C Programming Basics for High Performance Computing

1
C Programming Basics for High Performance
Computing

• A Short-Series Presentation by
• Evan Lee Turner

2
For the Beginner

• This presentation is created for the beginning
  programmer with little or no previous programming
  knowledge.
• Not everything is covered!
• Will emphasize how C works, what it can do, and
  how best to write efficient C code.
• Things you will not find in the manual
• Useful tools for HPC users
• Assume that you have taken the previous UNIX
  course

3
History of C

• Dennis Ritchie developed the C language with the
  UNIX OS in the early 1970s.
• UNIX and LINUX kernels are still developed in C
  and assembly today.

4
Why Use C?

• Scientific Applications
• C and FORTRAN remain the dominant programming
  forces in scientific computing because of the
  speed of execution. Performance libraries are
  written in C. Parallel pragma language
  extensions exist for C, C, and FORTRAN such as
  openMP and MPI.
• UNIX/LINUX Programming
• Since the OS and libraries are written in C there
  is a natural transition to write applications for
  Linux in C
• Interoperability with Fortran
• Traditionally, scientists will use C to glue
  together FORTRAN functions since C is easier to
  develop and maintain

5
My First Program

• include ltstdio.hgt
• int main()
• printf(hello world\n)
• return 1

6
Variables

• Several basic datatypes exist in C to represent
  data
• typename variablename
• Defines a variable name and type. All variables
  must be defined in C.
• Examples
• int myInteger
• char myCharacter

7
Integer Types

• Byte sizes for Lonestar Ranger.
• char c
• printf("u,u", (unsigned) sizeof c, (unsigned)
  sizeof (int))
• int - 4
• long - 8
• char - 1
• short - 2

8
Float Types

• The majority of scientific codes will use
  floating-point data in double precision
• float (single precision) 4 bytes
• double - 8 bytes
• long double 10 bytes

9
Character Types

• char myName10
• Create a string variable called myName which
  contains 10 elements
• strcpy(myName, Evan)
• myName Evan ! Invalid syntax

10
Assigning Values

• Why cant I assign a string to a value?
• Only scalar (a single value) can be assigned at a
  time. A string is an array of values, and each
  location must be assigned individually
• Values are assigned by the assignment operator
  .
• Valid syntax
• char myChar
• int x
• X 3
• myChar x
• - notice here we can use
  instead of

11
Mathematical Operations

• increment decrement , --
• result
• result result 1
• Normal mathematical operations , -, divide /
• capture a remainder for integer numbers
• result sum 4 / result is 2 /
• Normally when an operation occurs on variables of
  different type, the lesser type is promoted.

12
Working with Strings

• Must use built-in functions to work with string
  values
• strcpy(destination, source)
• copies one string to another
• strcpy(name1, name2)
• copies contents of name2 into name1
• int strcmp(string1, string2)
• compares string1 with string2, returns an
  integer less than, equal to, or greater than zero
  if s1 is found, respectively, to be less than,
  to match, or be greater than s2.

13
Strings and Arrays

• A string is just a one-dimensional array of
  characters
• An array is defined as
• datatype name size
• int myarray6
• myArray0 54
• Places the integer 54 in the first position of
  the array

14
Functions

• functions()
• A function can be thought as a mini-program where
  inputs are taken in and values returned
• returnValue MyFunction(input)
• Some functions are defined by the system such as
  printf(), scanf(), open(), or close(). User
  defined functions are declared in the calling
  program.

15
Program Flow

• loops
• for(i 0 i lt num i)
• while(thisValueIsTrue)
• do while(thisValueisTrue)
• Conditionals
• if (valueIsTrue) do this.. else do this..
  
• nested if else statements are possible

16
Constructing Logic

• Use in conditional statements (and loops)
• evaluation operators , gt, lt
• if (value1 gt value2)
• Logic and, or, not
• and operation, , !
• if (value1 gt value2 value3)
• Order of operation
• logic is processed first, then evaluations
• if((value1 gt value2) (value3))

17
Advanced Control

• continue
• used in a for() loop will cause the current
  iteration to stop and will start the next loop
  iteration
• break
• used to break outside of a loop or conditional
  prematurely.
• exit
• quits the program

18
Swtich Statements

• Useful for constructing argument parsing instead
  of using large if-else-if-else statements
• switch(letter)
• case A printf(A is for apple\n)
• break
• case B printf(B is for Bob Garza\n)
• break

