk.lock.base.jfr.md +0 -1040
- package/examples/output/java.jdk.lock.current.jfr.md +0 -1051
- package/examples/output/java.jdk.lock.diff.jfr.md +0 -339
- package/examples/output/javascript.bun.base.cpuprofile.md +0 -73
- package/examples/output/javascript.bun.base.heapsnapshot.md +0 -608
- package/examples/output/javascript.bun.base.jsc-heap-snapshot.json.md +0 -480
- package/examples/output/javascript.bun.current.cpuprofile.md +0 -73
- package/examples/output/javascript.bun.current.heapsnapshot.md +0 -608
- package/examples/output/javascript.bun.current.jsc-heap-snapshot.json.md +0 -480
- package/examples/output/javascript.bun.diff.cpuprofile.md +0 -54
- package/examples/output/javascript.bun.diff.heapsnapshot.md +0 -48
- package/examples/output/javascript.bun.diff.jsc-heap-snapshot.json.md +0 -62
- package/examples/output/javascript.chrome.base.cpuprofile.md +0 -163
- package/examples/output/javascript.chrome.base.heapprofile.md +0 -32
- package/examples/output/javascript.chrome.base.heapsnapshot.md +0 -649
- package/examples/output/javascript.chrome.current.cpuprofile.md +0 -149
- package/examples/output/javascript.chrome.current.heapprofile.md +0 -32
- package/examples/output/javascript.chrome.current.heapsnapshot.md +0 -647
- package/examples/output/javascript.chrome.diff.cpuprofile.md +0 -65
- package/examples/output/javascript.chrome.diff.heapprofile.md +0 -50
- package/examples/output/javascript.chrome.diff.heapsnapshot.md +0 -73
- package/examples/output/javascript.deno.base.cpuprofile.md +0 -122
- package/examples/output/javascript.deno.current.cpuprofile.md +0 -114
- package/examples/output/javascript.deno.diff.cpuprofile.md +0 -66
- package/examples/output/javascript.node.base.cpuprofile.md +0 -136
- package/examples/output/javascript.node.base.heapprofile.md +0 -56
- package/examples/output/javascript.node.base.heapsnapshot.md +0 -660
- package/examples/output/javascript.node.current.cpuprofile.md +0 -119
- package/examples/output/javascript.node.current.heapprofile.md +0 -52
- package/examples/output/javascript.node.current.heapsnapshot.md +0 -660
- package/examples/output/javascript.node.diff.cpuprofile.md +0 -70
- package/examples/output/javascript.node.diff.heapprofile.md +0 -34
- package/examples/output/javascript.node.diff.heapsnapshot.md +0 -39
- package/examples/output/javascript.pprof.cpu.base.pprof.md +0 -104
- package/examples/output/javascript.pprof.cpu.current.pprof.md +0 -120
- package/examples/output/javascript.pprof.cpu.diff.pprof.md +0 -63
- package/examples/output/javascript.pprof.heap.base.pprof.md +0 -86
- package/examples/output/javascript.pprof.heap.current.pprof.md +0 -57
- package/examples/output/javascript.pprof.heap.diff.pprof.md +0 -42
- package/examples/output/javascript.safari.base.webkit-timeline-recording.json.md +0 -388
- package/examples/output/javascript.safari.current.webkit-timeline-recording.json.md +0 -378
- package/examples/output/javascript.safari.diff.webkit-timeline-recording.json.md +0 -122
- package/examples/output/julia.pprof-jl.alloc.base.pprof.md +0 -124
- package/examples/output/julia.pprof-jl.alloc.current.pprof.md +0 -126
- package/examples/output/julia.pprof-jl.alloc.diff.pprof.md +0 -30
- package/examples/output/julia.pprof-jl.cpu.base.pprof.md +0 -177
- package/examples/output/julia.pprof-jl.cpu.current.pprof.md +0 -179
- package/examples/output/julia.pprof-jl.cpu.diff.pprof.md +0 -64
- package/examples/output/julia.pprof-jl.wall.base.pprof.md +0 -181
- package/examples/output/julia.pprof-jl.wall.current.pprof.md +0 -193
- package/examples/output/julia.pprof-jl.wall.diff.pprof.md +0 -64
- package/examples/output/julia.profile.base.heapsnapshot.md +0 -590
- package/examples/output/julia.profile.current.heapsnapshot.md +0 -590
- package/examples/output/julia.profile.diff.heapsnapshot.md +0 -288
- package/examples/output/kotlin.async-profiler.all.base.jfr.md +0 -920
- package/examples/output/kotlin.async-profiler.all.current.jfr.md +0 -925
- package/examples/output/kotlin.async-profiler.all.diff.jfr.md +0 -247
- package/examples/output/kotlin.async-profiler.alloc.base.jfr.md +0 -467
- package/examples/output/kotlin.async-profiler.alloc.current.jfr.md +0 -473
- package/examples/output/kotlin.async-profiler.alloc.diff.jfr.md +0 -122
- package/examples/output/kotlin.async-profiler.cpu.base.collapsed.md +0 -371
