[Python-Dev] Inherance of file descriptor and handles on Windows (PEP 446) (original) (raw)

Antoine Pitrou solipsis at pitrou.net
Sat Jul 27 00:23:59 CEST 2013

On Sat, 27 Jul 2013 00🔞40 +0200 Victor Stinner <victor.stinner at gmail.com> wrote:

2013/7/26 Antoine Pitrou <solipsis at pitrou.net>: > On Fri, 26 Jul 2013 22:17:47 +0200 >> """ >> On Linux, setting the close-on-flag has a low overhead on >> performances. Results of benchcloexec.py on Linux 3.6: >> >> - close-on-flag not set: 7.8 us >> - OCLOEXEC: 1% slower (7.9 us) >> - ioctl(): 3% slower (8.0 us) >> - fcntl(): 3% slower (8.0 us) >> """ > > You aren't answering my question: slower than what?

Ah, you didn't understand the labels. benchcloexec.py runs a benchmark on os.open(path, os.ORDONLY, cloexec=False) and os.open(path, os.ORDONLY, cloexec=True) with different implementation of making the file descriptor non-inheritable. close-on-flag not set: 7.8 us => C code: open(path, ORDONLY) OCLOEXEC: 1% slower (7.9 us) => C code: open(path, ORDONLY|CLOEXEC) => 1% slower than open(path, ORDONLY) ioctl(): 3% slower (8.0 us) => C code: fd=open(path, ORDONLY); ioctl(fd, FIOCLEX, 0) => 3% slower than open(path, ORDONLY) fcntl(): 3% slower (8.0 us) => C code: fd=open(path, ORDONLY); flags = fcntl(fd, FGETFD); fcntl(fd, FSETFD, flags | FDCLOEXEC) => 3% slower than open(path, ORDONLY)

Ok, so I think this it is a totally reasonable compromise.

People who bother about a 3% slowdown when calling os.open() can optimize the hell out of their code using Cython for all I care :-)

Regards

Antoine.

More information about the Python-Dev mailing list