PL/M 386 Programmer`s Guide - Micro

PL/M 386 Programmer`s Guide - Micro
PL/M 386
Programmer’s Guide
RadiSys Corporation
5445 NE Dawson Creek Drive
Hillsboro, OR 97124
(503) 615-1100
FAX: (503) 615-1150
www.radisys.com
07-0710-01
December 1999
EPC, iRMX, INtime, Inside Advantage, and RadiSys are registered trademarks of
RadiSys Corporation. Spirit, DAI, DAQ, ASM, Brahma, and SAIB are trademarks of
RadiSys Corporation.
Microsoft and MS-DOS are registered trademarks of Microsoft Corporation and Windows 95
is a trademark of Microsoft Corporation.
IBM and PC/AT are registered trademarks of International Business Machines Corporation.
Microsoft Windows and MS-DOS are registered trademarks of Microsoft
Corporation.
Intel is a registered trademark of Intel Corporation.
All other trademarks, registered trademarks, service marks, and trade names are property of
their respective owners.
December 1999
Copyright  1999 by RadiSys Corporation
All rights reserved.
ii
Quick Contents
Chapter 1.
Introduction
Chapter 2.
Language Elements
Chapter 3.
Data Declarations, Types, and Based Variables
Chapter 4.
Arrays and Structures
Chapter 5.
Expressions and Assignments
Chapter 6.
Flow Control Statements
Chapter 7.
Block Structure and Scope
Chapter 8.
Procedures
Chapter 9.
Built-In Procedures, Functions, and Variables
Chapter 10. Features Involving the Target CPU and Numeric
Coprocessor
Chapter 11. Compiler Invocations and Controls
Chapter 12. Sample Program
Chapter 13. Extended Segmentation Models
Chapter 14. Error and Warning Messages
Appendix A. PL/M Reserved Words and Predeclared Identifiers
Appendix B. PL/M Program Limits
Appendix C. Grammar of the PL/M Language
Appendix D. Differences Between PL/M Compilers
Appendix E. Character Set
Appendix F. Linking to Modules Written in Other Languages
Appendix G. Run-time Interrupt Processing
Appendix H. Run-time Support for PL/M Applications
Index
PL/M-386 Programmer's Guide
iii
Notational Conventions
The following notational conventions are used throughout this manual.
Monospace Type
indicates literal command syntax, and other actual
input/output.
italics
indicate variable expressions and filenames. Substitute a value
or a symbol.
directory
refers to a user-created directory. A forward slash (/) is used
for iRMX directory paths. A backward slash (\) is used for
DOS directory paths.
pathname
represents a fully-qualified reference to a file.
All numbers are decimal unless otherwise stated. Hexadecimal numbers include the
H radix character (for example, 0FFH). Binary numbers include the B radix character
(for example, 11011000B).
iv
Contents
1
Introduction
Product Definition ...........................................................................................
Compatible Assemblers, Debuggers, and Utilities...........................................
Advantages of Using the PL/M Language .......................................................
The Structure of a PL/M Program ...................................................................
Overview of PL/M Statements ........................................................................
Declaration Statements.............................................................................
Executable Statements..............................................................................
Built-in Procedures and Variables ............................................................
Overview of PL/M Expressions................................................................
Input and Output.......................................................................................
2
Language Elements
Character Set ...................................................................................................
Tokens, Separators, and the Use of Blanks......................................................
Identifiers and Reserved Words.......................................................................
Constants .........................................................................................................
Whole-number Constants .........................................................................
Floating-point Constants...........................................................................
Character Strings ......................................................................................
Comments........................................................................................................
3
1
2
4
5
6
6
6
7
7
8
9
11
12
12
13
13
15
15
Data Declarations, Types, and Based Variables
Variable Declaration Statements .....................................................................
Sample DECLARE Statements ................................................................
Results of Variable Declarations ..............................................................
Combining DECLARE Statements ..........................................................
Initializations ...................................................................................................
The Implicit Dimension Specifier ............................................................
Names for Execution Constants: the Use of DATA .................................
Types of Declaration Statements .....................................................................
PL/M-386 Programmer’s Guide
Contents
18
18
19
20
21
23
25
26
v
Compilation Constants (Text Substitution): The Use of LITERALLY ....
Declarations of Names for Labels ............................................................
Results of Label Declarations...................................................................
Declaration for Procedures.......................................................................
Data Types.......................................................................................................
Unsigned Binary Number Variables: Unsigned Arithmetic ....................
INTEGER Variables: Signed Arithmetic ................................................
Signed Arithmetic .............................................................................
REAL Variables: Floating-point Arithmetic ...........................................
Examples of Binary Scientific Notation ...................................................
POINTER Variables and Location References ........................................
The @ Operator.................................................................................
Storing Strings and Constants via Location References ....................
OFFSET Data Type and the Dot Operator ...............................................
SELECTOR Variables .............................................................................
Based Variables...............................................................................................
Location References and Based Variables ...............................................
The AT Attribute .............................................................................................
WORD32 | WORD16 Type Mapping..............................................................
Choosing WORD32 or WORD16 ...................................................................
4
Arrays and Structures
Arrays..............................................................................................................
Subscripted Variables...............................................................................
Structures.........................................................................................................
Arrays of Structures .................................................................................
Arrays Within Structures..........................................................................
Arrays of Structures With Arrays Inside the Structures ...........................
Nested Structures......................................................................................
References to Arrays and Structures................................................................
Fully Qualified Variable References ........................................................
Unqualified and Partially Qualified Variable References.........................
5
49
50
51
51
51
52
53
54
54
55
Expressions and Assignments
Operands..........................................................................................................
Constants .........................................................................................................
Whole-number Constants in Unsigned Context........................................
Whole-number Constants in Signed Context............................................
String Constants .......................................................................................
Variable and Location References...................................................................
Subexpressions ................................................................................................
vi
26
28
28
29
30
32
33
33
33
34
36
37
38
39
39
40
42
43
46
47
Contents
57
58
58
58
58
60
60
Compound Operands .......................................................................................
Arithmetic Operators .......................................................................................
The +, -, *, and / Operators.......................................................................
The MOD Operator ..................................................................................
Relational Operators ........................................................................................
Logical Operators ............................................................................................
Expression Evaluation .....................................................................................
Precedence of Operators: Analyzing an Expression ................................
Compound Operands Have Types ............................................................
Relational Operators Are Restricted.........................................................
Order of Evaluation of Operands..............................................................
Choice of Arithmetic: Summary of Rules ......................................................
Special Case: Constant Expressions ........................................................
Assignment Statements....................................................................................
Implicit Type Conversions .......................................................................
Constant Expression .................................................................................
Multiple Assignment ................................................................................
Embedded Assignments ...........................................................................
6
Flow Control Statements
DO and END Statements: DO Blocks ............................................................
Simple DO Blocks....................................................................................
DO CASE Blocks.....................................................................................
DO WHILE Blocks ..................................................................................
Iterative DO Blocks..................................................................................
END Statement................................................................................................
IF Statement ....................................................................................................
Nested IF Statements................................................................................
Sequential IF Statements ..........................................................................
GOTO Statements ...........................................................................................
The CALL and RETURN Statements .............................................................
7
60
61
61
64
65
67
69
69
71
72
72
73
76
78
78
81
81
82
85
87
88
90
91
94
94
96
98
99
100
Block Structure and Scope
Names Recognized Within Blocks ..................................................................
Restrictions on Multiple Declarations ......................................................
Extended Scope: The PUBLIC and EXTERNAL Attributes..........................
Scope of Labels and Restrictions on GOTOs ..................................................
PL/M-386 Programmer’s Guide
Contents
104
106
107
110
vii
8
Procedures
Procedure Declarations....................................................................................
Parameters................................................................................................
Typed Versus Untyped Procedures ..........................................................
Activating a Procedure: Function References and CALL Statements ............
Indirect Procedure Activation...................................................................
Code Examples.........................................................................................
Exit from a Procedure: The RETURN Statement...........................................
The Procedure Body ........................................................................................
Examples..................................................................................................
The Attributes: PUBLIC and EXTERNAL, INTERRUPT, REENTRANT...
Interrupts and the INTERRUPT Attribute................................................
Reentrancy and the REENTRANT Attribute ...........................................
9
Built-in Procedures, Functions, and Variables
Obtaining Information About Variables ..........................................................
The LENGTH Function............................................................................
The LAST Function .................................................................................
The SIZE Function ...................................................................................
Explicit Type and Value Conversions .............................................................
The PL/M-386 LOW, HIGH, and DOUBLE Functions...........................
The FLOAT Function...............................................................................
The FIX Function .....................................................................................
The INT Function.....................................................................................
The SIGNED Function.............................................................................
The UNSIGN Function ............................................................................
The Unsigned Binary Data Type Built-in Functions ................................
Signed Integer Data Type Built-in Function.............................................
REAL Built-in Functions .........................................................................
The SELECTOR Built-in Function ..........................................................
The POINTER Built-in Function .............................................................
The OFFSET Built-in Function................................................................
The ABS and IABS Functions..................................................................
Shift and Rotate Functions...............................................................................
Rotation Functions ...................................................................................
Logical-shift Functions.............................................................................
Algebraic-shift Functions .........................................................................
Concatenate Functions..............................................................................
String Manipulation Procedures and Functions ...............................................
The Copy String in Ascending Order Procedure ......................................
The Copy String in Descending Order Procedure ....................................
viii
115
116
118
119
120
122
123
125
125
127
128
129
Contents
135
135
136
136
137
145
146
146
147
147
148
149
149
150
150
150
151
151
152
152
153
154
155
156
157
157
The Compare String Function ..................................................................
The Find Element Functions.....................................................................
The Find String Mismatch Function.........................................................
The Translate String Procedure ................................................................
The Set String to Value Procedure ...........................................................
PL/M-386 Bit Manipulation Built-ins..............................................................
The Copy Bit String Procedure.................................................................
The Find Set Bit Function ........................................................................
Miscellaneous Built-ins ...................................................................................
The Move Bytes Procedure ......................................................................
The Time Delay Procedure.......................................................................
The Lock Set Function .............................................................................
The Lock Bit Functions ............................................................................
POINTER and SELECTOR-related Functions................................................
The Return POINTER Value Function.....................................................
The Return Segment Portion of POINTER Function ...............................
The Return Offset Portion of POINTER Function ...................................
The Set POINTER Bytes to Zero Variable...............................................
WORD16 Built-in Mapping ............................................................................