- package/examples/output/kotlin.async-profiler.cpu.base.jfr.md +0 -472
- package/examples/output/kotlin.async-profiler.cpu.current.collapsed.md +0 -370
- package/examples/output/kotlin.async-profiler.cpu.current.jfr.md +0 -445
- package/examples/output/kotlin.async-profiler.cpu.diff.collapsed.md +0 -124
- package/examples/output/kotlin.async-profiler.cpu.diff.jfr.md +0 -125
- package/examples/output/kotlin.async-profiler.live.base.jfr.md +0 -934
- package/examples/output/kotlin.async-profiler.live.current.jfr.md +0 -917
- package/examples/output/kotlin.async-profiler.live.diff.jfr.md +0 -245
- package/examples/output/kotlin.async-profiler.lock.base.jfr.md +0 -1
- package/examples/output/kotlin.async-profiler.lock.current.jfr.md +0 -1
- package/examples/output/kotlin.async-profiler.lock.diff.jfr.md +0 -1
- package/examples/output/kotlin.async-profiler.nativemem.base.jfr.md +0 -234
- package/examples/output/kotlin.async-profiler.nativemem.current.jfr.md +0 -234
- package/examples/output/kotlin.async-profiler.nativemem.diff.jfr.md +0 -65
- package/examples/output/kotlin.async-profiler.wall.base.jfr.md +0 -452
- package/examples/output/kotlin.async-profiler.wall.current.jfr.md +0 -457
- package/examples/output/kotlin.async-profiler.wall.diff.jfr.md +0 -125
- package/examples/output/kotlin.jdk.all.base.jfr.md +0 -1442
- package/examples/output/kotlin.jdk.all.current.jfr.md +0 -1343
- package/examples/output/kotlin.jdk.all.diff.jfr.md +0 -392
- package/examples/output/kotlin.jdk.alloc.base.jfr.md +0 -1418
- package/examples/output/kotlin.jdk.alloc.current.jfr.md +0 -1389
- package/examples/output/kotlin.jdk.alloc.diff.jfr.md +0 -383
- package/examples/output/kotlin.jdk.cpu.base.jfr.md +0 -1406
- package/examples/output/kotlin.jdk.cpu.current.jfr.md +0 -1374
- package/examples/output/kotlin.jdk.cpu.diff.jfr.md +0 -380
- package/examples/output/kotlin.jdk.live.base.jfr.md +0 -1378
- package/examples/output/kotlin.jdk.live.current.jfr.md +0 -1311
- package/examples/output/kotlin.jdk.live.diff.jfr.md +0 -381
- package/examples/output/kotlin.jdk.lock.base.jfr.md +0 -1427
- package/examples/output/kotlin.jdk.lock.current.jfr.md +0 -1374
- package/examples/output/kotlin.jdk.lock.diff.jfr.md +0 -386
- package/examples/output/php.excimer.wall.base.speedscope.json.md +0 -25
- package/examples/output/php.excimer.wall.current.speedscope.json.md +0 -25
- package/examples/output/php.excimer.wall.diff.speedscope.json.md +0 -29
- package/examples/output/python.py-spy.cpu.base.collapsed.md +0 -526
- package/examples/output/python.py-spy.cpu.base.speedscope.json.md +0 -37
- package/examples/output/python.py-spy.cpu.current.collapsed.md +0 -514
- package/examples/output/python.py-spy.cpu.current.speedscope.json.md +0 -37
- package/examples/output/python.py-spy.cpu.diff.collapsed.md +0 -122
- package/examples/output/python.py-spy.cpu.diff.speedscope.json.md +0 -31
- package/examples/output/python.py-spy.wall.base.collapsed.md +0 -527
- package/examples/output/python.py-spy.wall.current.collapsed.md +0 -518
- package/examples/output/python.py-spy.wall.diff.collapsed.md +0 -122
- package/examples/output/ruby.rbspy.cpu.base.collapsed.md +0 -90
- package/examples/output/ruby.rbspy.cpu.base.pprof.md +0 -105
- package/examples/output/ruby.rbspy.cpu.base.speedscope.json.md +0 -105
- package/examples/output/ruby.rbspy.cpu.current.collapsed.md +0 -90
- package/examples/output/ruby.rbspy.cpu.current.pprof.md +0 -115
- package/examples/output/ruby.rbspy.cpu.current.speedscope.json.md +0 -115
- package/examples/output/ruby.rbspy.cpu.diff.collapsed.md +0 -32
- package/examples/output/ruby.rbspy.cpu.diff.pprof.md +0 -44
- package/examples/output/ruby.rbspy.cpu.diff.speedscope.json.md +0 -44
- package/examples/output/rust.pprof-rs.cpu.base.pprof.md +0 -60
- package/examples/output/rust.pprof-rs.cpu.current.pprof.md +0 -60
- package/examples/output/rust.pprof-rs.cpu.diff.pprof.md +0 -24
package/docs/languages/c.md
CHANGED
|
@@ -1,7 +1,9 @@
|
|
|
1
1
|
# C/C++
|
|
2
2
|
|
|
3
3
|
C and C++ profiling uses [gperftools](https://github.com/gperftools/gperftools),
|
|
4
|
-
|
|
4
|
+
[Valgrind's Callgrind](https://valgrind.org/docs/manual/cl-manual.html),
|
|
5
|
+
[Linux `perf`](https://perfwiki.github.io/main/), or
|
|
6
|
+
[systing](https://github.com/josefbacik/systing).
