Multicore Migration Study in Automotive Powertrain Domain

Transcription

1 2 nd Workshop on Certifiable Multicore Avionics and Automotive Systems Multicore Migration Study in Automotive Powertrain Domain 21/Apr/2017 Hitachi, Ltd., Takeshi Fukuda Tasuku Ishigooka, Fumio Narisawa Hitachi Automotive Systems, Ltd., Toru Irie, Takafumi Suzuki

2 Contents 1. Outline of Hitachi 2. Background 3. Control Software and Issue 4. Our Method 5. Evaluation with Engine Control Software 6. Summary and Discussion 1

3 Hitachi corporate data Hitachi, Ltd. Incorporated 1 st February 1920 [Founded 1910] Head Office Marunouchi 1-6-6, Chiyoda-ku, Tokyo, JAPAN Hitachi Group (Consolidated) FY2015 Revenues JPY 10,343 billion R&D expenditure JPY billion [3.3% of revenue] No. of subsidiaries 1,056 companies [262 domestic; 794 overseas] No. of employees 335,244 [187,936 domestic; 147,308 overseas] Hitachi, Ltd All rights reserved.

4 Business portfolio Others (Logistics and Other services) 11% 3% Financial Services 19% Information & Telecommunication Systems Smart Life & Ecofriendly Systems Automotive Systems 9% 6% 14% Revenues (FY2015) 10,343 billion (JPY) 21% Social Infrastructure & Industrial Systems 7% 10% High Functional Materials & Components Construction Machinery Electronic Systems & Equipment

5 Business Description of Automotive Dep. System solutions that Hitachi Automotive Systems will provide for progress in Environment, Safety and Information field.

6 [Environment] Engine Management Systems Offering environment-friendly and high-efficiency engine management systems for reduction in CO2 and prevention of air pollution. Injector ECU: Engine control unit High-pressure fuel pump Variable Valve Event and Lift Control System Electronic throttle body Airflow sensor Ignition coil (Hitachi Automotive Systems Hanshin) Valve Timing Control Transmission control unit Oil pump Balancer Water pump Piston

7 Contents 1. Outline of Hitachi 2. Background 3. Control Software and Issue 4. Our Method 5. Evaluation with Engine Control Software 6. Summary and Discussion 6

8 Background: Trend of Exhaust Gas Regulation The performance requirements of automotive engine control are increasing, for instance to comply the exhaust emission regulations and reduce gasoline consumption s Air pollution problem Exhaust Gas Regulation about HC [g/km] 1980s 1990s 2000s 2010s North America Japan 53 regulation Muskie Act TLEV LEV Europe Euro3 New short term ULEV SULEV Euro4 New long term Global worming problem Euro5 Today Euro6 7

9 Performance Trend of Microcontroller for Powertrain Current electronic controller unit requests high performance microcontroller. Trend of Microcontroller for Powertrain Systems Clock [MHz] High performance trend Single Core Core * 2 Core * [Year] ECU: Electric Control Unit 8

10 Our study Our study: To migrate legacy source code to multicore platform. Comparison with other products PC, Smart phone Television Automotive engine Data Dependency Loosely coupled Loosely coupled Complicated 1 Task + Execution order Dataflow Mailer Text editor Web browser Broadcasting Recording Core1 Core2 Core3 Core1 Core Air-fuel ratio control Core1? Injection control 6 9 Core2? 9

11 Our study Our study: To migrate legacy source code to multicore platform. Comparison with other products A potion of dataflow of engine control software PC, Smart phone Television Automotive engine Data Dependency Loosely coupled Loosely coupled Complicated 1 Task + Execution order Dataflow mailer power point Web browser Broadcasting Recording Cite M. Panić, S. Kehr, E. Quiñones, B. Boddecker, J. Abella and F. J. Cazorla, RunPar: An Allocation Algorithm for Automotive Applications Exploiting Runnable Parallelism in Multicores, International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS 14), Oct. Hitachi, Ltd All rights reserved. Core1 Core2 Core3 Core1 Core2 Air-fuel ratio control Core1? Injection control 6 9 Core2? 10

