7. The Heap

Loading...

From the course by University of Colorado Boulder

Introduction to Embedded Systems Software and Development Environments

170 ratings

University of Colorado Boulder

Introduction to Embedded Systems Software and Development Environments

170 ratings

Welcome to the Introduction to Embedded Systems Software and Development Environments. This course is focused on giving you real world coding experience and hands on project work with ARM based Microcontrollers. You will learn how to implement software configuration management and develop embedded software applications. Course assignments include creating a build system using the GNU Toolchain GCC, using Git version control, and developing software in Linux on a Virtual Machine. The course concludes with a project where you will create your own build system and firmware that can manipulate memory. The second course in this 2 course series , Embedded Software and Hardware Architecture, will use hardware tools to program and debug microcontrollers with bare-metal firmware. Using a Texas Instruments MSP432 Development Kit, you will configure a variety of peripherals, write numerous programs, and see your work execute on your own embedded platform!

From the lesson

Memory Types, Segments and Management

Module 3 will begin to introduce important embedded concepts like the memory systems in their design. Learners will understand how the software to hardware mapping occurs for their designs including differentiating between your program code and your program data. Memory systems have many platform and architecture dependencies, and you will begin to learn about some of the fundamental concepts a software engineer needs to know to utilize all parts of an embedded system’s memory.

1. Introduction to Memory Organization6:58

2. Memory Architectures8:20

3. Memory Segments6:50

4. Data Memory8:26

5. Special Keywords (Const, Extern & Static)8:05

6. The Stack9:09

7. The Heap7:22

8. Code Memory8:34

Meet the Instructors

Alex Fosdick
Instructor
Electrical, Computer, and Energy Engineering

In this video, we'll go into more detail about the heap subsegment in data memory.

The heap is the last main segment of discussion for

the subsegments of the data segment.

We often call this memory dynamic memory.

Much like the stack, the heap is allocated and deallocated with software at runtime.

The difference is that this is done manually by the programmer

with the use of functions that manage the region.

The lifetime of heap data will exist for

as long as the software designer has not deallocated the data.

Unlike our data types which specify a size,

allocation in the heap can vary with each call.

The heap is a space of dynamic memory that is reserved in the data segment.

The heap is useful because it allows us to dynamically change the size

of allocated data.

Unlike other subsegments, static allocation reserves a set number of bytes.

With the heap, you do not have this limitation.

With each call to reserve memory, you can specify any number of bytes to reserve.

Additionally, you can resize a previous allocation.

Much like the stack, allocations and de-allocations in the heap

mean the compiler has to add extra instructions into your program for

data management in this region.

However, this is not done automatically, you have to use heap functions to directly

reserve and free data in the heap by calling special functions in the software.

Data that is reserved will remain reserved for as long as the programmer chooses.

Allocated data can live for longer than a function's scope but less than a program.

The process of reserving dynamic memory and freeing it is extra overhead for

a program to execute.

Meaning there is dynamic memory management in the form of software routines

that check for available space in the heap region to place your requested allocation.

This dynamic memory manager maintains a data structure to track the current state

of the heap space.

The way to use the heap on your embedded system is with the use of four functions.

These are malloc, calloc, realloc, and free.

Malloc and calloc reserve a contiguous block of memory by specifying the number

of bytes you want to reserve.

This can be placed anywhere in the heap depending on the current active

allocations as long as it does not overflow the heap region.

Calloc does the same thing as malloc.

But it will additionally initialize this memory space to 0s.

It's very important that once you have received a pointer from malloc or calloc,

that you do not change that address.

You must track that exact address in order to free your allocations later.

Realloc is the function that allows you to change the allocation size

of your allocated data.

This function requires the new size of memory that you wish to have

regardless if it is smaller or larger than the original.

It also requires the original heap pointer you allocated with malloc.

Realloc will return a pointer to the new memory region.

This new space could remain in the previous location or

be moved to a completely new spot.

Realloc will free the previous location automatically for you.

These heap functions require a raw byte count of memory and

will return a pointer to the beginning of the allocated region.

Therefore, if you're trying to allocate a data structure on the heap,

you need to do some math as to how many bytes your data structure occupies.

The C programming operator sizeof can help with this.

By giving it a type or a variable,

it can return the number of bytes this will occupy in memory.

These functions require a new type that we have not talked about yet called size_t.

For now, just think of this as an unsigned log int.

Free is a function that tells the dynamic memory manager that

the requested memory is free to go back into the available pile.

All you need to do is provide free with the pointer you are using to track your

memory space.

You must free your memory when you are done with it, otherwise,

the heap will have less space to work with.

All of these routines require the programmer

to keep track of the allocated memory with a pointer, or an address.

This pointer must have a scope as long as the memory's being used.

If you allow the pointer to be local and removed from the stack

before freeing the memory, that memory is lost for the remainder of the program.

There's no guarantee that when you malloc or realloc data,

it is even in the same space or there's even enough space to allocate it.

If there's not enough space available,

these functions will return a NULL pointer.

Since the heap region often becomes very segmented or fragmented as used and

free pieces of memory are filled throughout the area.

You can attempt to reserve memory, and although there might be

physically enough free memory in the heap, it might not be contiguous.

Requesting this memory would fail because your heap space has become too segmented,

over-allocating and freeing data frequently and inefficiently.

Here's an example where we have reserved 8 bytes of data on the heap.

Malloc returns a void pointer or

a generic pointer to the place in memory where this was allocated.

You do not want to operate with this generic pointer, so it is

important that you cast this piece of data to the type you wish to program with.

This code will perform some arbitrary operations on this piece of memory and

then eventually free the heap memory by calling free,

with the pointer we use to track the memory.

Here's another example that shows how your memory space gets fragmented.

Assume a heap space of 80 bytes with two allocations existing already.

The first one is 16 bytes, followed by a second one of 32 bytes.

Now let's say you free the 16 byte location but

now you want to make a 40 byte allocation.

There's a total of 48 bytes available but because of where the original 32 byte

allocation was put, you do not have enough contiguous memory space to place 40 bytes.

There are many negatives to using dynamic memory and some embedded engineers advise

against its use because the memory and execution performance hits.

Constantly allocating and freeing memory will add excessive execution overhead to

managing the heap region.

And heap can become fragmented with blocks of memory allocated and

freed interspersed with one another.

Also, failure to free a piece of memory is referred to as a memory leak.

If you lose track of the heap memory you have reserved by storing the reference of

data in a local pointer variable, the internal software that manages space

will think it's still needed even though you have now lost track of it.

The heap is a vital memory segment for software engineers and

some embedded systems.

However, as embedded system architectures scale up in complexity and

size, the likelihood of heap use on a system is a near guarantee.

The heap requires direct invocations to utilize the memory,

which leads to more overhead spent by the CPU to manage this region at runtime.

This overhead existed with the stack as well, but

the heap has the nice advantage of giving us a dynamic piece of memory that can have

a lifetime longer than a function but less than a program.

Explore our Catalog

Join for free and get personalized recommendations, updates and offers.

Coursera provides universal access to the world’s best education, partnering with top universities and organizations to offer courses online.

© 2019 Coursera Inc. All rights reserved.

Download on the App Store

Get it on Google Play