|
|
5
7
|
|
|
6
8
|
## CPU profiling
|
|
7
9
|
|
|
@@ -28,7 +30,7 @@ pprof --proto ./program cpu.prof > cpu.pprof
|
|
|
28
30
|
|
|
29
31
|
| Variable | Default | Description |
|
|
30
32
|
| ---------------------- | ------- | -------------------------------------------------------------------------------- |
|
|
31
|
-
| `CPUPROFILE` | — | Output filename
|
|
33
|
+
| `CPUPROFILE` | — | Output filename that enables profiling when set |
|
|
32
34
|
| `CPUPROFILE_FREQUENCY` | `100` | Samples per second |
|
|
33
35
|
| `CPUPROFILE_REALTIME` | — | Use wall-clock time (`ITIMER_REAL`) instead of CPU time (`ITIMER_PROF`) when set |
|
|
34
36
|
| `CPUPROFILESIGNAL` | — | Signal number to toggle profiling on/off at runtime |
|
|
@@ -61,7 +63,7 @@ g++ -O2 -Wl,--no-as-needed -ltcmalloc -o program program.cpp # C++
|
|
|
61
63
|
### CLI
|
|
62
64
|
|
|
63
65
|
```sh
|
|
64
|
-
# Generate heap profiles (written as <prefix>.
|
|
66
|
+
# Generate heap profiles (written as <prefix>.0001.heap, .0002.heap, …)
|
|
65
67
|
HEAPPROFILE=heap.prof ./program
|
|
66
68
|
|
|
67
69
|
# Convert a heap profile to pprof protobuf format
|
|
@@ -72,13 +74,13 @@ pprof --proto ./program heap.prof.0001.heap > heap.pprof
|
|
|
72
74
|
|
|
73
75
|
| Variable | Default | Description |
|
|
74
76
|
| ---------------------------------- | ------------ | --------------------------------------------------------------------------------------------- |
|
|
75
|
-
| `HEAPPROFILE` | — | Output filename prefix
|
|
77
|
+
| `HEAPPROFILE` | — | Output filename prefix that enables heap profiling when set |
|
|
76
78
|
| `HEAP_PROFILE_ALLOCATION_INTERVAL` | `1073741824` | Dump a profile after every this many bytes allocated (default 1 GB) |
|
|
77
79
|
| `HEAP_PROFILE_INUSE_INTERVAL` | `104857600` | Dump a profile each time the in-use high-water mark grows by this many bytes (default 100 MB) |
|
|
78
80
|
| `HEAP_PROFILE_TIME_INTERVAL` | `0` | Dump a profile every this many seconds (disabled by default) |
|
|
79
81
|
| `HEAPPROFILESIGNAL` | — | Signal number to dump a profile on demand |
|
|
80
82
|
| `HEAP_PROFILE_MMAP` | `false` | Also profile `mmap`/`mremap`/`sbrk` calls |
|
|
81
|
-
| `HEAP_PROFILE_ONLY_MMAP` | `false` | Profile only `mmap`/`mremap`/`sbrk
|
|
83
|
+
| `HEAP_PROFILE_ONLY_MMAP` | `false` | Profile only `mmap`/`mremap`/`sbrk`, skipping `malloc`/`new` |
|
|
82
84
|
|
|
83
85
|
### Programmatic API
|
|
84
86
|
|
|
@@ -93,11 +95,50 @@ HeapProfilerDump("checkpoint"); // optional mid-run profile dump
|
|
|
93
95
|
HeapProfilerStop();
|
|
94
96
|
```
|
|
95
97
|
|
|
98
|
+
## Callgrind profiling
|
|
99
|
+
|
|
100
|
+
Counts every executed instruction and records the call graph. Exact instead of
|
|
101
|
+
sampled, but runs the program ~20–100× slower. Requires Linux or another
|
|
102
|
+
Valgrind-supported platform.
|
|
103
|
+
|
|
104
|
+
Build with debug info so functions resolve to source files and lines:
|
|
105
|
+
|
|
106
|
+
```sh
|
|
107
|
+
gcc -O2 -g -o program program.c # C
|
|
108
|
+
g++ -O2 -g -o program program.cpp # C++
|
|
109
|
+
```
|
|
110
|
+
|
|
111
|
+
### CLI
|
|
112
|
+
|
|
113
|
+
```sh
|
|
114
|
+
# Profile, writing the callgrind format directly
|
|
115
|
+
valgrind --tool=callgrind --callgrind-out-file=program.callgrind ./program
|
|
116
|
+
|
|
117
|
+
# Optional: also simulate CPU caches and branch prediction
|
|
118
|
+
valgrind --tool=callgrind --cache-sim=yes --branch-sim=yes \
|
|
119
|
+
--callgrind-out-file=program.callgrind ./program
|
|
120
|
+
```
|
|
121
|
+
|
|
122
|
+
#### Flags
|
|
123
|
+
|
|
124
|
+
| Flag | Default | Description |
|
|
125
|
+
| ---------------------- | ------------------ | ----------------------------------------------------------------------- |
|
|
126
|
+
| `--callgrind-out-file` | `callgrind.out.%p` | Output filename (`%p` expands to the PID) |
|
|
127
|
+
| `--cache-sim` | `no` | Also count simulated cache hits/misses |
|
|
128
|
+
| `--branch-sim` | `no` | Also count branch mispredictions |
|
|
129
|
+
| `--dump-instr` | `no` | Record costs per instruction instead of per line |
|
|
130
|
+
| `--separate-threads` | `no` | Write one profile per thread |
|
|
131
|
+
| `--instr-atstart` | `yes` | Set to `no` to skip startup, then toggle with `callgrind_control -i on` |
|
|
132
|
+
|
|
96
133
|
## Heap checking
|
|
97
134
|
|
|
98
135
|
Compares the heap state before and after a region of code to detect memory
|
|
99
136
|
leaks. Useful for confirming a code path frees everything it allocates.
|
|
100
137
|
|
|
138
|
+
gperftools 2.17 removed the heap leak checker. On 2.17 and later
|
|
139
|
+
`HeapLeakChecker` is an empty stub and `HEAPCHECK` does nothing, so use
|
|
140
|
+
AddressSanitizer's `LeakSanitizer` (`-fsanitize=address`) instead.
