X10: an object-oriented approach to non-uniform cluster computing
OOPSLA '05 Proceedings of the 20th annual ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications
The 48-core SCC Processor: the Programmer's View
Proceedings of the 2010 ACM/IEEE International Conference for High Performance Computing, Networking, Storage and Analysis
X10 as a Parallel Language for Scientific Computation: Practice and Experience
IPDPS '11 Proceedings of the 2011 IEEE International Parallel & Distributed Processing Symposium
Exploring cross-layer power management for PGAS applications on the SCC platform
Proceedings of the 21st international symposium on High-Performance Parallel and Distributed Computing
Wait-Free message passing protocol for non-coherent shared memory architectures
EuroMPI'12 Proceedings of the 19th European conference on Recent Advances in the Message Passing Interface
| Hi-index | 0.00 |
The Single-Chip Cloud Computer (SCC) is an experimental processor created by Intel Labs. SCC is essentially a 'cluster-on-a-chip', so X10 with its support for places and remote asynchronous invocations is a natural fit for programming this platform. We report here on our experience porting X10 to the SCC, and show performance and scaling results for representative X10 benchmark applications. We compare results for our extensions to the SCC native messaging primitives in support of the X10 run-time, versus X10 on top of a prototype MPI API for SCC. The native SCC run-time exhibits better performance and scaling than the MPI binding. Scaling depends on the relative cost of computation versus communication in the workload used, since SCC is relatively underpowered for computation but has hardware support for message passing.