19
Output

• printf, the all-in-one output function for C.
• printf (hello world\n)
• will output hello world to screen with a newline
• printf(my name is s\n, myName)
• output my name is and the contents of string
  myName to screen with a newline

20
Conversion Characters

• Printing values of all data types is possible
  with conversion characters. Each conversion
  character tells printf what kind of data is
  expected to follow.
• Common Characters
• d, i, u integer
• x hexadecimal
• f floating point number
• s string
• c character

21
More printf Examples

• printf(My name is s and I am d years old\n,
  myName, myAge)
• Carriage control is provided by the backslash
• \n, newline
• \t tab
• \b backspace
• \\ produce a single backslash \

22
Advanced Output

• printf() is a highly robust function that can
  parse output effectively
• printf("\n\nHOST\t\tR15s\tR1m\tR15m\tPAGES
  \t\tMEM\tSWAP\tTEMP\n")
• printf("s\t5.1f\t5.1f\t5.1f\t5.1fP/s\t4.fM\t
  4.fM\t4.fM\n",
• hostsi.hostName,hostsi.liR15S,hos
  tsi.liR1M,hostsi.liR15M,
• hostsi.liPG,hostsi.liMEM,hosts
  i.liSWP,hostsi.liTMP)
• produces the output
• HOST R15s R1m R15m PAGES
  MEM SWAP TEMP
• c21-208 1.0 1.0 1.0 2.1P/s
  6468M 2021M 55776M

23
Input

• Input can be handled with the scanf function
  which behaves similarly to printf
• scanf (d f, integer, float)
• Variables must be handed to scanf with the
  address operator . In C the address operator
  gives a variables address in memory, not the
  value itself.

24
Working with scanf

• scanf will parse standard input and match cases
  that it finds.
• gets (string)
• grabs all of the string input on the screen and
  will save it in the array string
• Later in the presentation we will see a better
  way to do input processing

25
Working with Files

• Manipulating files is easy and fun with C!
• / definitions /
• define NAMELIMIT 40
• FILE fptr
• char filenameNAMELIMIT
• / in code /
• fptr fopen(filename, r)
• if ( fptr NULL)
• perror(error openning file)
• fcose (fptr)

26
Working with File Pointers

• File pointers can be manipulated in the same
  fashion as I/O from the keyboard and to the
  screen.
• Instead of fprint fscanf can read and write to a
  file pointer.
• int item
• fscanf (fptr, d, item) / read one value /
• printf( d\n, item) / print item to screen
  /

27
fprintf() and fscanf()

• Usage is exactly the same as printf and scanf
  except for specifying the file pointer in the
  beginning of the satement.
• fprintf(fptr, .)
• Incidentally, all input and output are file
  pointers that can be manipulated. STDOUT is the
  screen pointer. (STDIN is the input pointer)
• fprintf(STDOUT, Hi there!)
• printf(Hi there!) / equivalent statement
  /

28
fopen() arguments

• FILE fptr
• fptr fopen( myfile, r)
• - Open file for reading

29
Putting I/O together

• According to the MAN page for gets()
• BUGS
• Never use gets(). Because it is impossible to
  tell without knowing the data in advance how
  many characters gets() will read, and
  because gets() will continue to store
  characters past the end of the buffer, it is
  extremely dangerous to use. It has been used to
  break computer security. Use fgets() instead

30
Best Practice

• ALWAYS be suspect of the input given from users
  or other programs and use error cases!
• char fgets(char s, int size, FILE stream)
• Good error detection aids in debugging and test
  cases, program utility, and will extend program
  life.

31
Where to Look for Help

• The UNIX system manual
• man fprintf
• PRINTF(3)
  Linux Programmer's Manual
• NAME
• printf, fprintf, sprintf, snprintf,
  vprintf, vfprintf, vsprintf, vsnprintf -
  formatted output conversion
• SYNOPSIS
• include ltstdio.hgt
• int printf(const char format, ...)
• int fprintf(FILE stream, const char
  format, ...)
• int sprintf(char str, const char format,
  ...)
• int snprintf(char str, size_t size, const
  char format, ...)

