Page 150 - Embedded Microprocessor Systems Real World Design

The Development Language

It has been said that you cannot consider yourself a true embedded programmer
unless you can code in assembly language. While this may be an exaggeration, it is
true that many embedded systems have some code written in assembler. Part of this
is because high-level languages (HLL) often assume that things like the stack are
initialized. Since function calls in any language need the stack, which usually is in
RAM, you cannot do a function call to the function that initializes the stack or the
RAM chip select logic. The other reason is speed. The best code optimizers still are
not quite as good as a human at generating fast assembly code.
That said, high-level languages for embedded applications are becoming more
sophisticated, more efficient, and better able to handle the unique hardware that
embedded systems must control. The choice of development language, like every
other engineering tool, is driven by tradeoffs such as cost, ease of use, and utility.
I know that the following statement will make some software engineers throw
this book across the room in disgust, but a simple project may be well served with
just assembly language. Many small projects are not completed any faster by using
an HLL. Some manufacturers (Microchip and Atmel, for example) provide free
assemblers on their Web sites. But cost is not necessarily a factor in choosing assem-
bler over an HLL. Sometimes the issue is speed. Sometimes assembly language is
the only way to get everything to run fast enough. This is particularly true when
using a DSP or other hardware-intensive processor. And, if most of the code is
dedicated to flipping bits on 1/0 ports or loading timers, much of the design effort
goes into calculating things like the timer values. The actual code may be fairly
simple. Of course, the drawback to using assembler is that whoever must maintain
the code in the future must learn the architecture and instruction set of the
processor.
I once used assembly for a small microcontroller project because I could
download the assembler from the manufacturer’s Web site and get the project
going faster than I could get the paperwork through the purchasing department
to order a C compiler.
As the complexity of a project grows and labor becomes a larger part of the soft-
ware design, assembly language looks less attractive. The time spent debugging the
code becomes more of a factor, and the ability to write code for more than one
processor without learning a new language is important. However, some of this
advantage often is lost on simple microcontrollers. The high-level languages for
those devices often have limitations, such as lack of floating-point capability, that
limits code portability. In addition, special instructions are frequently needed for
controlling device-specific hardware (timers, 1/0 ports, and so on) that will not
port to another processor.

Software Design 131