Communications of the ACM - Special issue on parallelism
Graphinators and the duality of SIMD and MIMD
LFP '88 Proceedings of the 1988 ACM conference on LISP and functional programming
Data parallel simulation using time-warp on the connection machine
DAC '89 Proceedings of the 26th ACM/IEEE Design Automation Conference
MIND execution by SIMD computers
Journal of Information Processing
IEEE Transactions on Parallel and Distributed Systems
Probability and Statistics with Reliability, Queuing and Computer Science Applications
Probability and Statistics with Reliability, Queuing and Computer Science Applications
Asynchronous Problems on SIMD Parallel Computers
IEEE Transactions on Parallel and Distributed Systems
A Control-Parallel Programming Model Implemented on SIMD
Proceedings of the 5th International Workshop on Languages and Compilers for Parallel Computing
Shared Control - Supporting Control Parallelism Using a SIMD-like Architecture
Euro-Par '98 Proceedings of the 4th International Euro-Par Conference on Parallel Processing
| Hi-index | 0.00 |
We consider two models for the structure of the algorithm used for concurrent interpretation of MIMD code sequences on SIMD machines. The single-fetch model shares portions of the instruction execution among all the instructions, minimizing the interpreter length. Because the fetch of the next instruction is shared, only one instruction is executed at each PE. In contrast, the multiple-fetch model allows an instruction to be fetched after every instruction in the interpreter program. While the interpreter cycle is longer for this model, it is possible for each PE to interpret more than one instruction per cycle. In this paper, a mathematical characterization of the two models is developed and used to show that, for a given operation set, the relative performance of the two models depends on the granularity of the operations. If the granularity of the operations is large, the sharing of common portions of the interpreter cycle (single-fetch model) harms throughput.