deep

a Cross Development Platform for Java

User Tools

Site Tools


runtime_library:os:start

Operating System

You can program the processor either using interrupts in a foreground/background system or means of an operating system. Some classes will be linked and loaded in either case. This is true for the class Kernel for basic initializations, the class Heap for dynamic memory allocation and the classes for exception handling.
When choosing an operating system you can select a simple, robust and very efficient non-preemtive tasking system (class Task). The very popular MicroC/OSII will be also available on request. For any operating system the minimum task period is limited to the order of milliseconds. For fast control applications you can use the decrementer as a fast running timer and produce exceptions at a very high frequency. The basic modules are all written in Java. All the necessary hardware resources can be accessed through special built-in methods.

Class Constructors and the main() Method

When starting up a general java system, the static constructors of every class must be loaded and finally the main method has to be called. Our system does not support a main method. This has the advantage that you can have as many root level classes (or application classes) as you like. In deep projects you can see how to build several root level classes into the same project. An example of using a class constructor is shown in the following section.

Non-Preemptive Tasking System

The class Task offers a very efficient non-preemtive scheduler. The basic timing resolution is 1ms. You can create and install as many tasks as you like. Care must be taken that each of those tasks terminates after a reasonably short time span. Control is then taken over by the scheduler which chooses the next available task in a round-robin fashion.
There are two types of tasks: ready tasks have a period of 0. That is, they are scheduled as quickly as possible. Their repetition frequency depends highly on the processor load and the task code itself. Periodic tasks have a non-zero period. They get scheduled after the system time passes the assigned period of the task.
The following code shows how to create and install a task.

public class TestTask extends Task{
 
	public void action () {
		// put task code here
	}
 
	static {	
		Task task1 = new TestTask(); 
		task1.period = 500;	// call twice per second
		Task.install(task1);
	}
}

Interrupts

A processor might have some sort of a hardware priorization scheme when dealing with interrupts. Some of the potential internal and external interrupt sources use vectoring or get vectored to the same memory location and are dealt with by a software scheduler. You have to assign a priority level to each of the interrupts. A low level means a high priority. In order to achieve highest efficiency, interrupt nesting is switched off. That makes it necessary to write small interrupt service routines.

Internal Interrupts

Internal interrupts are all those who are caused by built-in peripheral modules, like time processor units, ADC and communication interfaces. You can use them by extending the class Interrupt, write your own action method and install this object into the interrupt scheduler by calling install. As parameter you have to indicate the interrupt object together with a level which has to be between 0 and 7. You can install any number of interrupt objects with the same level.

public class IntTest extends Interrupt {
 
	public void action() {
		// put your code here
	}
 
	static {
		IntTest int = new IntTest(); 
		int.enableRegAdr = registername; int.enBit = bitnumber;
		int.flagRegAdr = registername; int.flag = bitnumber;
		Interrupt.install(int, 5, true);
	}
}

External Interrupts

The mpc555 offers 8 external interrupt inputs. They are named IRQ0 to IRQ7. In order to use them you have to extend the class Interrupt, overwrite its action method and call the method install(). As parameters you have to give the newly allocated interrupt and the number of the interrupt pin which corresponds to the interrupt level. Care must be taken not to assign more than one interrupt objects to the same interrupt pin.

public class ExtIntTest extends Interrupt {
 
	public void action() {
		// put your code here
	}
 
	static {
		ExtIntTest int5 = new ExtIntTest(); 
		Interrupt.install(int5, 5, false);
	}
}

The voltage level on the external interrupt pins 5, 6 and 7 also select the operating when starting the device up. Do not override these levels during start-up!

Switch off Interrupts Globally

All the external and internal interrupts can be switched off and on again, e.g. for accessing shared resources. Keep the switch-off time as low as possible. This should always be done in the following manner.

void MyExampleMethod () {
	...
	US.PUTSPR(EID, 0);	// switch off interrupts globally
	...			// critical section
	US.PUTSPR(EIE, 0);	// interrupts reenabled 	
	...
}

Decrementer Exceptions

The decrementer is a timer with a resolution of 1μs. Therefore you can use it for very fast control applications. Simply extend the class decrementer and overwrite the action method with your specific code. Make sure to set the period in μs to some meaningful value. You then have to call the install method.

public class DecrementerTest extends Decrementer {
	static int count;
 
	public void action() {
		// put your code here
	}
 
	static {
		DecrementerTest d = new DecrementerTest();
		d.decPeriodUs = 1000000;	// period is 1 s
		Decrementer.install(d);
	}
}

Please make sure that your exception code is reasonably short and that the processor is able to finish prior to a new decrementer exception.

Floating Point Numbers in Exceptions

In order to use the built in floating point unit of the processor the compiler has to include the floating point registers when doing a context save. To minimize the associated overhead when not using floating point numbers in exceptions the user has to call the function ENABLE_FLOATS() from the package US if floats are used. The call of this method also enables the machine of using the floating point unit while an exception is processed. If this call is forgotten the machine will throw a floating point unavailable exception.

System Sanity Checks

  1. The stack size is set by the configuration. In order to check, whether this size is enough at all times for a running system, you can call a target command ch.ntb.inf.deep.runtime.mpc555.Kernel.checkStack. If the stack size was ever bigger than the maximum stack size this check fails and the LED on the 555 board starts blinking.
  2. The heap manager includes an automatic garbage collection. If the heap gets too fragmented or to large chunks of heap memory are requested and no free block can be found, the LED on the 555 board starts blinking and programm excecution stops.
runtime_library/os/start.txt · Last modified: 2016/02/25 13:33 (external edit)