Introduction to Concurrency

Concurrency: Execute two or more pieces of code at the same time Why? No choice: - Geographically distributed data - Interoperability of different machines - A piece of code mustserve many other client processes - To achieve reliability By Choice: to achieve speedup sometimes makes programming easier (e.g. UNIX pipes)

Possibilities for Concurrency

Architecture: Uniprocessor with: -I/O channel - I/o processor - DMA Network of uniprocessors Multiple CPUs Architecture: Multiprogramming, multiple process system Programs

Distributed programming Parallel programming

Definitions
Concurrent process execution can be: - interleaved,or - physically simultaneous Interleaved: Multiprogramming on uniprocessor Physically simultaneous: Uni-or multiprogramming on multiprocessor Process, thread, or task: Schedulable unit of computation Granularity: Process size or computation to communication ratio - Too small: excessive overhead - Too large: less concurrency

Precedence Graph
Consider writing a program as a set of tasks. Precedence graph: Specifies execution ordering among tasks S0: S1: S2: S3: S4: A:= X Y B:= X + Y C := Z + 1 C := A - B W:= C + 1 S0 S1 S3 S2

S4 Parallel zing compilers for computers with vector processors build dependency graphs

Cyclic Precedence Graphs

What does the following graph represent ? S1

S2

S3

Examples of Concurrency in Uniprocessors

Example 1: Unix pipes Motivations: -fast to write code -Fast to execute Example2: Buffering Motivation: -required when two asynchronous processes must communicate

Example3: Client/ Server model Motivation: - geographically distributed computing

Operating System Issues

Synchronization: What primitives should OS provide?
Communication: What primitives should OS provide to interface communication protocol?

Hardware support: Needed to implement OS primitives

Remote execution: What primitives should OS provide? - Remote procedure call(RPC) - Remote command shell Sharing address spaces: Makes programmer easier

Lightweight threads: Can a process creation be as cheap as a procedure call?

Parallel Language Constructs

FORK and JOIN
FORK L Starts parallel execution at the statement labeled L and at the statement following the fork JOIN Count Recombines Count concurrent computations Count:=Count-1; If (Count>0) Then

Quit;
Join is an atomic operation

Definition: Atomic Operation

If I am a process executing on a processor, and I execute an atomic operation, then all other processes executing on this or any other processor: Can see state of system before I execute or after I execute

but cannot see any intermediate state while I am executing

Example: bank teller /* Joe has $1000, split equally between savings and checking accounts*/ 1. Subtract $100 from Joes savings account 2. Add $100 to Joes checking account

Other processes should never read Joes balances and find he has 900 in both accounts.

Concurrency Conditions
Let Si denote a statement Read set of Si: R(Si) = {a1,a2,,an) Set of all variables referenced in Si Write set of Si: W(Si) = { b1, b2, , bm}, Set of all variables changed by si C := A - B R(C := A - B) = {A , B} W(C := A - B) = {C} Scanf(%d , &A) R(scanf(%d , &A))={} W(scanf(%d , &A))={A}

Bernsteins Conditions
The following conditions must hold for two statements S1 and S2 to execute concurrently with valid results: 1) R(S1) INTERSECT W(S2)={} 2) W(S1) INTERSECT R(S2)={} 3) W(S1) INTERSECT W(S2)={}

These are called the Berstein Conditions.

Fork and Join Example #1

S1: S2: S3: S4: A := X + Y B := Z + 1 C := A - B W :=C + 1 S1 S2

S3

Count := 2; FORK L1; A := X + Y; Goto L2; L1: B := Z + 1; L2: JOIN Count; C := A - B; W := C + 1;

S4

Structured Parallel Constructs

PARBEGIN / PAREND PARBEGIN Sequential execution splits off into several concurrent sequences PAREND PARBEGIN Statement 1; Statement 2; :::::::::::::::: Statement Nl PAREND; Parallel computations merge
PARBEGIN Q := C mod 25; Begin N := N - 1; T := N / 5; End; Proc1(X , Y); PAREND;

