WCET Testing For Embedded Applications Software

In the field of transportation,aviation industry and industrial production,embedded software has been widely used.Due to high requirement of embedded software for special use,reliability standard of embedded system must be very strict.Once an operation fails in aerospace or industrial control,it will lead to disastrous consequences.Rigorous reliability testing of embedded software plays an indispensable role in ensuring the smooth running of the whole software.Predictable of real-time embedded software is an important part of software reliability testing.And the real-time performance of software has important relation to the WCET(worst-case execution time)of itself.Because failure in estimating WCET is the reasons why most embedded software can't meet the real-time requirements.Therefore,study of program's WCET becomes a hot point of academic.The research of WCET divided into two areas:dynamic test and static analyze.And this thesis focuses on a study of WCET testing,the main work is as follows:The thesis gives a dynamic testing method based on associative process communication.Firstly,it uses abstract interpretation to do semantics analysis of program,then it generating nodes flow graph of program which similar to CFG(program flow diagram)through analyzing of execution path of state unchanged nodes.The WCET of the whole program can be found through the calculated of WCET of each nodes of directed acyclic graph by proposed APC algorithm.This thesis proposes two APC algorithms:algorithm1 is a high-density algorithm which test WCET values of each nodes using pipeline technology.Algorithm2 used to test the WCET value of nodes which contains function calls.APC test method tests all benchmarks of Malardalen many times.And the test results show that the proposed test method is precise and effective,and the test error is within the theoretical analysis.When testing the context switching time of the program,the gap between the proposed method test results and LMbench test tool is only 3?s.Moreover,the proposed method can test context switching time of almost all program slices.Given a WCET analysis framework based on MPA algorithm and Static Prediction.This framework firstly uses the MPA algorithm to do flow analysis and get the Min-Tree of each block.Then get more rigorous constraints through the analysis of inner loop variables of function by symbolic loop bound Computation and get the finally WCET expression through constraint of MPA algorithm and loop bound.Finally,the framework use static prediction to do absolute valuation of underlying instruction cycle of each basic block and calculates the final WCET value.The thesis use both the given framework and static prediction framework test a example program and the result shows the difference of those two framework is within 2%of the gap,and further tests showed that the difference are inversely related to the size of the two frameworks.When compared to the original framework,the given framework has advantages in smaller complexity and higher automation.