RTA-OSEK

Renesas SH2A with the Renesas Compiler

Features at a Glance

OSEK/VDX OS v2.2 Certified OS

RTOS overhead: 30 bytes RAM, 144
bytes ROM

Category 2 interrupt latency: 41
CPU cycles

Applications include: Engine Man-
agement, Transmission Control, In-
dustrial Equipment

RTA-OSEK The kernel is configured using an offline tool

provided with RTA-OSEK. Determining in ad-
RTA-OSEK provides an application design envi- vance which features are used allows memory
ronment that combines the smallest and fastest requirements to be minimized and API calls to
OSEK RTOS with an unique timing analysis tool. be optimized for greatest efficiency.
This datasheet discusses the Renesas SH2A port All tasks and ISRs in RTA-OSEK run on a single
of the RTA-OSEK kernel alone and should be stack – even extended tasks. This allows dramat-
read in conjunction with the Technical Product ic reductions in application stack space require-
Overview “Developing Embedded Real-Time ments.
Applications with RTA-OSEK” available from
ETAS. The RTA-OSEK kernel is designed to be scalable.
When a task uses queued activation or waits on
The kernel element of RTA-OSEK is a fixed prior- events, the additional RTOS overhead required
ity, pre-emptive real-time operating system that to support these features is paid by the task
is compliant to the OSEK/VDX OS standard ver- rather than by the system. This means that a ba-
sion 2.3 for all four conformance classes (BCC1, sic single activation task uses the same resources
BCC2, ECC1 and ECC2) and intra processor com- in a BCC1 system as it does in an ECC2 system.
munication using OSEK COM Conformance
Classes A and B (CCCA and CCCB). Compiler/Assembler/Linker

All CPU overheads of the kernel have low worst The libraries containing the code for the RTA-
case bounds and little variability in execution OSEK kernel have been built using the following
time. The kernel is particularly suited to systems tools:
with very tight constraints on hardware costs
and where run-time performance must be guar-
Renesas HEW4 SH Series Compiler v9.02
anteed.

Renesas HEW4 SH Series Assembler v7.01
Renesas HEW4 Optimizing Linker v9.04 Functionality

Memory Model The table below outlines the restrictions on the maxi-
mum number of operating system objects allowed by
RTA-OSEK supports the flat 32-bit address space memory RTA-OSEK.
model provided by the Renesas compiler.
BCC1 BCC2 ECC1 ECC2
ORTI Debugger Support
Max no of tasks 32 plus an idle task
ORTI is the OSEK Run-Time Interface that is supported by Max tasks per priority 1 32 1 32
RTA-OSEK for the following debuggers:
Max queued activations 1 255 1 255

Lauterbach TRACE32 Max events per task n/a n/a 32 32
Further information about ORTI for RTA-OSEK can be Max nested resources 255
found in the RTA-OSEK ORTI Guide. Max alarms Not limited by RTA-OSEK