|
|
141
|
+
|
|
101
142
|
Link with `-ltcmalloc`:
|
|
102
143
|
|
|
103
144
|
```sh
|
|
@@ -122,7 +163,7 @@ HEAPCHECK=draconian ./program
|
|
|
122
163
|
|
|
123
164
|
| Variable | Default | Description |
|
|
124
165
|
| ----------- | ------- | ------------------------------------------------------------------------------------ |
|
|
125
|
-
| `HEAPCHECK` | — | Leak-checking mode (`minimal`, `normal`, `strict`, or `draconian`)
|
|
166
|
+
| `HEAPCHECK` | — | Leak-checking mode, enabled when set (`minimal`, `normal`, `strict`, or `draconian`) |
|
|
126
167
|
|
|
127
168
|
### Programmatic API
|
|
128
169
|
|
|
@@ -138,10 +179,21 @@ if (!checker.NoLeaks()) abort(); // fails if any net allocation leaked
|
|
|
138
179
|
|
|
139
180
|
## System profiling
|
|
140
181
|
|
|
141
|
-
|
|
142
|
-
|
|
143
|
-
and
|
|
144
|
-
|
|
182
|
+
Samples on-CPU stacks and records a stack each time a thread sleeps, across user
|
|
183
|
+
and kernel code. Useful for costs outside the process: off-CPU waits, syscall
|
|
184
|
+
time, and contention across the whole node.
|
|
185
|
+
|
|
186
|
+
[systing](https://github.com/josefbacik/systing) is a Linux eBPF profiler. It
|
|
187
|
+
needs root (BPF) and a kernel with BTF (`/sys/kernel/btf/vmlinux`).
|
|
188
|
+
|
|
189
|
+
Build what you profile with frame pointers so systing's unwinder can walk the
|
|
190
|
+
stack, and with symbols for readable frames:
|
|
191
|
+
|
|
192
|
+
```sh
|
|
193
|
+
gcc -O2 -g -fno-omit-frame-pointer -o program program.c
|
|
194
|
+
```
|
|
195
|
+
|
|
196
|
+
### CLI
|
|
145
197
|
|
|
146
198
|
```sh
|
|
147
199
|
# Record a command (and its children) for 30 seconds
|
|
@@ -151,22 +203,57 @@ sudo systing --duration 30 --output profile.systing -- ./program args
|
|
|
151
203
|
sudo systing --duration 30 --output profile.systing --pid <pid>
|
|
152
204
|
```
|
|
153
205
|
|
|
154
|
-
|
|
155
|
-
|
|
206
|
+
#### Flags
|
|
207
|
+
|
|
208
|
+
| Flag | Default | Description |
|
|
209
|
+
| --------------------------------- | ---------- | ---------------------------------------------------------------- |
|
|
210
|
+
| `--duration` | — | Recording duration in seconds |
|
|
211
|
+
| `--output` | `trace.pb` | Output path, where `.systing` selects the export this tool reads |
|
|
212
|
+
| `--sample-freq` | `1000` | CPU stack-sampling rate in Hz |
|
|
213
|
+
| `--no-sleep-stack-traces` | off | Skip uninterruptible-sleep stacks |
|
|
214
|
+
| `--no-interruptible-stack-traces` | off | Skip interruptible-sleep stacks |
|
|
215
|
+
|
|
216
|
+
## Linux perf profiling
|
|
217
|
+
|
|
218
|
+
Samples the call stack through the kernel's own `perf_event` counters, across
|
|
219
|
+
user and kernel code. Useful for finding CPU hot spots, including the ones in
|
|
220
|
+
the kernel.
|
|
221
|
+
|
|
222
|
+
[Linux `perf`](https://perfwiki.github.io/main/) samples any native program and
|
|
223
|
+
writes `perf.data` itself. It needs a Linux kernel and a
|
|
224
|
+
`/proc/sys/kernel/perf_event_paranoid` low enough to permit sampling (`1` or
|
|
225
|
+
below for kernel stacks).
|
|
226
|
+
|
|
227
|
+
Build with frame pointers so perf's unwinder can walk the stack:
|
|
156
228
|
|
|
157
229
|
```sh
|
|
158
|
-
gcc -O2 -g -fno-omit-frame-pointer -o program program.c
|
|
230
|
+
gcc -O2 -g -fno-omit-frame-pointer -o program program.c # C
|
|
231
|
+
g++ -O2 -g -fno-omit-frame-pointer -o program program.cpp # C++
|
|
232
|
+
```
|
|
233
|
+
|
|
234
|
+
### CLI
|
|
235
|
+
|
|
236
|
+
```sh
|
|
237
|
+
# Sample a command at 999 Hz, recording call chains
|
|
238
|
+
perf record -F 999 -g -o perf.data -- ./program args
|
|
239
|
+
|
|
240
|
+
# Or sample a running process for 30 seconds
|
|
241
|
+
perf record -F 999 -g -o perf.data --pid <pid> -- sleep 30
|
|
159
242
|
```
|
|
160
243
|
|
|
161
|
-
|
|
244
|
+
#### Flags
|
|
245
|
+
|
|
246
|
+
| Flag | Default | Description |
|
|
247
|
+
| -------------- | ----------- | -------------------------------------------------------------------- |
|
|
248
|
+
| `-F` | `4000` | Samples per second |
|
|
249
|
+
| `-e` | `cycles` | Event to sample (`cpu-clock` where no hardware counter is available) |
|
|
250
|
+
| `-g` | off | Record call chains |
|
|
251
|
+
| `--call-graph` | `fp` | How to unwind: `fp`, `dwarf` (larger files), or `lbr` |
|
|
252
|
+
| `-o` | `perf.data` | Output path |
|
|
253
|
+
| `-a` | off | Sample every CPU on the system rather than one command |
|
|
162
254
|
|
|
163
|
-
|
|
164
|
-
|
|
165
|
-
| `--duration` | — | Recording duration in seconds |
|
|
166
|
-
| `--output` | `trace.pb` | Output path; `.systing` selects the profile export this tool reads |
|
|
167
|
-
| `--sample-freq` | `1000` | CPU stack-sampling rate in Hz |
|
|
168
|
-
| `--no-sleep-stack-traces` | off | Skip uninterruptible-sleep stacks |
|
|
169
|
-
| `--no-interruptible-stack-traces` | off | Skip interruptible-sleep stacks |
|
|
255
|
+
A `perf.data` file holds addresses rather than function names. See
|
|
256
|
+
`profiler-md --help perf` for how frames are named and how to symbolize them.
