profiler-md 0.1.3 → 0.2.0
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.mjs → cli/index.mjs} +794 -176
- package/dist/index.d.mts +16 -49
- package/dist/index.js +1 -1
- package/docs/formats/pprof.md +12 -0
- package/docs/formats/speedscope.md +11 -0
- package/docs/formats/v8-cpu-profile.md +13 -0
- package/docs/formats/v8-heap-profile.md +13 -0
- package/docs/formats/v8-heap-snapshot.md +10 -0
- package/docs/languages/c.md +136 -0
- package/docs/languages/go.md +187 -0
- package/docs/languages/java.md +117 -0
- package/docs/languages/javascript.md +185 -0
- package/docs/languages/php.md +31 -0
- package/docs/languages/python.md +112 -0
- package/docs/languages/ruby.md +67 -0
- package/docs/languages/rust.md +44 -0
- package/examples/bun.cpuprofile.md +497 -0
- package/examples/deno.cpuprofile.md +488 -0
- package/examples/node.cpuprofile.md +499 -0
- package/examples/node.heapprofile.md +326 -0
- package/examples/node.heapsnapshot.md +692 -0
- package/examples/node.pprof.md +450 -0
- package/examples/node.speedscope.json.md +416 -0
- package/examples/rust1.pprof.md +480 -0
- package/examples/rust2.pprof.md +480 -0
- package/package.json +10 -3
- package/readme.md +28 -3
|
@@ -0,0 +1,31 @@
|
|
|
1
|
+
# PHP
|
|
2
|
+
|
|
3
|
+
PHP profiling uses [phpspy](https://github.com/adsr/phpspy), a low overhead
|
|
4
|
+
sampling profiler for PHP processes.
|
|
5
|
+
|
|
6
|
+
## CPU profiling
|
|
7
|
+
|
|
8
|
+
Periodically samples the call stack. Useful for finding CPU hot spots.
|
|
9
|
+
|
|
10
|
+
```sh
|
|
11
|
+
# Profile a command (speedscope format)
|
|
12
|
+
phpspy -d 30 --output-format=speedscope -f cpu.speedscope -- php script.php
|
|
13
|
+
|
|
14
|
+
# Attach to a running process (speedscope format)
|
|
15
|
+
phpspy -d 30 --output-format=speedscope --pid <pid> -f cpu.speedscope
|
|
16
|
+
|
|
17
|
+
# Attach to a running process (callgrind format, for KCachegrind)
|
|
18
|
+
phpspy -d 30 --output-format=callgrind --pid <pid> > cpu.callgrind
|
|
19
|
+
```
|
|
20
|
+
|
|
21
|
+
### Flags
|
|
22
|
+
|
|
23
|
+
| Flag | Default | Description |
|
|
24
|
+
| -------------------- | ------- | ------------------------------------------------------------------ |
|
|
25
|
+
| `--pid` / `-p` | — | PID of the PHP process to profile |
|
|
26
|
+
| `--duration` / `-d` | — | Duration in seconds; profiles indefinitely when not set |
|
|
27
|
+
| `--sleep-us` / `-s` | `10000` | Sampling interval in microseconds |
|
|
28
|
+
| `--num-samples` | — | Maximum number of samples to collect |
|
|
29
|
+
| `--max-depth` / `-l` | `64` | Maximum stack frames per sample |
|
|
30
|
+
| `--output-format` | `raw` | Output format: `raw`, `callgrind`, `speedscope`, `valgrind`, `rep` |
|
|
31
|
+
| `--output-file` | stdout | Write output to this file instead of stdout |
|
|
@@ -0,0 +1,112 @@
|
|
|
1
|
+
# Python
|
|
2
|
+
|
|
3
|
+
Python profiling uses [py-spy](https://github.com/benford/py-spy), a sampling
|
|
4
|
+
profiler that attaches to a running process or wraps a script, and
|
|
5
|
+
[memray](https://github.com/bloomberg/memray) for memory profiling.
