1. technische universität dortmund fakultät für informatik informatik 12 Cyber-Physical Systems: (some) Opportunities and Challenges Peter Marwedel TU Dortmund (Germany) Informatik 12 These slides use Microsoft clip arts. Microsoft copyright restrictions apply. © Springer, 2010 2012 年 11 月 22 日

2. - 2 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Cyber-physical systems and embedded systems CPS = ES + physical environment Embedded systems ("computers in physical environments") Embedded systems ("small computers") Cyber-physical systems  Embedded systems (ES): information processing embedded into a larger product [Peter Marwedel, 2003]  Cyber-physical systems (CPS): integrations of computation with physical processes [Edward Lee, 2006].

3. - 3 - technische universität dortmund fakultät für informatik  Scope  Opportunities  Challenges Multidisciplinarity, timing predictability, A/D interface, energy efficiency, dependability, diversity of objectives  (Partial) solutions to challenges Timing predictability WCET-aware compilation, flexible fault handling, PAMONO Energy efficiency … PAMONO, scratch pads Educational challenges Concentrations, integrated programs, c ourse material  Summary  P.Marwedel, Informatik 12, 2012 Outline

4. - 4 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Opportunities ©: Microsoft Cliparts + P. Marwedel, 2011,  Transportation Automotive Avionics Rail  Medical Hospitals Support of handicapped Patient information © www.dobelle.com, ~2000

5. - 5 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Opportunities (2)  (Urban) living Telecommunication Consumer electronics Smart home Logistics Smart grid Public safety  Robotics  Military systems © Graphics: P. Marwedel, 2011, Microsoft

6. - 6 - technische universität dortmund fakultät für informatik  Scope  Opportunities  Challenges Multidisciplinarity, timing predictability, A/D interface, energy efficiency, dependability, diversity of objectives  (Partial) solutions to challenges Timing predictability WCET-aware compilation, flexible fault handling, PAMONO Energy efficiency … PAMONO, scratch pads Educational challenges Concentrations, integrated programs, c ourse material  Summary  P.Marwedel, Informatik 12, 2012 Outline

7. - 7 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Challenge (1): CPS design is multidisciplinary CS Physics Knowledge from many areas must be available, Walls between disciplines must be torn down EE medicine, statistics, ME, biology

8. - 8 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Challenge (2): timing predictability Timing: The system must react within a time interval dictated by the environment [Adopted from Bergé] t execute  Consequences frequently underestimated  Always! Not subject to statistical arguments (except possibly extremely small probabilities) © Graphics: Microsoft Cliparts

9. - 9 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Challenge (3): Impact of imprecise measurements and noise  Verification problem Implementation will not correspond to specification exactly  Boolean decision “specification is met/not met” has to be replaced by a level of confidence l, 0 ≤ l ≤ 1  New verification techniques required George Pappas: Science of Cyber-Physical Systems Bridging CS and Control, 2012 NSF CPS PI meeting, http://cps-vo.org/node/5876  Control loop problem: Shorter sampling intervals will possibly not lead to better approximation of the behavior of analog feedback. Non-periodic sampling may provide the best results! Gaps in corresponding theory. Paulo Tabuada: Is it about time for control?, 2012 NSF CPS PI meeting, http://cps-vo.org/node/5997

10. - 10 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Challenges (4): Energy/power efficiency  Need for energy/power efficient processing (“sweet corner”) © Graphics: H. De Man, Microsoft

11. - 11 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Challenge (5): Reliability/dependability Reliability: The system must be reliable © graphics: Microsoft + Triquint Inc (with permission)  Consequences frequently underestimated  Order of magnitude of probabilities: 10 -9 /hr = 1 FIT Failure rates for GaAs at 150°C [www.triquint.com/company/quality/ faqs/faq_11.cfm] Different devices Critical and non-critical systems connected  “Mixed criticality”  virtualization, …

