Introducing C++ Part 1

Hello again!

This post was initially meant to go straight on with the Unreal coding standards and the way they express C++ in the engine code.

I’ve made a little detour, as this series was meant to include both those new to C++ and those new to Unreal.

I don’t feel that the Unreal coding standard is much used outside of the Unreal Engine, so in the first posts that actually involves how to write C++ I’ve chosen to use my own coding standard.

All the content from the posts that contain C++ introductory material will also be posted to the C++ page on this blog, and it is also meant to be read by those that wish to learn C++ and not necessarily the Unreal Engine as well.

This might seem a bit clunky, and indeed it is, but this is the way I’ve decided to do it.

So this post and the next few posts will mainly contain C++ until I’ve presented what I believe is a minimum of required knowledge to get started with game development using any engine with a C++ API.

So, let’s get started.

C++ Introduction

C++ is a general purpose programming language with a bias towards systems programming that supports data abstraction, object oriented programming and generic programming.
C++ is a compiled, statically typed language, which means that the compiler must know the type of every object used in a program.

C++ is used in various fields of computer science including computer games.

It received a massive overhaul with the C++11 standard that introduced many new features, such as auto type deduction, lambdas and closures.

The development continues with the C++14 standard and compiler vendors have a great deal of work to do to make their compilers support all the latest features.

What makes C++ a good choice for mobile platforms and systems programming and games is that C++ produces very effective executable code that has low power consumption on mobile platforms combined with great performance in general.

This introduction will make use of some of the new features introduced in C++11, as these features makes C++ easier to understand and more safe to use in many ways.

Most popular compilers that I know of support most of the C++11 standard at the time of this writing, at least what we’ll present in this series of posts.

Why learn C++?

There are many reasons for learning any programming language. What makes C++ a good choice for those who plan on doing programming as a part of their careers?
First of all, C++ has been used in all kinds of systems since the language became an ISO Standard in 1998 and also before that.
Many of those systems that was created back then are still in heavy use today and demands maintenance programmers to keep them up to date.
A great deal of software that we use in our day to day lives today depends on C++ to keep running.

Many new software systems created today also uses C++ because of its universal properties, making it a safe, stable and long-term alternative for cross platform development of various systems, not tied to any manufacturer and not owned or governed by any company.

For more information on C++ as a language and links to various standards, compilers and libraries, visit:
http://www.stroustrup.com/C++.html

There is a ton of valuable information on that page, covering a vast amount of topics.

How to learn C++

If you’re reading this, my guess is that you either want to learn how to use C++ in your programs.
I do not claim that the introduction offered in this series of posts will be anywhere near sufficient to make you a C++ developer.

This is only meant as a starting point for learning more of the intricate nuances that C++ programming is all about.

I want to introduce you to some of the concepts that I found tricky to wrap my head around when I started learning about the C++ programming language, before the YouTube age and the massive amount of learning opportunities available today.

The fact that the sources for learning C++ has become so many and so diverse actually doesn’t make it that much easier, because you have to know what sources to trust as being good and that in it self is a rather hard task if you don’t know what to look for!

I’ve seen both paid and free resources available for learning C++ that range in quality from down right pathetic and plain wrong to excellent and well worth your precious time. Just remember: What you pay for is not necessarily what you get.

In these posts I’ve tried to extract knowledge from various books I wish I had when first learning C++, and turn it into easily digestible bits of information.

Some of the books I highly recommend and that I’ve used as inspiration for writing this series include:

Programming – Principles and Practice Using C++ 2nd Edition by Bjarne Stroustrup et. al
C++ Primer, 5th Edition by Lippman, Lajoie and Moo
Think Like a Programmer – An Introduction to Creative Problem Solving by V. Anton Spraul
Understanding and Using C Pointers by Richard Reese
Effective Modern C++ by Scott Meyers
Safe C++ by Vladimir Kushnir
The C++ Programming Language 4rd Edition by Bjarne Stroustrup

These are all books that are verified and accepted as good learning material by the C++ community at large and is a safe bet if you want to know more about the vast areas of programming and to find specific information about using C++ in the best ways possible.

