Informatica 3. Marcello Restelli. Laurea in Ingegneria Informatica Politecnico di Milano 9/15/07 10/29/07

Transcription

1 Informatica 3 Marcello Restelli 9/15/07 10/29/07 Laurea in Ingegneria Informatica Politecnico di Milano

2 Structuring the Computation Control flow can be obtained through control structure at instruction level (control in the small) conditional execution: if-then-else, case,... iteration: while, for, repeat,... low-level construct (GOTO and similar ones) control structure at execution-unit level (control in the large) routine calls exceptions handling control structures used in concurrent programming Example: non-deterministic choice 2

3 Exceptions Programs may go wrong when executing Examples: array index out of bounds, division by zero, square root of a negative integer, memory request exceeds available storage What is needed? erroneous conditions must be recognized response actions should be programmable Alternative, explicit monitoring code Examples anytime an array is accessed, the index is checked anytime a division occurs, the divider is checked 3

4 Terminology Exception undesirable, anomalous behavior Exception handler response code to execute when the exception is raised What is anomalous is defined by the programmer Exception is not necessarily a catastrophic error But simply: unable to proceed in a manner that leads to normal termination as specified by the programmer 4

5 Exception Definition in Programming Languages Which exceptions should be handled? Which programming units can rise an exception and how? How and where should exceptions be handled? How does the control flow pass form the exception to its handler? How do the control flow goes on after the exception has been handled? 5

6 Exception Handling in C++ Exceptions can raised by run-time environment (e.g.,divide by zero) explicitly by the program by a throw instruction Handler may be attached to any piece of code the code to be controlled is put in a try block when an exception occurs, the instruction throw transfers an object to the corresponding handler the exception is taken and handled by a catch 6

7 Example Division between two integers void division (int a, int b) { try { if (!b) throw b; cout << Division result: << a/b << endl; catch (int i) { cout << Division by zero << endl; 7

8 Effect of Throw Block execution abandoned + control transferred to appropriate handler Temporary object thrown used to initialize the variable named in the handler Unhandled exceptions (no handler) are propagated to the point of call of the routine within the block Explicit propagation by simply saying throw Execution then continues from the statement that follows the one to which the matched handler is attached 8

9 Try-catch Block At the same try block can be associated more than one catch, according to the typo of the thrown object try {...throw 1;...throw 'c'; catch (int i) {... catch (char c) {... A catch can intercept any exception try { if (!b) throw b; if (b==1) throw 'a'; if (b==2) throw 12.12; catch (int i){cout << Integer Exception << endl; catch (...){cout << Non-integer exception <<endl; 9

10 Exceptions and Interfaces Routine interfaces may list the exception they may raise Example void foo( ) throw (int, char); Good practice, possible anomalies anticipated and part of possible behavior If the routine throws an exception not listed in its interface, the routine unexpected() is called (by default an abort occurs) list of possible exceptions raised by a routine is not mandatory in the routine interface if no list provided, any exception can be propagated if empty list throw() provided, no exception propagated if after propagation no handler found, terminate() is automatically called; default behavior (redefinable) eventually aborts execution 10

11 Exceptions Thrown inside a Catch Block Inside the exception handling code it is possible to throw again exception using the instruction throw without parameters Current exception is passed to an outer try-catch block void main() { int a = 0;... try{ function(a);... catch (int i) {... void function(int b) { try{ if (!b) throw b;... catch (int i) {... throw; 11

12 Exceptions Handling in Java As in C++, exceptions are thrown as objects and handlers may be attached to blocks All exceptions that may be raised explicitly must be listed in interface (not those raised by the r_t_support) caller can choose among catch plus handle catch and throw list (in its interface) and propagate try block catch (except_type_1) handler_1;... catch (except_type_n) handler_n; finally final_block; /*the final block is executed at the end of the block whether exceptions are thrown or not, unless they are propagated*/ 12

13 Static Scope vs Dynamic Binding FILE 1 FILE2 class A {... extern void f(); void f() { class A { void foo() {... throw A(); try {... f();... catch (A a) { If f throws A in foo, the dynamic binding tries binding with class A defined in file 2, but static scope rules require type matching with class A in file 1 --> exception propagated 13

14 Concurrent Computing A new computation model Set of concurrent computations (tasks) that proceed in parallel and interact only occasionally Each process has its own thread of control thread: processes that share the same address space Parallelism may be physical multiprocessor systems distributed systems logical single-processor system 14

15 Parallelism and Cooperation Correct cooperation requires synchronization producer buffer consumer delay producer if buffer full delay consumer is buffer empty prohibit parallel modification of buffer length 15

16 Mutual Exclusion Some actions must be executed in mutual execution by two or more tasks I.e., they must be atomic (logically indivisible) operations E.g. update buffer length, what if Append t++; i=next_in(); buffer[i] = x; Remove t--; j=next_out; x=buffer[j]; 16

