C2115 Practical Introduction to Supercomputing 3rd Lesson -1C2115
Practical Introduction to
Supercomputing
Petr Kulhánek, Jakub Štěpán
kulhanek@chemi.muni.cz
National Centre for Biomolecular Research, Faculty of Science
Masaryk University, Kotlářská 2, CZ-61137 Brno
CZ.1.07/2.2.00/15.0233
3rd Lesson
C2115 Practical Introduction to Supercomputing 3rd Lesson -2-
Contents
 Computer Architecture
CPU, memory, graphics system, disks, network, pheripherials
 Numerical values representation in digital devices
whole numbers, real numbers
 From problem to result
algorithm, source codes, compilation, program running, programming
languages
C2115 Practical Introduction to Supercomputing 3rd Lesson -3Computer
architecture
 CPU, memory, graphics system, disks,
network, pheripherials
C2115 Practical Introduction to Supercomputing 3rd Lesson -4-
Overview
CPU Memory I/O
Controll bus
Data bus
Adress bus
1945 von Neumann architecture 1944 Hardward architecture
CPU
Program
memory
Data
memory
I/O
John von Neumann, hungarian
mathematician, worked in US
Harvard Mark I - completed computer
from relay, 24 bit instructions
ro rw
rw
• program may change itself
• program and data can not be
loaded simultaneously
• program may not change itself
• program and data may be loaded
simultaneously
C2115 Practical Introduction to Supercomputing 3rd Lesson -5Current
computers combine both models.
Overview
CPU Memory I/O
Controll bus
Data bus
Adress bus
1945 von Neumann architecture 1944 Hardward architecture
CPU
Program
memory
Data
memory
I/O
John von Neumann, hungarian
mathematician, worked in US
Harvard Mark I - completed computer
from relay, 24 bit instructions
ro rw
rw
• program may change itself
• program and data can not be
loaded simultaneously
• program may not change itself
• program and data may be loaded
simultaneously
C2115 Practical Introduction to Supercomputing 3rd Lesson -6Typical
computer scheme
CPU
North
bridge
South
bridge
USB
Mouse, keyboard
Real time clock SATA controller
hard drives
BIOS
Graphics
system
Memory
Memory controller
PCI Express bus fast access peripherals
network (ethernet)Sound
PCI bus
von Neumann architecture
North and south bridge
constitute chipset
C2115 Practical Introduction to Supercomputing 3rd Lesson -7Typical
computer scheme, II
CPU
North bridge
South
bridge
USB
Mouse, keyboard
Real time clock SATA controllers
hard drives
BIOS
graphics
system
Memory
Memory controller
PCI Express bus fast access peripherals
Network (ethernet)sound
PCI bus
Memory controller becomes part
of new processors.
C2115 Practical Introduction to Supercomputing 3rd Lesson -8-
CPU
Processor or CPU - Central Processing Unit is main part of computer; it is complicated
sequence circuite that processes machine code from operating memory. Machine code
consists from machine instructions of computer programs in memory.
www.wikipedia.org
CPU
Controller
ALU
ALU (arithmetic and logic unit),
processes arithmetic and logic
operations
Loads machine instructions and data
and prepares their processing in ALU
Sequence processing of machine instructions is
controlled by internal clock tick
C2115 Practical Introduction to Supercomputing 3rd Lesson -9Numerical
values
representation
 Whole numbers
 Real numbers
C2115 Practical Introduction to Supercomputing 3rd Lesson -10Whole
numbers
Smallest data unit in digital technique is one bit. Words are made from bits. Smallest word
is byte, that contains 8 bits.
One byte may contain numbers from 0 to 255.
128 64 32 16 8 4 2 1
0 1 0 1 0 1 1 1 = 87
Whole numbers with sign may be expressed too. In such a case one bit is dedicated to
sign, remaining bits for number. There are multiple possible implementations. Intel
architecture uses two’s complement, that leads to range from -128 to 127.
128 64 32 16 8 4 2 1
0 1 1 1 1 1 1 1 = 127
0 1 0 1 0 1 1 1 = 87
0 0 0 0 0 0 0 1 = 1
0 0 0 0 0 0 0 0 = 0
1 1 1 1 1 1 1 1 = -1
1 0 1 0 1 0 0 0 = -87
1 0 0 0 0 0 0 0 = -128
bit dedicated for
number sign
C2115 Practical Introduction to Supercomputing 3rd Lesson -11Whole
numbers, II
Whole numbers with larger range may be expressed by larger words typically consisting of
four bytes (32 bit word) or eight bytes (64 bit word).
Working with numbers is limited by range, it is not possible to express any large number,
avoid overflow of the value.