How to structure the material

I’ve seen many attempts to divide the beginning concepts of C++ into logical parts that makes it “easy” or easier to understand.
Some introduce functions before references and pointers, some introduce classes before general compound types and some choose to ignore the concepts of memory, pointers and references all together in the beginning chapters.

To me that’s all a bit too abstract. I prefer a more direct path.
Before learning to use functions I feel that it is imperative to know something about constants and references, as this will get more important when trying to write various functions that can both manipulate its arguments directly or being denied to change its
arguments as a whole.

I also feel that even though Object Oriented Programming (OOP) has many important uses it is also maybe the most abused concept in modern programming. This is a direct consequence of pure OOP languages such as Java that requires you to create a class just to do a simple “Hello World!” program.

In C++ OOP is an option but not a requirement. You can choose when to start introducing objects into your code and if you can do without it, why complicate matters if it doesn’t lead to anything other than unnecessary complexity?

Well, enough said.
Let’s get on with the actual introduction to programming in C++.

Variables and Basic Types

C++ defines a set of built-in primitive types that include the arithmetic types and a special type named “void”.
Void will be described later when we start talking about functions.

Arithmetic types are divided into “integral” types (numbers) and also include characters (char), like ‘A’ and Boolean (true/false) and floating point types (numbers with a fractional part), like 3.14.

Before continuing on, let’s detour into how these types are stored in a computer.

How machines represent the built-in types

As many knows, computers store data as a sequence of bits. each holding either the value 0 or 1. This is referred to as the binary number system.

That is a stream of bits, like 0001110011001110 …

To keep track of the stored values in memory, computers group chunks of “bits” and the smallest chunk of addressable memory is referred to as a “byte” and is usually made up of 8 bits.

The computer’s basic character set is used to define the size of a byte and in C++ a byte has “at least” as many bits needed to hold a character.
Most machines operate with 8 bit bytes and the ASCII basic character set.

The basic unit of storage is called a “word” and consists of a small amount of bytes grouped together. On 32-bit systems this is usually 4 bytes (4 * 8 = 32) and on 64-bit systems it is the double of that, 8 bytes (8 * 8 = 64).

You can visually represent a word like this:

byte-ordering

There is also a concept of Big-Endian and Little-Endian that describes the order of the bytes in memory and in transmission over a digital link. This is not something you need to concern yourself with when starting out, but it is an important topic later on.
The concept of “Big-Endian” vs. “Little-Endian” can be further investigated here:

https://en.wikipedia.org/wiki/Endianness

The following is a short description of the various arithmetic types.

C++ is designed to let the programmer exploit hardware specific optimizations and therefor provides a number of different types to squeeze the most out of any given hardware platform.

In practice, and when beginning to learn about C++, don’t get caught up in all the details.
The provided information is only provided for the sake of completeness.

Characters

There are a few different character types available, to accommodate for various needs and extended character sets.
The basic character type is called char and big enough to hold numerical values corresponding to the machine’s basic character set.
In addition there is wchar_t, char16_t and char32_t, all used for extended character sets.

char16_t and char32_t are intended for Unicode characters. Unicode is an example of an extended character set.
wchar_t is guaranteed to be large enough to hold any character in the machine’s largest character set, whatever that may be.

The character types has corresponding numerical values that maps to the different characters thay can represent,
like the character ‘A’ correspond to the decimal number 65, ‘B’ is 66 and so on.

Floating Types

These represent single (float), double (double) and extended precision (long double) values.
The standard only specifies a minimum number of significant digits but most compilers provide more precision than the required minimum.
Typically a float is represented in 32 bits giving it around 7 significant digits of precision, a double is stored in 64 bits with around 16 digit precision and a long double is stored in either 96 or 128 bits. The precision of all floating types are hardware dependent and the use of long doubles are largely dependent on the given hardware support.

