MATLAB XPC TARGET 4 - IO Manual do Utilizador descarregar pdf (Página 23)

Selecting a Target System

1-17

therefore important to choose a CPU from a class that is powerful enough to

provide sufficient performance headroom.

The general rule is to select the fastest CPU available for the chosen form

factor without violating the given system constraints. Choosing the fastest

CPU leads to the best performing system without adding significant cost.

However, because your system needs to work within the constraints that have

led you to already choose the form factor and I/O type, you can sometimes be

limited to the level, and therefore the power, of the available CPU. For

example, if you require the system to work within an extended temperature

range then you need a low power CPU. For example, this requirement limits

your available choices, and in this case you cannot use an Athlon CPU.

To better understand the complexity of your application and if the selected

target PC system will execute your application, use the following steps as a

guideline. Considerations include whether the selected target system with a

certain CPU class will execute the application at the given base sample time:

1 Calculate the total I/O driver latency.

Each I/O channel that is accessed by your application adds latency, thereby

increasing the achievable minimal base sample time. Because the I/O boards

are accessed through the peripheral bus, latency time depending on the

amount of I/O can quickly become larger than the time needed to execute the

algorithm (model) itself.

The supported I/O board list (see “Selecting a Target System Using the xPC

Target Interactive Hardware Selection Guide” on page 1-10) provides

latency information for the driver of each board. By knowing the number of

channels you use on each board, you can calculate the total I/O driver

latency. This total is a conservative estimate, but if the estimated latency is

already larger than the given base sample time, the real-time execution will

fail independent of CPU performance. In this case, contact your MathWorks

sales representative to discuss alternatives.

2 Subtract the total I/O driver latency estimate from the base sample time.

3 To provide headroom for background task switching and cache misses,

subtract an additional 25 μs. The resulting time is roughly the time

available to execute the Simulink-based algorithm (model).

1 2 ... 18 19 20 21 22 23 24 25 26 27 28 ... 51 52

Sem comentários

MATLAB XPC TARGET 4 - IO Manual do Utilizador Página 23