32
Importance of Documentation

• A necessity of programmers!
• C comments
• // comments a single line - GNU gcc
• / comment
• a block
• of text /

33
Using Brackets

• / What is wrong with this block? /
• while(true)
• for (i i lt 24 i 0)
• if ( i 5)
• printf(i is equal to 5)

We did not terminate this if() conditional with
another . The program would error on compile
time.
34
Which would be best?
/ better still?? / while(true) for (i i lt
24 i 0) if ( i 5)
printf(i is equal to 5)

• / better? /
• while(true)
• for (i i lt 24 i 0)
• if ( i 5)
• printf(i is equal to 5)

How about?
Or even?
while(true) for (i i lt 24 i 0) if (
i 5) printf(i is equal to 5)
35
less code is not always best

• The last example may not be the most readable,
  but neither is the first. The best documentation
  method is something in between the extremes.
• Unless you are experienced, it is best to
  terminate your blocks
• while(true)
• for (i i lt 24 i 0)
• if ( i 5)
• printf(i is equal to 5)
• // END FOR
• // END WHILE (true)

36
Self-Documenting Code

• Programmers create their own styles and
  conventions. I tend to name variables as
  firstwordSecondword and functions in Caps().
• myName, ourAddress, PrintTable()
• For the Love of UNIX and C make your variable
  names stand for something other than
  x1,xx1,zz2,t1,t2,t3
• Generally only loop counters and temp variables
  should be named with junk titles
• i,j,k, tmp,
• The term self-documentation refers to being able
  to read a code without any extra comments

37
Necessity of Good Style

• Self-Documentation and the position of code and
  brackets are all separate of comments, but are
  absolutely vital to clean, fast running code
• Poorly documented code with bad style will lend
  itself to bugs even if semantically correct

38
More Advanced Stuff

• Parsing your command arguments
• when you run a program on the command line,
  arguments that are passed to the program can be
  parsed. This is VERY useful for dynamic input
  such as input files.
• In order to capture your arguments start a C
  program with
• int main(int argc, char argv)

39
Example from lsuser.c
/ Process my argument List / if(argc gt 1)
if(strchr(argv1,'?')
strchr(argv1,'h')) PrintUsage()
if(strchr(argv1,'s'))
suppressNodes 1 if(strchr(argv1,'r'))
runningOnly 1
if(strchr(argv1,'a')) allJobs 1
if(strchr(argv1,'z'))
suppressNodesAll 1 if(strchr(argv1,'w'
))
40
System Calls

• system()
• Call a UNIX system program externally
• system(clear) / clears the screen /
• note that this will break an MPI program on
  lonestar
• system calls are expensive computationally
• most system calls can actually be accomplished by
  a similar C library. chmod, chown, chdir, rename

41
Retrieving Environment Variables

• include ltstdio.hgt
• include ltstdlib.hgt
• / To shorten example, not using argp /
• int main (int argc, char argv, char envp)
• char home, host
• home getenv("HOME")
• host getenv("HOSTNAME")
• printf ("Your home directory is s on s.\n",
  home, host)
• return 0

42
Signals

• Programs are given signals by the UNIX system to
  quit or exit, and they also create signals when
  errors occur
• It is useful to be able to catch signals in
  order to execute actions when bugs occur
• extremely advanced programming methods (read the
  manual before attempting this)

43
Continuing after Signals
/ Example program by pipas_at_linux.pte.hu
This program is uninterruptable with CtrlC,
uses signal handler http//www.metalshell.co
m/ / include ltstdio.hgt include
ltsignal.hgt include ltunistd.hgt / The signal
handler function / void handler( int signal )
printf("Signal handler...\n") psignal(
signal, "Signal ") /handler/
main() / Registering the handler, catching
SIGINT signals / signal( SIGINT, handler
) / Do nothing / while( 1 )
printf("Running...\n") sleep(10)
/while/ /main/
44
Power of Signals

• Graceful exits when things go wrong
• Program receives a kill signal while files are
  open
• print useful information such as an error code
• point fault locations such as functions or
  procedures

45
Manuals and References

• The C Programming Language by Brian W.
  Kernighan and Dennis M. Ritchie
• The C Bible
• Advanced Programming in the UINIX environment
  by Stevens
• The UNIX Bible
• UNIX System Programming by Haviland
• The UNIX Programming Environment by Kernighan
  and Pike
• http//www.metalshell.com/view/source/106/ -
  example of signal handling