10
158
159
160
161
162
163
163
163
165
165
166
166
168
169
169
169
169
170
170
Features Involving the Target CPU and Numeric
Coprocessor
Microprocessor Hardware-dependent Statements............................................
The ENABLE and DISABLE Statements ................................................
The CAUSE$INTERRUPT Statement .....................................................
The HALT Statement ...............................................................................
Microprocessor Hardware Flags......................................................................
Optimization and the Hardware Flags ......................................................
The CARRY, SIGN, ZERO, and PARITY Functions..............................
The PLUS and MINUS Operators............................................................
Carry-rotation Functions...........................................................................
The Decimal Adjust Function...................................................................
Microprocessor Hardware Registers................................................................
The Flags Register Access Variable .........................................................
The STACKPTR and STACKBASE Variables .......................................
Microprocessor Hardware I/O .........................................................................
The Find Value in Input Port Function .....................................................
The Access Output Port Array..................................................................
The Read and Store String Procedure.......................................................
The Write String Procedure......................................................................
The Hardware Protection Model......................................................................
The Task Register.....................................................................................
PL/M-386 Programmer’s Guide
Contents
171
171
172
172
173
173
174
174
174
175
175
175
176
177
177
177
178
179
179
179
ix
The TASK$REGISTER Variable .....................................................
The Global Descriptor Table Register......................................................
The SAVE$GLOBAL$TABLE Procedure .......................................
The RESTORE$GLOBAL$TABLE Procedure................................
The Interrupt Descriptor Table Register...................................................
The SAVE$INTERRUPT$TABLE Procedure .................................
The RESTORE$INTERRUPT$TABLE Procedure ..........................
The Local Descriptor Table Register........................................................
The LOCAL$TABLE Variable.........................................................
The Machine Status Register....................................................................
The MACHINE$STATUS Variable .................................................
The CONTROL$REGISTER, DEBUG$REGISTER, and
TEST$REGISTER Built-in Arrays ...............................................
The CLEAR$TASK$SWITCHED$FLAG Procedure ......................
Segment Information................................................................................
The GET$ACCESS$RIGHTS Function ...........................................
The GET$SEGMENT$LIMIT Function...........................................
Segment Accessibility ..............................................................................
The SEGMENT$READABLE Function ..........................................
The SEGMENT$WRITABLE Function ...........................................
Adjusting the Requested Privilege Level..................................................
The ADJUST$RPL Function ............................................................
The REAL Math Facility.................................................................................
Built-ins Supporting the REAL Math Unit ......................................................
The INIT$REAL$MATH$UNIT Procedure ............................................
The SET$REAL$MODE Procedure ........................................................
The GET$REAL$ERROR Function ........................................................
Saving and Restoring REAL Status..........................................................
The SAVE$REAL$STATUS Procedure...........................................
The RESTORE$REAL$STATUS Procedure ...................................
Interrupt Processing..................................................................................
The WAIT$FOR$INTERRUPT Procedure.......................................
WORD16 Mapping for Built-ins .....................................................................
Intel486 Processor Built-ins ............................................................................
11
184
186
186
186
187
188
188
188
189
189
190
193
193
193
194
194
195
196
196
196
197
197
Compiler Invocation and Controls
Invocation Syntax on iRMX Systems..............................................................
Invocation Examples and Sign-on/Sign-off Messages under the iRMX OS
Invocation Syntax on DOS Systems ................................................................
Invocation Examples and Sign-on/Sign-off Messages under DOS...........
File Usage under DOS and the iRMX OS .......................................................
Input Files.................................................................................................
x
179
180
181
181
182
182
183
183
183
184
184
Contents
199
201
202
202
203
203
Work Files ................................................................................................
Print Files ..........................................................................................
Object Files .......................................................................................
Executable Programs.........................................................................
Introduction to Compiler Controls...................................................................
Input Format Control................................................................................
Code Generation and Object File Controls ...............................................
Segmentation Controls .............................................................................
Listing Selection and Content Controls ....................................................
Listing Format Controls............................................................................
Source Inclusion Controls.........................................................................
Conditional Compilation Controls............................................................
Language Compatibility Control ..............................................................
Predefined Switches .................................................................................
Compiler Control Encyclopedia ......................................................................
CODE | NOCODE....................................................................................
COND | NOCOND...................................................................................
DEBUG | NODEBUG..............................................................................
EJECT ......................................................................................................
IF | ELSE | ELSEIF | ENDIF....................................................................
INCLUDE ................................................................................................
INTERFACE............................................................................................
LEFTMARGIN ........................................................................................
LIST | NOLIST ........................................................................................
MOD486...................................................................................................
OBJECT | NOOBJECT ............................................................................
OPTIMIZE ...............................................................................................
OVERFLOW | NOOVERFLOW .............................................................
PAGELENGTH .......................................................................................
PAGEWIDTH ..........................................................................................
PAGING | NOPAGING ...........................................................................
PRINT | NOPRINT ..................................................................................
RAM | ROM.............................................................................................
SAVE | RESTORE...................................................................................
SET | RESET............................................................................................
SMALL | COMPACT | MEDIUM | LARGE | FLAT ..............................
SMALL .............................................................................................
COMPACT .......................................................................................
MEDIUM ..........................................................................................
LARGE .............................................................................................
FLAT.................................................................................................
SUBTITLE...............................................................................................
SYMBOLS | NOSYMBOLS....................................................................
PL/M-386 Programmer’s Guide
Contents
203
204
204
206
207
212
212
212
213
213
213
214
217
217
218
218
218
219
219
219
221
222
227
227
228
228
229
244
244
245
245
245
246
246
247
248
248
249
251
251
251
252
252
xi
TITLE ......................................................................................................
TYPE | NOTYPE .....................................................................................
WORD32 | WORD16...............................................................................
XREF | NOXREF.....................................................................................
Program Listing...............................................................................................
Sample Program Listing...........................................................................
Symbol and Cross-reference Listing ........................................................
Compilation Summary .............................................................................
12
Sample Program
Introduction .....................................................................................................
FREQ Module .................................................................................................
OPEN Module .................................................................................................
PRINT Module ................................................................................................
Include Files ....................................................................................................
13
283
284
285
289
294
295
295
296
299
300
302
Error and Warning Messages
PL/M Program Error and Warning Messages..................................................
Fatal Command Tail and Control Error Messages...........................................
Fatal Input/Output Error Messages..................................................................
Fatal Insufficient Memory Error Messages .....................................................
Fatal Compiler Failure Error Messages...........................................................
Insufficient Memory Warning Messages.........................................................
xii
265
265
269
274
280
Extended Segmentation Models
Overview .........................................................................................................
Introduction .....................................................................................................
Segmentation Controls Architecture Overview ...............................................
Using Subsystems............................................................................................
Open Subsystems .....................................................................................
Closed Subsystems ...................................................................................
Communication Between Subsystems......................................................
Syntax..............................................................................................................
Placement of Segmentation Controls........................................................
Exporting Procedures ......................................................................................
Large Matrix Example .............................................................................
14
253
253
253
256
257
257
261
263
Contents
307
321
322
322
323
323
A
B
C
PL/M Reserved Words and
Predeclared Identifiers
Introduction .....................................................................................................
325
PL/M Program Limits
331
.............................................................
Grammar of the PL/M Language
Lexical Elements .............................................................................................
Character Sets...........................................................................................
Tokens ......................................................................................................
Delimiters.................................................................................................
Identifiers .................................................................................................
Numeric Constants ...................................................................................
Strings ......................................................................................................
PL/M Text Structure: Tokens, Blanks, and Comments ............................
Modules and the Main Program.......................................................................
Declarations.....................................................................................................
DECLARE Statement...............................................................................
Variable Elements ....................................................................................
Label Element ..........................................................................................
Literal Elements .......................................................................................
Factored Variable Element .......................................................................
Factored Label Element............................................................................
The Structure Type...................................................................................
Procedure Definition ................................................................................
Attributes..................................................................................................
AT .....................................................................................................
INTERRUPT.....................................................................................
Initialization ......................................................................................
Units ................................................................................................................
Basic Statements.......................................................................................
Assignment Statement.......................................................................
CALL Statement ...............................................................................
GOTO Statement...............................................................................
Null Statement...................................................................................
RETURN Statement..........................................................................
Microprocessor-dependent Statements..............................................
Scoping Statements ..................................................................................
Simple DO Statement........................................................................
DO-CASE Statement.........................................................................
DO-WHILE Statement......................................................................
Iterative DO Statement......................................................................
PL/M-386 Programmer’s Guide
Contents
336
336
336
336
336
337
337
337
338
339
339
339
340
340
340
340
340
341
341
341
341
341
342
342
342
342
342
342
342
343
344
344
344
344
344
xiii
END Statement .................................................................................
Procedure Statement..........................................................................
Conditional Clause ...................................................................................
DO Blocks................................................................................................
Simple DO Blocks.............................................................................
DO-CASE Blocks .............................................................................
DO-WHILE Blocks...........................................................................
Iterative DO Blocks...........................................................................
Expressions......................................................................................................
Primaries ..................................................................................................
Constants...........................................................................................
Variable References ..........................................................................
Location References..........................................................................
Operators..................................................................................................
Structure of Expressions...........................................................................
D
Differences Between PL/M Compilers
Differences between PL/M-86 and PL/M-80 ..................................................
Compatibility of PL/M-80 Programs and the PL/M-86 Compiler ...................
Differences between PL/M-286 and PL/M-86 ................................................
Compatibility of PL/M-86 Programs and the PL/M-286 Compiler .................
Differences between PL/M-386 and PL/M-286 ..............................................
Compatibility of PL/M-286 Programs and the PL/M-386 Compiler ...............
E
Character Set
F
Linking to Modules Written in Other Languages
Introduction .....................................................................................................
Calling Sequence .............................................................................................
Procedure Prologue .........................................................................................
Procedure Epilogue .........................................................................................
Register Usage.................................................................................................
Segment Name Conventions............................................................................
C Language Compatibility...............................................................................
Design Guidelines ...........................................................................................
Code Example ..........................................................................................
Compiling C and PL/M Modules.....................................................................
xiv
344
344
345
345
345
345
345
345
346
346
346
346
346
346
347
Contents
349
350
350
351
351
352
361
363
365
367
368
371
372
373
373
377
G
Run-time Interrupt Processing
General Information ........................................................................................
The Interrupt Descriptor Table ........................................................................
Procedures and Tasks ...............................................................................
Interrupt Procedure Prologue and Epilogue.....................................................