Hardware Environment Max standard resources 255

Max internal resources Not limited by RTA-OSEK
RTA-OSEK supports all variants of the Renesas SH-2A/
SH2A-FPU family including the SH7206, SH72513 and the Max application modes 65535
SH72546.
Note that OSEK specifies that queued activations in an
Interrupt Model ECC2 system are only possible for basic tasks. Where
tasks share a priority level, the maximum number of
RTA-OSEK supports 15 levels of interrupts. Suitable ini-
queued activations per priority level is 255.
tialization values for the Interrupt Priority Registers are
provided. RTA OSEK can also generate a vector table if The number of alarms, tasksets, schedules and schedule
required. arrivalpoints is only limited by available hardware re-
sources.
Floating Point Support
Memory Usage
RTA OSEK is designed to work on an SH2A CPU with fully
re-entrant software floating point libraries supplied by The memory overhead of RTA-OSEK is:
Renesas. This allows floating point to be used in tasks
and ISRs without the need to save and restore any addi- Memory Type Overhead (bytes)
tional context.
RAM 30
SH2A-FPU CPUs contain a hardware floating point arith- ROM/Flash 144
metic unit that is not part of a standard SH2A CPU. RTA-
OSEK supports the SH2A FPU hardware floating point In addition to the RTOS overhead, each object used by
unit. In order to ensure correct functionality of floating an application has the following memory requirements:
point code in tasks and Category 2 ISRs, "wrappers" are
supplied to save and restore the additional context. To Object RAM Bytes ROM Bytes
enable this functionality, configure the relevant tasks
and Category 2 ISRs as floating point using the RTA BCC1 task 0 36
OSEK GUI. BCC2 task 10 56
ECC1 task 60 60
Evaluation Board Support
ECC2 task 62 68
RTA-OSEK can be used with any Renesas SH2A/SH2A-FPU
Category 1 ISR 0 0
evaluation board. An example application is provided to
run on the Rensas ASKSH72546 evaluation board. This Category 2 ISR 0 48
application can be adapted for other target boards by Resource 0 20
adjusting the linker command file (to alter the RAM lo-
Internal Resource 0 0
cations) and one source file (if alternative output pins
are required). Event 0 4
Alarm 12 30
Counter 4 62
Object RAM Bytes ROM Bytes tasks use lightweight termination and no pre or post
task hooks were specified.
ScheduleTable 16 64
ScheduleTable Expiry 0 12 The execution time for every kernel API call is available
on request from ETAS.
Taskset (RW) 4 4
Taskset (RO) 0 4
Schedule 16 36
Arrivalpoint (RW) 12 12
Arrivalpoint (RO) 0 12 Interrupt A sserted
K L

In addition to these static memory requirements each

task priority and Category 2 interrupt has a stack over- RTA-OSEK activity
head (in addition to application stack usage). The single
stack model means that this overhead applies to each Category 1 ISR
priority level rather than to each task. Similarly, for Cat-
egory 2 interrupts this overhead applies for each unique Task
interrupt priority. The table below shows stack usage for
these objects.
Figure 1 - Category 1 interrupt with return to interrupted task
Object Stack Bytes
Task priority level 76
A B
Category 2 interrupt 44

RTA-OSEK activity
RTA-OSEK provides an optimization for task termination
if the user can guarantee that tasks only terminate from
their entry function. Tasks that terminate from else- Category 2 ISR
where are not eligible for this optimization and duly re-
Inte rrupt A sse rte d
quire 56 more stack bytes per priority level than
Task
indicated in the table above.

Performance Figure 2 - Category 2 interrupt with return to interrupted task

The following table gives the key kernel timings for op-
erating system behavior in CPU cycles.
A E
Task Type Basic Extended Ref ActivateTask(T2)
RTA-OSEK activity
Category 1 ISR Latency 29 29 K
TerminateTask()
Category 2 ISR Entry Latency 41 43 A Category 2 ISR
Category 2 ISR Exit Latency 241 359 E
Task T2 ready to run
Normal Termination 97 203 D Task T2
ChainTask 227 499 J Interrupt Asserted

Pre-emption 205 325 C Task T1

Triggered by alarm 313 435 F
Figure 3 - Category 2 interrupt activates a higher priority task
Schedule 187 305 Q
ReleaseResource 205 321 M
SetEvent n/a 531 S

All performance figures are for the non-optimized inter-

face to RTA-OSEK. Using the optimized interface will re-
sult in shorter execution times for some operations. All
 C D S

RTA-OSEK activity TerminateTask()

RTA-OSEK activity

Wa itEve nt(E 1)
Task T2 Task T2
SetE ve nt(T2,E1)
ActivateTask(T2)
Task T1 Task T1

Figure 4 - Task activates a higher priority task Figure 8 - Activation by SetEvent(

F M

RTA-OSEK activity RTA-OSEK activity

TerminateT ask()
ReleaseResource(R1)
Task T2 Task T2

A larm activate s T2
Task T1 Task T1

Figure 5 - Alarm activates task Figure 9 - ReleaseResource()

Benchmarks
J The following sections shows benchmarks for RTA-OSEK
memory usage for BCC1, BCC2, ECC1 and ECC2 conform-
ant applications. The applications have the following
RTA-OSEK activity ChainTask(T1) framework:

