[Boost-users] graph theoretical problem

Dear All, does the boost graph library provide a solution to the problem I'm trying to solve: I want to distribute a set of tasks to multiple CPUs/Threads. The tasks have predecessor and successor dependencies. There are no loops. More often than not, a task consists of only very few operations. So if I brute force program this, the resulting code is busy only with scheduling -- the real code does not show up in the profiling results. I think this can be solved by collapsing some tasks into a single task to minimize scheduling/communication. I've got a cost function associated with every task. Is this some well known algorithm which is already available in boost graph? Kind Regards Peter Foelsche