Interrupt Tasks.................................................................................................
Exception Conditions in REAL Arithmetic .....................................................
Invalid Operation Exception.....................................................................
Denormal Operand Exception ..................................................................
Zero Divide Exception .............................................................................
Overflow Exception..................................................................................
Underflow Exception................................................................................
Precision Exception ..................................................................................
Writing a Procedure to Handle REAL Interrupts.............................................
H
379
380
380
381
384
386
388
389
389
389
390
390
391
Run-time Support for PL/M Applications
Numeric Coprocessor Support Libraries..........................................................
PL/M Support Libraries...................................................................................
Index
PL/M-386 Programmer’s Guide
397
398
407
Contents
xv
Tables
Table 1-1. Assemblers, Debuggers, and Utilities..........................................................
Table 2-1. PL/M Special Characters.............................................................................
Table 3-1. Declaration Elements...................................................................................
Table 3-2. Data Types...................................................................................................
Table 3-3. WORD32 | WORD16 Data Type Mapping .................................................
Table 5-1. Operator Precedence....................................................................................
Table 5-2. Summary of Expression Rules for PL/M-386..............................................
Table 5-3. Implicit Type Conversions in Assignment Statements for PL/M-386 .........
Table 9-1. Value and Type Conversions for PL/M-386................................................
Table 11-1. Compiler Controls .....................................................................................
Table 11-1. Compiler Controls (continued) ..................................................................
Table 11-1. Compiler Controls (continued) ..................................................................
Table 11-4. WORD32 | WORD16 Data Type Mapping ...............................................
Table 11-5. WORD32 | WORD16 Built-in Mapping....................................................
Table 11-5. WORD32 | WORD16 Built-in Mapping (continued) ................................
Table 13-1. Segmentation Controls and Memory Partitions .........................................
Table 13-2. Segmentation Controls, Register Addresses and Pointer Values ...............
Table 13-3. Intel386 and Intel486 Microprocessor-specific ES Register Segmentation
Controls, Register Addresses and Pointer Values ........................................
Table E-1. Character Set...............................................................................................
Table E-1. Character Set (continued)............................................................................
Table E-1. Character Set (continued)............................................................................
Table E-1. Character Set (continued)............................................................................
Table E-1. Character Set (continued)............................................................................
Table F-1. Stack Representation for PL/M Parameters.................................................
Table F-2. Summary of the Intel386 Microprocessor Register Usage ..........................
Table F-3. Registers Used to Hold Simple Data Types ................................................
Table F-4. Summary of PL/M-386 Segment Names.....................................................
Table G-1. Exception and Response Summary.............................................................
xvi
Contents
2
10
17
31
46
61
74
79
138
209
210
211
254
255
256
286
287
289
353
354
355
356
358
363
368
370
371
387
Figures
Figure 1-1. 32-bit Protected Mode iRMX Application Development ...........................
Figure 3-1. Successive Byte References of a Structure.................................................
Figure 7-1. Inclusive Extent of Blocks..........................................................................
Figure 7-2. Sample Program Modules Illustrating Valid GOTO Usage .......................
Figure 7-3. Sample Program Modules Illustrating Valid GOTO Transfers ..................
Figure 10-1. The Hardware Flags Register ...................................................................
Figure 10-2. The REAL Error Byte ..............................................................................
Figure 10-3. The REAL Mode Word ............................................................................
Figure 11-1. Sample Program Using Conditional Compilation (SET control)..............
Figure 11-2. Sample Program Showing the NOCOND Control ...................................
Figure 11-3. Sample Program Showing the OPTIMIZE(0) Control .............................
Figure 11-4. Sample Program Showing the OPTIMIZE(1) Control .............................
Figure 11-5. Sample Program Showing the OPTlMIZE(2) Control .............................
Figure 11-6. Sample Program Showing the OPTIMIZE(3) Control .............................
Figure 11-7. Program Listing........................................................................................
Figure 11-8. Code Listing (continued) ..........................................................................
Figure 11-9. Cross-reference Listing ............................................................................
Figure 11-10. Compilation Summary............................................................................
Figure 12-1. Source Code for FREQ Module ...............................................................
Figure 12-2. Source Code for OPEN Module ...............................................................
Figure 12-3. Source Code for PRINT Module ..............................................................
Figure 12-4. Include File -- defns.inc............................................................................
Figure 12-5. Include File -- udi.inc ...............................................................................
Figure F-1. Stack Layout at Point Where a Non-interrupt Procedure is Activated .......
Figure F-2. Stack Layout During Execution of a Non-interrupt Procedure Body .........
Figure G-1. Stack Layout at Point Where an Interrupt Procedure Gains Control .........
Figure G-2. Stack Layout during Execution of Interrupt Procedure Body ....................
Figure G-3. Tag Word Format ......................................................................................
Figure G-4. Memory Layout of the REAL Save Area in Protected Mode for the
386 Microprocessor .....................................................................................
3
44
105
112
113
176
190
191
215
216
230
233
239
242
258
260
261
263
266
270
275
280
281
364
365
382
383
393
PL/M-386 Programmer’s Guide
xvii
Contents
395
xviii
Contents
Introduction
1
This chapter introduces the PL/M-386 compiler and explains the process of
developing software for execution by an Intel386™ or Intel486™
microprocessor-based system.
Product Definition
The PL/M-386 compiler is a software tool that translates PL/M source code into a
relocatable object module. The object modules (in OMF386 format) are compatible
with all other OMF386-producing translators, such as ASM386, iC-386, and Fortran386.
The PL/M-386 compiler translates a PL/M source text file into an object module and
a listing file. Parts or all of a program can be compiled in a single compilation.
Object modules can then be linked or bound in various combinations to form
different applications. These applications can be run on a DOS host or as part of the
iRMX® Operating System (OS) or its Human Interface layer.
The PL/M-386 compiler also provides a listing output, error messages, and a number
of compiler controls that aid in developing and debugging programs.
✏
Note
For information on invoking the compiler, see Chapter 11,
Compiler Invocation and Controls. That chapter covers invocation
on both iRMX and DOS-hosted systems.
PL/M-386 Programmer's Guide
Chapter 1
1
Compatible Assemblers, Debuggers, and Utilities
Table 1-1 shows the compatible Intel assemblers, debuggers, and utilities. Figure 1-1
shows the role of these software tools in developing an iRMX application.
Table 1-1. Assemblers, Debuggers, and Utilities
Tool
Name for Intel386 and
Intel486 Processors
assembler
ASM386
C compiler
iC-386
FORTRAN compiler
Fortran-386
debuggers
System Debugger (SDB)
System Debug Monitor (SDM)
Soft-Scope III Debugger
binder
BND386
librarian
LIB386
cross-reference
MAP386
These tools support modular application development. Refer to the following
publications for further information:
2
•
ASM386 - ASM386 Macro Assembler Operating Instructions and ASM386
Assembly Language Reference
•
iC-386 - iC-386 Compiler User's Guide
•
Fortran-386 - Fortran 386 Compiler User's Guide
•
SDB and SDM - iRMX System Debugger Reference
•
Soft-Scope III Debugger - Soft-Scope Debugger User's Guide
•
BND386 - Intel386 Family Utilities User's Guide
•
LIB386 - Intel386 Family Utilities User's Guide
•
MAP386 - Intel386 Family Utilities User's Guide
®
Chapter 1
Introduction
Create and Maintain
Libraries With
LIB386
Bound
Modules
Translate
With
Bind Object
Files With
ASM386
Write Source
File With
iC-386
Text Editor
Source
Code
BND386
Linkable
OMF386
Object
Code
Using
load
Option
Debug Application
Software on Target
With
Soft-Scope
III
iRMX Target
System
PL/M-386
Fortran-386
Using
noload
Option
Executable
iRMX
32-bit
Program
Load for On-target
Assembly-language
Debugging Using SDB/SDM
With
Bootstrap
Loader
Correct Errors Found During Translation
Correct Errors Found During Debugging
Combine Into
iRMX III Operating
System With
Load for On-target
Symbolic Debugging With
ICU386
Bootstrap
Loader
Build
Application
System With
BLD386
Submit
File
Soft-Scope
III
iRMX
Target System
Load for Cross-hosted
Symbolic Debugging
With
iRMX
Application
System with
32-bit First
Level
or I/O Job
Soft-Scope
III
(BLD386 Automatically
Invoked by Submit File)
Load into Emulation and Analysis
Tools for Cross-hosted Debugging
DOS
Correct Errors Found During Debugging
= ICU-configurable systems only.
W-3359
Figure 1-1. 32-bit Protected Mode iRMX Application Development
PL/M-386 Programmer's Guide
Chapter 1
3
Advantages of Using the PL/M Language
PL/M programs are portable, which means that they are easily transferred from one
microprocessor to another. When using PL/M, you need not be concerned with the
instruction set of the target processor. Additionally, there is no need to be concerned
with other details of the target processor, such as register allocation or assigning the
proper number of bytes for each data item. The PL/M-386 compiler does these
functions automatically. PL/M keywords and phrases are close to natural English,
and many operations (including arithmetic and Boolean operations) can be combined
into expressions. This enables the execution of a sequence of operations with just
one program statement. Data types and data structures have functional attributes.
For instance, in PL/M, the program can be written in terms of Boolean expressions,
characters, and data structures, in addition to bytes, words, and integers.
Coding programs in a high-level language rather than assembly language involves
thinking closer to the level used when planning the overall system design. Following
is a list of the advantages of using PL/M, and the applications for which PL/M is best
suited:
•
PL/M block structure and control constructs aid and encourage structured
programming.
•
PL/M has facilities for data structures such as structured arrays and
pointer-based dynamic variables.
•
PL/M is a typed language. The compiler does data type compatibility checking
during compilation to help detect logic errors in programs.
•
PL/M data structuring facilities and control statements are designed in a logically
consistent way. Thus, PL/M is a good language for expressing algorithms for
systems programming.
•
PL/M is a standard language used for application development on Intel systems.
PL/M programs are compatible across the Intel386 and Intel486 family of
microprocessors.
•
PL/M was designed for programmers (generally systems programmers) who
need access to the microprocessor's features such as indirect addressing and
direct I/O for optimum use of all system resources.
In comparison with other languages, PL/M has more features than BASIC and is a
more general-purpose language than either FORTRAN (best suited for scientific
applications) or COBOL (designed for business data processing). PL/M accesses the
microprocessor hardware features more easily than C. Additionally, in comparison to
C, PL/M offers the ability to nest procedures and the program structure is easier to
maintain.
