Skip to content

bpo-30919: shared memory allocation performance regression in multiprocessing #2708

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
pitrou merged 5 commits into from
Jul 23, 2017
Merged

bpo-30919: shared memory allocation performance regression in multiprocessing #2708

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
8834c32
Fix #30919: shared memory allocation performance regression in multip…
pitrou Jul 14, 2017
c2ba94e
Merge branch 'master' of https://github.com/python/cpython into mp_he…
pitrou Jul 17, 2017
a9d1686
Change strategy for Arena directory choice
pitrou Jul 17, 2017
982d304
Merge branch 'master' of https://github.com/python/cpython into mp_he…
pitrou Jul 23, 2017
3bb7c2c
Add blurb
pitrou Jul 23, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 16 additions & 10 deletions Lib/multiprocessing/heap.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,26 +60,32 @@ def __setstate__(self, state):
else:

class Arena(object):
if sys.platform == 'linux':
_dir_candidates = ['/dev/shm']
else:
_dir_candidates = []

def __init__(self, size, fd=-1):
self.size = size
self.fd = fd
if fd == -1:
self.fd, name = tempfile.mkstemp(
prefix='pym-%d-'%os.getpid(), dir=util.get_temp_dir())
prefix='pym-%d-'%os.getpid(),
dir=self._choose_dir(size))
os.unlink(name)
util.Finalize(self, os.close, (self.fd,))
with open(self.fd, 'wb', closefd=False) as f:
bs = 1024 * 1024
if size >= bs:
zeros = b'\0' * bs
for _ in range(size // bs):
f.write(zeros)
del zeros
f.write(b'\0' * (size % bs))
assert f.tell() == size
os.ftruncate(self.fd, size)
self.buffer = mmap.mmap(self.fd, self.size)

def _choose_dir(self, size):
# Choose a non-storage backed directory if possible,
# to improve performance
for d in self._dir_candidates:
st = os.statvfs(d)
if st.f_bavail * st.f_frsize >= size: # enough free space?
return d
return util.get_temp_dir()

def reduce_arena(a):
if a.fd == -1:
raise ValueError('Arena is unpicklable because '
Expand Down
4 changes: 4 additions & 0 deletions Misc/NEWS.d/next/Library/2017-07-23-11-33-10.bpo-30919.5dYRru.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,4 @@
Fix shared memory performance regression in multiprocessing in 3.x.

Shared memory used anonymous memory mappings in 2.x, while 3.x mmaps actual
files. Try to be careful to do as little disk I/O as possible.