Signed and Unsigned Types

Except for bool and the extended character types, the integral types may be “signed” (can contain negative values) or “unsigned”
(can only contain values 0 or above).
One way to remember this is that to use a negative number, you have to use a sign (the – sign), hence you need a signed type.

The types int, short, long and long long are all signed types and can contain values below 0. (negative numbers).

We can get the corresponding unsigned type by adding the word “unsigned” before the name of the type, as in unsigned int.
In the case of int you can use the abbreviated form of just unsigned, as in:

unsigned count = 0;

Signed and Unsigned Characters

To complicate things a bit more, there are three different types of characters.
char, signed char and unsigned char are all distinct basic character types.
There are actually only two representations of characters in memory, signed or unsigned, but the plain char uses one of these depending on the compiler you use.

The 8-bit unsigned char can hold values from 0..255 (inclusive) while the signed char can hold values from -127..127.
Most modern machines use representations that allow -128..127, but the standard only guarantees -127..127 (inclusive).

What to use when?

As mentioned earlier, all though C++ provides a staggering amount of arithmetic types, you’ll find yourself using only a handful of these on a day to day basis.

Some rules for what to use when can include:

  • Use an unsigned type when you KNOW that the values should never be negative
  • Use regular int for integer arithmetic. (short is usually to small and long usually has the same size as int)
  • Use long long if you KNOW that the value will exceed the minimum guaranteed size of an int
  • Use double for floating point computations
  • Don’t use char or bool for arithmetic expressions. Only to hold characters or true / false respectively

Next time I’ll cover some information regarding type conversion!

Stay tuned.

INTRODUCTION TO GAME DEVELOPMENT WITH UNREAL 4 AND C++ – PART 2

This section will explain what a project in Unreal is about, and some of the terminology used in the Unreal Engine Documentation, such as objects, classes, actors and pawns.

Projects

So, what exactly does a UE4 project consist of and what is going on when you create one?

A project consists of everything you want to include in your game, all content and all code for interacting with the objects in this content and can be seen as a self-containing entity from which your game is created.

When you create a project you include the source code for the engine in the project and basically you’re working on a separate version of the engine for every game you create.

This is why the start-up of a new UE4 project is taking so long, as the source code must be compiled before you can use the engine in your game. The next time you open your project, things will start a lot faster.

From the Project Page in the Unreal Editor, you can define the basic properties for your project, such as whether you want to start off with a Blueprint or a C++ based solution.

As mentioned earlier, this choice does only concern how the initial gameplay is defined within your project and can be changed around later.

Also, the Project Page defines what Build Target you’re developing for, as in Desktop / Console, Mobile or any other target you plan on developing for.

This is also a setting that can be changed later, in the Project Properties page in the editor after creation.

You can chose between Scale-able 2D / 3D and Maximum Quality graphics. This setting is also configurable after initial creation and is mostly relevant if you create for mobile platforms.

The project I’ll be creating in this series is called “TheMadMan”, and will be a basic UE4 C++ project that contains no starter content.

blog_unreal_cpp_00

The assets used in the project can be downloaded from GitHub along with the source code and project files if you care to follow along.

I’ll cover setting up a GitHub repository for your own projects in a separate post, where I’ll also give you the address for my project files used in this series.

The different templates you can choose from are examples of different gameplay modes and will be discussed later. For now I’ll go with “Basic Code”.

When I hit “Create Project”, the UE4 Editor starts working with setting up the source code for my project and Visual Studio launches to work through the code produced by the editor.

The editor will detect changes to the code made in Visual Studio and reload any changes back into the editor, and changes initiated in the UE4 Editor will update in Visual Studio.

The default “map” or “level” or “scene” is created in the UE4 Editor. The terminology I’ll use for this is a “level” as this is what it is often referred to in the UE4 Documentation and I personally feel that it gives more meaning than “map”.

The level that starts up with your project is by default called “Untitled”, and you can see this in the lower right of the viewport and on the tab in the upper left corner of the screen.

You can save the level with a different name by choosing File → Save As, or using Ctrl+Shift+S and giving it a new name.

