Subject: Re: [boost] [Parallel Algorithms] Queries regarding the Boost.Thread / Parallel algorithms
From: Vicente J. Botet Escriba (vicente.botet_at_[hidden])
Date: 2014-03-19 14:59:32
Le 19/03/14 01:37, NAVEEN NAMASHIVAYAM a écrit :
> Hi,
Hi,
> I am interested in the implementation of parallel algorithms. From given
> documentation in
> http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3429.pdf
> I could understand the basic crux of the project. But the part which I was
> not able to figure from the documentation are these;
>
> 1. The type of network layer or the communication layer to be used. Can we
> make use of GASNet or ARMCI as the communication layer and build of ideas
> on it? Or do we have any specific library to work on this?
I don't see where you will need such a communication layer. IMHO, the
proposal is restricted to one process only. I think that going further
is out of the scope of the project (if you are interested in parallelism
inter process/, you maybe you can see HPX.
>
> 2. And the implementation of parallel_for and parallel_for_each, are we
> trying to mimic the available functionality from the parallel programming
> models like OpenMP? How far are these functions different from them.
I don't know OpenMP :(
>
> 3. Should we analyze about compiler optimizations. Because once these
> external parallel libraries are called; this necessarily means all the
> compiler optimizations within this parallel region should be excluded or
> else the complexity of the parallel library would be increased.
Could you elaborate?
HTH,
Vicente