summaryrefslogtreecommitdiffstats
path: root/doc/user/preface.texi
blob: c1565877c5e08e32b76209df5487c97d18285142 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
@c
@c  COPYRIGHT (c) 1988-1999.
@c  On-Line Applications Research Corporation (OAR).
@c  All rights reserved.
@c
@c  $Id$
@c

@ifinfo
@node Preface, Overview, Top, Top
@end ifinfo
@unnumbered Preface

In recent years, the cost required to develop a
software product has increased significantly while the target
hardware costs have decreased.  Now a larger portion of money is
expended in developing, using, and maintaining software.  The
trend in computing costs is the complete dominance of software
over hardware costs.  Because of this, it is necessary that
formal disciplines be established to increase the probability
that software is characterized by a high degree of correctness,
maintainability, and portability.  In addition, these
disciplines must promote practices that aid in the consistent
and orderly development of a software system within schedule and
budgetary constraints.  To be effective, these disciplines must
adopt standards which channel individual software efforts toward
a common goal.

The push for standards in the software development
field has been met with various degrees of success.  The
Microprocessor Operating Systems Interfaces (MOSI) effort has
experienced only limited success.  As popular as the UNIX
operating system has grown, the attempt to develop a standard
interface definition to allow portable application development
has only recently begun to produce the results needed in this
area.  Unfortunately, very little effort has been expended to
provide standards addressing the needs of the real-time
community.  Several organizations have addressed this need
during recent years.

The Real Time Executive Interface Definition (RTEID)
was developed by Motorola with technical input from Software
Components Group.  RTEID was adopted by the VMEbus International
Trade Association (VITA) as a baseline draft for their proposed
standard multiprocessor, real-time executive interface, Open
Real-Time Kernel Interface Definition (ORKID).  These two groups
are currently working together with the IEEE P1003.4 committee
to insure that the functionality of their proposed standards is
adopted as the real-time extensions to POSIX.

This emerging standard defines an interface for the
development of real-time software to ease the writing of
real-time application programs that are directly portable across
multiple real-time executive implementations.  This interface
includes both the source code interfaces and run-time behavior
as seen by a real-time application.  It does not include the
details of how a kernel implements these functions.  The
standard's goal is to serve as a complete definition of external
interfaces so that application code that conforms to these
interfaces will execute properly in all real-time executive
environments.  With the use of a standards compliant executive,
routines that acquire memory blocks, create and manage message
queues, establish and use semaphores, and send and receive
signals need not be redeveloped for a different real-time
environment as long as the new environment is compliant with the
standard.  Software developers need only concentrate on the
hardware dependencies of the real-time system.  Furthermore,
most hardware dependencies for real-time applications can be
localized to the device drivers.

A compliant executive provides simple and flexible
real-time multiprocessing.  It easily lends itself to both
tightly-coupled and loosely-coupled configurations (depending on
the system hardware configuration).  Objects such as tasks,
queues, events, signals, semaphores, and memory blocks can be
designated as global objects and accessed by any task regardless
of which processor the object and the accessing task reside.

The acceptance of a standard for real-time executives
will produce the same advantages enjoyed from the push for UNIX
standardization by AT&T's System V Interface Definition and
IEEE's POSIX efforts.  A compliant multiprocessing executive
will allow close coupling between UNIX systems and real-time
executives to provide the many benefits of the UNIX development
environment to be applied to real-time software development.
Together they provide the necessary laboratory environment to
implement real-time, distributed, embedded systems using a wide
variety of computer architectures.

A study was completed in 1988, within the Research,
Development, and Engineering Center, U.S. Army Missile Command,
which compared the various aspects of the Ada programming
language as they related to the application of Ada code in
distributed and/or multiple processing systems.  Several
critical conclusions were derived from the study.  These
conclusions have a major impact on the way the Army develops
application software for embedded applications. These impacts
apply to both in-house software development and contractor
developed software.

A conclusion of the analysis, which has been
previously recognized by other agencies attempting to utilize
Ada in a distributed or multiprocessing environment, is that the
Ada programming language does not adequately support
multiprocessing.  Ada does provide a mechanism for
multi-tasking, however, this capability exists only for a single
processor system.  The language also does not have inherent
capabilities to access global named variables, flags or program
code.  These critical features are essential in order for data
to be shared between processors.  However, these drawbacks do
have workarounds which are sometimes awkward and defeat the
intent of software maintainability and portability goals.

Another conclusion drawn from the analysis, was that
the run time executives being delivered with the Ada compilers
were too slow and inefficient to be used in modern missile
systems.  A run time executive is the core part of the run time
system code, or operating system code, that controls task
scheduling, input/output management and memory management.
Traditionally, whenever efficient executive (also known as
kernel) code was required by the application, the user developed
in-house software.  This software was usually written in
assembly language for optimization.

Because of this shortcoming in the Ada programming
language, software developers in research and development and
contractors for project managed systems, are mandated by
technology to purchase and utilize off-the-shelf third party
kernel code.  The contractor, and eventually the Government,
must pay a licensing fee for every copy of the kernel code used
in an embedded system.

The main drawback to this development environment is
that the Government does not own, nor has the right to modify
code contained within the kernel.  V&V techniques in this
situation are more difficult than if the complete source code
were available. Responsibility for system failures due to faulty
software is yet another area to be resolved under this
environment.

The Guidance and Control Directorate began a software
development effort to address these problems.  A project to
develop an experimental run time kernel was begun that will
eliminate the major drawbacks of the Ada programming language
mentioned above. The Real Time Executive for Multiprocessor Systems
(RTEMS) provides full capabilities for management of tasks,
interrupts, time, and multiple processors in addition to those
features typical of generic operating systems.  The code is
Government owned, so no licensing fees are necessary.  RTEMS has
been implemented in both the Ada and C programming languages.
It has been ported to the following processor families:

@itemize @bullet
@item Intel i80386 and above 
@item Intel i80960
@item Motorola MC68xxx 
@item Motorola MC683xx
@item MIPS
@item PowerPC
@item SPARC
@item Hewlett Packard PA-RISC
@item Hitach SH
@item AMD A29K
@item UNIX
@end itemize

Support for other processor families, including RISC, CISC, and DSP, is
planned.  Since almost all of RTEMS is written in a high level language,
ports to additional processor families require minimal effort.

RTEMS multiprocessor support is capable of handling
either homogeneous or heterogeneous systems.  The kernel
automatically compensates for architectural differences (byte
swapping, etc.) between processors.  This allows a much easier
transition from one processor family to another without a major
system redesign.

Since the proposed standards are still in draft form,
RTEMS cannot and does not claim compliance.  However, the status
of the standard is being carefully monitored to guarantee that
RTEMS provides the functionality specified in the standard.
Once approved, RTEMS will be made compliant.

This document is a detailed users guide for a
functionally compliant real-time multiprocessor executive.  It
describes the user interface and run-time behavior of Release
@value{RTEMS-RELEASE} of the @value{LANGUAGE} interface
to RTEMS.