applicationspecifictopologyawaremappingforthreedi应用程序特定的拓扑映射三迪.pptxVIP

会计学 1 ApplicationspecificTopologyawareMappingforThreeDi应用程序特定的拓扑映射三迪 2 The network latency for wormhole routing is (Lf/B)*D + L/B Lf = Length of each flit, B = bandwidth D = number of hops, L = length of message Lionel M. Ni and Philip K. McKinley, “A Survey of Wormhole Routing Techniques in Direct Networks”, Computer, Volume 26, Issue 2, pages 62-76, 1993 第1页/共22页 3 Message Latencies NN = Near Neighbor, RND = Random 第2页/共22页 4 Hardware Latencies Blue Gene/L Near neighbor: 1 µs Worst case: 7 µs Blue Gene/P Near neighbor: 1 µs Worst case: 5 µs Corresponding differences for MPI messages 第3页/共22页 5 Topology-aware mapping Problem: Given a object communication graph and a processor graph, find an optimal mapping Minimizes communication Ensure load balance Metric for communication traffic Hop-bytes = number of links (hops) traversed X message size 第4页/共22页 6 Machine Topology Information required at runtime No. of processors in the allocated partition No. of processors along each dimension Physical coordinates of each processor 第5页/共22页 7 第6页/共22页 8 Communication Graph Static 3D Stencil: regular communication graph Dynamic Molecular dynamics application Changes as atoms migrate from one processor to another 第7页/共22页 9 Static Graph - 3D Stencil 第8页/共22页 10 Performance 第9页/共22页 11 Hop counts 第10页/共22页 12 Dynamic Graph - NAMD Molecular Dynamics (MD) application Simulation box is a 3D cell full of atoms 第11页/共22页 13 第12页/共22页 14 Load Balancing in NAMD Measurement-based (Charm++) Principle of persistence Patches are statically mapped Orthogonal recursive bisection Computes can be migrated Load balancing framework gathers the communication information Goal Minimize communication Maximize load balance 第13页/共22页 15 第14页/共22页 16 Old strategy Greedy approach Pick the heaviest compute Place it on a processor with one of the patches OR On a processor which already has a compute for this patch 第15页/共22页 17 第16页/共22页 18 Hop-bytes