Modern Fortran: Building Efficient Parallel Applications

Modern Fortran
contents
foreword
preface
acknowledgments
about this book
Who should read this book
A bit of Fortran history
How this book is organized: a roadmap
About the code
Requirements
Get involved
liveBook discussion forum
about the author
about the cover illustration
Part 1—Getting started with Modern Fortran
1 Introducing Fortran
1.1 What is Fortran?
1.2 Fortran features
1.3 Why learn Fortran?
1.4 Advantages and disadvantages
1.4.1 Side-by-side comparison with Python
1.5 Parallel Fortran, illustrated
1.6 What will you learn in this book?
1.7 Think parallel!
1.7.1 Copying an array from one processor to another
1.8 Running example: A parallel tsunami simulator
1.8.1 Why tsunami simulator?
1.8.2 Shallow water equations
1.8.3 What we want our app to do
1.9 Further reading
Summary
2 Getting started: Minimal working app
2.1 Compiling and running your first program
2.2 Simulating the motion of an object
2.2.1 What should our app do?
2.2.2 What is advection?
2.3 Implementing the minimal working app
2.3.1 Implementation strategy
2.3.2 Defining the main program
2.3.3 Declaring and initializing variables
2.3.4 Numeric data types
2.3.5 Declaring the data to use in our app
2.3.6 Branching with an if block
2.3.7 Using a do loop to iterate
2.3.8 Setting the initial water height values
2.3.9 Predicting the movement of the object
2.3.10 Printing results to the screen
2.3.11 Putting it all together
2.4 Going forward with the tsunami simulator
2.5 Answer key
2.5.1 Exercise: Cold front propagation
2.6 New Fortran elements, at a glance
2.7 Further reading
Summary
Part 2—Core elements of Fortran
3 Writing reusable code with functions and subroutines
3.1 Toward higher app complexity
3.1.1 Refactoring the tsunami simulator
3.1.2 Revisiting the cold front problem
3.1.3 An overview of Fortran program units
3.2 Don’t repeat yourself, use procedures
3.2.1 Your first function
3.2.2 Expressing finite difference as a function in the tsunami simulator
3.3 Modifying program state with subroutines
3.3.1 Defining and calling a subroutine
3.3.2 When do you use a subroutine over a function?
3.3.3 Initializing water height in the tsunami simulator
3.4 Writing pure procedures to avoid side effects
3.4.1 What is a pure procedure?
3.4.2 Some restrictions on pure procedures
3.4.3 Why are pure functions important?
3.5 Writing procedures that operate on both scalars and arrays
3.6 Procedures with optional arguments
3.7 Tsunami simulator: Putting it all together
3.8 Answer key
3.8.1 Exercise 1: Modifying state with a subroutine
3.8.2 Exercise 2: Writing an elemental function that operates on both scalars and arrays
3.9 New Fortran elements, at a glance
3.10 Further reading
Summary
4 Organizing your Fortran code using modules
4.1 Accessing a module
4.1.1 Getting compiler version and options
4.1.2 Using portable data types
4.2 Creating your first module
4.2.1 The structure of a custom module
4.2.2 Defining a module
4.2.3 Compiling Fortran modules
4.2.4 Controlling access to variables and procedures
4.2.5 Putting it all together in the tsunami simulator
4.3 Toward realistic wave simulations
4.3.1 A brief look at the physics
4.3.2 Updating the finite difference calculation
4.3.3 Renaming imported entities to avoid name conflict
4.3.4 The complete code
4.4 Answer key
4.4.1 Exercise 1: Using portable type kinds in the tsunami simulator
4.4.2 Exercise 2: Defining the set_gaussian subroutine in a module
4.5 New Fortran elements, at a glance
4.6 Further reading
Summary
5 Analyzing time series data with arrays
5.1 Analyzing stock prices with Fortran arrays
5.1.1 Objectives for this exercise
5.1.2 About the data
5.1.3 Getting the data and code
5.2 Finding the best and worst performing stocks
5.2.1 Declaring arrays
5.2.2 Array constructors
5.2.3 Reading stock data from files
5.2.4 Allocating arrays of a certain size or range
5.2.5 Allocating an array from another array
5.2.6 Automatic allocation on assignment
5.2.7 Cleaning up after use
5.2.8 Checking for allocation status