4
Chapter 1
Introduction
The Structure of a PL/M Program
PL/M is a block-structured language; every statement in a program is part of at least
one block. A block is a well-defined group of statements that begins with a DO
statement or a procedure declaration and ends with an END statement.
A module is a labeled simple DO-block. A module must begin with a labeled DO
statement and end with an END statement. Between the DO statement and the END
statement other statements provide the definitions of data and processes that make up
the program. These statements are said to be part of the block, contained within the
block, or nested within the block. A module can contain other blocks but is never
itself contained within another block. See Chapter 6 for a description of DO-blocks.
Every PL/M program consists of one or more modules, separately compiled, each
consisting of one or more blocks. The two kinds of blocks are DO-blocks and
procedure definition blocks.
A procedure definition block is a set of statements beginning with a procedure
declaration and ending with an END statement. Other declarations and executable
statements can be placed between these points, and are used later when the procedure
is actually invoked or called into execution. The definition block is a further
declaration of everything the procedure will use and do.
PL/M-386 Programmer's Guide
Chapter 1
5
Overview of PL/M Statements
The two types of statements in PL/M are declarations and executable statements. All
PL/M statements end with a semicolon (;).
Declaration Statements
The following is a simple example of a declaration statement:
DECLARE WIDTH BYTE;
This statement introduces the identifier WIDTH and associates it with the contents of 1
byte (8 bits) of memory. Now, rather than having to know the memory address of
this byte, you can refer to it by the name WIDTH.
A group of statements intended to perform a function (i.e., a subprogram or
subroutine) can be given a name by declaring them to be a procedure:
ADDER_UPPER: PROCEDURE (BETA) BYTE;
The statements that define the procedure follow the semicolon. This block of PL/M
statements is invoked from other points in the program, and may involve passing
parameters to the program. When a procedure has finished executing, control is
returned immediately to the main program. This capability is the major feature
enabling modular program construction.
Executable Statements
The following is an example of an executable statement:
CLEARANCE = WIDTH + 2;
The two identifiers, CLEARANCE and WIDTH, must be declared prior to this executable
statement, which produces machine code to retrieve the WIDTH value from memory.
Once the WIDTH value is obtained, 2 is added to it and the sum is stored in the
memory location for CLEARANCE.
For most purposes, it is unnecessary to think in terms of memory locations when
programming in PL/M. CLEARANCE and WIDTH are variables, and the assignment
statement assigns the value of the expression WIDTH + 2 to the variable
CLEARANCE. The compiler automatically generates all the machine code necessary
to retrieve data from memory, to evaluate the expression retrieved, and to store the
result in the proper location.
6
Chapter 1
Introduction
Executable statements are discussed in the following chapters:
Assignment Statement
Chapter 5
CALL Statement
Chapter 8
CAUSE$INTERRUPT Statement
Chapter 10
DISABLE Statement
Chapter 10
DO CASE Statement
Chapter 6
DO WHILE Statement
Chapter 6
ENABLE Statement
Chapter 10
END Statement
Chapter 6
Executable Functions
Chapter 9
GOTO Statement
Chapter 6
HALT Statement
Chapter 10
IF Statement
Chapter 6
Iterative DO Statement
Chapter 6
Nested IF Statement
Chapter 6
RETURN Statement
Chapter 8
Simple DO Statement
Chapter 6
Built-in Procedures and Variables
PL/M provides a variety of built-in procedures and variables. These include
functions such as shifts and rotations, data type conversions, executable functions,
block I/O, real math, and string manipulation (see Chapters 9 and 10).
Overview of PL/M Expressions
A PL/M expression is made up of operands and operators, and resembles a
conventional algebraic expression.
Operands include numeric constants (such as 3.78 or 105) and variables (as well as
other types discussed in Chapters 3 and 5). The operators include + and - for
addition and subtraction, * and / for multiplication and division, and MOD for
modular arithmetic.
As in an algebraic expression, elements of a PL/M expression can be grouped with
parentheses.
PL/M-386 Programmer's Guide
Chapter 1
7
An expression is evaluated using unsigned binary arithmetic, signed integer
arithmetic, and/or floating-point arithmetic, depending on the types of operands in the
expression (see Chapters 3 and 5).
Input and Output
PL/M does not provide formatted I/O capabilities like those of FORTRAN, BASIC,
or COBOL. However, PL/M does provide built-in functions for direct I/O that do not
require operating system run-time support. The PL/M-386 compiler has built-in
functions which allow for single-byte, half-word or word I/O, as well as for block I/O
(for strings of bytes, half-words, or single-words). For detailed information on these
I/O functions, see Chapter 10.
■■■
8
Chapter 1
Introduction
Language Elements
2
PL/M-386 programs are free-form, meaning there are no restrictions on where you
place a statement on a line. You can use as many blanks (spaces) as necessary to
format your program for readability.
Character Set
The PL/M-386 source program character set is the following subset of the ASCII
character set:
A..Z
a..z
0..9
and the following special characters:
= .
/ ( ) + - ' * , < > : ; @ $ _
and the blank (space), tab, carriage-return and line-feed characters. (Appendix E
indicates if each ASCII character is a member of the PL/M character set and, if so,
the hexadecimal value.)
PL/M does not distinguish between uppercase and lowercase letters, except in string
constants. For example, the variable names xyz and XYZ are the same. (In this
manual, all PL/M syntax is uppercase, by convention.)
PL/M-386 Programmer's Guide
Chapter 2
9
Special characters have particular meaning in PL/M, as explained throughout this
manual. Table 2-1 summarizes the meaning of special characters in PL/M.
Table 2-1. PL/M Special Characters
Symbol
Name
Use
=
equal sign
Two distinct uses:
(1) assignment operator
(2) relational test operator
:=
assign
embedded assignment operator
@
at-sign
location reference operator
.
dot
Three distinct uses:
(1) decimal point
(2) structure member qualification
(3) address operator
/
slash
division operator
/*
*/
beginning-of-comment delimiter
end-of-comment delimiter
(
)
left parenthesis
right parenthesis
left delimiter of lists, subscripts, some expressions
right delimiter of lists, subscripts, some expressions
+
-
plus
minus
addition or unary plus operator
subtraction or unary minus operator
'
apostrophe
string delimiter
*
asterisk
Two distinct uses:
(1) multiplication operator
(2) implicit dimension specifier
<
>
<=
>=
<>
less than
greater than
less or equal
greater or equal
not equal
relational test operator
relational test operator
relational test operator
relational test operator
relational test operator
:
;
colon
semicolon
label terminator
statement terminator
,
comma
list element delimiter
_
underscore
significant character in identifier
$
dollar sign
Two distinct uses:
(1) non-significant character embedded within number
of identifier
(2) significant as the first character on a control line in a
source file
10
Chapter 2
Language Elements
The PL/M compiler treats multiple contiguous blanks in PL/M source programs as
single blanks, by ignoring all the blanks except the first one.
The compiler produces an error or warning message whenever it encounters a
character other than those described above in a source program.
In addition to the source character set, PL/M allows the use of special character sets
(such as Kanji characters), located from 0080H through 00FFH (excluding 0081H).
Tokens, Separators, and the Use of Blanks
The smallest meaningful unit of a PL/M statement is a token. Every token belongs to
one of the following classes:
•
Identifiers
•
Reserved words
•
Simple delimiters (all of the special characters, except the dollar sign, are simple
delimiters)
•
Compound delimiters (combinations of two special characters):
<>, <=, >=, :=, /*, */
•
Numeric constants
•
Character string constants
It is usually clear where one token ends and the next one begins. For example, in the
assignment statement:
EXACT=APPROX*(HEIGHT-3)/SCALE;
EXACT, APPROX, HEIGHT, and SCALE are identifiers, 3 is a numeric constant, and all
the other characters are simple delimiters.
If a delimiter (simple or compound) does not naturally occur between two tokens,
you must separate them with one or more blank(s).
A comment can also be used as a separator.
Blanks can be inserted around any token without changing the meaning of the PL/M
statement. Thus, the assignment statement:
EXACT = APPROX * ( HEIGHT - 3 ) / SCALE;
is equivalent to:
EXACT=APPROX*(HEIGHT-3)/SCALE;
PL/M-386 Programmer's Guide
Chapter 2
11
Identifiers and Reserved Words
Identifiers name variables, procedures, symbolic constants, and statements.
Statement identifiers are called labels. Identifiers can be up to 31 characters long.
The first character must be alphabetic or the underscore (_), and the remaining
characters may be alphabetic, numeric, or the underscore.
You can use the dollar sign character to improve the readability of an identifier or
constant, but the dollar character is not meaningful to the compiler. An identifier or
constant containing a dollar sign is equivalent to the same identifier without the dollar
sign. Note that you must not use a dollar character in a procedure name within a
subsystem definition. See Chapter 13.
Examples of valid identifiers are:
INPUT_COUNT
X
GAMM
LONGIDENTIFIERNUMBER3
LONG$$$IDENTIFIER$$$NUMBER$$$3
_MAIN
INPUT$COUNT
INPUTCOUNT
The long identifiers are identical to the compiler. INPUT$COUNT and INPUTCOUNT
are interchangeable, but are different from INPUT_COUNT.
Identifiers must be distinct from reserved words. If you want to use PL/M built-in
procedures and variables, the identifiers in your source program must be distinct from
the built-ins' predefined identifiers. Appendix A lists the reserved words and
predefined identifiers.
Constants
A constant is a value that does not change during a program's execution. The three
types of constants are whole-number constants, floating-point constants, and
character strings.
12
Chapter 2
Language Elements
Whole-number Constants
Whole-number constants can be binary, octal, decimal, or hexadecimal numbers.
Specify the base of these constants by appending a B, Q, D, or H suffix. The compiler
interprets numbers without a base suffix as decimal numbers. When they encounter
characters that are invalid in the specified (or assumed) base, the compiler produces
appropriate messages. If a constant contains characters invalid in the designated
number base, it will be flagged as an error.
In PL/M-386, a whole-number constant can be an 8-bit, 16-bit or 32-bit value. It can
also be a 64-bit value. The range of whole-number constants is non-negative. (The
minus sign in front of a whole-number constant is not part of the constant.)
The first character of a hexadecimal number must be a numeric digit to avoid looking
like an identifier. For example, write the hexadecimal form of the decimal value 163
as 0A3H (rather than A3H); otherwise the compiler will interpret it as an identifier.