12 Contents 1. Outline of Hitachi 2. Background 3. Control Software and Issue 4. Our Method 5. Evaluation with Engine Control Software 6. Summary and Discussion 11

13 Dataflow of Control Software Dataflow of control software roughly consists of sensing, calculating, and actuating. E.g. Dataflow of fuel injection Airflow sensor ECU Notes (1) (2) (3) Task in AD convertor (1)Sensing airflow (2)Calculating air volume (3)Fitting fuel volume (4)Calculating timing of fuel injection (4) Injector Task in microcontroller Dataflow 12

14 Cyclic Execution of Control Software (1) (2) (3) (4) Tasks are executed periodically with each cyclic time or executed sporadically based on event like an engine rotation speed. Time chart(single core processor) AD convertor (1)250μ (2)1ms task ECU (3)2ms task (4)Interrupt based on rpm T Wait time by fixed priority scheduling T+1 T+2 T+3 T+4 [ms] 13

15 Parallelization Issue of parallelization A large number of inter-core communication data causes high frequent inter-core synchronization between cores. Single core 10ms task Multicore Core 1 10ms task A Inter-core synchronization Core 2 10ms task B 14

16 Issue of parallelization (Cont d) Issue: Long wait time is frequently happened when one core is interrupted by other high priority tasks due to frequent inter-core synchronization. 1ms task 1ms task Multicore Core 1 10ms task A Wait time Inter-core synchronization Interrupt Core 2 10ms task B Wait time Wait time 15

17 Issue of parallelization (Cont d) Issue: Long wait time is frequently happened when one core is interrupted by other high priority tasks due to frequent inter-core synchronization. 1ms task 1ms task Our Goal Multicore In order to Core achieve 1 10ms high task A parallelization, our goal is to reduce a number of inter-core synchronization. Wait time Interrupt Inter-core synchronization Core 2 10ms task B Wait time Wait time 16

18 Contents 1. Outline of Hitachi 2. Background 3. Control Software and Issue 4. Our Method 5. Evaluation with Engine Control Software 6. Summary and Discussion 17

19 Overview of our method We proposed parallelization method with performance requirements of system control. Feature of our method is to identify the inter-core data which doesn t need inter-core synchronization according to the requirements of Data Delay Time. Strategy of core allocation Requests for data delay Control SW for single core (1)Program analysis Dataflow (2)Detecting inter-core data inter-core communication data (3)Categorizing data Data with synchronization intra-core communication data Data without synchronization 18

20 Data delay time Data delay time is one kind of end-to-end path latency, especially, from start time of sensing task to start time of other task which has data dependence with sensing task. With synchronization between cores Time chart 1 Data delay Data delay AD convertor (1)250μ Core 1 (2)1ms task Synchronization between cores Synchronization between cores Core 2 (3)2ms task T T+1 T+2 T+3 T+4 [ms] Wait time Wait time 19

21 Data delay time (cont d) Without synchronization between cores Time chart 2 Data delay Data delay AD convertor (1)250μ Core 1 (2)1ms task Core 2 (3)2ms task T T+1 T+2 T+3 T+4 No wait time [ms] 20

22 Idea Our idea: we select the parallelization option which is without synchronization between cores when we can allow a WCRT(Worst Case Response Time) of data delay time. With synchronization Without synchronization Input data Same timing data Previous task data Wait time Long Zero Parallelization Low High Data delay time Short Long WCRT of data delay time Short Long N 1 i=1 (R τ i + nt τ i+1 r(τ i+1 ) + S(τ N ) Cite Balsini, A., Melani, A., Buonocunto, P. and Natale, Ki, M.: FMTV 2016: Where is the Actual Challenge?, 3 rd Challenge on Formal Method for Timing Verification (FMTV), International Workshop on Analyais Tools and Methodologies for Embedded and Real- Time Systems (WATERS 2016), (2016) 21

23 Contents 1. Outline of Hitachi 2. Background 3. Control Software and Issue 4. Our Method 5. Evaluation with Engine Control Software 6. Summary and Discussion 22