5.2.9 Catching allocation and deallocation errors
5.2.10 Implementing the CSV reader subroutine
5.2.11 Indexing and slicing arrays
5.3 Identifying risky stocks
5.4 Finding good times to buy and sell
5.5 Answer key
5.5.1 Exercise 1: Convenience (de)allocator subroutines
5.5.2 Exercise 2: Reversing an array
5.5.3 Exercise 3: Calculating moving average and standard deviation
5.6 New Fortran elements, at a glance
5.7 Further reading
Summary
6 Reading, writing, and formatting your data
6.1 Your first I/O: Input from the keyboard and output to the screen
6.1.1 The simplest I/O
6.1.2 Reading and writing multiple variables at once
6.1.3 Standard input, output, and error
6.2 Formatting numbers and text
6.2.1 Designing the aircraft dashboard
6.2.2 Formatting strings, broken down
6.2.3 Format statements in legacy Fortran code
6.3 Writing to files on disk: A minimal note-taking app
6.3.1 Opening a file and writing to it
6.3.2 Opening a file
6.3.3 Writing to a file
6.3.4 Appending to a file
6.3.5 Opening files in read-only or write-only mode
6.3.6 Checking whether a file exists
6.3.7 Error handling and closing the file
6.4 Answer key
6.4.1 Exercise: Redirect stdout and stderr to files
6.5 New Fortran elements, at a glance
Summary
Part 3—Advanced Fortran use
7 Going parallel with Fortran coarrays
7.1 Why write parallel programs?
7.2 Processing real-world weather buoy data
7.2.1 About the data
7.2.2 Getting the data and code
7.2.3 Objectives
7.2.4 Serial implementation of the program
7.3 Parallel processing with images and coarrays
7.3.1 Fortran images
7.3.2 Getting information about the images
7.3.3 Telling images what to do
7.3.4 Gathering all data to a single image
7.4 Coarrays and synchronization, explained
7.4.1 Declaring coarrays
7.4.2 Allocating dynamic coarrays
7.4.3 Sending and receiving data
7.4.4 Controlling the order of image execution
7.5 Toward the parallel tsunami simulator
7.5.1 Implementation strategy
7.5.2 Finding the indices of neighbor images
7.5.3 Allocating the coarrays
7.5.4 The main time loop
7.6 Answer key
7.6.1 Exercise 1: Finding the array subranges on each image
7.6.2 Exercise 2: Writing a function that returns the indices of neighbor images
7.7 New Fortran elements, at a glance
7.8 Further reading
Summary
8 Working with abstract data using derived types
8.1 Recasting the tsunami simulator with derived types
8.2 Defining, declaring, and initializing derived types
8.2.1 Defining a derived type
8.2.2 Instantiating a derived type
8.2.3 Accessing derived type components
8.2.4 Positional vs. keyword arguments in derived type constructors
8.2.5 Providing default values for derived type components
8.2.6 Writing a custom type constructor
8.2.7 Custom type constructor for the Field type
8.3 Binding procedures to a derived type
8.3.1 Your first type-bound method
8.3.2 Type-bound methods for the Field type
8.3.3 Controlling access to type components and methods
8.3.4 Bringing it all together
8.4 Extending tsunami to two dimensions
8.4.1 Going from 1-D to 2-D arrays
8.4.2 Updating the equation set
8.4.3 Finite differences in x and y
8.4.4 Passing a class instance to diffx and diffy functions
8.4.5 Derived type implementation of the tsunami solver
8.5 Answer key
8.5.1 Exercise 1: Working with private components
8.5.2 Exercise 2: Invoking a type-bound method from an array of instances
8.5.3 Exercise 3: Computing finite difference in y direction.
8.6 New Fortran elements, at a glance
8.7 Further reading
Summary
9 Generic procedures and operators for any data type
9.1 Analyzing weather data of different types
9.1.1 About the data
9.1.2 Objectives
9.1.3 Strategy for this exercise
9.2 Type systems and generic procedures
9.2.1 Static versus strong typing
9.3 Writing your first generic procedure
9.3.1 The problem with strong typing
9.3.2 Writing the specific functions
9.3.3 Writing the generic interface
9.3.4 Results and complete program
9.4 Built-in and custom operators
9.4.1 What’s an operator?
9.4.2 Things to do with operators
9.4.3 Fortran’s built-in operators