Examples of valid whole-number constants are:
12AH 2 33Q 1010B 55D 0BF3H 65535 777O 3EACH 0F76C05H
Examples of invalid whole-number constants are:
12AF
Hexadecimal digits used without an H suffix, and invalid in the default
decimal interpretation.
12AD
The final D could be a suffix but the A is not a decimal digit. If
hexadecimal is intended, a final H is needed.
11A2B
A and 2 are not valid binary digits. If hexadecimal is intended, a final H
is necessary.
2ADGH
G is not a valid hexadecimal digit.
For example, the maximum whole-number 16-bit constant is:
2**16-1 = 1111$1111$1111$1111B = 177777Q = 65535D = 0FFFFH
The maximum whole-number 32-bit constant is:
2**32-1 = 1111$1111$1111$1111$1111$1111$1111$1111B
= 37777777777Q
= 4294967295D
= 0FFFFFFFFH
Floating-point Constants
The presence of a decimal point in a decimal constant creates a floating-point
constant. Floating-point constants are represented in REAL precision (see Duty
Types). Only decimal real constants are allowed.
PL/M-386 Programmer's Guide
Chapter 2
13
At least one decimal digit (e.g., 0) must precede the decimal point. A fractional part
is optional after the decimal point, as is the base-ten exponent, which is indicated by
the letter E. This exponent must have at least one digit. Note that no fractional
exponents are possible.
In PL/M-386, the range is -2**(+128) to -2**(-126), zero, +2**(-126) to +2**(+128).
This range is approximately -3.4 x 10**38 to -8.4 x 10**(-37), zero, and 8.4 x
10**(-37) to 3.4 x 10**38.
The following are examples of valid floating-point constants:
5.30
176.0
1.88
3.14159
16.
53.0E-1
1.760E2
0.188E1
314159.E-5
1.6E+1
222.2
2.222E+2
Note that plus signs do not change the meaning of exponents.
The following are examples of invalid floating-point constants:
6
No decimal point
1.3AH
Hexadecimal not allowed in floating-point constants
10.011B
Binary not allowed
7.52Q
Octal not allowed
4.8E1AH/2 Only decimal constants in exponents; no hexadecimal, no expressions,
no fractions
14
Chapter 2
Language Elements
Character Strings
Character strings are printable ASCII characters enclosed within apostrophes. There
are two types of character strings: 1) string constants and 2) character constants. A
string constant is used to initialize variables or to pass a pointer. The maximum
length of a string constant is 255. A character constant is used in expressions, and its
value should fit into a double or machine word (32 bits). A string used as a character
constant can contain from one to four characters.
To include an apostrophe in a string, write it as two apostrophes (e.g., the string
'''Q' comprises 2 characters, an apostrophe followed by a Q). Values 0080H
through 00FFH (excluding 0081H) can be used in a quoted character string. Spaces
are allowed but line-feeds are not. The compiler represents character strings in
memory as ASCII codes, one 7-bit character code to each 8-bit byte, with a
high-order zero bit. Strings of length 1 translate to single-byte values. Character
constants of length 2 translate to 16-bit values, and those of length 3 or 4 translate to
32-bit values. For example:
'A' is equivalent to 41H
'AG' is equivalent to 4147H
'AGR' is equivalent to 414752H
'AGRX' is equivalent to 41475258H
Therefore, character constants can be used as 8-bit, 16-bit, or 32-bit values.
Character constants longer than 4 characters exceed the 32-bit capacity.
See also Appendix E, Character Set.
Comments
In PL/M, a comment is a sequence of characters delimited on the left by the character
pair /* and on the right by the character pair */. These delimiters instruct the
compiler to ignore any text between them and to consider such text as not part of the
program.
A comment can contain any printable ASCII or special character and can also include
space, carriage-return, line-feed, and tab characters. If you embed a comment in a
character string constant, it becomes part of the constant. A comment can appear
anywhere that a blank character can appear except embedded within a token.
The following is an example of a PL/M comment:
/*This procedure copies one structure to another.*/
In this manual, comments are presented in lowercase to distinguish them visually
from program code, which is presented in uppercase.
■■■
PL/M-386 Programmer's Guide
Chapter 2
15
16
Chapter 2
Language Elements
Data Declarations, Types,
and Based Variables
3
In PL/M-386, you can declare symbolic names for variables, constants, procedures
and statements (labels). For each symbolic name, there must be one declaration at
the beginning of the block containing the name, or in an outer, enclosing block. A
declaration consists of an identifier, type, attributes and/or location. Multiple
declarations of a name in a block are invalid. Required and optional declaration
elements are shown in Table 3-1.
Table 3-1. Declaration Elements
Declaration
Must Use
Can Use
Variable
Names
BYTE, INTEGER, CHARINT,
SHORTINT, LONGINT,
OFFSET, WORD, QWORD,
HWORD, DWORD, REAL,
STRUCTURE, ADDRESS*
linkage attributes:** PUBLIC or
EXTERNAL; or location
attributes: AT (location
reference) variable initialization
attribute: INITIAL (value-list)
Constant
Names
type, as above, and constant
initialization attribute: DATA
(value-list)
linkage attributes as above
Label Names
LABEL
linkage attributes as above
Macros
LITERALLY 'string'
* ADDRESS is equivalent to the OFFSET data type.
** Placement is important (see Variable Declaration Statements).
The declaration of a variable or constant identifier must precede use of the identifier
in an executable statement. Although it is not good programming practice, you can
call a reentrant procedure before defining it. You can either explicitly declare a
statement label, or implicitly declare it by attaching it to an executable statement with
a colon character.
PL/M-386 Programmer's Guide
Chapter 3
17
Variable Declaration Statements
A DECLARE statement is a nonexecutable statement that introduces some object or
collection of objects, associates names (and sometimes values) with them, and
allocates storage if necessary. The most important use of DECLARE is for declaring
variables.
A variable can be a scalar (i.e., a single quantity), an array, or a structure.
A scalar variable is a single object whose value may not be known at compile time
and may change during the execution of the program.
An array is a list of scalars of the same data type, referred to by one identifier and
distinguished by the subscript associated with each scalar.
A structure is an aggregate of scalars, arrays and/or structures with the same main
identifier. The members of a structure are differentiated from each other by their
own member-identifiers or field names. For example, EMPLOYEES.NAME would
refer to the NAME field within the structure EMPLOYEES.
Sample DECLARE Statements
Note that when using linkage (PUBLIC/EXTERNAL) and initialization
(DATA/INITIAL) attributes, the order of declaration is critical. Place linkage
attributes before the initialization attribute, and after the type declaration.
For example:
DECLARE a$p BYTE PUBLIC INITIAL(4);
The following statements declare scalars:
DECLARE APPROX REAL;
DECLARE (OLD, NEW) BYTE;
DECLARE POINT WORD, VAL12 BYTE;
The first example declares a single scalar variable of type REAL, with the identifier
APPROX.
The second example declares two scalars, OLD and NEW, both of type BYTE. This
kind of statement is called a factored declaration, which is similar to the sequence:
DECLARE OLD BYTE;
DECLARE NEW BYTE;
A factored declaration (for structures and arrays) guarantees that the bytes will be
contiguously located in memory, which may be useful in real-time applications (see
also Combining DECLARE Statements). Separate declaration statements do not
guarantee this.
18
Chapter 3
Data Declarations, Types, and Based Variables
The third example declares two scalars of different types: POINT is of type WORD,
and VAL12 is of type BYTE.
The following statements declare arrays:
DECLARE DOMAIN (128) BYTE;
DECLARE GAMMA (19) DWORD;
The first example declares the array DOMAIN, with 128 scalar elements of type BYTE.
These elements are distinguishable by subscripting the name DOMAIN, using the range
0 to 127 for the subscripts. For example, the third element of DOMAIN can be referred
to as DOMAIN(2). The first element of every array has subscript 0.
The second example declares the array GAMMA, with 19 scalar elements of type
DWORD. The subscripts for this array can range from 0 to 18.
The third example declares a structure with two scalar members:
DECLARE RECORD STRUCTURE (KEY BYTE, INFO WORD);
The two members are a BYTE scalar that can be referred to as RECORD.KEY and a
WORD scalar that can be referred to as RECORD.INFO. The word named by
RECORD.INFO is the second and third bytes of this structure.
Structures are discussed in further detail in Chapter 4.
Results of Variable Declarations
Valid variable declarations result in the following:
•
The name is given a unique address.
•
The variable is considered to have the attributes declared.
All subsequent uses of the variable in the block where it is declared refer to the same
address (except for based variables, discussed in Based Variables).
A valid variable declaration also requires all references to the variable to conform to
the rules for the current attributes (i.e., those attributes having priority in the current
block). Thus, the compiler can flag a large variety of errors caused by incompatible
references within the current block. The variable reference must be consistent with
the variable declaration.
PL/M-386 Programmer's Guide
Chapter 3
19
Combining DECLARE Statements
A separate DECLARE statement is not required for each declaration. For example,
instead of writing the two DECLARE statements:
DECLARE CHR BYTE INITIAL ('A');
DECLARE COUNT INTEGER;
Both declarations can be written in a single DECLARE statement, as follows:
DECLARE CHR BYTE INITIAL ('A'), COUNT INTEGER;
This declare statement contains two declaration elements, separated by a comma. A
declaration element is the text for declaring one identifier (or one factored list of
identifiers). Every DECLARE statement contains at least one declaration element. If a
DECLARE statement contains more than one declaration element, they are separated
by commas.
Most of the examples shown previously have only one declaration element in each
DECLARE statement. In the preceding example, the text CHR BYTE INITIAL
('A') is one declaration element; the text COUNT INTEGER is another.
Another way of combining declaration elements is called a factored declaration as
indicated above in this section. For example, the non-factored declarations:
DECLARE A BYTE, B BYTE;
DECLARE C WORD, D WORD;
DECLARE E DWORD, F DWORD;
can be combined as:
DECLARE (A,B) BYTE, (C,D) WORD, (E,F) DWORD;
In each factored declaration, the allocated locations are contiguous. Elements
declared in a nonfactored declaration statement are not necessarily contiguous.
Use factored declarations if the order in which variables are allocated is important.
Variables declared in a factored declaration (i.e., variables within a parenthesized list
that are not based, are not used as parameters, or are not EXTERNAL), are stored
contiguously in the order specified. (If a based variable occurs in a parenthesized list,
the variable is ignored when storage is allocated.)
The declaration elements in a single DECLARE statement are independent of each
other, as if they were declared in separate DECLARE statements.