12. - 12 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Challenge (6): Many more objectives  average case delay  safety  security  EMC characteristics  radiation hardness  cost, size  weight  thermal behavior  environmental friendliness  extendability … © Graphics: P. Marwedel, 2011; Microsoft Threatened by increased openness

13. - 13 - technische universität dortmund fakultät für informatik  Scope  Opportunities  Challenges Multidisciplinarity, timing predictability, A/D interface, energy efficiency, dependability, diversity of objectives  (Partial) solutions to challenges Timing predictability WCET-aware compilation, flexible fault handling, PAMONO Energy efficiency … PAMONO, scratch pads Educational challenges Concentrations, integrated programs, c ourse material  Summary  P.Marwedel, Informatik 12, 2012 Outline

14. - 14 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Let’s look at timing! Definition of worst case execution time: WCET EST © Graphics: R. Wilhelm + Microsoft Cliparts WCET EST must be 1.safe (i.e. ≥ WCET) and 2.tight (WCET EST -WCET ≪ WCET EST ) Deadline

15. - 15 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Current Trial-and-Error Based Development 1.Specification of CPS/ES system 2.Generation of Code (ANSI-C or similar) 3.Compilation of Code 4.Execution and/or simulation of code, using a (e.g. random) set of input data 5.Measurement-based computation of “estimated worst case execution time” (WCET meas ) 6.Adding safety margin m on top of WCET meas : WCET hypo := (1 + m )* WCET meas 7.If “WCET hypo ” > deadline: change some detail, go back to 1 or 2.

16. - 16 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Integration of WCET estimation and compilation Computing WCET EST after code generation is too late. Why not consider WCET EST as an objective function already in the compiler?  Integration of aiT and compiler

17. - 17 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 WCET EST -oriented optimizations  Extended loop analysis (CGO 09)  Instruction cache locking (CODES/ISSS 07, CGO 12)  Cache partitioning (WCET-Workshop 09)  Procedure cloning (WCET-WS 07, CODES 07, SCOPES 08)  Procedure/code positioning (ECRTS 08, CASES 11 (2x))  Function inlining (SMART 09, SMART 10)  Loop unswitching/invariant paths (SCOPES 09)  Loop unrolling (ECRTS 09)  Register allocation (DAC 09, ECRTS 11))  Scratchpad optimization (DAC 09)  Extension towards multi-objective optimization (RTSS 08)  Superblock-based optimizations (ICESS 10)  Surveys (Springer 10, Springer 12)

18. - 18 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Relative WCET EST with I-Cache Locking 5 Bench- marks/ARM920T/Postpass-Opt (ARM920T) [S. Plazar et al.] H. Falk, Heiko, S. Plazar, H. Theiling: Compile Time Decided Instruction Cache Locking Using Worst-Case Execution Paths, Intern. Conf. on Hardware/Software Codesign and System Synthesis (CODES+ISSS), 2007

19. - 19 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Register Allocation 100% = WCET EST using Standard Graph Coloring (highest degree) H. Falk, Heiko: WCET-aware Register Allocation based on Graph Coloring, 46th Design Automation Conference (DAC), 2009

20. - 20 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Property of the PREDATOR consortium. Confidential. WCET EST -aware scratch pad allocation Setup  Bosch Democar: Runnable IgnitionSWCSync  WCET EST -aware scratch pad allocation of program code by WCC, including fully automated WCET EST analyses using aiT  WCET EST reduced to about 50%, compared to gcc. © P. Marwedel, 2011

21. - 21 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Tradeoff between reliability and timeliness  Traditional error correction requires resources, including time  Conflicting with timing requirements of real-time systems ➁ ➀ © ➂

22. - 22 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Flexible Error Correction  In some cases, errors have no effect  In some cases, timing requirements are dominating & error correction may be compromised  In other cases, error correction requirements are dominating & timing requirements may be compromised Imprecise outputNo effect System crash unused variable i unused variable j Register 0 Register 1 Register 2 Register 3  Classification requires application knowledge