When the Unreal Editor starts your project it will look in the Project Settings to determine what level (map) that loads as default into the editor.

Even if you save your map with a new name, it will load the Untitled map when you start your project if you don’t specify otherwise in the Edit → Project Settings → Maps & Modes → Default Maps → Editor Startup Map.

You can see your newly saved level (map) in the Content Browser, which is a central component of the UE4 Editor.

ue4_new_project

A project is stored with a .uproject extension

Although a project is often referenced by the .uproject file associated with it, they are actually two separate files that exist alongside each other. The .uproject is a reference file used to create, open, or save a file, whereas the Project contains all of the files and folders associated with it.

You can create any number of different Projects which can all be maintained and developed in parallel. Both the Engine (and Editor) can easily switch between them, which will allow you to work on multiple games at once or have several test projects in addition to your main game Project open at any time.

Objects

The basic building blocks of the Unreal Engine are Objects of various kinds. The base that most elements in the engine gets functionality from or more correctly “inherits” from is the Object class, UObject.

From C++, UObject is the base class of all other Unreal Engine Objects and contains essential functionality for other objects you create, such as garbage collection and property support (through UProperty).

Actors

An actor is any object that you can add to your map (aka level). You could say that anything that can be “acted on” is an Actor. Actors support basic properties such as transformation (movement), rotation and scale in 3D space and actors can be made to spawn (get created) and destroyed by your gameplay code.

In C++ AActor is the base class for all actors. Actors can be static meshes, cameras, lights, particle emitters or skeletal meshes to name a few.

Actors will include “Components” that define their behavior and what they can do, like a light actor will include a component that defines what type of light it is, like a spot light will have a spot light component that defines how light is emitted and so on.

The actor will have access to and can use the functionality provided by the component.

Classes

Classes define the properties (state) and behavior of actors and objects in your game.
Classes in C++ defines a hierarchy where a given class can inherit from their parent classes and also pass information down to new child classes (sub class) that inherits from it.

I’ll have much more to say about classes when I start writing some actual code.
One thing worth mentioning is that you can also use Blueprints to define Unreal classes using the editor.

Pawns

A Pawn is a sub-class of Actor that can act as a character in your level and can be controlled by a human player or as an NPC (Non Player Character) controlled by the game’s AI. (Artificial Intelligence).

A player controlled by a human player is said to be “possessed” and an NPC is said to be “unpossessed”.

Characters

A Character is a sub-class of a Pawn that’s intended to be used as a player character.

It includes collision properties (how it reacts to other objects in the environment) and input bindings for bipedal (two legs) movement among other things that can be controlled by the player.

PlayerControllers

The PlayerController Class is used to take player input and translate that into interactions in the game and every game has at least one PlayerController in it. A PlayerController often possesses a Pawn or Character as a representation of the player in a game.

There are also many other terms that I’ll cover as the need shows up, but this is the fundemental terminology you’ll need to understand to follow tutorials on beginning game development in Unreal Engine 4.

What’s up next?

The next post will cover some more details on how to write C++ code in the “Unreal Style”, so that your code will blend easier with the existing code and the code that gets created by the Unreal Editor.

Unreal has their own coding standard that is slightly different from most mainstream coding standards used in C++ programming, so it deserves some attention.

Unreal Engine 4 vs Unity 5

Oh, great. Another post about Unreal vs. Unity…

This is not a Unreal vs Unity, “what is best” type of comparison, just my thoughts on why I decided to go all in with Unreal 4 instead of settling for the arguably more popular (and more forgiving) Unity 5 game engine for learning game programming.

First of all, let me point out that there is no clear cut way of saying that one option is “better” than the other, it all depends on your needs, prior knowledge and how much time you have available for learning new stuff.

My focus with this series of blog posts is to cover beginning game development using C++, so with that criteria, the choice between the two is self explanatory.

But if Unity is so popular and easy to use, why choose something more complex if all you want is to create some fun games? And if C# is so much more forgiving and easier for most people to understand, why use C++ in all its complex glory?

