REAL-TIME OPERATING SYSTEMS AND STANDARDS - Hard Real-Time Computing Systems: Predictable Scheduling Algorithms and Applications - page 441

The implementation of EDF is done in a new conformance class, called “ EDF ”. The
EDF conformance class retains the same static approach of the OSEK/VDX standard,
with a slightly simplified API (basically, the API does not include the TerminateTask
primitive). The configuration of the EDF scheduler is then specified in the OIL file
used by ERIKA Enterprise. For example, an EDF conformance class with a timer tick
of 25 nanoseconds is specified as follows:
KERNEL_TYPE = EDF { TICK_TIME = "25 ns ";};
A relative deadline of 10 milliseconds can be specified (in the OIL file) as
TASK myTask1 {
REL_DEADLINE = "10 ms ";
};
The OIL compiler converts the milliseconds in the proper number of ticks, using the
tick duration specified in the TICK TIME variable. The rest of the OSEK/VDX kernel
primitives are basically the same, since all system calls do not have periodicity and
priorities as parameters.
Finally, based on the EDF scheduler, ERIKA Enterprise also supports resource reser-
vation (see Section 9.3). The implementation was carried out within the FRESCOR
European Project 13 and consisted in the development of the IRIS scheduler [MLBC04]
on top of the EDF scheduler. As a result, ERIKA Enterprise is able to support resource
reservation under EDF with reclaiming of unused computation time and with a foot-
print of about 10Kb Flash.
MULTICORE SUPPORT
Another feature that distinguishes ERIKA Enterprise from other real-time kernels is
the support for multicore systems; that is, the possibility for an application to be stati-
cally partitioned on the various cores available on modern architectures.
Under ERIKA Enterprise each concurrent task is statically linked at build time to
a given CPU. Each CPU includes a separate copy of the operating system and of
the device drivers that are present on the particular CPU. The partitioned approach is
typically forced by modern multicore system-on-a-chip, since it is often not efficient
to migrate the execution of a thread from one CPU to another.
13 FRESCOR Project: http://www.frescor.org