23. - 23 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Analysis of H.264 video decoder (~ 3500 lines of C code)  Fraction of memory space Resolution Memory size of reliable data Memory size of unreliable data 176 x 14490 kB (55%)74 kB (45%) 352 x 288223 kB (43%)297 kB (57%) 1280 x 7201 585 kB (37%)2 700 kB (63%)  Impact of uncorrected errors on the quality of service: Injection into unreliable memory240 faults / sec80 faults / sec Average PSNR of frames with errors40.89 dB50.57 dB Deteriorated reliability/QoS, timeliness maintained!

24. - 24 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Detection of nano objects using plasmon effect (Optical effect of objects ≪ wavelength of light) Timeliness in (bio-) virus detection Laser Image sequence CCD Prisma Nano object Sensor

25. - 25 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Timeliness in (bio-) virus detection (2)  Low-end Mobile GPU meets realtime constraint up to 1024x256 pixels  Faster GPUs give room for larger usable sensor area Frame Rate (fps)Image Size 1024x1281024x2561024x5121024x1024 CPU: i7-26002320.116.111.5 CPU: i7-620M (mobile)168.54.62.4 GPU: 3100M (mobile)6032.6167.8 GPU: GTX 48060 26.8 GPU: GTX 560Ti60 40.3 Target: 30 frames per second at 1024x256 pixels

26. - 26 - technische universität dortmund fakultät für informatik  Scope  Opportunities  Challenges Multidisciplinarity, timing predictability, A/D interface, energy efficiency, dependability, diversity of objectives  (Partial) solutions to challenges Timing predictability WCET-aware compilation, flexible fault handling, PAMONO Energy efficiency … PAMONO, scratch pads Educational challenges Concentrations, integrated programs, c ourse material  Summary  P.Marwedel, Informatik 12, 2012 Outline

27. - 27 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Energy-efficient execution on graphics processor (GPU) C. Timm, A. Gelenberg, P. Marwedel, F. Weichert: Energy Considerations within the Integration of General Purpose GPUs in Embedded Systems. Intern. Conf. on Advances in Distributed and Parallel Computing, 2010 C. Timm, F. Weichert, P. Marwedel, H. Müller: Design Space Exploration Towards a Realtime and Energy-Aware GPGPU-based Analysis of Biosensor Data. Computer Science - Research and Development, ENA-HPC, 2011 current clamp Energy per frame CPU3.26 J5.84 J10.52 JReduced to Energy per frame GPU0.93 J1.56 J2.76 Javg 27%

28. - 28 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Scratch pad memories (SPM): Fast, energy-efficient, timing-predictable Address space ARM7TDMI cores, well- known for low power consumption scratch pad memory 0 FFF.. Example Small; no tag memory SPMs are small, physically separate memories mapped into the address space; Selection is by an appropriate address decoder (simple!) SPM select

29. - 29 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Migration of data & instructions, global optimization Which memory object (array, loop, etc.) to be stored in SPM? Non-overlaying (“static”) allocation: Gain g k and size s k for each object k. Maximise gain G =  g k, respecting size of SPM SSP   s k. Solution: knapsack algorithm. Overlaying (“dynamic”) allocation: Moving objects back and forth processor scratch pad memory, capacity SSP main memory ? for i.{ } for j..{ } while... repeat … call... array... int... array … Example:

30. - 30 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Dynamic set of multiple applications MEM CPU SPM Manager SPM Manager App. 2 App. 1 App. n SPM App. 3 App. 2 App. 1 t Address space: SPM ? Compile-time partitioning of SPM not feasible  Introduction of SPM-manager  Runtime decisions, but compile-time supported [R. Pyka, Ch. Faßbach, M. Verma, H. Falk, P. Marwedel: Operating system integrated energy aware scratchpad allocation strategies for multi-process applications, SCOPES, 2007]

