Re: GNAT for MS Visual Studio

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On Sat, 06 Oct 2007 13:32:47 -0700, in comp.lang.ada you wrote:

> On 6 Pa , 20:30, "Dmitry A. Kazakov" <mail...@dmitry-kazakov.de>
> wrote:
> 
>>>    type Whatever is tagged record
>>>       ...
>>>    end record;
>>
>>> You can definitely read an object of type Whatever from multiple tasks.
>>> You don't have to use a protected object.
>>
>>> You can also read the memory pointed to any access type from multiple
>>> tasks. You don't have to use a protected object.
>>
>> AFAIK, ARM is silent about this. The most close reference is 9(8):
>>
>> "In addition, tasks can communicate indirectly by reading and updating
>> (unprotected) shared variables, presuming the access is properly
>> synchronized through some other kind of task interaction."
> 
> Note that introducing some causal ordering (a.k.a. signalling) is
> enough to make it safe.
>
> Consider a main task that populates the shared object *and then*
> starts N tasks that concurrently read the shared data (the main task
> does not interfere with the data anymore). There is a causal
> relationship between populating and reading which is guaranteed by
> task activation. See 9.10/4. The tasks can safely read the data.
> 
> In other words, it is perfectly imaginable to set up a set of tasks
> that concurrently read the same data without any additional hassle
> like protected objects. So - Pascal is right that this is in principle
> possible.
> At the same time this is explicitly forbidden for containers.

I don't see any difference for containers. When reads are synchronized (by
whatever means) then everything is OK, so would be container reading.

[ For protected objects, it seems that functions do not signal, so I guess
it might be illegal to read a container component of a protected object
from its function unless the compiler chooses another implementation for
access from protected functions. This looks weird to me.]

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de