02 variables

7/27/2019 02 Variables

1/37

Structure Variables Comments Floats Simple Expressions

2. Variables, Identifiers, and Expressions

24. Juni 2011


Einfuhrung in die ProgrammierungIntroduction to C/C++, Tobias Weinzierl page 1 of 35
http://www5.in.tum.de/wiki/index.php/Dr._rer._nat._Tobias_Weinzierlhttp://www5.in.tum.de/wiki/index.php/Dr._rer._nat._Tobias_Weinzierl


2/37


Outline

A Movie

Structure of Simple C Codes

Variables & Identifiers

Comments & Documentation

Built-in Datatypes

Excursus: Floating Point Precision

Simple Expressions: Assignments and

Basic Arithmetics




3/37


The Golden Age of (Super)Computing

Chris Johnson,

head of the Scientific Computing &Imaging Institute in Salt Lake City, and

member of the PITAC committee. At TUM, theres similar institutes

(organised as consortia) such as theMunich Centre of AdvancedComputing.




4/37


2.1.Structure

# i n c l u d e

i n t main ( ) {s t d : : c ou t


5/37


An Addendum: C vs. C++

C is (basically) subset of C++

Some features were deprecated or augmented (see comments)

Some functions were replaced, in particular the output functions

i n t a=10; double b=20;s t d : : c ou t


6/37


2.2.Variables




7/37


Declaring a Variable

Computer runs through code step-by-step

Declaration= define a name/alias for a memory location

Definition= find a well-suited (free) place in memory

Variable then is alias for this storage point we work with names instead of addresses




8/37


Identifiers

W. Savitch: A C++ identifier must start with either a letter or

the underscore symbol, and the remaining characters must allbe letters, digits, or underscore symbols. C++ identifiers are casesensitive and have no limit to their length. (p. 7)

An identifier is (unique) name of a variable

Must not equal a keyword

Should have a name with a precise meaning(UNIX-style and Hungarian notation today are considered to be a bad smell)

Examples:

userNumber, UserNumber, userNumber, usernumber, user number

usno, usrn, usrn, nusr, usr1, usr1no iUserNumber, intUserNumber




9/37


Built-in Datatypes

A variable corresponds to memory location and holds a value.

What type of value?char, int, float, anddouble, and so forth.

Compiler cares for memory layout.




10/37


What Value Does a Variable Have By Default?




11/37


Scope of Built-in Datatypes

E. W. Dijkstra: Once a person has understood the way variables

are used in programming, he has understood the quintessenceof programming. (Notes on Structured Programming)

name size range memory footprintbool {,} 1 Byte (?)char 0 . . . 255 1 Byteshort int 32, 768 . . . 32, 767 2 Bytesint 2, 147, 483, 648 . . . 2, 147, 483, 647 4 Bytes (?)long int 2, 147, 483, 648 . . . 2, 147, 483, 647 4 Bytes (?)float 1037 . . . 1038 4 Bytes

double 10

307 . . . 10308 8 Bytes




12/37


Long and Unsigned Modifier

Long Integers

l on g i n t a ; / / s iz e ? i n t b ; / / s iz e ?

C/C++ standard defines at leastnbits.

On 64 bit architectures,int and long inttypically are the same.

Some 32 bit architectures support 64 bit integers due to a tailored compiler. Others dont (such as the RZGs BlueGene/P system with the IBM compiler).

Recommendation: Avoid modifiers such as longand short.

Unsigned Integers

u ns ig ne d i n t a ; / / s iz e ?

i n t b ; / / s iz e ?

That sign consumes one bit of the representation.

If you are sure you dont need the sign, you can squeeze out an additional bit,

however, be careful with this!




13/37


What is an Overflow?

An excursus into binary number systems:

00000000b = 010

00000001b = 110

00000010b = 210

00000011b = 310

ALU internal (for an inc): 00110111b+ 00000001b 00110111b 00110110b 00110100b 00111000b

Your turn: Increment 11111111b(it is an unsigned integer)!



St t V i bl C t Fl t Si l E i


14/37


What is an Overflow?

C++ does not check for overflows(unlike all the interpreted languagessuch as Java and C# that are slower inturn).

C++ does not check for underflows

(unlike all the interpreted languagessuch as Java and C# that are slower inturn).

Now, how would you define/encodeyour signed integer in terms of bitsand what do the operations look like?





15/37


Codes for Signed Integers

1-2:

00000001b- 00000001b- 00000001b 00000001b- 00000001b 00000000b- 00000001b 11111111b

-1+1:

11111111b+ 00000001b- 00000001b

11111111b 11111110b 10000000b 00000000b

Still, the first bit is the sign, but an overflow is just running through zero.