31. - 31 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Comparison of SPMM to Caches for SORT  Baseline: Main memory only  SPMM peak energy reduction by 83% at 4k Bytes scratchpad SPM SizeΔ 4-way 102474,81% 204865,35% 409664,39% 819265,64% 1638463,73%

32. - 32 - technische universität dortmund fakultät für informatik  Scope  Opportunities  Challenges Multidisciplinarity, timing predictability, A/D interface, energy efficiency, dependability, diversity of objectives  (Partial) solutions to challenges Timing predictability WCET-aware compilation, flexible fault handling, PAMONO Energy efficiency … PAMONO, scratch pads Educational challenges Concentrations, integrated programs, c ourse material  Summary  P.Marwedel, Informatik 12, 2012 Outline

33. - 33 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 CS, EE or physics concentration vs. separate CPS/ES program ConcentrationSeparate program Well-known degrees+- Enough headroom for teaching integrated CPS/ES material -+ Headroom for more depth in physics, mechanical engineering,.. -+ Effort for introductionsmalllarge No. of faculty members requiredmoderatelarger Building community?difficulteasier Is it feasible?For ES ok, for CPS questionable Yes, but there are also constraints

34. - 34 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Concentration on ES at TU Dortmund: approach  Integrated as a concentration into CS curriculum Pro- gramming algorithmscomputer organization Math education EE fundamentals first course on ES/CPS (in German) lab control systems DSPmachine vision real-time systems project applications graduate level thesis undergraduate OS & networks middle- ware Lab + courses for minor degree

35. - 35 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Test Evaluation & verification Optimization ES/CPS Course content Specification & modeling HW components System software (RTOS, …) Design repository Application mapping Design Application knowledge

36. - 36 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Course material: textbook(s) Several editions/translations:  1st edition English German, 2007 Chinese, 2006 Macedonian, 2010  2nd edition, with CPS English, Dec. 2010 German and (translated) Chinese edition TBA  Courseware at //ls12-www.cs. tu-dortmund.de/ ~marwedel (click on ES Textbook) Peter Marwedel Peter Marwedel Peter Marwedel © Graphics: Springer, Science Publishers, Ad Verbum Peter Marwedel Peter Marwedel

37. - 37 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Recorded lectures available on youtube http://www.youtube.com/user/cyphysystems

38. - 38 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Concentration on ES/CPS at TU Dortmund: Teaching elements, experience  ES/CPS Course modeling, ES hardware, RTOS, metrics, mapping, optimizations; Lab & assignments during course  Additional teaching elements Project, optionally in this area Bachelor thesis, optionally in this area, optionally in industry Courses at the graduate level Software of ubiquitous systems Design evaluation,..  Experience: ES education as a concentration works, but has inherent limitations

39. - 39 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Planned integrated B.Sc. program at TU Dortmund 30 credits = 1 term Preliminary information © P. Marwedel, W. Rhode

40. - 40 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Planned integrated M.Sc. Program, TU Dortmund 30 credits = 1 term Preliminary information © P. Marwedel, W. Rhode

41. - 41 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 EMSIG, the successor to Artist  Contact point for education  Goals: leverage experiences gained in the Artist network & stimulate future cooperation  see http://emsig.embedded-systems-portal.org  Running WESE workshop on education

42. - 42 - technische universität dortmund fakultät für informatik  Scope  Opportunities  Challenges Multidisciplinarity, timing predictability, A/D interface, energy efficiency, dependability, diversity of objectives  (Partial) solutions to challenges Timing predictability WCET-aware compilation, flexible fault handling, PAMONO Energy efficiency … PAMONO, scratch pads Educational challenges Concentrations, integrated programs, c ourse material  Summary  P.Marwedel, Informatik 12, 2012 Summary Many opportunities, many challenges, many open questions, more research!

43. - 43 - technische universität dortmund fakultät für informatik  P.Marwedel, Informatik 12, 2012 Q&A!