20
Chapter 3
Data Declarations, Types, and Based Variables
Initializations
Initialization guarantees that the variables being initialized have a particular value
before program execution begins. Every constant should be initialized. Variables
can also be initialized. There are no default values for constants or variables. Of
course, variables can be initialized by an assignment statement such as the following:
PI = 3.1415927;
VAR13 = 10;
/* PI must first be declared REAL */
/* VAR13 must be declared earlier */
However, in PL/M-386, the compiler can set up these values during the compilation
rather than using both instruction space and execution time to initialize variables in
the program.
There are two kinds of compile-time initializations: INITIAL, used with variables,
and DATA, used for constants. (DATA is explained in greater detail later in this
section.) In both initializations, the initialization attribute is placed after the type in
the declaration. For example:
DECLARE FAMILY WORD INITIAL (2);
Additionally, when using a linkage attribute (PUBLIC/EXTERNAL), place the linkage
attribute after the type declaration and before the initialization attribute.
INITIAL causes initialization to occur during program loading for variables that have
storage allocated for them. Such variables can subsequently be changed during
execution (just as any other variable). These variables will not be reinitialized on a
program restart.
The following rules apply to both INITIAL and DATA:
•
INITIAL and DATA cannot be used together in the same declaration.
•
INITIAL can occur only in declarations at the outer level of a module. DATA,
however, can occur in declarations at any level.
•
No initializations are permitted with based variables, formal parameters (see
Chapter 8), or with the EXTERNAL attribute (see Chapter 7).
•
Either INITIAL or DATA can follow use of the AT attribute. However, if this use
of INITIAL or DATA causes multiple initializations, the result cannot be
predicted.
PL/M-386 Programmer's Guide
Chapter 3
21
•
The initializing value should fit into the space allocated by the data type. The
only exception is initialization of HWORD when the offset is derived with a dot
operator. For example:
DECLARE HH HWORD INITIAL (.B)
In this case, the real offset is truncated to give the lower 16 bits. A warning
message is issued when an OFFSET value is truncated.
The general form of the INITIAL attribute is as follows:
INITIAL (value-list)
Where:
value-list
is a sequence of values separated by commas.
Values are taken one at a time from the value list and used to initialize the individual
scalars being declared. The initialization is performed in the same manner as an
assignment. Initial values for members of an array or structure must be specified
explicitly. For character string constants, the characters are taken one at a time to
initialize an 8-bit scalar, two at a time to initialize a 16-bit scalar, four at a time to
initialize a 32-bit scalar, and eight at a time to initialize a 64-bit scalar.
The expressions used with the INITIAL attribute have the following restrictions:
22
•
For real variables only: An expression, which can contain a unary + or operator, can only be a single floating-point constant which can be used to
initialize a REAL scalar only.
•
For POINTER variables only: A restricted expression can be a location reference
formed with the @ operator, which must refer to a variable already declared or to
a constant list.
•
For all other types (except SELECTOR): A restricted expression can be a constant
expression containing no operators except + or -. A constant expression has only
whole-number constants as operands (e.g., 2048, 256+5), as explained in Chapter
5. The constant expression is evaluated as if it were being assigned to the scalar
being initialized, using the rules described in Chapter 5.
•
For OFFSET or WORD variables only: A constant expression containing only the +
and - operators, and operands that can be whole-number constants and/or "."
location references. If the expression contains a "." location reference, only the
+ operator can precede it. Any combination of + and - operators can follow the
"." location reference. For example: 5+.xyz-10.
Chapter 3
Data Declarations, Types, and Based Variables
The declaration:
DECLARE THRESHOLD BYTE INITIAL (48);
declares the BYTE scalar THRESHOLD and initializes the scalar to a value of 48.
The declaration:
DECLARE EVEN (5) BYTE INITIAL (2, 4, 6, 8, 10);
declares the BYTE array EVEN and initializes its five scalar elements to 2, 4, 6, 8, and
10, respectively.
The declaration:
DECLARE COORD STRUCTURE (HIGH$BOUND WORD,
VALUE (3) BYTE,
LOW$BOUND BYTE) INITIAL (302, 3, 6, 12, 0);
declares the structure COORD and initializes it as follows:
COORD.HIGH$BOUND to 302
COORD.VALUE(0) to 3
COORD.VALUE(1) to 6
COORD.VALUE(2) to 12
COORD.LOW$BOUND to 0
If a string occurs in the value list, it is taken apart from left to right and the pieces are
stored in the scalars being initialized. One character is stored in each BYTE scalar,
two characters in each WORD scalar, and four in each DWORD scalar. For
example:
DECLARE GREETING (5) BYTE AT (@HI) INITIAL ('HELLO');
causes GREETING(0) to be initialized with the ASCII code for H, GREETING(1)
with the ASCII code for E, and so on.
All the examples shown previously have had value lists that match up one-for-one
with the scalars being declared. The value list can have fewer elements than are
being declared. Thus:
DECLARE DATUM (100) BYTE INITIAL (3, 5, 7, 8);
will work. The first four elements of the array DATUM are initialized with the four
elements in the value list, and the remainder of the array is left uninitialized.
However, the value list cannot have more elements than are being declared.
The Implicit Dimension Specifier
Often, when initializing an array, you want the array to have the same number of
elements as the value list. This can be done conveniently by using the implicit
PL/M-386 Programmer's Guide
Chapter 3
23
dimension specifier in place of an ordinary dimension specifier (a parenthesized
constant). The implicit dimension specifier has the form:
(*)
Also use the implicit dimension specifier to define an external or based array whose
precise number of elements is either unknown or insignificant. Thus the declaration:
DECLARE FAREWELL(*) BYTE PUBLIC INITIAL ('GOODBYE, NOW');
declares a public BYTE array, FAREWELL, with enough elements to contain the string
'GOODBYE, NOW' (namely 12), and initializes the array elements with the characters
of the string. To reference this array in another program module, declare it as
follows:
DECLARE FAREWELL(*) BYTE EXTERNAL;
See Chapter 7 for more information about PUBLIC and EXTERNAL attributes.
Note that the INITIAL and DATA value-lists must not be present when the
implicit dimension specifier is used with an external array; otherwise, INITIAL and
DATA value-lists are required. Also, the LENGTH, LAST, and SIZE built-ins
cannot be used on an external array that was declared with the implicit dimension
specifier.
The following is an example of an implicit dimension in a based declaration:
DECLARE X BASED P(*) BYTE;
The implicit dimension specifier cannot be used after the parenthesized list of
identifiers in a factored declaration (unless it is declared EXTERNAL). Additionally,
an implicit dimension specifier cannot be used to specify an array that is a member of
a structure.
The implicit dimension specifier can be used with any value list; it is not restricted to
strings.
24
Chapter 3
Data Declarations, Types, and Based Variables
Names for Execution Constants: the Use of DATA
A variable is the name of a single data item intended to be used and altered by a
program. If the variable is not altered during execution, it is a constant.
For example, the formula for the circumference of a circle (R x 2 x pi) or (radius x 2 x
pi) could be written in PL/M as:
C = R * 2.0 * 3.14159;
in which C and R would be variables. The declarations for C and R would have to
precede the executable statement, and could appear as:
DECLARE (C, R) REAL;
If pi is used often enough, simplify writing of statements by using PI to declare a
symbolic name with that value as follows:
DECLARE PI REAL DATA (3.1415927);
An array of constants requires a list of values. For example:
DECLARE FIBONACCI(9) BYTE DATA (0,1,1,2,3,5,8,13,21);
The form and use of the DATA initialization is identical to that of INITIAL except for
the following differences:
•
DATA causes storage to be allocated in the program's constant data segment. The
content and meaning of the name cannot be changed during execution. The
name should never appear on the left-hand side of an assignment statement. This
is not the case with INITIAL.
•
DATA initializations can be used in declarations at any block level in the
program. INITIAL can occur only at the module level, that is, inside the
DO-block that is the module itself, and outside any sub-blocks that the module
may contain.
•
If the keyword DATA is used in a PUBLIC declaration when compiling with the
ROM option, DATA must also be used in the EXTERNAL declaration of program
modules that reference it. However, no value-list can be used since the data
is defined elsewhere. INITIAL cannot be combined with EXTERNAL.
•
Use of the AT attribute forces a name to be associated with a specific memory
location, which can defeat the purpose of the DATA initialization. This will not
happen with INITIAL unless the variables and locations are explicitly redefined
using multiple ATs.
•
If the first declaration has a data initialization, then the variable that is AT that
location is also referred to as DATA, i.e., cannot have a value assigned into it.
PL/M-386 Programmer's Guide
Chapter 3
25
Types of Declaration Statements
Compilation Constants (Text Substitution):
The Use of LITERALLY
If the program is large enough to have many declarations, declaring a compilation
constant will save time at the keyboard, as follows:
DECLARE DCL LITERALLY 'DECLARE';
Thereafter, during compilation, every time DCL appears alone (not as part of a word),
the full string DECLARE will be substituted by the compiler. Subsequent declarations
can be written as follows:
DCL AREA REAL;
DCL SIZE WORD;
A declaration using the reserved word LITERALLY defines a parameterless macro for
expansion at compile-time. Declare an identifier to represent a character string,
which will then be substituted for each occurrence of the identifier in subsequent text.
This expansion will not take place in strings or constants. The form of the
declaration is:
DECLARE identifier LITERALLY 'string';
Where:
identifier
is any valid PL/M identifier.
string
is a sequence of arbitrary characters (limited by the size of the
symbol table) from the PL/M set (except an apostrophe).
An apostrophe can be included in a string by writing it as two consecutive
apostrophes.
26
Chapter 3
Data Declarations, Types, and Based Variables
The following example illustrates another use of LITERALLY:
DECLARE TRUE LITERALLY '0FFH', FALSE LITERALLY '0';
DECLARE ROUGH BYTE;
DECLARE (X, Y, DELTA, FINAL) REAL;
. . .
ROUGH = TRUE;
DO WHILE ROUGH;
X = SMOOTH (X, Y, DELTA);
/* SMOOTH is a procedure declared elsewhere. */
IF (X-FINAL) < DELTA THEN
ROUGH = FALSE;
END;
. . .
This example of a LITERALLY declaration defines the Boolean values TRUE and
FALSE in a manner consistent with the way PL/M handles relational operators (see
Chapter 5). Literal substitution for fixed values makes a program more readable.
