Subject | RE: [IB-Architect] Re: Events Improvement |
---|---|
Author | Jim Starkey |
Post date | 2001-05-11T13:54:30Z |
At 02:39 AM 5/11/01 -0400, Claudio Valderrama C. wrote:
manager) was based on shared memory guarded by semaphores in which
all "pointers" had to be relative and relocated on reference.
Since the super server runs in a single process, there is no
need for the share memory and relative pointers. I haven't
looked to see if there is separate implementation for super-server,
but I would be very surprised to find on. This is probably
another of the many instances of the super-server paying a
complexity and performance tax for sharing a code base with
classic (recognize a familiar theme here?).
The super server still needs synchronization primitives to control
thread access to shared data structure, but the thread services
are a great deal cheaper than SysV semaphores.
Jim Starkey
>The original implementation of the event manager (like the lock
>Can you explain that phrase in more detail, please? I'm not sure I
>understand how this piece of code works. Does SS still suffer from the
>limitations of the classic's event manager, for example?
>
manager) was based on shared memory guarded by semaphores in which
all "pointers" had to be relative and relocated on reference.
Since the super server runs in a single process, there is no
need for the share memory and relative pointers. I haven't
looked to see if there is separate implementation for super-server,
but I would be very surprised to find on. This is probably
another of the many instances of the super-server paying a
complexity and performance tax for sharing a code base with
classic (recognize a familiar theme here?).
The super server still needs synchronization primitives to control
thread access to shared data structure, but the thread services
are a great deal cheaper than SysV semaphores.
Jim Starkey