Good questions. The answer is not obvious, but for me, being able to access the game engine’s source code is of great value. I’m not a fan of obscurity, and if something is not doing what I want it to do, I want to be able to know why it is behaving as it does and to have the ability to change it. This is something you can’t do with a closed source engine like Unity, at least not at the time of this writing.

C# is a very productive language to code in, and it is easier to learn than C++, no doubt about that.
But C++ is the “lingua de franca” or common standard programming language in many areas of computing including games, also it is a platform neutral programming language.
C++ is also the common base that other languages such as Java and C# is influenced from, making it a very useful language to know in the world of programming.

Arguably, C++ also tends to produce faster executable code than other higher level languages, and when your programs get more complex, it is vital to be able to churn out all the juice your processors has to offer. Learning to code in C++ gives you the ability to do just that. It has a good mixture of low-level and high-level features making it ideal for something like games.

So, all though Unity and C# may be a better match for newcomers to programming, I do believe that investing the time to learn C++ and Unreal gives a better value in the long term.

Also, there are some annoyances associated with using Unity if you don’t pay for the Pro version, like not having the ability to customize the start screen you get when starting your games. It will say “Made with Unity Personal Edition” or something like that, and that doesn’t exactly portray a professional game as I see it. It is also a constant reminder of your limited freedom and the fact that there is better options available.

madewithunity

This may or may not be an issue if what you are developing is an indie game anyway, but I like to be in control and Unreal gives the developer more control.
There is no “Pro” version of Unreal, just the source code that is shared by all users of the engine, and that’s kind of how I like it to be.

The reason I chose to create these posts is because I’m now in a position in life where I have some time on my hands, and I’ve chosen to use this time to learn more about game development.

I already know a fair amount of programming, both using C# and C++ from previous jobs I’ve had and from my personal interests through the years which has involved programming in some form or the other.

For people new to both programming and game development, there is a great deal to learn and I will not be able to cover anywhere near what you need to learn as this would amount to around a three year full-time study at some university for being able to produce professional quality games.

But, I do hope that the posts about C++ and Unreal will give some insight into how this works and possibly making it easier for someone to grasp the overall picture of how creating games in this environment can work out.

unreal-logo-small

More to come…

Using LINQ and EntityFramework, Measuring and analyzing the LINQ effectiveness

Some Entity Framework Components

When you use something like the extensive “Entity Framework” or any other ORM (Object Relational Mapping) tool, the process of writing sophisticated database driven applications becomes a lot easier.
It eliminates most of the data access code used to be required to create object-based data applications, and contains tools that makes it possible to create object-oriented data applications from an existing database, by using .edmx files that describes the CLR (Common Language Run-time) objects and these can be “extracted” from any existing relational database that the Entity Framework has a driver for.

This enables developers to program directly against data types defined at the conceptual level, presented as regular object instances.

But.
The responsibility to create clean queries and dispose the resources that’s being used, are just as important as it was in more “old fashioned” database connected applications that does all the SQL manually.

Many people assume that, as long as the application compiles and runs without complaining loudly, everything is probably done correctly, right?
This is .Net we’re talking about, so if something was wrong I would probably get a warning or something, right?

Sadly life is not that simple in this case.

The risk of introducing data-access “anti-patterns” in your code that may decrease the performance of your application and greatly slow down your organization’s database server is highly possible, and  actually quite likely to happen if you’re new to ORM and the Entity Framework in particular.

How Can You Measure the LINQ query’s Efficiency?

To reduce the chances of creating inefficient ORM based applications, we need some way to easily (or it won’t be used) analyze exactly what goes on in the EF brain when it executes our code.

This post focuses on a product called LINQ Insight by DevArt, that addresses this problem (and a lot of other LINQ related problems) and may be downloaded for free from the DevArt site:

Download LINQ Insight Standard

NB! The FULL functionality is free for 30 days, before it changes to the more restricted, but highly useful Free Edition.

Data access anti-patterns decreases performance greatly and must be avoided.

