[Date Prev][Date Next][Thread Prev][Thread Next]
[Date Index]
[Thread Index]
- Subject: RE: lua dictionary mechanism
- From: "Samson, Lyndon (GEIS)" <Lyndon.Samson@...>
- Date: Wed, 14 Oct 1998 16:38:32 +0100
You want some sort of shared memory space,
Under Win32 threads can have there own copys of global variables, called
thread local storage (TLS)
I think you could declare
extern lua_State *lua_state;
as TLS and have each thread running an independant lua VM, each with its own
variables.
You would have to
lua_setstate (lua_State *st);
to your shared state before any thread accessed the shared area, and this
make the threads own state inacessable, not very
useful...hmm
You will probably end up having to write some C extension functions, called
shared_var_get,shared_var_set which copy
variables between the shared state and the local state.
What would be nice ( on a low level ) is if we could map areas of the VM
address space so they were shared between states,
sort of like what MVS does.
As for UNIX...
lyndon
> -----Original Message-----
> From: Dan Marks [SMTP:dmarks@dionysus.phs.uiuc.edu]
> Sent: Wednesday, October 14, 1998 2:31 PM
> To: Multiple recipients of list
> Subject: lua dictionary mechanism
>
>
> Is it possible to use the lua dictionary mechanism to keep a data
> space that is separate from the global variables in lua? I would like
> to create a separate variable namespace that is accessible from
> several currently running lua threads. To do this, I wanted to create
> a new kind of lua_tag that would have the settable/gettable redefined
> to retrieve the objects from this new table, so that many lua instances
> could retrieve this data from the same namespace.
>
> I was thinking of creating a separate lua interpreter instance that would
> be for just storing global variables, and then I would just use a
> semaphore to block so that only one other thread could access it at once.
> But then I have to duplicate the variable instance to prevent lua_Objects
> from this instance to being passed back to other threads, which would
> no longer synchronize on the semaphore used to prevent concurrent access.
>
> Maybe this will just have to wait for a multithreaded lua. ?
>
> Dan
>
> Dan Marks
> dmarks@uiuc.edu