Merge pull request #1120 from gevent/faster-stack

Speed up Greenlet creation on CPython

CHANGES.rst

@@ -87,13 +87,15 @@
 
 - Greenlet objects now keep track of their spawning parent greenlet
   and the code location that spawned them, in addition to maintaining
-  a "spawn tree local" mapping. Based on a proposal from PayPal and
-  comments by Mahmoud Hashemi and Kurt Rose. See :issue:`755` and
-  :pr:`1115`. As always, feedback is appreciated.
+  a "spawn tree local" mapping. This adds some runtime overhead in
+  relative terms, but absolute numbers are still relatively small.
+  Based on a proposal from PayPal and comments by Mahmoud Hashemi and
+  Kurt Rose. See :issue:`755` and :pr:`1115`. As always, feedback is
+  appreciated.
 
 - The :mod:`gevent.greenlet` module is now compiled with Cython to
   offset any performance decrease due to :issue:`755`. Please open
-  issues for any compatibility concerns. See :pr:`1115`.
+  issues for any compatibility concerns. See :pr:`1115` and :pr:`1120`.
 
 - Greenlet objects now have a ``minimal_ident`` property. It functions
   similarly to ``Thread.ident`` or ``id`` by uniquely identifying the

src/gevent/_greenlet.pxd

@@ -4,6 +4,8 @@ cimport cython
 from gevent.__ident cimport IdentRegistry
 cdef bint _greenlet_imported
 cdef bint _PYPY
+cdef sys_getframe
+cdef sys_exc_info
 
 cdef extern from "greenlet/greenlet.h":
 
@@ -25,6 +27,23 @@ cdef inline void greenlet_init():
         PyGreenlet_Import()
         _greenlet_imported = True
 
+cdef extern from "Python.h":
+
+    ctypedef class types.CodeType [object PyCodeObject]:
+        pass
+
+cdef extern from "frameobject.h":
+
+    ctypedef class types.FrameType [object PyFrameObject]:
+        cdef CodeType f_code
+        cdef int f_lineno
+        # We can't declare this in the object, because it's
+        # allowed to be NULL, and Cython can't handle that.
+        # We have to go through the python machinery to get a
+        # proper None instead.
+        # cdef FrameType f_back
+
+
 cdef void _init()
 
 cdef class SpawnedLink:
@@ -42,18 +61,18 @@ cdef class FailureSpawnedLink(SpawnedLink):
 @cython.final
 @cython.internal
 cdef class _Frame:
-    cdef readonly object f_code
+    cdef readonly CodeType f_code
     cdef readonly int f_lineno
-    cdef public _Frame f_back
+    cdef readonly _Frame f_back
 
 
 @cython.final
-@cython.locals(
-               previous=_Frame,
-               first=_Frame,
-               next_frame=_Frame)
-cdef _Frame _extract_stack(int limit, _Frame f_back)
+@cython.locals(frames=list,frame=FrameType)
+cdef inline list _extract_stack(int limit)
 
+@cython.final
+@cython.locals(previous=_Frame, frame=tuple, f=_Frame)
+cdef _Frame _Frame_from_list(list frames)
 
 
 cdef class Greenlet(greenlet):
@@ -61,7 +80,10 @@ cdef class Greenlet(greenlet):
     cdef readonly args
     cdef readonly object spawning_greenlet
     cdef public dict spawn_tree_locals
-    cdef readonly _Frame spawning_stack
+
+    # This is accessed with getattr() dynamically so it
+    # must be visible to Python
+    cdef readonly list _spawning_stack_frames
 
     cdef list _links
     cdef tuple _exc_info

src/gevent/greenlet.py

@@ -2,13 +2,15 @@
 # cython: auto_pickle=False,embedsignature=True,always_allow_keywords=False
 from __future__ import absolute_import, print_function, division
 
-import sys
+from sys import _getframe as sys_getframe
+from sys import exc_info as sys_exc_info
 from weakref import ref as wref
 
 
 from greenlet import greenlet
 
 from gevent._compat import reraise
+from gevent._compat import PYPY as _PYPY
 from gevent._tblib import dump_traceback
 from gevent._tblib import load_traceback
 from gevent.hub import GreenletExit
@@ -18,8 +20,7 @@ from gevent.hub import get_hub
 from gevent.hub import iwait
 from gevent.hub import wait
 from gevent.timeout import Timeout
-
-_PYPY = hasattr(sys, 'pypy_version_info')
+from gevent._util import Lazy
 