When you write applications using for example LINQ to Entities with the Entity Framework, what goes on behind the scenes is something like:

1) The Entity Framework “Mapped” Objects are created with a ObjectQuery() method, run against an existing ObjectContext instance.

The ObjectContext instance is a container for in-memory objects and metadata extracted from the relational database.
It manages the object identities and states, keeps a pointer to original and current values of the objects properties, and tracks changes made to each object in its cache store.

The ObjectContext provides the connection and metadata information that is required to compose and execute queries, and should therefore be set up in advance.

2) Entity Framework components like the ObjectQuery class are compiled behind the scenes, and generate “regular” SQL statements that does the “actual” query against the database.

The queried data is returned as entity objects, most often as anonymous class instances, and the application works with these objects instead of directly querying the SQL database.
This is an important point to understand.
Using LINQ in this way, to offload the database server is a fundamental aspect of ORM frameworks.

Constructing an ObjectQuery Instance

The ObjectQuery returns a collection of zero or more typed entities.  The ObjectQuery generic class implements the IQueryable generic interface, whose methods enable LINQ queries to be incrementally built. You can also let the compiler infer the type of entities by using the C# var keyword, which is something that simplifies specific query situations considerably.

Instances of the ObjectQuery generic class, which implements the generic IQueryable interface, serve as the data source for LINQ to Entities queries. In a query, you specify exactly the information that you want to retrieve from the data source. A query can also specify how that information should be sorted, grouped, and shaped before it is returned. In LINQ, a query is stored in a variable. This query variable takes no action and returns no data; it only stores the query information. After you create a query you must execute that query to retrieve any data.

// Call ctor with query for "persons" and the named ObjectContext
var personsQuery = new ObjectQuery("Persons", context);

3) You can get access to the entities via the reference and collection properties, and this in turn produces additional SQL queries to get the necessary data.

4) When your application finishes working out what entities are needed, the “mapping” process of EF is more or less done.

BUT: if there are any changes to the queried entities, the application will have to save them to database somehow.
It does this by transforming your LINQ, like SubmitChanges() to “regular” SQL statements, like SELECT, UPDATE etc.

All of this happens without you being able to “see” exactly what’s going on at a “lower” level. That’s kinda the idea, to abstract away the unnecessary details from the user, and let you think about designing the flow of the application, not the technicalities used to make things work.

Though all these actions are performed by Entity Framework run-time, and are transparent for the application that works with entities, they may be complex and affect application performance greatly. That’s why you probably need a good Entity Framework profiler to create efficient applications using Entity Framework for data access. Entity Framework profiler in LINQ Insight really profiles ORM events and allows you to see how much time LINQ statement or SubmitChanges call takes to execute as a whole even if it generated multiple SQL statements.

Example Application

Let’s show how to profile Entity Framework applications on a simple example. We have used the widely used NorthwindEF sample database for our example, and generated the model for it with EDM wizard using default settings. The application is a simple console application with the following code:

18 Reasons to Use C# / .Net

18 Reasons to Use C# / .Net

(some of this also explain why C# > Java)

This list is by no means complete, but may be a helper for those that wonder if they should jump on the old Java Wagon or the shiny C#/.Net Wagon with dual suspension and enhanced body construction…