24 Overview of evaluation We migrated whole legacy engine control software to multicore ECU which has two cores. And evaluated it with HILS. Core 1: Cyclic tasks (E. g. 1ms, 2ms, 10ms tasks) Core2: Sporadic tasks (regarding engine rotation speed) Periodic Task API Crank Event Task API Periodic BSW OS Inter -Core COM OS Event (rpm) Core1 Core2 Engine control SW included aprox control signals ECU two cores ECU (Engine Controller Unit) HILS (Hardware-in-the-loop Simulation) 23

25 Periodic BSW Periodic Task API Core1 Single Core OS Inter -Core COM Crank Event Task API Single Core OS Core2 Event (Crank) Total amount of Inter-core communication data We found that a total amount of inter-core communication data is approx. 600 in our process step two. コア1: 周期タスクコア (1,2,4,10,20,100ms タスク ) コア2: クラセンコア (Quick REF タスク ) Strategy of core allocation Requests for data delay Requests for input data set Control SW for single core (1)Program analysis Dataflow (2)Detecting inter-core data inter-core communication data (3)Selecting data Data with synchronization Function Data depencency intra-core communication data Data without synchronization Dataflow(portion) the result of program analysis List of Inter-core communication data the result of detecting inter-core data 24

26 Periodic BSW Periodic Task API Core1 Single Core OS Inter -Core COM Crank Event Task API Single Core OS Core2 Event (Crank) The results Our method indicated that more than 90% out of approx. 600 inter- core communication data don t need synchronization. コア1: 周期タスクコア (1,2,4,10,20,100ms タスク ) コア2: クラセンコア (Quick REF タスク ) Strategy of core allocation Requests for data delay Requests for input data set Control SW for single core (1)Program analysis Dataflow (2)Detecting inter-core data inter-core communication data (3)Selecting data Data with synchronization Total amount of controls data intra-core communication data # of control data with program analysis Our method Data without synchronization 25

27 CPU load[%] Evaluation with HILS(Hardware-in-the-loop Simulator) Result of HILS evaluation indicated that our method is (1)useful for parallelizing engine control software (2) (3) 55% 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% CPU load vs. Engine rotation speed Increasing depends on rpm Constant CPU load Multi core (Core 1) Multi core (Core 2) Engine rotation speed [rpm] 26

28 CPU load[%] Evaluation with HILS(Hardware-in-the-loop Simulator) Result of HILS evaluation indicated that our method is (1)useful for parallelizing engine control software (2)able to distribute CPU load to cores (3) 55% 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% CPU load vs. Engine rotation speed CPU load 44% CPU load 16% Max. CPU load 51% Single core Multi core (Core 1) Multi core (Core 2) Engine rotation speed [rpm] 27

29 CPU load[%] Evaluation with HILS(Hardware-in-the-loop Simulator) Result of HILS evaluation indicated that our method is (1)useful for parallelizing engine control software (2)able to distribute CPU load to cores (3)Keeping hard real-time deadline. e.g. fuel infection 55% 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% CPU load vs. Engine rotation speed CPU load 44% CPU load 16% Max. CPU load 51% Single core Multi core (Core 1) Multi core (Core 2) Engine rotation speed [rpm] 28

30 Contents 1. Outline of Hitachi 2. Background 3. Control Software and Issue 4. Our Method 5. Evaluation with Engine Control Software 6. Summary and Discussion 29

31 Summary Proposed method Parallelization method for control software The feature of our method is -identify the inter-core data which doesn t need core synchronization according to the requirement of Data Delay Time Evaluation We apply our method to whole legacy engine control software. Results More than 90% of data out of approximately 600 inter-core communication data don t need synchronization mechanism. The evaluation results with HILS indicated that the parallelized software satisfied requirements of real-time performance. 30

32 Discussion Discussion Process and Method to decide requirement time of data delay with control engineer is required. It is tough issue to decide it which doesn t cause a bad effect for the performance of systems. It is a nonsense if multicore migration decrease engine performance like fuel efficiency. Precise WCRT analysis method is required. The result of WCRT analysis for data delay time is too pessimistic. 31

33 Thank you!