 
 __all__ = [
@@ -108,26 +109,28 @@ class _Frame(object):
         self.f_lineno = f_lineno
         self.f_back = None
 
-    f_globals = property(lambda _self: None)
+    @property
+    def f_globals(self):
+        return None
 
-def _extract_stack(limit, f_back):
+def _Frame_from_list(frames):
     previous = None
-    frame = sys._getframe()
-    first = None
+    for frame in reversed(frames):
+        f = _Frame(*frame)
+        f.f_back = previous
+        previous = f
+    return previous
 
-    first = previous = _Frame(frame.f_code, frame.f_lineno)
-    limit -= 1
-    frame = frame.f_back
+def _extract_stack(limit):
+    frame = sys_getframe()
+    frames = []
 
     while limit and frame is not None:
         limit -= 1
-        next_frame = _Frame(frame.f_code, frame.f_lineno)
-        previous.f_back = next_frame
-        previous = next_frame
+        frames.append((frame.f_code, frame.f_lineno))
         frame = frame.f_back
 
-    previous.f_back = f_back
-    return first
+    return frames
 
 
 _greenlet__init__ = greenlet.__init__
@@ -274,9 +277,16 @@ class Greenlet(greenlet):
                 # Its children get separate locals.
                 spawner.spawn_tree_locals = self.spawn_tree_locals
 
+        self._spawning_stack_frames = _extract_stack(self.spawning_stack_limit)
+        self._spawning_stack_frames.extend(getattr(spawner, '_spawning_stack_frames', []))
 
-        self.spawning_stack = _extract_stack(self.spawning_stack_limit,
-                                             getattr(spawner, 'spawning_stack', None))
+    @Lazy
+    def spawning_stack(self):
+        # Store this in the __dict__. We don't use it from the C
+        # code. It's tempting to discard _spawning_stack_frames
+        # after this, but child greenlets may still be created
+        # that need it.
+        return _Frame_from_list(self._spawning_stack_frames)
 
     def _get_minimal_ident(self):
         reg = self.parent.ident_registry
@@ -476,9 +486,9 @@ class Greenlet(greenlet):
 
         .. versionadded:: 1.1
         """
-        exc_info = self._exc_info
-        if exc_info is not None and exc_info[0] is not None:
-            return (exc_info[0], exc_info[1], load_traceback(exc_info[2]))
+        ei = self._exc_info
+        if ei is not None and ei[0] is not None:
+            return (ei[0], ei[1], load_traceback(ei[2]))
 
     def throw(self, *args):
         """Immediately switch into the greenlet and raise an exception in it.
@@ -708,7 +718,7 @@ class Greenlet(greenlet):
             try:
                 result = self._run(*self.args, **self.kwargs)
             except: # pylint:disable=bare-except
-                self._report_error(sys.exc_info())
+                self._report_error(sys_exc_info())
                 return
             self._report_result(result)
         finally:
@@ -801,7 +811,7 @@ class Greenlet(greenlet):
             try:
                 link(self)
             except: # pylint:disable=bare-except
-                self.parent.handle_error((link, self), *sys.exc_info())
+                self.parent.handle_error((link, self), *sys_exc_info())
 
 
 class _dummy_event(object):
@@ -828,7 +838,7 @@ def _kill(glet, exception, waiter):
         glet.throw(exception)
     except: # pylint:disable=bare-except
         # XXX do we need this here?
-        glet.parent.handle_error(glet, *sys.exc_info())
+        glet.parent.handle_error(glet, *sys_exc_info())
     if waiter is not None:
         waiter.switch()
 
