利用常規式軌跡模擬器進行設計空間探索之精準度增進

Making Ordinary Trace-driven Simulators Accurate for Design Space Exploitation

王泰元(Tai-Yuan Wang)

指導教授 ： 金仲達

國立清華大學/電機資訊學院/資訊工程學系/碩士(2014年)

https://doi.org/10.6843/NTHU.2014.00492

摘要

隨著晶片網路複雜度的增加，設計空間探索變得更加的困難。因此，軌跡式模擬器所能提供的模擬速度在設計空間探索當中是不可或缺的。然而，軌跡式模擬所產生的時序卻常常與實際情形相去甚遠，主要原因是在傳統的軌跡式模擬當中，時序是固定的，也無法基於模擬結果做動態的調節。具有處理相依性軌跡的軌跡式模擬器可以解決這個問題，但是這樣的模擬器需要緊密的連結輸入的軌跡，而許多傳統的模擬器以批次的方式模擬軌跡，若要達到上述的目的則必須對模擬器做許多更改及了解。
這篇論文以一個迴路的方式解決這個問題，除了輸入軌跡格式外，不需要對模擬器有多餘的修正或了解，我們使用此方法模擬晶片系統網路，最高可以達到28.75%的精準度增進，相較於全系統模擬器也只有平均3.73%的誤差。

關鍵字

軌跡式模擬 ； 設計 ； 效能 ； 晶片網路

並列摘要

The raising complexity of system architectures makes it more difficult to exploit the design space. For design space exploitation, trace-driven simulation is becoming indispensable due to its simplicity and speed. Unfortunately, the timing produced by trace-driven simulation is often inaccurate by a large margin. This is because the event timing in traditional trace-driven simulators is fixed and cannot be adjusted dynamically according to results out of the simulators. Dependency-aware simulators solve the accuracy problem, but they require close interactions between the input traces and the simulators. Many traditional simulators process trace events in batch, which require non-negligible modification efforts to make them interactive. This thesis solves the problem by adopting an iterative strategy. There is no need to modify the simulators, except perhaps in reading and reformatting the input traces. Our evaluations using Gem5 and BookSim for Network-on-Chip (NoC) simulation show that this technique can achieve up to 28.75% improvement in timing accuracy compared to the traditional trace-driven simulators, and result in only 3.73% error in average in terms of network latency compared to full-system simulators.

並列關鍵字

Trace based simulation ； Design ； Performance ； on-chip networks

參考文獻

[1] N. Agarwal, T. Krishna, Li-Shiuan Peh, and N.K. Jha. Garnet: A detailed on-chip network model inside a full-system simulator. In Performance Analysis of Systems and Software, 2009. ISPASS 2009. IEEE International Symposium on, pages 33–42, April 2009.

[2] Jung Ho Ahn, Sheng Li, O. Seongil, and N.P. Jouppi. Mcsima+: A manycore simulator with application-level+ simulation and detailed microarchitecture modeling. In Performance Analysis of Systems and Software (ISPASS), 2013 IEEE International Symposium on, pages 74–85, April

[3] Nathan Binkert, Bradford Beckmann, Gabriel Black, Steven K. Reinhardt, Ali Saidi, Arkaprava Basu, Joel Hestness, Derek R. Hower, Tushar Krishna, Somayeh Sardashti, Rathijit Sen, Korey Sewell, Muhammad Shoaib, Nilay Vaish, Mark D. Hill, and David A. Wood. The gem5 simulator. SIGARCH Comput. Archit. News, 39(2):1–7, August 2011.

[4] D. R. Butenhof. Programming with POSIX threads. 1997.

[5] S. Fytraki and D. Pnevmatikatos. Resim, a trace-driven, reconfigurable ilp processor simulator.