|
|
6
|
+
|
|
7
|
+
## CPU profiling
|
|
8
|
+
|
|
9
|
+
Periodically samples the call stack. Useful for finding CPU hot spots.
|
|
10
|
+
|
|
11
|
+
```sh
|
|
12
|
+
# pprof format (wrap a script)
|
|
13
|
+
py-spy record -f pprof -o cpu.pprof -- python script.py
|
|
14
|
+
|
|
15
|
+
# pprof format (attach to a running process)
|
|
16
|
+
py-spy record -f pprof -o cpu.pprof --pid <pid>
|
|
17
|
+
|
|
18
|
+
# Speedscope format
|
|
19
|
+
py-spy record -f speedscope -o profile.speedscope -- python script.py
|
|
20
|
+
|
|
21
|
+
# Include native extensions written in Cython or C
|
|
22
|
+
py-spy record -f pprof -o cpu.pprof --native -- python script.py
|
|
23
|
+
```
|
|
24
|
+
|
|
25
|
+
## Wall-clock profiling
|
|
26
|
+
|
|
27
|
+
Samples wall-clock time rather than CPU time. Useful for I/O-bound code where
|
|
28
|
+
threads spend time sleeping or waiting.
|
|
29
|
+
|
|
30
|
+
```sh
|
|
31
|
+
# Wrap a script
|
|
32
|
+
py-spy record -f pprof -o profile.pprof --idle -- python script.py
|
|
33
|
+
|
|
34
|
+
# Attach to a running process
|
|
35
|
+
py-spy record -f pprof -o profile.pprof --idle --pid <pid>
|
|
36
|
+
```
|
|
37
|
+
|
|
38
|
+
## Memory profiling
|
|
39
|
+
|
|
40
|
+
Tracks heap allocations. Useful for finding allocation hot spots and memory
|
|
41
|
+
leaks.
|
|
42
|
+
|
|
43
|
+
### CLI
|
|
44
|
+
|
|
45
|
+
```sh
|
|
46
|
+
# Record a memory profile (wrap a script)
|
|
47
|
+
memray run -o mem.bin script.py
|
|
48
|
+
|
|
49
|
+
# Attach to a running process
|
|
50
|
+
memray attach -o mem.bin <pid>
|
|
51
|
+
|
|
52
|
+
# Convert to pprof format
|
|
53
|
+
memray pprof mem.bin
|
|
54
|
+
```
|
|
55
|
+
|
|
56
|
+
### Programmatic API
|
|
57
|
+
|
|
58
|
+
```python
|
|
59
|
+
import memray
|
|
60
|
+
|
|
61
|
+
with memray.Tracker("mem.bin"):
|
|
62
|
+
# Code to profile...
|
|
63
|
+
pass
|
|
64
|
+
```
|
|
65
|
+
|
|
66
|
+
## Call stack dump
|
|
67
|
+
|
|
68
|
+
Captures the current call stacks of all threads in a running process. Useful for
|
|
69
|
+
diagnosing hangs, deadlocks, or unexpected blocking.