@@ -863,7 +873,7 @@ def _killall3(greenlets, exception, waiter):
             try:
                 g.throw(exception)
             except: # pylint:disable=bare-except
-                g.parent.handle_error(g, *sys.exc_info())
+                g.parent.handle_error(g, *sys_exc_info())
             if not g.dead:
                 diehards.append(g)
     waiter.switch(diehards)
@@ -875,7 +885,7 @@ def _killall(greenlets, exception):
             try:
                 g.throw(exception)
             except: # pylint:disable=bare-except
-                g.parent.handle_error(g, *sys.exc_info())
+                g.parent.handle_error(g, *sys_exc_info())
 
 
 def killall(greenlets, exception=GreenletExit, block=True, timeout=None):

src/greentest/bench_spawn.py

 """Benchmarking spawn() performance.
 """
 from __future__ import print_function, absolute_import, division
-import sys
-import os
-import time
+
+import perf
 
 try:
     xrange
 except NameError:
     xrange = range
 
-if hasattr(time, "perf_counter"):
-    curtime = time.perf_counter  # 3.3
-elif sys.platform.startswith('win'):
-    curtime = time.clock
-else:
-    curtime = time.time
 
-N = 100000
+N = 10000
 counter = 0
 
 
 def incr(sleep, **_kwargs):
     global counter
     counter += 1
-    sleep(0)
 
 
 def noop(_p):
     pass
 
+class Options(object):
+
+    # TODO: Add back an argument for that
+    eventlet_hub = None
+
+    loops = None
+
+    def __init__(self, sleep, join, **kwargs):
+        self.kwargs = kwargs
+        self.sleep = sleep
+        self.join = join
+
+class Times(object):
+
+    def __init__(self,
+                 spawn_duration,
+                 sleep_duration=-1,
+                 join_duration=-1):
+        self.spawn_duration = spawn_duration
+        self.sleep_duration = sleep_duration
+        self.join_duration = join_duration
 
-def _report(name, delta):
-    print('%8s: %3.2f microseconds per greenlet' % (name, delta * 1000000.0 / N))
 
-def test(spawn, sleep, kwargs):
-    start = curtime()
+def _test(spawn, sleep, options):
+    global counter
+    counter = 0
+    before_spawn = perf.perf_counter()
     for _ in xrange(N):
-        spawn(incr, sleep, **kwargs)
-    _report('spawning', curtime() - start)
+        spawn(incr, sleep, **options.kwargs)
+
+    before_sleep = perf.perf_counter()
+    if options.sleep:
         assert counter == 0, counter
-    start = curtime()
         sleep(0)
-    _report('sleep(0)', curtime() - start)
+        after_sleep = perf.perf_counter()
         assert counter == N, (counter, N)
+    else:
+        after_sleep = before_sleep
+
+
+    if options.join:
+        before_join = perf.perf_counter()
+        options.join()
+        after_join = perf.perf_counter()
+        join_duration = after_join - before_join
+    else:
+        join_duration = -1
+
+    return Times(before_sleep - before_spawn,
+                 after_sleep - before_sleep,
+                 join_duration)
 
+def test(spawn, sleep, options):
+    all_times = [_test(spawn, sleep, options)
+                 for _ in xrange(options.loops)]
+
+    spawn_duration = sum(x.spawn_duration for x in all_times)
+    sleep_duration = sum(x.sleep_duration for x in all_times)
+    join_duration = sum(x.sleep_duration for x in all_times
+                        if x != -1)
+
+    return Times(spawn_duration, sleep_duration, join_duration)
 
 def bench_none(options):
-    kwargs = options.kwargs
-    start = curtime()
-    for _ in xrange(N):
-        incr(noop, **kwargs)
-    assert counter == N, (counter, N)
-    _report('noop', curtime() - start)
+    options.sleep = False
+    def spawn(f, sleep, **kwargs):
+        return f(sleep, **kwargs)
+    from time import sleep
+    return test(spawn,
+                sleep,
+                options)
 
 
 def bench_gevent(options):
-    import gevent
-    print('using gevent from %s' % gevent.__file__)
     from gevent import spawn, sleep
-    test(spawn, sleep, options.kwargs)
+    return test(spawn, sleep, options)
 
 
 def bench_geventraw(options):
-    import gevent
-    print('using gevent from %s' % gevent.__file__)
     from gevent import sleep, spawn_raw
-    test(spawn_raw, sleep, options.kwargs)
+    return test(spawn_raw, sleep, options)
 
 
 def bench_geventpool(options):
-    import gevent
-    print('using gevent from %s' % gevent.__file__)
     from gevent import sleep
     from gevent.pool import Pool
     p = Pool()
-    test(p.spawn, sleep, options.kwargs)
-    start = curtime()
-
-    p.join()
-
-    _report('joining', curtime() - start)
+    if options.join:
+        options.join = p.join
+    times = test(p.spawn, sleep, options)
+    return times
 
