QEMU是一個可以模擬多個平台的系統模擬器。QEMU雖然可以模擬多核心，但他只能循序執行方式模擬CPU。所以沒有發揮底層多核心的機器架構。PQEMU是設計成多執行緒的QEMU，讓模擬器充份發輝多核心硬體的較能。目前PQEMU只能模擬目標ARM 平台。我們把模擬目標x86_64平台移植到PQEMU內。x86_64平台可以支援到255核心。由於PQEMU是多個執行緒同時執行。由於目前的PQEMU對中斷控制器沒有做有較保護，所以當有2個或以上的執行緒同時對中斷控制器操作時會發生競賽情況。所以為了避免發生競賽情況我們設計了2種設計方案，PQEMU-BUS和PQEMU-LOCKS。為了評估這設計，我們在Dell PowerEdge R910機器上分別用QEMU，PQEMU-BUS和PQEMU-LOCKS來模擬x86_64平台，並測量中斷延遲。由於QEMU模擬CPU超過16 cores時啟動Linux作業系統失敗，所以QEMU部份只量測到16 cores。PQEMU-BUS和PQEMU-LOCKS部份量測到255 cores。從測量結果顯示, 在模擬16 cores下PQEMU-BUS和QEMU-LOCKS的外部裝置中斷延遲比QEMU分別減少6.08%和45.92%。PQEMU-BUS和QEMU-LOCKS的CPU之間的中斷延遲比QEMU分別減少99.23%和99.71%。在模擬255 cores下，PQEMU-LOCKS的外部裝置中斷延遲和CPU之間的中斷延遲比PQEMU-BUS分別減少30.49%和43.44%。結果可看出這2個設計不只能避免競賽情況，而且對中斷延遲的減少也有幫助。從結果也可看出PQEMU-LOCKS的設計比PQEMU-BUS設計更能減少中斷延遲。

QEMU is a system emulator that can emulate multiple platforms. Although QEMU can simulate the multi-core, but is round robin sequential execution in single thread, so can’t use advantage of underlying multi-core hardware. PQEMU is designed the multi-thread QEMU, so can use advantage of underlying multi-core hardware. Currently PQEMU emulator support target ARM platform. We ported target x86_64 platform to PQEMU. X86_64 platform support up to 255 cores emulation. PQEMU all threads concurrently execution. Currently PQEMU do not protect interrupt controller, so 2 or more threads to access interrupt controller, can cause race condition. We will implement two kinds of design avoid race condition, in PQEMU. These two kinds of design are PQEMU-BUS and PQEMU-LOCKS. To evaluate the design, we emulate a target x86_64 platform by separately running QEMU, PQEMU-BUS and PQEMU-LOCKS on Dell PowerEdge R910 machine, and measure them interrupt latency. QEMU emulating more than 16 cores, will not booting Linux operating system. So, QEMU measure to 16 cores. PQEMU-BUS and PQEMU-LOCKS measure to 255 cores. The experimental results show that in emulation 16 cores, PQEMU-BUS and PQEMU-LOCKS external interrupt latency had separately decreased 6.08% and 45.92% compared to QEMU. PQEMU-BUS and PQEMU-LOCKS interprocessor interrupt latency had separately decreased 99.23% and 99.71% compared to QEMU. In emulation 255 cores, PQEMU-LOCKS external interrupt latency and interprocessor interrupt latency had separately decreased 30.49% and 43.44% compared to PQEMU-BUS. The results show that the two designs avoid the race condition, and the decreased interrupt latency. The results show, PQEMU-LOCKS interrupt latency less than PQEMU-BUS.

[1] J.H. Ding, P.C. Chang, W.C. Hsu and Y.C. Chung, "PQEMU: A Parallel System Emulator Based on QEMU". Parallel and Distributed Systems (ICPADS), 2011 IEEE 17th International Conference on , vol., no., pp.276-283, 7-9 Dec. 2011
[2] Intel 64 and IA-32 Architectures Software Developer’s Manual, Volume 3A: System Programming Guide, Part 1, Chapter 10 Advanced Programmable Interrupt Controller(APIC)
[3] R. Love, "Linux Kernel Development". Third Edition, Chapter 11, 2010
[4] F. Bellard, "QEMU, a fast and portable dynamic translator". In Proc. of the USENIX Annual Technical Conference, pages 41–46, April 2005.
[5] M. Rosenblum, S.A. Herrod, E. Witchel, and A. Gupta, "Complete computer system simulation: the SimOS approach," Parallel & Distributed Technology: Systems & Applications, IEEE , vol.3, no.4, pp.34-43, 1995
[6] P.S. Magnusson, M. Christensson, J. Eskilson, D. Forsgren, G. Hallberg, J. Hogberg, F. Larsson, A. Moestedt, B. Werner, "Simics: A full system simulation platform", IEEE Computer , vol.35, no.2, pp.50-58, Feb 2002
[7] E. Witchel and R. Rosenblum, "Embra: fast and flexible machine simulation". In Proc. of the SIGMETRICS ’96 Conference on Measurement and Modeling of Computer Systems. 68-78, 1996
[8] Z.Wang, R. Liu, Y. Chen, X.Wu, H. Chen, W. Zhang, and B. Zang, "COREMU:a scalable and portable parallel full-system emulator". In Proc. PPoPP, 2011
[9] hackbench. http://people.redhat.com/mingo/cfs-scheduler/tools/hackbench.c
[10] R.F. Cmelik and D. Keppel, “Shade: a fast instruction set simulator for execution profiling,” Technical Report UWCSE-93-06-06, Dept. Computer Science and Engineering, University of Washington, 1993