Perform ance Estim ation for Em bedded System s w ith Data and - PowerPoint PPT Presentation

Perform ance Estim ation for Em bedded System s w ith Data and Control Dependencies Paul Pop, Petru Eles, Zebo Peng Department of Computer and I nformation Science Linköpings universitet Sweden 1 of 12 May 3 , 2 0 0 0

Motivation and Characteristics Perform ance estim ation. � Worst case delay on the system execution time. Characteristics: � Distributed hard real-time applications. � Heterogeneous system architectures. � Fixed priority preemptive scheduling. � Systems with data and control dependencies. Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 2 o f 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Schedulability Analysis Schedulability test : � The worst case response time of each process is compared to its deadline . Process m odels : � Independent processes; � Data dependencies: release jitter , offsets , phases ; � Control dependencies : modes , periods , recurring tasks . Message: � The pessimism of the analysis can be significantly reduced by considering the conditions during the analysis. Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 3 o f 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Conditional Process Graph Subgraph corresponding P 0 P 0 to D ∧ C ∧ K P 1 P 1 P 11 P 11 P 1 P 11 D D P 2 P 2 P 3 P 3 P 2 P 3 C C C P 12 P 12 P 13 P 12 P 13 P 6 P 6 P 6 K K P 4 P 5 P 4 P 5 P 14 P 14 P 16 P 16 P 14 P 16 P 8 P 8 P 9 P 9 P 15 P 8 P 9 P 15 P 7 P 7 P 17 P 17 P 17 P 10 P 10 P 10 � First processor � Second processor � ASIC P 18 P 18 Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 4 of 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Problem Form ulation I nput � An application modelled using conditional process graphs (CPG). � Each CPG in the application has its own independent period . � Each process has a worst case execution time , a deadline , and a priority . � The system architecture and mapping of processes are given. Output � Worst case response times for each process. Perform ance estim ation for system s m odelled using conditional process graphs. Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 5 o f 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Exam ple P 0 P 0 P 1 27 P 1 P 9 P 9 C C P 6 30 P 6 Worst Case Delays P 2 CPG 30 P 2 No conditions Conditions P 3 25 P 3 P 10 P 11 25 32 P 10 P 11 24 P 4 P 4 Γ 1 120 100 P 7 22 P 7 P 12 P 12 Γ 2 82 82 19 P 5 P 5 Γ 2 : 150 P 8 P 8 Γ 1 : 20 0 Deadline: 110 Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 6 of 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Task Graphs w ith Data Dependencies � K. Tindell: Adding Time-Offsets to Schedulabilty Analysis, Research Report Offset: fixed interval in time between the arrival of sets of tasks. Can reduce the pessimism of the schedulability analysis. Drawback: how to derive the offsets? � T. Yen, W. Wolf: Performance Estimation for Real-Time Distributed Embedded Systems, IEEE Transactions On Parallel and Distributed Systems Phase (similar concept to offsets). Advantage: gives a framework to derive the phases. Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 7 o f 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Schedulability Analysis for Task Graphs DelayEstimate (task graph G, system S) for each pair (P i, P j ) in G maxsep[P i, P j ]= ∞ end for worst case response times and step = 0 upper bounds for the offsets repeat LatestTimes(G) EarliestTimes(G) lower bounds for the offsets for each P i ∈ G MaxSeparations(P i ) end for until maxsep is not changed or step < limit return the worst case delay d G of the graph G end DelayEstimate maximum separation: maxsep[P i, P j ]=0 if the execution of the two processes never overlaps Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 8 of 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Schedulability Analysis for CPGs, 1 Tw o extrem e solutions: � Ignoring Conditions (IC) Ignore control dependencies and apply the schedulability analysis for the (unconditional) task graphs. � Brute Force Algorithm (BF) Apply the schedulability analysis after each of the CPGs in the application have been decomposed in their constituent unconditional subgraphs. Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 9 of 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Schedulability Analysis for CPGs, 2 I n betw een solutions: � Conditions Separation (CS) Similar to Ignoring Conditions but uses the knowledge about the conditions in order to update the maxsep table: maxsep[P i , P j ] = 0 if P i and P j are on different conditional paths. � Relaxed Tightness Analysis (two variants: RT1, RT2) Similar to the Brute Force Algorithm , but tries to reduce the execution time by removing the iterative tightening loop (relaxed tightness) in the DelayEstimation function. Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 10 of 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Experim ental Results 10 0 Average Percentage Deviation [ % ] 8 0 Ignoring Conditions 6 0 Conditions Separation 4 0 Relaxed Tightness 1 20 Relaxed Tightness 2 Brute Force 0 8 0 16 0 24 0 320 4 0 0 Number of processes Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 11 of 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0

Conclusions � Perform ance estim ation for hard real-time systems with control and data dependencies. � Modelling using conditional process graphs that capture both the flow of data and that of control. � Heterogeneous architectures, fixed priority scheduling . � Five approaches to the schedulability analysis of such systems. � Extensive experiments and a real-life example show that: The pessimism of the analysis can be significantly reduced by considering the conditions during the analysis. Pe rfo rm an ce Estim atio n fo r Em be dde d Syste m s w ith Data an d Co n tro l De pe n de n cie s 12 of 12 Paul Po p, Pe tru Ele s, Ze bo Pe n g May 3 , 2 0 0 0