 
 
 def bench_eventlet(options):
-    try:
-        import eventlet
-    except ImportError:
-        if options.ignore_import_errors:
-            return
-        raise
-    print('using eventlet from %s' % eventlet.__file__)
     from eventlet import spawn, sleep
     from eventlet.hubs import use_hub
     if options.eventlet_hub is not None:
         use_hub(options.eventlet_hub)
-    test(spawn, sleep, options.kwargs)
-
-
-
-def bench_all():
-    from time import sleep
-    error = 0
-    names = sorted(all())
-
-    for func in names:
-        cmd = '%s %s %s --ignore-import-errors' % (sys.executable, __file__, func)
-        print(cmd)
-        sys.stdout.flush()
-        sleep(0.01)
-        if os.system(cmd):
-            error = 1
-            print('%s failed' % cmd)
-        print('')
-    for func in names:
-        cmd = '%s %s --with-kwargs %s --ignore-import-errors' % (sys.executable, __file__, func)
-        print(cmd)
-        sys.stdout.flush()
-        if os.system(cmd):
-            error = 1
-            print('%s failed' % cmd)
-        print('')
-    if error:
-        sys.exit(1)
+    return test(spawn, sleep, options)
 
 
 def all():
     result = [x for x in globals() if x.startswith('bench_') and x != 'bench_all']
-    try:
-        result.sort(key=lambda x: globals()[x].func_code.co_firstlineno)
-    except AttributeError:
-        result.sort(key=lambda x: globals()[x].__code__.co_firstlineno)
+    result.sort()
     result = [x.replace('bench_', '') for x in result]
     return result
 
 
-def all_functions():
-    return [globals()['bench_%s' % x] for x in all()]
-
 
 def main():
-    import argparse
-    parser = argparse.ArgumentParser()
-    parser.add_argument('--with-kwargs', default=False, action='store_true')
-    parser.add_argument('--eventlet-hub')
-    parser.add_argument('--ignore-import-errors', action='store_true')
-    parser.add_argument('benchmark', choices=all() + ['all'])
-    options = parser.parse_args()
-    if options.with_kwargs:
-        options.kwargs = {'foo': 1, 'bar': 'hello'}
-    else:
-        options.kwargs = {}
-    if options.benchmark == 'all':
-        bench_all()
+    def worker_cmd(cmd, args):
+        cmd.extend(args.benchmark)
+
+    runner = perf.Runner(add_cmdline_args=worker_cmd)
+    runner.argparser.add_argument('benchmark',
+                                  nargs='*',
+                                  default='all',
+                                  choices=all() + ['all'])
+
+    def spawn_time(loops, func, options):
+        options.loops = loops
+        times = func(options)
+        return times.spawn_duration
+
+    def sleep_time(loops, func, options):
+        options.loops = loops
+        times = func(options)
+        return times.sleep_duration
+
+    def join_time(loops, func, options):
+        options.loops = loops
+        times = func(options)
+        return times.join_duration
+
+    args = runner.parse_args()
+
+    if 'all' in args.benchmark or args.benchmark == 'all':
+        args.benchmark = ['all']
+        names = all()
     else:
-        function = globals()['bench_' + options.benchmark]
-        function(options)
+        names = args.benchmark
+
+    names = sorted(set(names))
+
+    for name in names:
+        runner.bench_time_func(name + ' spawn',
+                               spawn_time,
+                               globals()['bench_' + name],
+                               Options(False, False),
+                               inner_loops=N)
+
+        if name != 'none':
+            runner.bench_time_func(name + ' sleep',
+                                   sleep_time,
+                                   globals()['bench_' + name],
+                                   Options(True, False),
+                                   inner_loops=N)
+
+    if 'geventpool' in names:
+        runner.bench_time_func('geventpool join',
+                               join_time,
+                               bench_geventpool,
+                               Options(True, True),
+                               inner_loops=N)
+
+    for name in names:
+        runner.bench_time_func(name + ' spawn kwarg',
+                               spawn_time,
+                               globals()['bench_' + name],
+                               Options(False, False, foo=1, bar='hello'),
+                               inner_loops=N)
 
 if __name__ == '__main__':
     main()