Theres more negative values than positive ones.





16/37


Syntax vs. Semantics

Syntax= rules for the constructing of sentences in the languages

Semantics= meaning of sentences, words, or fragments.

The syntax is (more or less) prescribed by C, but

the semantics is what you had in mind, i.e. the reader has to reconstruct it fromyour program.

So, use meaningful names, and,

if that is not sufficient, add documentation, documentation, and documentation.

If code is not documented, it does not exist!





17/37


2.3.Comments

# i n c l u d e

/ T h i s i s t h e famous main o p e r a t i on . Here t he a p p l i c a t i o n s t a r t s ./

i n t main ( ) { / / hey dude , h er e we go / w r i t e yourname t o t h e c o ns ol e /s t d : : c ou t


18/37


Facts About Comments

Sad But True

C code is difficult to understandit aint literate programming (Knuths TEXexperience)

Documentation (design drafts, Ph.D. theses, papers, webpages) never isup-to-date (quick-fix disease)

After a few days, the author is not familiar with the code anymore

Every developer has a different style of writing and a different style of thinking

Ill comment when it is runningthat is a lie!

document everything in the code, otherwise its lost or wont be read!

Tools

Best Practices





19/37

p p

Documentation Tools

Screenshot from Doxygen

Sad But True document everything in the code,

otherwise its lost or wont be read!

Tools

Idea: Extract documentation fromsource code

Format: Webpages, PDF, TEX

Content: Diagrams,documentation text, mathematicalformulas, images

Tools: JavaDoc, Doxygen (bothfor C/C++)

Best Practices





20/37

p p

Documentation Best Practices

Sad But True

document everything in the code, otherwise its lost or wont be read! Tools

Best Practices

Document everything in the code (first place)beforeyou write or alter thecode.

If you have to work with a third-party code, document it in the first place andsend it back to the authors. Theyll love you!

Add a description before each complicated part (operations, methods, . . . ).

Make yourself familiar with the tools and add formulas and images to yourdocumentation.

Document what, why, rationale, alternatives, and bugs fixed.





21/37

Screenshot

isPassive

Peano - Source Directory and Architecture grid AbstractVertex isPassive

Search

Help

grid

Directories

adapter

checkpoint

configuration

integration-tests

multicore

records

runners

statistics

tests

Classes

AbstractCell

AbstractLeafEvent

AbstractNewEvent

AbstractPersistentRefined.

AbstractRefinedEvent

AbstractRootEvent

AbstractStackBasedRefined.

AbstractVertex

Block

BlockContainer

Event

EventHelper

GridCell

GridVertex

LeafEvent

NewEvent

PassiveLeafEventPassiveNewEvent

PassivePersistentRefinedE.

PassiveRefinedEvent

PersistentRefinedEvent

RefinedEvent

RootEvent

StaticLeafEvent

Description Source Call Graph

peano::grid::AbstractVertex::isPassive ( Const Public Method )Author: Tobias Weinzierl

Is Vertex Passive.

Syntax / parameters

Return value

Description

Is Vertex Passive.

If a vertex is passive, the event handler is not called for this vertex. A vertex is set passive by the NewEvent::createVertex() operation if the vertex and thewhole surrounding support are outside of the domain.

Be careful to makeany vertex outside the computational domain a passive vertex. If you want to plot boundary cells, e.g., you have to ensure that all thevertices are plotted before you plot your cells. Yet, the touch operations are not invoked for passive vertices.

In the parallelmode, each vertex is set passive that does not hold the actual node as adjacent element. Thus, vertices outside the computational partition are

set passive although they might be inside the computational domain. This switch is implemented within derivePersistentVertexDataFromCoarseGrid() .

Peano-Sources - isPassive http://www5.in.tum.de/peano/src/src/grid/isPassive3949054044_member_description.html

1 von 3 07.06.2010 16:07





22/37

2.4.Floats

# i n c l u d e

i n t main ( ) { / / hey dude , h er e we go double a ;double b ;double c ;a = 0 .1 45 e07;b = 2 3. 24 e09 ;c = a + b ;

s t d : : c ou t


23/37

A GedankenexperimentFixed-point notation

A number in a computer has to have afinite number of bits.

This means a finite number of digits.

Lets assume we have four digits in theformxx.yy.

a = 0 1 . 5 0 ;b = 0 0 . 5 0 ;c = a / b ;d = c / 3 ;





24/37

Floating-point notation

A dynamic scheme (with a header, e.g.) cannot be fast (although Maple

e.g. supports this if we need arbitary number of significant/valid digits). Define number into significant digitsand exponent.