Fork and Join Example #2

S1

S2
S4 S6

S3

S5

S7 Up to three tasks may concurrently execute

S1; Count := 3; FORK L1; S2; S4; FORK L2; S5; Goto L3; L2: S6; Goto L3 L1: S3; L3: JOIN Count S7;

Parbegin / Parend Examples

Begin PARBEGIN A := X + Y; B := Z + 1; PAREND; C := A - B; W := C + 1; End
Begin S1; PARBEGIN S3 BEGIN S2; S4; PARBEGIN S5; S6; PAREND; End; PAREND; S7; End;

Comparison
Unfortunately, the structured concurrent statement is not powerful enough to model all precedence graphs.
S1 S1 S1

S1 S1 S1 S1

Comparison(contd)
Fork and Join code for the modified precedence graph:
S1; Count 1:=2; FORK L1; S2; S4; Count2:=2; FORK L2; S5 Goto L3; L1: S3; L2: JOIN Count1; S6; L3: JOIN Count2; S7;

Comparison(contd)
There is no corresponding structured construct code for the same graph However, other synchronization techniques can supplement Also, not all graphs need implementing for real-world problems

Overview
System Calls - fork( ) - wait( ) - pipe( )

- write( )
- read( ) Examples

Process Creation
Fork( ) NAME fork() create a new process SYNOPSIS # include <sys/types.h> # include <unistd.h> pid_t fork(void) RETURN VALUE success parent- child pid child- 0 failure -1

Fork() system call- example

#include <sys/types.h> #include <unistd.h> #include <stdio.h> Main() { printf([%ld] parent process id: %ld\n, getpid(), getppid()); fork(); printf(\n[%ld] parent process id: %ld\n, getpid(), getppid()); }

Fork() system call- example

[17619] parent process id: 12729 [17619] parent process id: 12729 [2372] parent process id: 17619

Fork()- program structure

#include <sys/types.h> #include <unistd.h> #include <stdio.h> Main() { pid_t pid; if((pid = fork())>0){ /* parent */ } else if ((pid==0){ /*child*/ } else { /* cannot fork* } exit(0); }

Wait() system call

Wait()- wait for the process whose pid reference is passed to finish executing
SYNOPSIS #include<sys/types.h> #include<sys/wait.h> pid_t wait(int *stat)loc)

The unsigned decimal integer process ID for which to wait

RETURN VALUE success- child pid failure- -1 and errno is set

Wait()- program structure

#include <sys/types.h> #include <unistd.h> #include <stdlib.h> #include <stdio.h> Main(int argc, char* argv[]) { pid_t childPID; if((childPID = fork())==0){ /*child*/ } else { /* parent* wait(0); } exit(0); }

Pipe() system call

Pipe()- to create a read-write pipe that may later be used to communicate with a process well fork off.
SYNOPSIS

Int pipe(pfd)
int pfd[2]; PARAMETER Pfd is an array of 2 integers, which that will be used to save the two file descriptors used to access the pipe RETURN VALUE: 0 success; -1 error.

Pipe() - structure
/* first, define an array to store the two file descriptors*/ Int pipe[2];
/* now, create the pipe*/ int rc = pipe (pipes); if(rc = = -1) { /* pipe() failed*/ Perror(pipe); exit(1); } If the call to pipe() succeeded, a pipe will be created, pipes[0] will contain the number of its read file descriptor, and pipes[1] will contain the number of its write file descriptor.

Write() system call

Write() used to write data to a file or other object identified by a file descriptor. SYNOPSIS #include <sys/types.h> Size_t write(int fildes, const void * buf, size_t nbyte); PARAMETER fildes is the file descriptor, buf is the base address of area of memory that data is copied from, nbyte is the amount of data to copy RETURN VALUE The return value is the actual amount of data written, if this differs from nbyte then something has gone wrong

Read() system call

Read() read data from a file or other object identified by a file descriptor SYNOPSIS #include <sys/types.h> Size_t read(int fildes, void *buf, size_t nbyte); ARGUMENT fildes is the file descriptor, buf is the base address of the memory area into which the data is read, nbyte is the maximum amount of data to read. RETURN VALUE The actual amount of data read from the file. The pointer is incremented by the amount of data read.