|
|
70
|
+
|
|
71
|
+
### CLI
|
|
72
|
+
|
|
73
|
+
```sh
|
|
74
|
+
# Dump call stacks of a running process (text output)
|
|
75
|
+
py-spy dump --pid <pid>
|
|
76
|
+
|
|
77
|
+
# Include local variables in each frame
|
|
78
|
+
py-spy dump --pid <pid> --locals
|
|
79
|
+
```
|
|
80
|
+
|
|
81
|
+
### Programmatic API
|
|
82
|
+
|
|
83
|
+
```python
|
|
84
|
+
import faulthandler
|
|
85
|
+
|
|
86
|
+
faulthandler.dump_traceback() # all threads to stderr
|
|
87
|
+
```
|
|
88
|
+
|
|
89
|
+
## CLI flags
|
|
90
|
+
|
|
91
|
+
### `py-spy`
|
|
92
|
+
|
|
93
|
+
| Flag | Default | Description |
|
|
94
|
+
| ------------------- | ------------ | ---------------------------------------------------------- |
|
|
95
|
+
| `-f` / `--format` | `flamegraph` | Output format: `flamegraph`, `raw`, `speedscope`, `pprof` |
|
|
96
|
+
| `-o` / `--output` | — | Output file path |
|
|
97
|
+
| `-d` / `--duration` | — | Duration in seconds (default: until program exits) |
|
|
98
|
+
| `-r` / `--rate` | `100` | Sampling rate in Hz |
|
|
99
|
+
| `--idle` | off | Include stack traces for idle (sleeping/waiting) threads |
|
|
100
|
+
| `--native` | off | Profile native extensions written in Cython or C |
|
|
101
|
+
| `--nonblocking` | off | Don't pause the process to collect samples (less accurate) |
|
|
102
|
+
| `--subprocesses` | off | Also profile subprocesses of the target |
|
|
103
|
+
|
|
104
|
+
### `memray`
|
|
105
|
+
|
|
106
|
+
| Flag | Default | Description |
|
|
107
|
+
| --------------------------- | ------- | --------------------------------------------------- |
|
|
108
|
+
| `-o` / `--output` | — | Output file path |
|
|
109
|
+
| `--native` | off | Also capture native (C extension) stack frames |
|
|
110
|
+
| `--follow-fork` | off | Also profile child processes after `fork()` |
|
|
111
|
+
| `--aggregate` | off | Aggregate allocations to reduce output file size |
|
|
112
|
+
| `--trace-python-allocators` | off | Trace Python allocator calls for finer-grained data |
|
|
@@ -0,0 +1,67 @@
|
|
|
1
|
+
# Ruby
|
|
2
|
+
|
|
3
|
+
Ruby profiling uses [rbspy](https://rbspy.github.io), a sampling profiler that
|
|
4
|
+
attaches to a running Ruby process.
|
|
5
|
+
|
|
6
|
+
## CPU profiling
|
|
7
|
+
|
|
8
|
+
Periodically samples the call stack. Useful for finding CPU hot spots.
|
|
9
|
+
|
|
10
|
+
### CLI
|
|
11
|
+
|
|
12
|
+
```sh
|
|
13
|
+
# Profile a command (speedscope format)
|
|
14
|
+
rbspy record --format speedscope -o profile.speedscope -- ruby script.rb
|
|
15
|
+
|
|
16
|
+
# Attach to a running process
|
|
17
|
+
rbspy record --format speedscope -o profile.speedscope --pid <pid>
|
|
18
|
+
|
|
19
|
+
# Profile for a fixed duration
|
|
20
|
+
rbspy record --format speedscope -o profile.speedscope --duration 30 --pid <pid>
|
|
21
|
+
|
|
22
|
+
# Profile including child processes
|
|
23
|
+
rbspy record --format speedscope -o profile.speedscope --subprocesses --pid <pid>
|
|
24
|
+
```
|
|
25
|
+
|
|
26
|
+
## Memory profiling
|
|
27
|
+
|
|
28
|
+
Records RSS memory alongside call stack samples. Useful for correlating memory
|
|
29
|
+
growth with specific code paths.
|
|
30
|
+
|
|
31
|
+
### CLI
|
|
32
|
+
|
|
33
|
+
```sh
|
|
34
|
+
# Record CPU and memory together
|
|
35
|
+
rbspy record --memory --format speedscope -o profile.speedscope --pid <pid>
|
|
36
|
+
|
|
37
|
+
# Profile a command with memory tracking
|
|
38
|
+
rbspy record --memory --format speedscope -o profile.speedscope -- ruby script.rb
|
|
39
|
+
```
|
|
40
|
+
|
|
41
|
+
## Snapshot
|
|
42
|
+
|
|
43
|
+
Captures a single stack trace from a running process. Useful for quickly
|
|
44
|
+
checking what a process is currently executing without starting a full
|
|
45
|
+
recording.