|
|
170
257
|
|
|
171
258
|
## Tips
|
|
172
259
|
|
|
@@ -174,20 +261,20 @@ gcc -O2 -g -fno-omit-frame-pointer -o program program.c
|
|
|
174
261
|
|
|
175
262
|
gperftools symbolizes addresses from the binary's and its shared libraries'
|
|
176
263
|
symbol tables, so build and link what you profile with symbols (`-g`, not
|
|
177
|
-
stripped). If the hot code
|
|
178
|
-
|
|
264
|
+
stripped). If the hot code is in a stripped shared library, statically link an
|
|
265
|
+
unstripped build instead.
|
|
179
266
|
|
|
180
267
|
### Linking
|
|
181
268
|
|
|
182
|
-
Modern toolchains link with `--as-needed` by default,
|
|
183
|
-
`-lprofiler`/`-ltcmalloc` when
|
|
184
|
-
|
|
185
|
-
|
|
186
|
-
|
|
187
|
-
|
|
188
|
-
`-lprofiler`/`-ltcmalloc` works there.
|
|
269
|
+
Modern toolchains link with `--as-needed` by default, so the linker drops
|
|
270
|
+
`-lprofiler`/`-ltcmalloc` when the program references no symbol from them. That
|
|
271
|
+
is the case when you enable profiling through environment variables instead of
|
|
272
|
+
the programmatic API. Force the link with `-Wl,--no-as-needed`, or preload the
|
|
273
|
+
library at runtime with `LD_PRELOAD`. The programmatic API (`ProfilerStart`,
|
|
274
|
+
etc.) references the symbols, so plain `-lprofiler`/`-ltcmalloc` works there.
|
|
189
275
|
|
|
190
276
|
### Converting
|
|
191
277
|
|
|
192
|
-
`pprof --proto` requires the Go `pprof`. gperftools
|
|
193
|
-
`pprof`/`google-pprof`
|
|
278
|
+
`pprof --proto` requires the Go `pprof`. gperftools stopped shipping a `pprof`
|
|
279
|
+
of its own in 2.17, and the Perl `pprof`/`google-pprof` that earlier releases
|
|
280
|
+
and the distro packages built from them install has no protobuf output.
|
package/docs/languages/csharp.md
CHANGED
|
@@ -26,8 +26,8 @@ stacks at roughly 100 Hz.
|
|
|
26
26
|
# Attach to a running process, written directly as speedscope
|
|
27
27
|
dotnet-trace collect --process-id <pid> --format speedscope
|
|
28
28
|
|
|
29
|
-
# Launch and trace a command from startup (
|
|
30
|
-
#
|
|
29
|
+
# Launch and trace a command from startup (`dotnet exec` traces your app rather
|
|
30
|
+
# than the dotnet CLI's own child processes)
|
|
31
31
|
dotnet-trace collect --format speedscope -- dotnet exec app.dll
|
|
32
32
|
|
|
33
33
|
# Attach for a fixed duration (dd:hh:mm:ss)
|
|
@@ -44,7 +44,9 @@ dotnet-trace convert trace.nettrace --format speedscope
|
|
|
44
44
|
| `-p` / `--process-id` | — | PID of the process to trace |
|
|
45
45
|
| `-n` / `--name` | — | Name of the process to trace |
|
|
46
46
|
| `--duration` | — | Duration to run, in `dd:hh:mm:ss` format |
|
|
47
|
-
| `-o` / `--output` |
|
|
47
|
+
| `-o` / `--output` | see below | Output file path |
|
|
48
48
|
| `--format` | `nettrace` | Output format: `nettrace`, `speedscope`, `chromium` |
|
|
49
49
|
| `--profile` | — | Preset provider configuration (e.g. `gc-verbose`) |
|
|
50
50
|
| `--providers` | — | Comma-separated EventPipe providers to enable |
|
|
51
|
+
|
|
52
|
+
Without `-o`, the trace is written to `<appname>_<yyyyMMdd>_<HHmmss>.nettrace`.
|
package/docs/languages/elixir.md
CHANGED
|
@@ -3,17 +3,17 @@
|
|
|
3
3
|
Elixir and Erlang profiling uses
|
|
4
4
|
[eflambe](https://github.com/Stratus3D/eflambe).
|
|
5
5
|
|
|
6
|
-
##
|
|
6
|
+
## Wall-clock profiling
|
|
7
7
|
|
|
8
8
|
Traces function calls and writes them as collapsed stacks. Useful for finding
|
|
9
|
-
|
|
9
|
+
hot spots, including time spent sleeping, which eflambe records as its own
|
|
10
|
+
frame.