Make the representation unique, i.e.ab.cd 104 =a.bcd 105. This is anormalisation.

In a binary system, base is 2 and first digit before comma always is 0 (or 1respectively).

So, we need one bit for the sign, sbits for the significant digits, one bit for the signof the exponent, andebits for the exponent.

Type Sign Exponent Significand Total bits

Single 1 8 23 32Double 1 11 52 64

This is a at least standard! And theres two additional/special values: nanand inf.





25/37

Normalisation & round-off errors

# i n c l u d e

i n t main ( ) {/ / hey dude , here we go

double a ;double b ;double c ;a = 0 .1 45 e07;b = 2 3. 24 e09 ;c = a + b ;s t d : : c ou t


26/37

Normalisation & round-off errors

C1 =N

i=1

i

C2 =

N/2

i=1

(i+N i)

Which variant is the better one?

What means stability in this context?

Why is the condition number importantwithin this context?

We have to study all our algorithms in de-

tail!





27/37

Some remarks on performance

A Flop is a floating point operation.

A MFlop is a . . . ? How many Flops does the SuperMUC provide?

If one of these GPGPU guys tells you something about performance, ask himabout which precision he is using!





28/37

2.5.Simple Expressions

The Assignment Statement

The assignment operator takes expressionfrom the right-hand side, evaluates it, andstores the result in the variable on the left-hand side.From a mathematical point of view, choo-sing = as assignment operator is a poorchoice. Pascal, e.g., uses the := operator,which, from the authors point of view, is abetter symbol.A declaration can directly be combined withan assignemnt.





29/37

Arithmetic Expressions (for Numbers)

c=10+2; assign variablecthe value 12.

a=10; b=a; a=2; afterwards,aholds 2 andbholds 10, i.e. a=b.

c=10*2; assign variablecthe value 20.

b=3; a=(b+2)*10; assign variablebthe value 3.abecomes 50.

a=2; a=a*2; this aint a fixpoint formula, i.e.aholds the value 4 afterwards.

a=3; a++; incrementoperator makesahold 4.

(unary operator)

a=3; a--; decrementoperator.

a=2; a+=4; shortcut fora=a+4.

a=2; a-=4; shortcut fora=a-4.

a=2; a*=4; shortcut fora=a*4.

a=a/2; divide value ofa by 2.

a=4; a/=2; shortcut fora=a/2.





30/37

Further Expressions

For booleans

bool a=true; bool a=false;

bool a=0;

bool a=1;

bool a=2;

a=!a;

a=!(a | a);

For characters

char x=a;

char x=40;

Comparsions

bool x=a>4;

bool x=a>b;

bool x=a==b;

bool x=a!=b;





31/37

Fancy Initialisation

: In C++, you can initialise all variables with brackets!

int a = 10; int a(10);

2. Variables, Identifiers, and ExpressionsEinfuhrung in die ProgrammierungIntroduction to C/C++, Tobias Weinzierl page 31 of 35



32/37

Initialisation Pitfalls IMultiple Declarations

int a;

int b;

int a,b; Shortcut int a,b =2 ; What does this

mean?




33/37

Initialisation Pitfalls IMultiple Declarations

int a;

int b;

int a,b; Shortcut int a,b =2 ; What does this

mean?

either declare one variable per line or use C++ initialisation with brackets.




34/37

Initialisation Pitfalls IIImplicit Type Conversion

double a;

double a=0.3;

double a=10.0 / 4.0;

double a=10.0 / 4;

double a=10 / 4;




35/37

Initialisation Pitfalls IIImplicit Type Conversion

double a;

double a=0.3;

double a=10.0 / 4.0;

double a=10.0 / 4;

double a=10 / 4;

If you use floating point arithmetics always write .0 for natural numbers. If you are interested in the underlying techniques, study the field oftype inference. If you wanna have a more dynamic feeling, use a language with a dynamic typesystem.




36/37

Initialisation Pitfalls IIIAssignment Operators

i n t a = 3

bool b = (a==3);

bool c = (a=3);i n t d = 0

bool e = (d==0);

bool f = (d=0);




37/37

Initialisation Pitfalls IVFreaky Collaborators

int a = 3; int b = ++a; (that is want we want)int c = 3; int d = c++; (that is something different)

int e = (d+=a)++;(also very funny)int f = ((d=4)+a)--;

C/C++ offer many strange constructs. This is good luck for posers, freaks, andpeople giving lectures. Others should avoid them.


02 variables

Documents