File: jit-hooks.rst

package info (click to toggle)
pypy3 7.0.0%2Bdfsg-3
  • links: PTS, VCS
  • area: main
  • in suites: buster
  • size: 111,848 kB
  • sloc: python: 1,291,746; ansic: 74,281; asm: 5,187; cpp: 3,017; sh: 2,533; makefile: 544; xml: 243; lisp: 45; csh: 21; awk: 4
file content (82 lines) | stat: -rw-r--r-- 2,704 bytes parent folder | download | duplicates (3)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
JIT hooks
=========

There are several hooks in the ``pypyjit`` module that may help you with
understanding what's pypy's JIT doing while running your program. There
are three functions related to that coming from the ``pypyjit`` module:


.. function:: set_compile_hook(callable, operations=True)

    Set a compiling hook that will be called each time a loop is compiled.

    The callable will be called with the ``pypyjit.JitLoopInfo`` object.
    Refer to it's documentation for details.

    Note that jit hook is not reentrant. It means that if the code
    inside the jit hook is itself jitted, it will get compiled, but the
    jit hook won't be called for that.

    if operations=False, no list of operations will be available. Useful
    if the hook is supposed to be very lighweight.

.. function:: set_abort_hook(hook)

    Set a hook (callable) that will be called each time there is tracing
    aborted due to some reason.

    The hook will be invoked with the siagnture:
    ``hook(jitdriver_name, greenkey, reason, oplist)``

    Reason is a string, the meaning of other arguments is the same
    as attributes on JitLoopInfo object

.. function:: enable_debug()

    Start recording debugging counters for ``get_stats_snapshot``

.. function:: disable_debug()

    Stop recording debugging counters for ``get_stats_snapshot``

.. function:: get_stats_snapshot()

    Get the jit status in the specific moment in time. Note that this
    is eager - the attribute access is not lazy, if you need new stats
    you need to call this function again. You might want to call
    ``enable_debug`` to get more information. It returns an instance
    of ``JitInfoSnapshot``

.. class:: JitInfoSnapshot

    A class describing current snapshot. Usable attributes:

    * ``counters`` - internal JIT integer counters

    * ``counter_times`` - internal JIT float counters, notably time spent
      TRACING and in the JIT BACKEND

    * ``loop_run_times`` - counters for number of times loops are run, only
      works when ``enable_debug`` is called.

.. class:: JitLoopInfo

   A class containing information about the compiled loop. Usable attributes:

   * ``operations`` - list of operations, if requested

   * ``jitdriver_name`` - the name of jitdriver associated with this loop

   * ``greenkey`` - a key at which the loop got compiled (e.g. code position,
     is_being_profiled, pycode tuple for python jitdriver)

   * ``loop_no`` - loop cardinal number

   * ``bridge_no`` - id of the fail descr

   * ``type`` - "entry bridge", "loop" or "bridge"

   * ``asmaddr`` - an address in raw memory where assembler resides

   * ``asmlen`` - length of raw memory with assembler associated