|
|
10
11
|
|
|
11
12
|
eflambe's default `brendan_gregg` format is collapsed stacks. Add `eflambe` to
|
|
12
13
|
your dependencies, then profile a specific function.
|
|
13
14
|
|
|
14
|
-
|
|
15
|
-
|
|
16
|
-
large input can produce hundreds of MB.
|
|
15
|
+
Each line's count is one microsecond of wall time on its stack, not one sample,
|
|
16
|
+
because eflambe derives the counts from the trace's timestamps.
|
|
17
17
|
|
|
18
18
|
### Elixir
|
|
19
19
|
|
|
@@ -37,9 +37,17 @@ eflambe:apply({my_module, my_function, [Arg1, Arg2]}, [{output_format, brendan_g
|
|
|
37
37
|
|
|
38
38
|
### Options
|
|
39
39
|
|
|
40
|
-
| Option | Default | Description
|
|
41
|
-
| ------------------ | --------------- |
|
|
42
|
-
| `output_format` | `brendan_gregg` | Output format: `brendan_gregg` (collapsed) or `svg`
|
|
43
|
-
| `output_directory` |
|
|
44
|
-
| `open` | — | Auto-open the result in `speedscope` or `hotspot`
|
|
45
|
-
| `return` |
|
|
40
|
+
| Option | Default | Description |
|
|
41
|
+
| ------------------ | --------------- | -------------------------------------------------------------------------------------------------------- |
|
|
42
|
+
| `output_format` | `brendan_gregg` | Output format: `brendan_gregg` (collapsed) or `svg` |
|
|
43
|
+
| `output_directory` | `./` | Directory for the output file |
|
|
44
|
+
| `open` | — | Auto-open the result in `speedscope` or `hotspot` |
|
|
45
|
+
| `return` | see description | Return value: `value`, `flamegraph`, or `filename`. `capture` returns the filename and `apply` the value |
|
|
46
|
+
|
|
47
|
+
## Tips
|
|
48
|
+
|
|
49
|
+
### Overhead
|
|
50
|
+
|
|
51
|
+
Keep the call small. eflambe traces every function call instead of sampling, so
|
|
52
|
+
output grows with the number of calls. A single invocation over a large input
|
|
53
|
+
can produce hundreds of MB.
|
|
@@ -0,0 +1,141 @@
|
|
|
1
|
+
# Fortran
|
|
2
|
+
|
|
3
|
+
Fortran profiling uses [gperftools](https://github.com/gperftools/gperftools),
|
|
4
|
+
whose CPU profiler and tcmalloc heap profiler attach to any dynamically linked
|
|
5
|
+
native program, including one built by `gfortran`, `ifx`, or `flang`.
|
|
6
|
+
|
|
7
|
+
## CPU profiling
|
|
8
|
+
|
|
9
|
+
Periodically samples the call stack. Useful for finding CPU hot spots.
|
|
10
|
+
|
|
11
|
+
Keep frame pointers so the profiler's unwinder can walk the stack, and debug
|
|
12
|
+
info so its frames contain source locations:
|
|
13
|
+
|
|
14
|
+
```sh
|
|
15
|
+
gfortran -O2 -g -fno-omit-frame-pointer -o program program.f90
|
|
16
|
+
```
|
|
17
|
+
|
|
18
|
+
### CLI
|
|
19
|
+
|
|
20
|
+
```sh
|
|
21
|
+
# Generate a CPU profile
|
|
22
|
+
CPUPROFILE=cpu.prof LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libprofiler.so ./program
|
|
23
|
+
|
|
24
|
+
# Convert to pprof protobuf format
|
|
25
|
+
pprof --proto ./program cpu.prof > cpu.pprof
|
|
26
|
+
```
|
|
27
|
+
|
|
28
|
+
#### Environment variables
|
|
29
|
+
|
|
30
|
+
| Variable | Default | Description |
|
|
31
|
+
| ---------------------- | ------- | -------------------------------------------------------------------------------- |
|
|
32
|
+
| `CPUPROFILE` | — | Output filename that enables profiling when set |
|
|
33
|
+
| `CPUPROFILE_FREQUENCY` | `100` | Samples per second |
|
|
34
|
+
| `CPUPROFILE_REALTIME` | — | Use wall-clock time (`ITIMER_REAL`) instead of CPU time (`ITIMER_PROF`) when set |
|
|
35
|
+
| `CPUPROFILESIGNAL` | — | Signal number to toggle profiling on/off at runtime |
|
|
36
|
+
|
|
37
|
+
### Programmatic API
|
|
38
|
+
|
|
39
|
+
Declare the C entry points through `iso_c_binding` and link `-lprofiler`:
|
|
40
|
+
|
|
41
|
+
```fortran
|
|
42
|
+
use iso_c_binding, only: c_char, c_int, c_null_char
|
|
43
|
+
interface
|
|
44
|
+
function profiler_start(filename) bind(C, name='ProfilerStart')
|
|
45
|
+
import :: c_char, c_int
|
|
46
|
+
character(kind=c_char), intent(in) :: filename(*)
|
|
47
|
+
integer(c_int) :: profiler_start
|
|
48
|
+
end function
|
|
49
|
+
subroutine profiler_stop() bind(C, name='ProfilerStop')
|
|
50
|
+
end subroutine
|
|
51
|
+
end interface
|
|
52
|
+
|
|
53
|
+
if (profiler_start('cpu.prof'//c_null_char) == 0) error stop 'ProfilerStart failed'
|
|
54
|
+
|
|
55
|
+
! Code to profile...