17 Type of Tasks Threads They share the same address space They can communicate via shared memory Processes They do not share the same address space They communicate via messages 17

18 Java Threads Class Thread public run() the code that will be executed simultaneously with the other threads in a program public start() performs special initialization for the thread and then calls run() public sleep(...) public interrupt (or notify) public boolean isalive(...) true if the thread is running, false if the thread is finished public join()/join(...) wait for the second thread to complete before proceeding public wait()/wait(int milliseconds) similar to sleep, but releases the object lock released by a notify or by the expiration of the time segment 18

19 Java Threads New The object has been created, but it hasn t been started yet, so it cannot run. Runnable The thread can be run when the time-slicing mechanism has CPU cycles available for the thread Dead A thread dies by returning from its run( ) method. Now deprecated, stop( ) destroy( ) never been implemented Blocked The thread could be run, but there s something that prevents it The scheduler skip over it and not give it any CPU time, until the thread reenters the runnable state 19

20 Java Threads Yielding If you know that you ve accomplished what you need to in your run( ) method, you can give a hint to the thread scheduling mechanism that you ve done enough and that some other thread might as well have the CPU. This hint (and it is a hint there s no guarantee your implementation will listen to it) takes the form of the yield( ) method. Sleeping Another way you can control the behavior of your threads is by calling sleep( ) to cease execution for a given number of milliseconds. 20

21 The State Diagrams of Threads born sleep quantum expired notify yield, interrupt,... ready running start run I/O completed send I/O wait sleep complete waiting sleeping dead blocked 21

22 Synchronized Methods A synchronized method in Java guarantees access in mutual exclusion to a shared resources Threads (possibly synchronized) methods are suspended and resumed through wait and notify methods A shared object that exploits such synchronization mechanisms may implement the concept of monitor (shared, passive object) Monitor first introduced in concurrent Pascal (by Hoare and Brinch-Hansen) The risk of deadlock in concurrent programming: P1 is blocked waiting to access resource A P2 owns A but is blocked waiting to access resource B B is owned by P1 22

23 Example: Producer and Consumer public class Main { public static void main (String args[]) { // create shared buffer Buffer B = new Buffer (100); // create consumer Consumer C = new Consumer (B); // create producer Producer P = new Producer (B); // start consumer and producer C.start(); P.start(); 23

24 Example: Producer class Producer extends Thread { private Buffer buff; //construct a producer object Producer (Buffer b) { buff = b; public void run() { //produce things and put in buff buff.put (...); //buff is a share... 24

25 Example: Synchronized Buffer public class Buffer { private int n; // size of buffer private int [] contents; // contents of buffer private int in, out = 0; // indexes of where to read from/write to private int total = 0; // number of items in the buffer Buffer (int size) { n = size; contents = new int [n]; 25

26 Example: Synchronized Buffer public synchronized void put (int item) { while (!(total < n)) // wait until there is space try { wait(); catch (InterruptedException e) { contents [in] = item; System.out.println("Buffer: write at " + in + " item " + item); if (++in == n) in = 0; total++; notify(); // wake up any blocked threads 26

27 Example: Synchronized Buffer public synchronized int get () { int temp; // wait till there is something while (!(total > 0)) try { wait(); catch (InterruptedException e) { temp = contents[out]; System.out.println("Buffer: read from + out + " item " + temp); if (++out == n) out = 0; total--; notify(); // wake up any blocked threads return temp; 27

28 Task Synchronization in Ada Guardians and rendez-vous The Ada style of designing concurrent systems In Ada a shared object is active (whereas a monitor is passive) it is managed by a guardian process which can accept rendezvous requests from tasks willing to access the object Guardian Guardian.GET(c);... Guardian.PUT(c); Task using the resource Resource 28

29 A Guardian Task loop select when NOT_FULL accept PUT (C: in CHAR); This is the body of PUT; the client calls it as if it were a normal procedure end ; or when NOT_EMPTY accept GET (C: out CHAR); This is the body of GET; the client calls it as if it were a normal procedure end ; end select ; end loop ; 29

30 Real-time Software (hints) Case where some processes exist in the environment Example put operation in a shared buffer invoked by a plant sending data to a controller plant cannot be suspended if buffer full design must ensure that producer never finds the buffer full this constrains the speed of the consumer process in the controller 30

31 Distributed Software Issues to consider module-machine binding intermodule communication e.g., remote procedure call or message passing access to shared objects may require replication for efficiency reasons 31

32 Client-server Architecture The most popular distributed architecture Server modules provide services to client modules Clients and servers may reside on different machines Issues Binding modules to machines static vs. dynamic (migration) Inter-module communication e.g., RPC IDL to define interface of remote procedures Replication and distribution 32

33 Middleware Layer residing between the network operating system and the application Helps building network applications Provides useful services Name services, to find processes or resources on the network Communication services, such as message passing or RPC 33