  • No pointers required! C# programs typically have no need for direct pointer
    manipulation (BUT you are free to drop down to that level if absolutely
    necessary)
  • Automatic memory management through garbage collection. Given this, C# does
    not support a delete keyword.
  • Formal syntactic constructs for classes, interfaces, structures, enumerations, and
    delegates.
  • The C++-like ability to overload operators for a custom type, without the
    complexity (e.g., making sure to “return *this to allow chaining” is not your
    problem).
  • Support for attribute-based programming. This brand of development allows you
    to annotate types and their members to further qualify their behavior. For
    example, if you mark a method with the [Obsolete] attribute, programmers will
    see your custom warning message print out if they attempt to make use of the
    decorated member.
  • The ability to build generic types and generic members. Using generics, you are
    able to build very efficient and type-safe code that defines numerous placeholders
    specified at the time you interact with the generic item.
  • Support for anonymous methods, which allow you to supply an inline function
    anywhere a delegate type is required.
  • The ability to define a single type across multiple code files (or if necessary, as an
    in-memory representation) using the partial keyword.
  •  LINQ: Support for strongly typed queries used to interact with various forms
    of data.
  • Support for anonymous types that allow you to model the shape of a type rather
    than its behavior.
  • The ability to extend the functionality of an existing type (without subclassing)
    using extension methods.
  • Inclusion of a lambda operator (=>), which even further simplifies working with
    .NET delegate types.
  • An object initialization syntax, which allows you to set property values at the
    time of object creation.
  • Support for optional method parameters, as well as named method arguments.
  • Support for dynamic lookup of members at runtime via the dynamic keyword.
    This provides a unified approach to invoking members
    on the fly, regardless of which framework the member implemented (COM,
    IronRuby, IronPython, or via .NET reflection services)
  • Working with generic types is much more intuitive, given that you can easily map
    generic data to and from general System.Object collections via covariance and
    contravariance.
  • More and more companies and projects are going for the Open Source Mono Project’s implementation to create true cross-platform applications, both on desktop and mobile platforms.
  • The current version of C#
    provides a pair of new keywords (async and await), which massively simplify multithreaded and
    asynchronous programming.

You’ll be the judge! Both Languages has their charm, but personally (I learned Java many years ago) I strongly prefer C#, as you probably reasoned from this article…

Parallel Computing With CUDA Extensions (Part 2)

Parallel Computing With CUDA Extensions Part 2

A “kernel” in CUDA terms can be thought of as a series of instructions to be carried out
by a computation unit on a GPU. Basically a regular program.

1. You write kernels / code as if it where regular serial (top to bottom) programs, like
those designed to run on a single thread.

2. You tell the GPU to launch this code, and it will do so on multiple threads. (you can define how many)

But won’t this lead to the same code being executed multiple times?
Yes, it will, but that will be explained later.
If you write code to output “Hello World” and ask the GPU to run this on 50 threads,
you’ll be greeted back 50 times, and it will all happen in parallel.

But let’s say that your program code contains a vector of 50 floats.
For each of those floats you want something done, the same way, as fast as possible..
You tell the GPU to spawn 50 threads to work on this kernel (program code).

Inside the kernel when run on the GPU, each thread can connect to one vector member, and have full control over which thread works with what member of the vector.

Each thread starts doing the work as instructed in the code received from the CPU.

As an example, let’s say that when running the code on a CPU only, the CPU would have to traverse the vector members one by one, do the job needed, and continue on to the next member to do the same.
The total execution time for the task would vary based on how busy the CPU was and other factors, but let’s assume that we have 50 members that each needs 10 ms to be processed.

Easily this would take 50 x 10 ms (500 ms) to complete, as we work in a non-parallel way.
If we compare this to how the execution would be done in a GPU assisted way,
the time needed to process each element might be a bit higher, because of the general fact that the working unit will not be as fast as a regular CPU thread, so let’s say 20 ms per member.

The difference is that because these tasks are all started in parallel, they would finish processing the whole vector of 50 members in just 20 ms compared to the CPU, that would need to use 10 ms x 50 members, giving us 500 ms!

To not loose focus, it might help
to visualize situations in programming that could benefit from being able to do several equal tasks at the same time.

One thing that comes to my mind is in image editing applications. When you have an image consisting of millions of pixels, there will be several thousand pixels that share the same characteristics / properties, like color and brightness.
If you where to write a function to lighten or change the color of all those equal pixels, you’d basically have a job that could benefit from being executed simultaneously, rather than doing the same thing to each pixel in a linear fashion.

Usually, when programming using only the CPU, launching and running threads in parallel is considered an expensive and cumbersome activity.
The whole point of using the GPU as a processing unit for “regular” tasks is that it’s very good at certain things, like these two:

1. Launch a lot of threads (and “a lot” is MANY, think thousands)
2. To actually run these threads in parallel

So GPU’s makes perfect candidates for doing the kind of processing that’s lacking in regular CPU’s.

