[Boost-users] [graph] Example parallel BGL

[Sensei]

Dear all,

I'd like to play with Parallel BGL (with MPI), and I'd like to see some
example code. I've followed the Overview and the Adjacency List example,
but I cannot compile my code.

Right now, I'd like to create a graph, and then fill in with some nodes.
Then, access the nodes belonging to a single process, but I wish I had
an error there

My error concerns the definition of the graph: No template named
'bsp_process_group' in namespace 'boost::parallel'; did you mean
'boost::process_group'? The code follows.

What am I doing wrong here? I may have forgotten some header... Is there
any distributed (complete) example code that I might read?

Thanks & Cheers!



#include <iostream>
#include <string>
#include <vector>

#include <boost/mpi.hpp>
#include <boost/graph/use_mpi.hpp>
#include <boost/graph/distributed/mpi_process_group.hpp>
#include <boost/graph/distributed/adjacency_list.hpp>

int main(int argc, const char * argv[])
{
boost::mpi::environment env;
boost::mpi::communicator comm;

typedef boost::adjacency_list<boost::vecS,

boost::distributedS<boost::parallel::bsp_process_group, boost::vecS>,
boost::directedS> graph;
typedef boost::erdos_renyi_iterator<boost::minstd_rand, graph>
generator;

if (comm.rank() == 0)
{
boost::add_edge(1, 2);
boost::add_edge(2, 2);
boost::add_edge(2, 3);
boost::add_edge(3, 4);
boost::add_edge(4, 5);
boost::add_edge(5, 1);

// access ***only rank 0 nodes***
}
return 0;
}
_______________________________________________
Boost-users mailing list
Boost...@lists.boost.org
http://lists.boost.org/mailman/listinfo.cgi/boost-users

[Marcin Zalewski]

On Thu, Jul 30, 2015 at 9:51 AM Sensei <sens...@gmail.com> wrote:

Dear all,

I'd like to play with Parallel BGL (with MPI), and I'd like to see some
example code. I've followed the Overview and the Adjacency List example,
but I cannot compile my code.

You can find some examples here:

https://github.com/boostorg/graph_parallel/tree/master/example

You can also look at tests:

https://github.com/boostorg/graph_parallel/tree/master/test

[Sensei]

On 8/3/15 12:18pm, Marcin Zalewski wrote:
> You can find some examples here:
>
> https://github.com/boostorg/graph_parallel/tree/master/example
>
> You can also look at tests:
>
> https://github.com/boostorg/graph_parallel/tree/master/test

Thanks, Marcin, I finally can compile my code!

I wrote a simple graph based on the named graph example (*), in order to
learn, but my code seems to avoid nodes distribution, while the example
does it perfectly.

(*)
https://github.com/boostorg/graph_parallel/blob/master/test/named_vertices_test.cpp


In my source I think I made the same steps, but I'm obviously
mistaken... my example every processor has every node, and every edge
if I comment the if statement.

I think I'm using the wrong boost::add_vertex/edge here, but honestly, I
am a very newbie on the distributed graph side.

Thanks for any pointers you can give me!

#include <iostream>
#include <string>
#include <vector>

#include <cmath>
#include <list>
#include <functional>

#include <boost/mpi.hpp>
#include <boost/graph/use_mpi.hpp>
#include <boost/graph/distributed/mpi_process_group.hpp>
#include <boost/graph/distributed/adjacency_list.hpp>

//using namespace boost;
//using boost::graph::distributed::mpi_process_group;

int main(int argc, const char * argv[])
{
boost::mpi::environment env;
boost::mpi::communicator comm;

std::cout << comm.rank() << std::endl;

typedef boost::adjacency_list<boost::vecS,

boost::distributedS<boost::graph::distributed::mpi_process_group,
boost::vecS>,
boost::directedS, std::string> graph;

graph g;

std::vector<std::string> names { "alpha", "beta", "gamma", "delta" };

std::vector<graph::vertex_descriptor> descs;

for(auto p : names) descs.push_back(boost::add_vertex(p, g));

if (comm.rank() == 0)
{
boost::add_edge(descs[0], descs[0], g);
boost::add_edge(descs[0], descs[1], g);
boost::add_edge(descs[1], descs[2], g);
boost::add_edge(descs[2], descs[3], g);
}

boost::synchronize(g);


BGL_FORALL_VERTICES(v, g, graph)
{
std::cout << "V @ " << comm.rank() << " " << g[v] << std::endl;
}

BGL_FORALL_EDGES(e, g, graph)
{
std::cout << "E @ " << comm.rank() << " " << boost::source(e,
g).local << " -> " << boost::target(e, g).local << " srccpu " <<
e.source_processor << " dstcpu " << e.target_processor << std::endl;
}

return 0;
}


***OUTPUT***
5
4
2
1
3
0
V @ 1 alpha
V @ 1 beta
V @ 1 gamma
V @ 1 delta
E @ 1 0 -> 0 srccpu 1 dstcpu 1
E @ 1 0 -> 1 srccpu 1 dstcpu 1
E @ 1 1 -> 2 srccpu 1 dstcpu 1
E @ 1 2 -> 3 srccpu 1 dstcpu 1
V @ 3 alpha
V @ 3 beta
V @ 3 gamma
V @ 3 delta
E @ 3 0 -> 0 srccpu 3 dstcpu 3
E @ 3 0 -> 1 srccpu 3 dstcpu 3
E @ 3 1 -> 2 srccpu 3 dstcpu 3
E @ 3 2 -> 3 srccpu 3 dstcpu 3
V @ 2 alpha
V @ 2 beta
V @ 2 gamma
V @ 2 delta
E @ 2 0 -> 0 srccpu 2 dstcpu 2
E @ 2 0 -> 1 srccpu 2 dstcpu 2
E @ 2 1 -> 2 srccpu 2 dstcpu 2
E @ 2 2 -> 3 srccpu 2 dstcpu 2
V @ 5 alpha
V @ 5 beta
V @ 5 gamma
V @ 5 delta
E @ 5 0 -> 0 srccpu 5 dstcpu 5
E @ 5 0 -> 1 srccpu 5 dstcpu 5
E @ 5 1 -> 2 srccpu 5 dstcpu 5
E @ 5 2 -> 3 srccpu 5 dstcpu 5
V @ 4 alpha
V @ 4 beta
V @ 4 gamma
V @ 4 delta
E @ 4 0 -> 0 srccpu 4 dstcpu 4
E @ 4 0 -> 1 srccpu 4 dstcpu 4
E @ 4 1 -> 2 srccpu 4 dstcpu 4
E @ 4 2 -> 3 srccpu 4 dstcpu 4
V @ 0 alpha
V @ 0 beta
V @ 0 gamma
V @ 0 delta
E @ 0 0 -> 0 srccpu 0 dstcpu 0
E @ 0 0 -> 1 srccpu 0 dstcpu 0
E @ 0 1 -> 2 srccpu 0 dstcpu 0
E @ 0 2 -> 3 srccpu 0 dstcpu 0