|
|
46
|
+
|
|
47
|
+
### CLI
|
|
48
|
+
|
|
49
|
+
```sh
|
|
50
|
+
rbspy snapshot --pid <pid>
|
|
51
|
+
|
|
52
|
+
# Don't pause the process while reading its stack
|
|
53
|
+
rbspy snapshot --nonblocking --pid <pid>
|
|
54
|
+
```
|
|
55
|
+
|
|
56
|
+
## CLI flags
|
|
57
|
+
|
|
58
|
+
| Flag | Default | Description |
|
|
59
|
+
| ------------------- | ------------ | ------------------------------------------------------------------------------------------------------ |
|
|
60
|
+
| `-p` / `--pid` | — | PID of a running process to attach to |
|
|
61
|
+
| `-d` / `--duration` | — | Duration in seconds; records until Ctrl-C if omitted |
|
|
62
|
+
| `-r` / `--rate` | `100` | Samples per second |
|
|
63
|
+
| `-f` / `--format` | `flamegraph` | Output format: `flamegraph`, `speedscope`, `callgrind`, `summary`, `summary_by_line`, `raw_call_stack` |
|
|
64
|
+
| `-o` / `--file` | — | Output file path |
|
|
65
|
+
| `-m` / `--memory` | — | Also record RSS memory usage alongside stack samples |
|
|
66
|
+
| `--subprocesses` | — | Also profile child processes |
|
|
67
|
+
| `--nonblocking` | — | Don't pause the process while reading its stack (lower overhead, may miss frames) |
|
|
@@ -0,0 +1,44 @@
|
|
|
1
|
+
# Rust
|
|
2
|
+
|
|
3
|
+
Rust CPU profiling uses [pprof-rs](https://github.com/tikv/pprof-rs), a sampling
|
|
4
|
+
profiler that integrates directly into your binary.
|
|
5
|
+
|
|
6
|
+
## CPU profiling
|
|
7
|
+
|
|
8
|
+
Periodically samples the call stack. Useful for finding CPU hot spots.
|
|
9
|
+
|
|
10
|
+
Add to `Cargo.toml`:
|
|
11
|
+
|
|
12
|
+
```toml
|
|
13
|
+
[dependencies]
|
|
14
|
+
pprof = { version = "...", features = ["prost-codec", "frame-pointer"] }
|
|
15
|
+
```
|
|
16
|
+
|
|
17
|
+
The `frame-pointer` feature is required on macOS (the default stack unwinder
|
|
18
|
+
crashes with SIGBUS). Also add the following rustflag to make frame-pointer
|
|
19
|
+
unwinding reliable:
|
|
20
|
+
|
|
21
|
+
```sh
|
|
22
|
+
RUSTFLAGS="-C force-frame-pointers=yes" cargo build ...
|
|
23
|
+
```
|
|
24
|
+
|
|
25
|
+
```rust
|
|
26
|
+
use pprof::protos::Message;
|
|
27
|
+
|
|
28
|
+
let guard = pprof::ProfilerGuardBuilder::default()
|
|
29
|
+
.frequency(1000)
|
|
30
|
+
.build()?;
|
|
31
|
+
|
|
32
|
+
// Code to profile...
|
|
33
|
+
|
|
34
|
+
let mut buf = Vec::new();
|
|
35
|
+
guard.report().build()?.pprof()?.encode(&mut buf)?;
|
|
36
|
+
std::fs::write("cpu.pprof", &buf)?;
|
|
37
|
+
```
|
|
38
|
+
|
|
39
|
+
### `ProfilerGuardBuilder` options
|
|
40
|
+
|
|
41
|
+
| Method | Default | Description |
|
|
42
|
+
| -------------------- | ------- | ---------------------------------------- |
|
|
43
|
+
| `.frequency(hz)` | `100` | Samples per second |
|
|
44
|
+
| `.blocklist(frames)` | `[]` | Stack frames to exclude from the profile |
|