For those learning about programming, maybe as a student or on their own, I seriously believe that there will be heavy demand for competent C/C++ programmers that knows how to program using GPU assistance soon, and also into the unforeseeable future.

C and C++ might be lower-level than the languages you find most comfortable to use, but the truth is that even though these statically typed compiled languages has experienced a drop in general interest the last ten years, they’re now on the rise again thanks to technologies like this and because of the importance of power consumption / watts per cycle on modern handheld devices.

C++ is the most efficient language to use for low power consumption devices (if done right) compared to any other high-level language in existence today, and many large companies invests huge sums of money to the driving forces behind these languages now.

The future is mobile and the future is (hopefully) green.
To achieve this, we also need to start making software that’s green and environmentally friendly.

I hope this article has made you more interested in learning about GPU assisted processing using tools such as CUDA or OpenCL.

There’s more in the world than an Apple.

Parallel Computing With CUDA Extensions (Part 1)

cuda_spotlight

Parallel Computing With CUDA Extensions (Part 1)

First, let’s see how to rate a CPU in a parallel way of thinking.

Let’s say we have an eight Core Intel CPU.
That’s:

With eight cores, you can execute 8 operations (Wide AVX vector operations) per core,
and each core has support for running two threads in parallel via Intel “HyperThreading” technology, so you get:

8 cores * 8 operations/core * 2 threads and end up with what’s called
“128-Way Parallelism”

For more about AdvancedVectoreXtentions (AVX) in CPU’s, check this page.

Programming without taking advantage of ANY multithreading / parallel processing
techniques, means that for each program you run, you use

2/128 = 1/64 of your CPU’s total resources (including the automatic “HyperThreading”).

In an ordinary C/C++ program you can only run code that uses the CPU as
the computing resource.
If people really took advantage of their cores and threading capabilities, this would
probably be enough for most regular applications, but for applications that does a lot of
heavy calculations, like video / image processing or 3D graphics it’s way better if you could
offload some of these tasks to the simpler (in terms of instructions), but well capable GPU(‘s) in your machine.

One way to do this is through the use of CUDA extensions.

In this model, the CPU is considered the “HOST” and each GPU is a “DEVICE”
in your system that can be used for doing calculations.
When such a program is compiled, instructions for both the HOST and any DEVICE
is created.
In CUDA the GPU/DEVICE is seen as a “CO-PROCESSOR” to the CPU/HOST.
The processor also assumes that the HOST and DEVICE has access to separate physical
memory where they can store data.
The DEVICE memory is typically a very high-speed block of memory, faster than the one
on the HOST.

The HOST is “In charge” in CUDA and sends messges to the DEVICE telling it what to do.
The HOST keeps track of:

Moving data:
1. From CPU memory -> GPU memory
2. Grom GPU memory -> CPU memory
CUDA’s version of C’s memcpy() is cudaMemcpy()
3. Allocating GPU memory
Again CUDA uses cudaMalloc() instead of malloc()
4. Launch “kernel” on GPU (in CUDA, the HOST launches “kernels” on the DEVICE)

A Typical flow in a CUDA Application would be something like:

1. CPU runs cudaMalloc on GPU
2. CPU copies input data from CPU->GPU with cudaMemcpy
3. CPU launches the transfered “kernels” on GPU (kernel launch)
4. CPU copies results back with cudaMemcpy

So, what is this “Kernel” stuff all about?

Guess we’ll find out in part 2 of this series…

I’m creating a new Blog these days!

While I’ll still maintain the thoughts of a Madman, I realize that its content starts to get a little bit fragmented. Therefore I’m creating a new Blog dedicated especially to C# and .Net, while also covering general application development topics and tools. The new creation lives over at Blogger at the following address:

“The C# Aficionado”

http://tcatitles.blogspot.com

If programming and C# is your thing, please consider following the Blog. 😉

Have a great weekend!