Communication Overhead Estimation on Multicores

1. Communication Overhead Estimation on Multicores S. M. Farhad The University of Sydney Joint work with Yousun Ko Bernd Burgstaller Bernhard Scholz

2. Outline • Motivation • Multicore trend • Stream programming • Profiling communication overhead • Related works 2

3. 512 PicoChip AMBRIC 256 CISCO CSR1 128 NVIDIA G80 Larrabee 64 Unicore 32 Homogeneous Multicore RAZA XLR Cavium Heterogeneous Multicore 16 RAW Cell Niagara 8 AMD Fusion Opteron 4P BCM 1480 4 Core2Quad Xeon Xbox 360 Power6 Power4 PA8800 PA8800 4004 8008 8080 4004 8008 8080 2 Opteron CoreDuo Core2Duo 8086 286 386 486 Pentium P2 P3 P4 1 Core Athlon Itanium Itanium2 1975 1980 1985 1990 1995 2000 2005 2010 Motivation Stream Programming CUDA Courtesy: Scott’08 X10 Peakstream Fortress C/C++/Java # cores/chip Accelerator Ct C T M Rstream Rapidmind 3

4. Stream Programming Paradigm Programs expressed as stream graphs Streams: Infinite sequence of data elements Actors: Functions applied to streams Stream Actor Stream 4

5. Properties of Stream Program AtoD FMDemod • Regular and repeating computation • Independent actors with explicit communication • Producer / Consumer dependencies Splitter LPF1 LPF2 LPF3 HPF1 HPF2 HPF3 Joiner Adder Speaker 5

6. StreamIt Language filter pipeline • An implementation of stream prog. • Hierarchical structure • Each construct has single input/output stream may be any StreamIt language construct splitjoin parallel computation splitter joiner feedback loop splitter joiner 6

7. How to Estimate the Communication Overhead?

8. Problems to Measure Communication Overhead • Reasons: • Multicores are non-communication exposed architecture • Complex cache hierarchy • Cache coherence protocols • Consequence: • Cannot directly measure the communication cost • Estimate the communication cost by measuring the execution time of actors

9. Measuring the Communication Overhead of an Edge Processor 1 Processor 1 Processor 2 i k No communication cost With communication cost k i

10. Processor 1 Processor 2 Processor 1 Processor 2 A A 1 B 1 2 B 2 C C 3 D 3 D 4 4 E E 5 F Even edges across partition Odd edges across partition How to Minimize the Required Number of Experiments Requires 2+1 Exps A 1 B Graph Coloring 2 C Pipeline

11. Obs. 1: There is no loop of three actors in a stream graph Processor 1 Processor 2 l i k

12. P-1 P-2 P-3 P-4 Obs. 2: There is no interference of adjacent nodes between edges A B C D E F For blue color edges

13. Remove Interference • Convert to a line graph • Add interference edges • Use vertex coloring algorithm A AB AB BC BC B BD BD CE CE C D DE DE E EF EF F Line graph Stream graph

14. Processor Leveling Graph A B A C D B, C, D, E E F F For blue colored edge Processor leveling graph

15. Coloring the Processor Labelling Graph Processor 1 Processor 2 A A A B, C, D, E B, C, D, E B, C, D, E F F F

16. Measuring the Communication Cost Processor 1 Processor 2 A A B B, C, D, E C D E F F For blue colored edge

17. Profiling Performance

18. Related Works [1] Static Scheduling of SDF Programs for DSP [Lee ‘87] [2] StreamIt: A language for streaming applications [Thies ‘02] [3] Phased Scheduling of Stream Programs [Thies ’03] [4] Exploiting Coarse Grained Task, Data, and Pipeline Parallelism in Stream Programs [Thies ‘06] [5] Orchestrating the Execution of Stream Programs on Cell [Scott ’08] [6] Software Pipelined Execution of Stream Programs on GPUs [Udupa‘09] [7] Synergistic Execution of Stream Programs on Multicores with Accelerators [Udupa ‘09] [8] Orchestration by approximation [Farhad ‘11] 18

19. Questions?