Loading...
Please wait, while we are loading the content...
Similar Documents
Performance Evolution
| Content Provider | CiteSeerX |
|---|---|
| Author | Pai, Vijay S. |
| Abstract | Looking back over the past 15 years, the architectural community has seen considerable consolidation in choices of quantitative performance evaluation methods. Previous generations of architecture conferences featured a gamut of evaluation schemes for new architectural proposals, including mean-value analysis and other analytical models, trace-driven simulation of various system components, execution-driven simulation based on application instrumentation, instruction-level emulators, and detailed microarchitectural simulators. Workloads also were highly divergent, often test cases hand-crafted to expose some particular architectural feature. In contrast, the overwhelming majority of papers published in major architectural conferences today use variants on a few widely-distributed detailed microarchitectural simulators (e.g., Simplescalar and, to a lesser extent, RSIM), and workloads focus primarily on SPEC, with some contributions from Olden, SPLASH, and TPC. Even publications on emerging areas of concern such as technology and power have standardized to a small number of simulators. This paper contends that such consolidation has unwarranted side-effects on the |
| File Format | |
| Access Restriction | Open |
| Subject Keyword | Mean-value Analysis Major Architectural Conference Today Instruction-level Emulator Architecture Conference Architectural Community Overwhelming Majority Widely-distributed Detailed Microarchitectural Simulator Execution-driven Simulation Trace-driven Simulation Detailed Microarchitectural Simulator Particular Architectural Feature Quantitative Performance Evaluation Method Evaluation Scheme Various System Component Application Instrumentation Even Publication New Architectural Proposal Previous Generation Analytical Model Considerable Consolidation |
| Content Type | Text |
| Resource Type | Article |