Why our speedup for batching less than nonbatching? Shouldn't the load balancing help?
llcoolj
i think the overall speedup for batching might be more, but this graph is 'relative to 1 thread', for which speedup might be less due to communication costs?
Why our speedup for batching less than nonbatching? Shouldn't the load balancing help?
i think the overall speedup for batching might be more, but this graph is 'relative to 1 thread', for which speedup might be less due to communication costs?