🚀 go-pugleaf

"Remy Lebeau \(TeamB\)" <no.spam@no.spam.com> wrote:

>
>> Really because one variable can be changed to some variables or
>> structure via refactoring the using semephors and critical sections
>> preferable. If think that better way will be if reading thread creates
>> another thread that will have sole task - listen/watch for
>> variables.
>
>That just complicates the design even more, and doesn't even solve the
>problem at all.  You still have to synchronize, and now you've introduced
>another thread that still has to synchronize as well
>

You can't track all changes of variables and do another work 
in the one thread.

>> This thread (in my opinion) must use semaphor for watching for
>variable(s).
>
>You just contradicted yourself.  First you said that synchronization is not
>needed, but then you offer an approach that uses synchronization.  

I didn't contradicted. I talked about easist case.

Intel has operation that executed as primitives and in easist 
case no need to use critical sections. You said before about 
Win32 functions that represent this operations (InterlockedExchange, InterlockedExchangeAdd). 

>But it
>still won't work.  A critical section or semaphor wil not work unless ALL
>threads accessing the shared data use them.  Otherwise they do not provide
>any protection at all.

Off course, it is no problem there and I said about it in one 
my message before.
Vladimir.