What sort of synchronization mechanisms does D provide for multiple threads?  Is there any standard implementation of structures like locks / memory barriers? Java has the methods wait() and notify() built in each object, and C/C++ has many libraries which can get the job done.  What is there in D/Phobos?

"mclysenk" <mclysenk_member@pathlink.com> wrote in message news:cordtd$fbo$1@digitaldaemon.com...
> What sort of synchronization mechanisms does D provide for multiple
> threads?  Is
> there any standard implementation of structures like locks / memory
> barriers?
> Java has the methods wait() and notify() built in each object, and C/C++
> has
> many libraries which can get the job done.  What is there in D/Phobos?
>
>

The D language has basic critical section support using the "synchronized"
statement or attribute. It doesn't have wait/notify or other types of locks.
The "volatile" statement is for memory barriers.
A port of the Java 1.5 concurrency classes is available at
http://home.comcast.net/~benhinkle/mintl/locks.html
It has a reentrant lock with condition variables, countdown, cyclic-barrier,
exchanger, semaphore and a read-write lock.

-Ben

In article <cot4e8$2s7u$1@digitaldaemon.com>, Ben Hinkle says...
>
>
>"mclysenk" <mclysenk_member@pathlink.com> wrote in message news:cordtd$fbo$1@digitaldaemon.com...
>> What sort of synchronization mechanisms does D provide for multiple
>> threads?  Is
>> there any standard implementation of structures like locks / memory
>> barriers?
>> Java has the methods wait() and notify() built in each object, and C/C++
>> has
>> many libraries which can get the job done.  What is there in D/Phobos?
>>
>>
>
>The D language has basic critical section support using the "synchronized"
>statement or attribute. It doesn't have wait/notify or other types of locks.
>The "volatile" statement is for memory barriers.
>A port of the Java 1.5 concurrency classes is available at
>http://home.comcast.net/~benhinkle/mintl/locks.html
>It has a reentrant lock with condition variables, countdown, cyclic-barrier,
>exchanger, semaphore and a read-write lock.

Note that "synchronized" just indicates that the compiler should not optimize past those barriers.  AFAIK there are currently no hardware-level barrier instructions in place, so lockless synchronization methods make a handy supplement.


Sean

> Note that "synchronized" just indicates that the compiler should not optimize

I think you meant "volatile" instead of "synchronized", yes?

> past those barriers.  AFAIK there are currently no hardware-level barrier instructions in place, so lockless synchronization methods make a handy supplement.
>
>
> Sean
>
> 

In article <cotr0v$m1l$1@digitaldaemon.com>, Ben Hinkle says...
>
>> Note that "synchronized" just indicates that the compiler should not optimize
>
>I think you meant "volatile" instead of "synchronized", yes?

Oops.  Yes I did.  That's what I get for posting before having my morning coffee.


Sean