9.4.4 Operator precedence
9.4.5 Writing custom operators
9.4.6 Redefining built-in operators
9.5 Generic procedures and operators in the tsunami simulator
9.5.1 Writing user-defined operators for the Field type
9.6 Answer key
9.6.1 Exercise 1: Specific average function for a derived type
9.6.2 Exercise 2: Defining a new string concatenation operator
9.7 New Fortran elements, at a glance
Summary
10 User-defined operators for derived types
10.1 Happy Birthday! A countdown app
10.1.1 Some basic specification
10.1.2 Implementation strategy
10.2 Getting user input and current time
10.2.1 Your first datetime class
10.2.2 Reading user input
10.2.3 Getting current date and time
10.3 Calculating the difference between two times
10.3.1 Modeling a time interval
10.3.2 Implementing a custom subtraction operator
10.3.3 Time difference algorithm
10.3.4 The complete program
10.4 Overriding operators in the tsunami simulator
10.4.1 A refresher on the Field class
10.4.2 Implementing the arithmetic for the Field class
10.4.3 Synchronizing parallel images on assignment
10.5 Answer key
10.5.1 Exercise 1: Validating user input
10.5.2 Exercise 2: Leap year in the Gregorian calendar
10.5.3 Exercise 3: Implementing the addition for the Field type
10.6 New Fortran elements, at a glance
Summary
Part 4—The final stretch
11 Interoperability with C: Exposing your app to the web
11.1 Interfacing C: Writing a minimal TCP client and server
11.1.1 Introducing networking to Fortran
11.1.2 Installing libdill
11.2 TCP server program: Receiving network connections
11.2.1 IP address data structures
11.2.2 Initializing the IP address structure
11.2.3 Checking IP address values
11.2.4 Intermezzo: Matching compatible C and Fortran data types
11.2.5 Creating a socket and listening for connections
11.2.6 Accepting incoming connections to a socket
11.2.7 Sending a TCP message to the client
11.2.8 Closing a connection
11.3 TCP client program: Connecting to a remote server
11.3.1 Connecting to a remote socket
11.3.2 Receiving a message
11.3.3 The complete client program
11.4 Some interesting mixed Fortran-C projects
11.5 Answer key
11.5.1 Exercise 1: The Fortran interface to ipaddr_port
11.5.2 Exercise 2: Fortran interfaces to suffix_detach and tcp_close
11.6 New Fortran elements, at a glance
11.7 Further reading
Summary
12 Advanced parallelism with teams, events, and collectives
12.1 From coarrays to teams, events, and collectives
12.2 Grouping images into teams with common tasks
12.2.1 Teams in the tsunami simulator
12.2.2 Forming new teams
12.2.3 Changing execution between teams
12.2.4 Synchronizing teams and exchanging data
12.3 Posting and waiting for events
12.3.1 A push notification example
12.3.2 Posting an event
12.3.3 Waiting for an event
12.3.4 Counting event posts
12.4 Distributed computing using collectives
12.4.1 Computing the minimum and maximum of distributed arrays
12.4.2 Collective subroutines syntax
12.4.3 Broadcasting values to other images
12.5 Answer key
12.5.1 Exercise 1: Hunters and gatherers
12.5.2 Exercise 2: Tsunami time step logging using events
12.5.3 Exercise 3: Calculating the global mean of water height
12.6 New Fortran elements, at a glance
12.7 Further reading
Summary
Appendix A—Setting up the Fortran development environment
A.1 Editing Fortran source files
A.2 Setting up the Fortran compiler
Linux
macOS
Windows
A.3 Setting up the MPI library (Message Passing Interface)
A.4 Setting up OpenCoarrays
Linux
macOS
Using OpenCoarrays
A.5 Building a Docker image
Appendix B—From calculus to code
B.1 The advection equation explained
B.1.1 Discretizing the derivatives
B.1.2 Casting the derivatives into code
Appendix C—Concluding remarks
C.1 Tsunami simulator: The complete code
C.1.1 Main program: tsunami.f90
C.1.2 The Field module: mod_field.f90
C.1.3 The I/O module: mod_io.f90
C.1.4 The parallel module: mod_parallel.f90
C.2 Going forward with the tsunami simulator
C.3 Neural networks and deep learning
C.4 Online resources
C.5 Compilers
C.6 Books
index
Symbols
A
B
C
D
E
F
G
H
I
K
L
M
N
O
P
Q
R
S
T
U
V
W
Y