|
|
56
|
+
|
|
57
|
+
call profiler_stop()
|
|
58
|
+
```
|
|
59
|
+
|
|
60
|
+
```sh
|
|
61
|
+
gfortran -O2 -g -fno-omit-frame-pointer -Wl,--no-as-needed -lprofiler \
|
|
62
|
+
-o program program.f90
|
|
63
|
+
```
|
|
64
|
+
|
|
65
|
+
## Memory profiling
|
|
66
|
+
|
|
67
|
+
Periodically dumps the live heap, capturing allocation stack traces. Useful for
|
|
68
|
+
finding memory leaks and allocation hot spots.
|
|
69
|
+
|
|
70
|
+
`allocate` and the temporaries the compiler creates for allocatable and
|
|
71
|
+
automatic arrays go through `malloc`, so tcmalloc records them.
|
|
72
|
+
|
|
73
|
+
### CLI
|
|
74
|
+
|
|
75
|
+
```sh
|
|
76
|
+
# Generate heap profiles (written as <prefix>.0001.heap, .0002.heap, …)
|
|
77
|
+
HEAPPROFILE=heap.prof LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libtcmalloc.so ./program
|
|
78
|
+
|
|
79
|
+
# Convert a heap profile to pprof protobuf format
|
|
80
|
+
pprof --proto ./program heap.prof.0001.heap > heap.pprof
|
|
81
|
+
```
|
|
82
|
+
|
|
83
|
+
#### Environment variables
|
|
84
|
+
|
|
85
|
+
| Variable | Default | Description |
|
|
86
|
+
| ---------------------------------- | ------------ | --------------------------------------------------------------------------------------------- |
|
|
87
|
+
| `HEAPPROFILE` | — | Output filename prefix that enables heap profiling when set |
|
|
88
|
+
| `HEAP_PROFILE_ALLOCATION_INTERVAL` | `1073741824` | Dump a profile after every this many bytes allocated (default 1 GB) |
|
|
89
|
+
| `HEAP_PROFILE_INUSE_INTERVAL` | `104857600` | Dump a profile each time the in-use high-water mark grows by this many bytes (default 100 MB) |
|
|
90
|
+
| `HEAP_PROFILE_TIME_INTERVAL` | `0` | Dump a profile every this many seconds (disabled by default) |
|
|
91
|
+
| `HEAPPROFILESIGNAL` | — | Signal number to dump a profile on demand |
|
|
92
|
+
| `HEAP_PROFILE_MMAP` | `false` | Also profile `mmap`/`mremap`/`sbrk` calls |
|
|
93
|
+
|
|
94
|
+
### Programmatic API
|
|
95
|
+
|
|
96
|
+
Declare the C entry points and link `-ltcmalloc`:
|
|
97
|
+
|
|
98
|
+
```fortran
|
|
99
|
+
use iso_c_binding, only: c_char, c_null_char
|
|
100
|
+
interface
|
|
101
|
+
subroutine heap_profiler_start(prefix) bind(C, name='HeapProfilerStart')
|
|
102
|
+
import :: c_char
|
|
103
|
+
character(kind=c_char), intent(in) :: prefix(*)
|
|
104
|
+
end subroutine
|
|
105
|
+
subroutine heap_profiler_stop() bind(C, name='HeapProfilerStop')
|
|
106
|
+
end subroutine
|
|
107
|
+
end interface
|
|
108
|
+
|
|
109
|
+
call heap_profiler_start('heap.prof'//c_null_char)
|
|
110
|
+
|
|
111
|
+
! Code to profile...
|
|
112
|
+
|
|
113
|
+
call heap_profiler_stop()
|
|
114
|
+
```
|
|
115
|
+
|
|
116
|
+
## Tips
|
|
117
|
+
|
|
118
|
+
### Names
|
|
119
|
+
|
|
120
|
+
`gfortran` mangles a module procedure to `__<module>_MOD_<procedure>` and a
|
|
121
|
+
free-standing procedure to `<procedure>_`. The program body appears as `MAIN__`.
|
|
122
|
+
A procedure the compiler specialized or partly inlined gets a suffix such as
|
|
123
|
+
`.part.0` or `.constprop.0`.
|
|
124
|
+
|
|
125
|
+
### Optimization
|
|
126
|
+
|
|
127
|
+
`-O0` frames are easy to read, but the timings differ from a release build's.
|
|
128
|
+
Profile the `-O2` or `-O3` build you ship, and expect inlining to attribute some
|
|
129
|
+
samples to the caller.
|
|
130
|
+
|
|
131
|
+
### Threads
|
|
132
|
+
|
|
133
|
+
An OpenMP program runs its worker threads in the same process, so the profile
|
|
134
|
+
covers all of them. Set `OMP_NUM_THREADS=1` to profile the serial path first,
|
|
135
|
+
because a parallel region spreads one hot loop across every worker's stack.
|
|
136
|
+
|
|
137
|
+
### Converting
|
|
138
|
+
|
|
139
|
+
`pprof --proto` requires the Go `pprof`. gperftools stopped shipping a `pprof`
|
|
140
|
+
of its own in 2.17, and the Perl `pprof`/`google-pprof` that earlier releases
|
|
141
|
+
and the distro packages built from them install has no protobuf output.
|
package/docs/languages/go.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
# Go
|
|
2
2
|
|
|
3
|
-
Go
|
|
3
|
+
Go profiling uses pprof, built into the runtime as the
|
|
4
4
|
[`runtime/pprof`](https://pkg.go.dev/runtime/pprof) package and the
|
|
5
5
|
[`net/http/pprof`](https://pkg.go.dev/net/http/pprof) HTTP handler.
