[Python-Dev] Making the GIL faster & lighter on Windows (original) (raw)
Curt Hagenlocher curt at hagenlocher.org
Wed May 27 14:59:48 CEST 2009
- Previous message: [Python-Dev] Making the GIL faster & lighter on Windows
- Next message: [Python-Dev] Making the GIL faster & lighter on Windows
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
On Wed, May 27, 2009 at 4:24 AM, Nick Coghlan <ncoghlan at gmail.com> wrote:
CriticalSections are first come first served on Windows, just like a regular mutex.
"Starting with Windows Server 2003 with Service Pack 1 (SP1), threads waiting on a critical section do not acquire the critical section on a first-come, first-serve basis."
http://msdn.microsoft.com/en-us/library/ms682530(VS.85).aspx
Windows critical sections use events for kernel-level synchronization. The user-mode code basically consists of an interlocked instruction inside the spin loop. When the likelihood of contention is low, a critical section should be a big win because it won't need to switch into the kernel. I suspect that contention will be frequent for the GIL
A good description of pre-Vista Windows critical sections can be found here: http://msdn.microsoft.com/en-us/magazine/cc164040.aspx
-- Curt Hagenlocher curt at hagenlocher.org
- Previous message: [Python-Dev] Making the GIL faster & lighter on Windows
- Next message: [Python-Dev] Making the GIL faster & lighter on Windows
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]