| msg261539 - (view) |
Author: Daniel Stokes (Daniel Stokes) * |
Date: 2016-03-11 03:15 |
| Adding the SIO_LOOPBACK_FAST_PATH option to socket.ioctl on Windows allows Windows users to enable the loopback fast path option available on Windows 8+. This allows for much better TCP loopback performance on Windows. For more information on TCP Loopback Fast Path, see: * https://msdn.microsoft.com/en-us/library/windows/desktop/jj841212%28v=vs.85%29.aspx * http://blogs.technet.com/b/wincat/archive/2012/12/05/fast-tcp-loopback-performance-and-low-latency-with-windows-server-2012-tcp-loopback-fast-path.aspx |
|
|
| msg264326 - (view) |
Author: Daniel Stokes (Daniel Stokes) * |
Date: 2016-04-26 20:21 |
| The "Lifecycle of a Patch" document recommends pinging an issue after a month of no review. I do not see any special ping option, so I am assuming that means to post a comment. To help with code review I tried to keep this change very similar to a previous similar change: http://bugs.python.org/issue6971 |
|
|
| msg264335 - (view) |
Author: Berker Peksag (berker.peksag) *  |
Date: 2016-04-27 00:53 |
| Thanks for pinging, Daniel. I'm not a Windows user so I can't test the patch, but I left some review comments on Rietveld: http://bugs.python.org/review/26536/ |
|
|
| msg265805 - (view) |
Author: Daniel Stokes (Daniel Stokes) * |
Date: 2016-05-18 04:52 |
| Thank you for the review. I have made the changes suggested in the code and I have uploaded a new patch. |
|
|
| msg265845 - (view) |
Author: Daniel Stokes (Daniel Stokes) * |
Date: 2016-05-19 04:52 |
| I have added a zip containing a simple server and client to test the patch with. With loopback fast path enabled, I get 3-5x more throughput on my Windows 10 computer with an AMD 900 series chipset. |
|
|
| msg265847 - (view) |
Author: Berker Peksag (berker.peksag) * |
Date: 2016-05-19 05:36 |
| Thanks for the test script. I left some minor comments about loopback_fast_path_v2.patch on Rietveld. |
|
|
| msg265906 - (view) |
Author: Daniel Stokes (Daniel Stokes) * |
Date: 2016-05-20 01:53 |
| I have uploaded a new patch with the suggested changes, thanks for the review. |
|
|
| msg268738 - (view) |
Author: Roundup Robot (python-dev)  |
Date: 2016-06-17 19:52 |
| New changeset f8957c755c7a by Steve Dower in branch 'default': Issue #26536: socket.ioctl now supports SIO_LOOPBACK_FAST_PATH. Patch by Daniel Stokes. https://hg.python.org/cpython/rev/f8957c755c7a |
|
|
| msg268739 - (view) |
Author: Steve Dower (steve.dower) * |
Date: 2016-06-17 19:53 |
| All looked good to me, and as far as I could see all of Berker's feedback was addressed, so consider it in! |
|
|
| msg268777 - (view) |
Author: Martin Panter (martin.panter) * |
Date: 2016-06-18 09:02 |
| This is failing on Windows 7 buildbots: http://buildbot.python.org/all/builders/AMD64%20Windows7%20SP1%203.x/builds/7849/steps/test/logs/stdio ====================================================================== ERROR: test_sio_loopback_fast_path (test.test_socket.GeneralModuleTests) ---------------------------------------------------------------------- Traceback (most recent call last): File "C:\buildbot.python.org\3.x.kloth-win64\build\lib\test\test_socket.py", line 1226, in test_sio_loopback_fast_path s.ioctl(socket.SIO_LOOPBACK_FAST_PATH, True) OSError: [WinError 10045] The attempted operation is not supported for the type of object referenced |
|
|
| msg268801 - (view) |
Author: Roundup Robot (python-dev) |
Date: 2016-06-18 13:09 |
| New changeset e8b10ed64e63 by Berker Peksag in branch 'default': Issue #26536: Skip test_sio_loopback_fast_path under Windows 7 https://hg.python.org/cpython/rev/e8b10ed64e63 |
|
|
| msg268802 - (view) |
Author: Berker Peksag (berker.peksag) * |
Date: 2016-06-18 13:40 |
| It looks like SIO_LOOPBACK_FAST_PATH is also defined in older Windows versions. I updated the test to skip if SIO_LOOPBACK_FAST_PATH is defined and exc.winerror is WSAEOPNOTSUPP. Buildbot is now green: http://buildbot.python.org/all/builders/AMD64%20Windows7%20SP1%203.x/builds/7855 Please reopen if I did something wrong. |
|
|
| msg268803 - (view) |
Author: Roundup Robot (python-dev) |
Date: 2016-06-18 13:42 |
| New changeset 213c240cce0a by Berker Peksag in branch 'default': Issue #26536: Use spaces instead of tabs https://hg.python.org/cpython/rev/213c240cce0a |
|
|
| msg268805 - (view) |
Author: Steve Dower (steve.dower) * |
Date: 2016-06-18 14:02 |
| That'll handle the test fine, though I wonder whether we should try and conditionally define the constant at runtime? Probably at least want to add a note to the docs about Win 8 being the minimum, especially since so many people are still on 7. |
|
|
| msg268806 - (view) |
Author: Berker Peksag (berker.peksag) * |
Date: 2016-06-18 14:18 |
| > That'll handle the test fine, though I wonder whether we should try and conditionally define the constant at runtime? +1. Is it safe to use IsWindows8OrGreater()? |
|
|
| msg268816 - (view) |
Author: Steve Dower (steve.dower) * |
Date: 2016-06-18 17:17 |
| That's the best function to use. |
|
|