|
|
6
6
|
|
|
@@ -38,12 +38,12 @@ defer pprof.StopCPUProfile()
|
|
|
38
38
|
|
|
39
39
|
## Memory profiling
|
|
40
40
|
|
|
41
|
-
|
|
41
|
+
The heap profile variants are:
|
|
42
42
|
|
|
43
|
-
- **heap** (`inuse_space` / `inuse_objects`): live allocations at each sample
|
|
43
|
+
- **heap** (`inuse_space` / `inuse_objects`): live allocations at each sample,
|
|
44
44
|
useful for finding memory leaks
|
|
45
45
|
- **allocs** (`alloc_space` / `alloc_objects`): all allocations since program
|
|
46
|
-
start
|
|
46
|
+
start, useful for finding allocation hot spots
|
|
47
47
|
|
|
48
48
|
`runtime.MemProfileRate` controls sampling granularity (default: one sample per
|
|
49
49
|
512 KB allocated). Set to `1` to capture every allocation (expensive):
|
|
@@ -93,7 +93,7 @@ leaks or deadlocks.
|
|
|
93
93
|
### CLI
|
|
94
94
|
|
|
95
95
|
```sh
|
|
96
|
-
# Full stack traces (debug=2
|
|
96
|
+
# Full stack traces (`debug=2` writes text, and `debug=0` the pprof binary)
|
|
97
97
|
curl -o goroutine.pprof 'http://localhost:6060/debug/pprof/goroutine'
|
|
98
98
|
|
|
99
99
|
# Human-readable text dump
|
|
@@ -107,6 +107,27 @@ f, _ := os.Create("goroutine.pprof")
|
|
|
107
107
|
pprof.Lookup("goroutine").WriteTo(f, 0)
|
|
108
108
|
```
|
|
109
109
|
|
|
110
|
+
## Goroutine leak profiling
|
|
111
|
+
|
|
112
|
+
Captures stack traces of leaked goroutines: goroutines blocked on a channel,
|
|
113
|
+
mutex, or other concurrency primitive that no runnable goroutine can reach.
|
|
114
|
+
Requires Go 1.27, though Go 1.26 supports it behind
|
|
115
|
+
`GOEXPERIMENT=goroutineleakprofile`. Writing the profile first runs a garbage
|
|
116
|
+
collection cycle that detects the leaks.
|
|
117
|
+
|
|
118
|
+
### CLI
|
|
119
|
+
|
|
120
|
+
```sh
|
|
121
|
+
curl -o goroutineleak.pprof 'http://localhost:6060/debug/pprof/goroutineleak'
|
|
122
|
+
```
|
|
123
|
+
|
|
124
|
+
### Programmatic API
|
|
125
|
+
|
|
126
|
+
```go
|
|
127
|
+
f, _ := os.Create("goroutineleak.pprof")
|
|
128
|
+
pprof.Lookup("goroutineleak").WriteTo(f, 0)
|
|
129
|
+
```
|
|
130
|
+
|
|
110
131
|
## Block profiling
|
|
111
132
|
|
|
112
133
|
Captures stack traces that led to blocking on synchronization primitives
|
|
@@ -115,7 +136,7 @@ diagnosing synchronization bottlenecks.
|
|
|
115
136
|
|
|
116
137
|
Disabled by default. Call `runtime.SetBlockProfileRate(rate)` in your program
|
|
117
138
|
before the code to profile. This applies whether you collect via HTTP or the
|
|
118
|
-
programmatic API. Rate is in nanoseconds
|
|
139
|
+
programmatic API. Rate is in nanoseconds. `1` captures every blocking event, and
|
|
119
140
|
higher values sample less.
|
|
120
141
|
|
|
121
142
|
```go
|
|
@@ -146,8 +167,8 @@ goroutine tries to acquire it. Useful for finding lock contention hot spots.
|
|
|
146
167
|
|
|
147
168
|
Disabled by default. Call `runtime.SetMutexProfileFraction(rate)` in your
|
|
148
169
|
program before the code to profile. This applies whether you collect via HTTP or
|
|
149
|
-
the programmatic API. `1` captures every contended event
|
|
150
|
-
`1/n` events.
|
|
170
|
+
the programmatic API. `1` captures every contended event, and `n` samples
|
|
171
|
+
roughly `1/n` events.
|
|
151
172
|
|
|
152
173
|
```go
|
|
153
174
|
runtime.SetMutexProfileFraction(1)
|
|
@@ -170,8 +191,8 @@ pprof.Lookup("mutex").WriteTo(f, 0)
|
|
|
170
191
|
|
|
171
192
|
## Thread creation profiling
|
|
172
193
|
|
|
173
|
-
Captures stack traces
|
|
174
|
-
|
|
194
|
+
Captures the stack traces where the program created new OS threads. Useful for
|
|
195
|
+
finding code that creates too many threads.
|
|
175
196
|
|
|
176
197
|
### CLI
|
|
177
198
|
|
|
@@ -188,5 +209,7 @@ pprof.Lookup("threadcreate").WriteTo(f, 0)
|
|
|
188
209
|
|
|
189
210
|
## Tips
|
|
190
211
|
|
|
212
|
+
### Locations
|
|
213
|
+
|
|
191
214
|
Build with `-trimpath` (e.g. `go test -trimpath ...`) so locations are
|
|
192
215
|
package-relative (`encoding/json/encode.go`) instead of absolute build paths.
|