LITERALLYs can also be used to declare quantities that are fixed for one
compilation, but are subject to change from one compilation to the next. Consider
the following example:
DECLARE BUFFER$SIZE LITERALLY '32';
DECLARE PRINT$BUFFER(BUFFER$SIZE) WORD;
. . .
PRINT$BUFFER(BUFFER$SIZE - 10) = 'G';
. . .
A future change to BUFFER$SIZE can be made in one place, at the first declaration,
and the compiler will propagate the change throughout the program during
compilation. This eliminates the need to search the program for the occurrences of
32 that are BUFFER$SIZE references and not some other reference to 32.
PL/M-386 Programmer's Guide
Chapter 3
27
Declarations of Names for Labels
A label marks the location of an instruction. Labels are permitted only on executable
statements, not on declarations.
A name can be declared as a label both explicitly and implicitly. Explicit label
declarations are used mainly to enable module-to-module references (see Chapter 7).
The three explicit label declarations have the following formats:
DECLARE PART3 LABEL;
DECLARE START1 LABEL PUBLIC; /* for intermodule reference */
DECLARE PHASE2 LABEL EXTERNAL; /* for intermodule reference */
The rules for explicit label declarations are discussed in detail in Chapter 7.
In implicit label declarations (used more commonly than explicit label declarations),
the name is placed at the very beginning of the executable statement to which the
name is supposed to point. For example:
START2: ALPHA = 127;
This statement defines the label START2 as pointing to the location of the PL/M
instruction shown. If this block has no explicit declaration of START2, such as the
following:
DECLARE START2 LABEL;
then the compiler takes the definition of START2 as an implicit declaration as well as
a definition, as if the declaration had occurred at the start of the last simple DO or
procedure statement. If there is an explicit declaration, then the actual placement of
the label remains simply a definition.
Labels are used to indicate significant instructions or the starting point of instruction
sequences. Labels can be useful reference points for understanding the parts of a
program, or targets for the transfer of control during execution (as discussed under
GOTO and CALL in Chapter 6).
Results of Label Declarations
Valid label declarations result in the following:
28
•
The declared name can be used to point to an executable instruction.
•
The use of the declared name as a variable in its block is disallowed.
•
If the label is also defined in its block by appearing in an executable statement,
the address of that statement will be assigned as the value of the label.
Chapter 3
Data Declarations, Types, and Based Variables
Declaration for Procedures
To declare a procedure, give its name with a statement of the form:
name: PROCEDURE
followed optionally by parameters, type and/or attributes. The definition of the
procedure then follows. The procedure definition is the set of statements declaring
items used in the procedure (including any parameters) and the executable statements
of the procedure itself. The definition ends with an END statement, optionally
including the procedure name.
The complete declaration of a procedure includes all the statements from the
PROCEDURE statement through the END statement. This definition/declaration must
appear before the procedure name is used in an executable statement, just as variable
and constant names must be declared before their use.
The only exceptions arise when the full definition may appear in another separately
compiled module where it is declared PUBLIC, or when a procedure has been
declared REENTRANT. A PUBLIC procedure can be used (called) only if the calling
module meets the following requirements:
1.
The procedure has been declared with the EXTERNAL attribute (so the linker or
binder will search for it).
2.
Each formal parameter the procedure uses has been declared so the compiler can
verify correct usage when this module invokes the procedure. End this local
declaration with an END statement.
For example:
SUMMER: PROCEDURE (A, B) EXTERNAL;
DECLARE A WORD, B BYTE;
END SUMMER;
See Chapter 7 for details on intermodule references. See Chapter 8 for details on
procedure definition and use.
PL/M-386 Programmer's Guide
Chapter 3
29
Data Types
Data types apply not only to variables, but to every value processed by a PL/M
program. This includes values returned by procedures as well as values calculated by
processing expressions. Data type specifications determine the value an object can
have, how this value is stored in memory, and the operations that can be used on the
value.
The PL/M-386 compiler recognizes five classes of data, each of which has one or
more data types.
There are several unsigned binary number types: BYTE (8-bit number), HWORD
(16-bit number), WORD (32-bit number), and DWORD (64-bit number). The OFFSET
type is a 32-bit number that represents the offset portion of a pointer, which has its
own type: POINTER. (The POINTER type itself is also recognized.) Note that the
compiler controls WORD32 and WORD16 automate mapping 32- and 16-bit types.
These controls are discussed in Chapter 11.
There are four signed integer data types: INTEGER (32-bit number); CHARINT (8-bit
number); SHORTINT (16-bit number).
PL/M-386 recognizes the floating-point data type REAL, for signed 32-bit numbers.
Throughout this manual, the data types are referenced according to the data type
class. Table 3-2 summarizes the data type classes for the Intel386 and Intel486
microprocessors. See the sections at the end of this chapter for a discussion on the
PL/M-386 compiler's WORD32|WORD16 mapping.
✏
30
Note
Although the PL/M-386 compiler assumes a 32-bit word it also
accepts PL/M-286 code as input. PL/M-286 code can take
advantage of the 32-bit data type provided by the Intel386 and
Intel486 microprocessors when compiled with the PL/M-386
compiler.
Chapter 3
Data Declarations, Types, and Based Variables
Table 3-2. Data Types
Data Type and Value
Unsigned Binary
Number
Description
BYTE
8-bit number ranging from 0 to 255.
Occupies one byte of memory.
HWORD
Occupies two contiguous bytes of memory.
The least significant 8 bits are stored in the lower address.
WORD
32-bit number ranging from 0 to 4,294,967,295.
Occupies two contiguous HWORDs of memory.
The least significant 16 bits are stored in the lower address.
DWORD
64-bit number ranging from 0 to (2**64) -1.
Occupies two contiguous WORDs of memory.
The least significant 32 bits are stored in the lower address.
OFFSET
32-bit number that represents the offset portion of a POINTER.
(ADDRESS supported by PL/M-80 and PL/M-86/286, is equivalent to
OFFSET.)
SHORTINT
16-bit number from -32768 to +32767 occupies contiguous bytes of
memory. The least significant 8 bits are stored in the low address.
Internally stored in two's complement notation.
INTEGER
32-bit number ranging from -2,147,483,648 to +2,147,483,647.
Occupies four contiguous bytes of memory. The least significant 16
bits are stored in the low address. Internally stored in two's
complement notation. WORD32's LONGINT is equivalent to
INTEGER.
CHARINT
8-bit number ranging from -128 to +127. Occupies one byte of
memory. Internally stored in two's complement notation.
Real Numbers
Description
REAL
Signed, floating-point number. Occupies four contiguous bytes of
memory.
Pointers
Description
POINTER
The value is the address of the memory storage location.
Consists of a segment selector portion and an offset portion.
Selectors
Description
SELECTOR
The value is equivalent to the segment selector portion of a POINTER.
Can be used as the base of a based variable.
PL/M-386 Programmer's Guide
Chapter 3
31
Unsigned Binary Number Variables: Unsigned Arithmetic
Unsigned arithmetic is used to perform any arithmetic operation on unsigned binary
number variables. All of the PL/M operators can be used with these data types.
Arithmetic and logical operations on such variables yield a result of one of the
unsigned binary number types, depending on the operation and the operands.
Relational operations always yield a true or false result of type BYTE.
With unsigned arithmetic, if a large value is subtracted from a smaller one, the result
is the two's complement of the absolute difference between the two values. For
example, if a BYTE value of 1 (00000001 binary) is subtracted from a BYTE value of
0 (00000000 binary), the result is a BYTE value of 255 (11111111 binary).
Also, the result of a division operation is always truncated (rounded down) to a whole
number. For example, if an HWORD value of 7 (0000000000000111 binary) is divided
by a BYTE value of 2 (00000010 binary), the result is an HWORD value of 3
(0000000000000011 binary).
When declaring a variable that may be used to hold or produce a negative result, it is
advisable to make the variable either a signed integer or real. If the variable is
supposed to hold or produce a non-integer, it must be declared as REAL. Use of the
appropriate data types will reduce the occurrences of incorrect results from arithmetic
operations (see Chapter 5).
32
Chapter 3
Data Declarations, Types, and Based Variables
INTEGER Variables: Signed Arithmetic
The sign bit is 0 if the INTEGER value is positive or zero, and 1 if the value is
negative. The magnitude is given in two's complement notation.
Signed Arithmetic
For the Intel386 and Intel486 microprocessors, arithmetic operations on signed
variables use 32-bit signed arithmetic to hold signed intermediate or final results.
Thus, addition and subtraction always produce mathematically correct results if
overflow does not occur. (See also the OVERFLOW control in Chapter 11.)
Relational operations are signed arithmetic comparisons that yield a true or false
result of type BYTE.
However, as with unsigned binary number operands, division produces only an
INTEGER result. The result is rounded toward zero (i.e., down if the result is
positive, up if the result is negative).
Only the arithmetic and relational operators can be used with signed operands.
Logical operators are not allowed except for constant expressions within cast
parentheses (see Chapter 5).
REAL Variables: Floating-point Arithmetic
The value of a REAL variable is a signed floating-point number that occupies four
contiguous bytes of memory, which may be viewed as 32 contiguous bits in the
single precision format. The bits are divided into fields as follows:
SIGN
SIGNIFICAND
EXPONENT
31 30
24 23
16 15
8 7
0
OSD567
The byte with the lowest address contains the least significant 8 bits of the
significand, and the byte with the highest address contains the sign bit and the most
significant 7 bits of the exponent field.
The sign bit is 0 if the REAL value is positive or zero, and 1 if the REAL value is
negative.
The exponent field contains a value offset by 127. In other words, the actual
exponent can be obtained from the exponent field value by subtracting 127. This
field is all 0s if the REAL value is zero.
PL/M-386 Programmer's Guide
Chapter 3
33
The significand contains the binary digits of the fractional part of the REAL value
when this part is represented in binary scientific notation. This field is all 0s if the
REAL value is zero.
Operations on REAL operands use signed floating-point arithmetic to yield a result of
type REAL. The implementation guarantees that the result of each operation is the
closest floating-point number to the mathematical real-number result (if overflow or
underflow does not occur). The relational operators and the arithmetic operators +, -,
*, and / can be used with REAL operands: the MOD operator and the logical operators
are not allowed. Arithmetic operations yield a result of type REAL and relational
operations yield a true or false result of type BYTE.
The PL/M compiler extends the utility of the REAL data types by holding
intermediate results in the numeric coprocessor's temporary-real format (80-bit).
This format preserves 64 bits of precision and the full range of representable
numbers. The exponent in this format is 15 bits instead of 8 in the single precision
format.
