We present a novel modeling methodology which enables the generation of a high-performance, cycle-accurate simulator from a cycle-level specification of the target design. We describe Arete, a full-system multicore processor simulator, developed using our modeling methodology. We provide details on Arete's resource-efficient and high-performance implementation on multiple FPGA platforms, and the architectural experiments performed using it.We present clear evidence that the use of simplified models in architectural studies can lead to wrong conclusions. Through two experiments performed using both cycle-accurate and simplified models, we show that on one hand there are substantial quantitative and qualitative differences in results, and on the other, the results match quite well. (Copies available exclusively from MIT Libraries, libraries.mit.edu/docs - [email protected])
Recommendations
Per-thread cycle accounting in multicore processors
Special Issue on High-Performance Embedded Architectures and CompilersWhile multicore processors improve overall chip throughput and hardware utilization, resource sharing among the cores leads to unpredictable performance for the individual threads running on a multicore processor. Unpredictable per-thread performance ...