32 bit whole number unsigned: 0 to 4.294.967.295
32 bit whole number signed : −2.147.483.648 to 2.147.483.647
64 bit whole number unsigned : 0 to 18.446.744.073.709.551.615
64 bit whole number signed : −9.223.372.036.854.775.808 to
9.223.372.036.854.775.807
C2115 Practical Introduction to Supercomputing 3rd Lesson -12Real
numbers
Real numbers are expressed in floating point format:
mantissa
exponent
In digital technology real numbers are usually expressed according to standard IEEE 754.
type width mantissa exponent
single precision 32 23 8
double precision 64 52 11
...
2
1
2
1
2
1
2
1
44332211
mmmmQ 
m1,m2,m3 are mantissa bits
C2115 Practical Introduction to Supercomputing 3rd Lesson -13Real
numbers, II
type range Precision
single precision ±1,18×10−38 to ±3, 4×1038 Approximately 7 decimal
places
double precision ±2,23×10−308 to ±1,80×10308 Approximately 15 decimal
places
Special combination of mantissa and exponent special values may be expressed:
0 positive zero
-0 negative zero
NaN not a number, e. g. division by zero
+Inf positive infinity (number too large to express)
-Inf negative infinity (negative number too large to express)
Working with real numbers is limited by spread of rounding error. Logical compating
operators equal to and not equal to are not appropriate to be used with real numbers
(except to comparing with zero).
C2115 Practical Introduction to Supercomputing 3rd Lesson -14Exercise
LII.1
1. Variable of signed char type(byte with sign) contains number 127. What will be walue
if we increment value by one?
2. What will be result of sum of two real numbers in double precission with values:
0,1346978.10-12
1,2312657.106
C2115 Practical Introduction to Supercomputing 3rd Lesson -15From
problem to result
 Algorithm
 Source codes, compilation
 Running program
 Programming languages
C2115 Practical Introduction to Supercomputing 3rd Lesson -16From
problem to result ...
Problem Result
C2115 Practical Introduction to Supercomputing 3rd Lesson -17From
problem to result ...
Problem Result
algorithm
source
code
program
calculation
C2115 Practical Introduction to Supercomputing 3rd Lesson -18From
problem to result ...
Problem Result
algorithm
source
code
program
calculation
Flowcharts
Language selection,
human optimising
Compilation,
machine optimising
Alocation and
effective
resource usage
C2115 Practical Introduction to Supercomputing 3rd Lesson -19From
problem to result ...
Problem Result
algorithm
source
code
program
calculation
hardware
Flowcharts
Language selection,
human optimising
Compilation,
machine optimising
Alocation and
effective
resource usage
C2115 Practical Introduction to Supercomputing 3rd Lesson -20From
problem to result ...
algorithm
source
code
program
calculation
hardware
Solving problems using computers (supercomputers) it is necessary to evaluate number of
aspects including used hardware and architecture.
Flowcharts
Language selection,
human optimising
Compilation,
machine optimising
Alocation and
effective
resource usage
C2115 Practical Introduction to Supercomputing 3rd Lesson -21Lecture
focus ...
algorithm
source
code
program
calculation
hardware
Basic problem with creation of programs
adreesing complex problems - parallelisation
Solving problems using computers (supercomputers) it is necessary to evaluate number of
aspects including used hardware and architecture.
C2115 Practical Introduction to Supercomputing 3rd Lesson -22Lecture
focus ...
algorithm
source
code
program
calculation
hardware
Effective running of
computational chemistry
applications (MetaCentrum,
small clusters)
Solving problems using computers (supercomputers) it is necessary to evaluate number of
aspects including used hardware and architecture.
C2115 Practical Introduction to Supercomputing 3rd Lesson -23Programs
vs. Scripts
Source code
program
input
output
Compilation
script interpreter
input
output
Program is machine instruction file
processed directly by processor. It is
created by procedure called compilation
from source code.
Compiled languages:
C/C++
Fortran
Script is text file containing commands
and special constructions, these are
processed by interpreter of scripting
language.
Skripting languages:
bash
gnuplot
awk
JavaScript
PHP
C2115 Practical Introduction to Supercomputing 3rd Lesson -24Programs
vs. Scripts, ...
No recompilation
Program can generate and
run self running code
Poor optimization
Slower processing
Easy optimization
Fast processing
Recompilation needed
Self run code not available
Source code
program
input
output
Compilation
script interpreter
input
output
C2115 Practical Introduction to Supercomputing 3rd Lesson -25Programs
vs Scripts
Programs dedicated to demanding scientific-technical calculations are allways written in
compiled programming languages. These include:
 Fortran
 C/C++
Scripting languages are not used in such calculations at all, or only in supporting parts of
calculation, that are not demanding.
C2115 Practical Introduction to Supercomputing 3rd Lesson -26Exercise
LII.2
1. Determine rate of programs in languages Fortran, C/C++ and others, that are listed on
page:
http://en.wikipedia.org/wiki/List_of_quantum_chemistry_and_solid_state_physics_software
Plot result in sector graph.