Task T2
8 tasks plus the idle task

All basic tasks are lightweight tasks

Task T1

1 Category 2 ISR with a 10ms minimum inter-arrival
time
Figure 6 - Task chaining

1 Counter

Q
7 or 8 alarms, all attached to the same counter

No resources or internal resources

RTA-OSEK activity TerminateTask()

No hooks
ActivateTask(T2)
Task T2
No schedules
Schedule()

Task T1
No tasksets

Built using standard status

Figure 7 - Schedule() call
The following table shows the task priority configura-
tion for each benchmark application: ECC1

The ECC1 application uses 7 basic tasks and 1 extended

Stack (bytes)

Period (ms)

task with unique priorities. Task H is the extended task

Task/ISR

and it waits on a single event that is set by basic tasks A-

G.
BCC1

BCC2

ECC1

ECC2
This application has the following overheads:
ISR1 10 10 IPL1 IPL1 IPL1 IPL1
A 10 10 8 8 8 8 Memory Usage Bytes
B 20 20 7 7 7 7 OS ROM 2800
C 30 20 6 6 6 6 OS RAM 954
D 40 30 5 5 5 5 comprising RAM data 206
E 50 50 4 4 4 4 comprising RAM stack 748
F 60 80 3 3 3 3
G 70 100 2 2 2 2 ECC2
H 80 150 1 1 1 2 The ECC2 application uses 6 basic tasks and 2 extended
Idle 10 - idle idle idle idle tasks. Tasks G and H are the extended tasks and share a
priority. The extended tasks wait on a single event that
is set by tasks A-F.
The overhead figures give the ROM and RAM required
for RTA-OSEK in addition to that required by the appli- This application has the following overheads:
cation. The RAM figure is shown split into RAM data and
RAM stack. Memory Usage Bytes

BCC1 OS ROM 3436

OS RAM 1108
The BCC1 application uses 8 basic tasks with unique pri-
comprising RAM data 276
orities. This application has the following overheads:
comprising RAM stack 832
Memory Usage Bytes
OS ROM 1946 Stack Optimization
OS RAM 786
Using stack optimization with the benchmark example
comprising RAM data 146 identifies that the following tasks can share internal re-
comprising RAM stack 640 sources:

Tasks A, B and C
BCC2

Tasks D, E and F
The BCC2 application uses 8 basic tasks with unique pri-
orities.
Tasks G and H

Tasks A-G are attached to 7 alarms. Task H is activated The benefit of this optimization is shown in the follow-
multiple times from Task A and has maximum queued ing table:
activation count of 255.
Total Stack Space (bytes) BCC1 BCC2 ECC1 ECC2
This application has the following overheads:
Non-optimized 1020 996 1128 1212
Memory Usage Bytes OS Overhead 640 616 748 832
OS ROM 2234 Application Overhead 380 380 380 380
OS RAM 758 Optimized 440 440 548 548
comprising RAM data 142 OS Overhead 260 260 368 268
comprising RAM stack 616 Application Overhead 180 180 180 180
Notes

Contact Addresses

ETAS GmbH
70469 Stuttgart, Germany
Phone +49 711 89661-0
Fax +49 711 89661-106
[email protected]

ETAS S.A.S.
94588 Rungis Cedex, France
Phone +33 1 567000-50
Fax +33 1 567000-51
[email protected]

ETAS Ltd.
Derby DE21 4SU
Great Britain
Phone +44 1332 253770
Fax +44 1332 253779
[email protected]

ETAS Inc.
Ann Arbor, MI 48103, USA
Phone +1 888 ETAS INC
Fax +1 734 997-9449
[email protected]

ETAS K.K.
Yokohama 220-6217, Japan
Phone +81 45 222-0900
Fax +81 45 222-0956
[email protected]

ETAS Korea Co., Ltd.

Seoul 137-889, Korea
Phone +82 2 5747-016
Fax +82 2 5747-120
[email protected]

ETAS (Shanghai) Co., Ltd.

Shanghai 200120, P.R. China
Phone +86 21 5037 2220
Fax +86 21 5037 2221
[email protected]

ETAS Automotive India Pvt. Ltd.

Bangalore 560 068, India
Phone +91 80 4191 2585
Fax +91 80 4191 2586
[email protected]

www.etas.com

Subject to change (06/2009)

ETAS/COM_Fi/07.2008