The increased exponent range greatly reduces the likelihood of underflow and
overflow, and eliminates roundoff as a source of error until the final assignment of
the result is performed. Underflow, overflow, and roundoff errors are probable for
intermediate computations as well as in the final result. For example, an intermediate
underflow result might later be multiplied by a very large factor, providing a final
result of acceptable magnitude.
Examples of Binary Scientific Notation
1.
Consider the following binary number (which is equivalent to the decimal value
10.25):
1010.01B
The dot (.) in this number is a binary point. The same number can be
represented as:
1.01001B * 2**3
This is binary scientific notation, with the binary point immediately to the right
of the most significant digit. The digits 01001 are the fractional part, and 3 is the
exponent. This value would be represented in the single precision format as
follows:
34
•
The sign bit would be 0, because the value is positive.
•
The exponent field would contain the binary equivalent of 127+3=130.
•
The leftmost digits of the fraction field would be 01001, and the remainder
of this field would be all 0s.
Chapter 3
Data Declarations, Types, and Based Variables
The complete 32-bit representation would be:
0 10000010 010010 00000000000000000
and the contents of the four contiguous memory bytes would be as follows:
highest address:
lowest address:
01000001
00100100
00000000
00000000
Note that the most significant digit is not actually represented, because by
definition it is a 1 unless the REAL value is zero. If the REAL value is zero,
the entire 32-bit representation is all 0s.
2.
Consider the fraction 1/16, or 0.0625. In binary, it is:
1.0000B * 2**(-4)
In single precision format, the actual exponent -4 would be represented as 123
(127-4), and the fraction field would contain all 0s.
In the single precision format, the largest possible value for a valid exponent
field is 254, which corresponds to an actual exponent of 127. Therefore, the
largest possible absolute value for a positive or negative REAL value is:
1.11111111111111111111111B * 2**127
or approximately 3.37 * 10**38.
The lowest permissible exponent field value for a non-zero REAL value is 1,
which corresponds to an actual exponent of -126. Therefore, the smallest
possible absolute value for a positive or negative REAL value is:
1.0B * 2**(-126)
or approximately 8.43 * 10**(-37).
PL/M-386 Programmer's Guide
Chapter 3
35
POINTER Variables and Location References
The value of a POINTER variable is the address of the microprocessor's storage
location and consists of a segment selector portion (see Chapter 9) and an offset
portion.
The bits are divided as follows:
SELECTOR
INDEX
40 39
47
T
RPL
I
34 33 32 31
OFFSET
24 23
0
OSD577
POINTER variables are important as bases for based variables.
Only the relational operators for equality and inequality (= or < >) can be used with
POINTER operands, yielding a true or false result of type BYTE. No arithmetic or
logical operations are allowed (see Chapter 5).
A POINTER can be viewed as a structure of SELECTOR and OFFSET rather than a
scalar. Therefore, arithmetic with POINTERS (e.g., PTR+1) is illegal.
The value of a POINTER variable can be created or changed in the following ways:
36
•
The variable can be initialized when declared, using INITIAL or DATA with an
address created with .
•
The variable can be assigned an address created via the @ operator (described in
the following section). This is the most commonly used method.
•
The variable can be assigned the value of a POINTER variable or function
(including NIL, described in Chapter 9).
•
The variable can be assigned a value generated by the BUILD$PTR function (also
described in Chapter 9).
•
POINTER type conversion (cast). Changing from one value to another is
different from the POINTER built-in function (see Chapter 9).
•
In SMALL RAM model, the POINTER is actually the offset portion only. In this
case, all operations on the PL/M-386 OFFSET data type can be used, including
arithmetic.
Chapter 3
Data Declarations, Types, and Based Variables
The @ Operator
A location reference is formed with the @ operator. A location reference has a value
of type POINTER, that is, a location address. An important use of location references
is to supply values for POINTER variables.
The basic form of a location reference is as follows:
@ variable-ref
Where:
variable-ref is the name of a variable.
The value of this location reference is the actual run-time location of the variable.
The variable-ref may also refer to an unqualified array or structure name. The
pointer value is the location of the first element or member of the array or structure.
Consider the following declarations:
DECLARE RESULT REAL;
DECLARE XNUM(100) BYTE;
DECLARE RECORD STRUCTURE (KEY BYTE,
INFO(25) BYTE,
HEAD POINTER);
DECLARE LIST(128) STRUCTURE (KEY BYTE,
INFO(25) BYTE,
HEAD POINTER);
The @RESULT is the location of the REAL scalar RESULT, and @XNUM(5) is the
location of the 6th element of the array XNUM. @XNUM is the location of the beginning
of the array, that is, the location of the first element (element 0).
The RECORD STRUCTURE declares a byte called KEY followed by 25 bytes called
INFO(0), INFO(1), and so on, followed by the POINTER variable named HEAD.
Because KEY, INFO, and HEAD are all declared part of the RECORD structure, their
contents must be referred to as RECORD.KEY, RECORD.INFO(0), . . . ,
RECORD.INFO(24), and RECORD.HEAD.
Refer to the addresses of these elements of the RECORD structure by using the @
operator. @RECORD.HEAD is the location of the POINTER scalar RECORD.HEAD and
@RECORD is the location of the structure, which is the same as that of the BYTE scalar
RECORD.KEY. @RECORD.INFO is the location of the first element of the 25-byte
array RECORD.INFO, whereas @RECORD.INFO(7) is the location of the 8th element
of the same array.
PL/M-386 Programmer's Guide
Chapter 3
37
LIST is an array of structures. The location reference @LIST(5).KEY is the location
of the scalar LIST(5).KEY. Note that @LIST.KEY is illegal because it does not
identify a unique location (i.e., the KEY of which LIST).
The location reference @LIST(0).INFO(6) is the location of the scalar
LIST(0).INFO(6). Also, @LIST(0).INFO is the location of the first element of
the same array (i.e., the location of the array itself).
A special case exists when the identifier used as variable-ref is the name of a
procedure. This procedure must be declared at the outer level of the program
module. No actual parameters can be given (even if the procedure declaration
includes formal parameters). The value of the location reference in this case is the
location of the entry point of the procedure. (See Chapter 8 and Appendices F
and G.)
Storing Strings and Constants via Location References
Another form of location reference is the following:
@(constant list)
Where:
constant list
is a sequence of one or more BYTE constants separated by commas and
enclosed by parentheses.
When this type of location reference is made, space is allocated for the constants.
The constants are stored in this space (contiguously, in the order given by the list),
and the value of the location reference is the location of the first constant. If RAM is
specified on the compiler invocation command, constants are stored in the DATA
segment. If ROM is specified on the compiler invocation command, constants are
placed in the CODE segment (see Chapters 11 and 13).
Values in the constant list are treated as if they were in a BYTE array initialization
list.
Strings can be included in the list. For example, if the operand:
@('NEXT VALUE')
appears in an expression, it causes the string 'NEXT VALUE' to be stored in memory
(one character per byte, thus occupying 10 contiguous bytes of storage). The value of
the operand is the location of the first of these bytes; in other words, it is a pointer to
the string.
38
Chapter 3
Data Declarations, Types, and Based Variables
OFFSET Data Type and the Dot Operator
A dot operator is provided for compatibility with PL/M-80 programs. The dot
operator (.) is similar to the @ operator, but produces an address of type WORD. This
address represents an offset in the current data segment (for variables) or in the
current code segment (for procedures). Use this address with caution, because it can
produce unexpected results in a PL/M program that contains more than one data
segment or more than one code segment.
In a PL/M-386 program, wherever WORD can be used, OFFSET can also be used. The
main difference between the two types is in casting.
To create or change the value of an OFFSET variable, it can be assigned an OFFSET
variable or function, or assigned the result of the built-in function OFFSET$OF, or
OFFSET type conversion, or the dot operator (see Chapter 9).
SELECTOR Variables
The value of a SELECTOR variable is equivalent to the segment selector portion of a
POINTER, and can also be used as the base of a based variable.
In PL/M-386, the bits of the SELECTOR portion of a POINTER are shown below:
T
RPL
I
INDEX
47
41 40
34 33 32
OSD578
The sections of this diagram are discussed in detail in Chapter 10.
Only the logical and relational operators for equality and inequality (=, <, > and <>)
can be used with SELECTOR operands, yielding a true or false result of type BYTE.
No arithmetic operations are allowed (see Chapter 5).
To create or change the value of a SELECTOR variable it can be assigned a SELECTOR
variable or function, or assigned the result of the built-in function SELECTOR$OF or
SELECTOR type conversion (see Chapter 9).
The results of the @ and dot operators cannot be assigned directly to SELECTOR
variables. They must first be converted to the SELECTOR type with the built-in
functions SELECTOR$OF and SELECTOR.
PL/M-386 Programmer's Guide
Chapter 3
39
Based Variables
Sometimes, the address of a variable is not known until the program is actually run.
For instance, if a procedure is written to swap two bytes and this procedure is called
from various places in the code, the addresses of the two bytes are not known when
writing the procedure definition.
For this type of manipulation, PL/M uses based variables. A based variable is one
that is pointed to by another variable called its base. This means the base contains the
address of the desired (based) variable. A variable is made BASED by inserting in
its declaration the word BASED and the identifier of the base (which must already
have been declared).
A based variable is not allocated storage by the compiler. At different times during
program execution the based variable may actually refer to different places in
memory, because the variable's base may be changed by the program.
To declare an address based variable, first declare its base, which must be of type
POINTER, SELECTOR, WORD, or OFFSET. Next, declare the based variable itself as
follows:
DECLARE I BYTE;
DECLARE ITEM$PTR POINTER;
DECLARE ITEM BASED ITEM$PTR BYTE;
In these declarations, a reference to ITEM is, in effect, a reference to the BYTE value
pointed to by the current value of ITEM$PTR. Thus, the sequence:
ITEM$PTR = @I;
ITEM = 77H;
loads the BYTE value of 77 (hex) into the variable I.
PL/M supports more than one level of based variable, so variables can be based on
based variables.
For example, the following declarations are valid:
DECLARE PTR1 POINTER;
DECLARE PTR2 BASED PTR1 POINTER;
DECLARE STR1 BASED PTR2 STRUCTURE (
X REAL,
Y REAL);
40
Chapter 3
Data Declarations, Types, and Based Variables
The following restrictions apply to bases:
•
No initializations are permitted with based variables.
•
The base must be of type POINTER, SELECTOR, WORD, or OFFSET. However,
use a base of type OFFSET or WORD with caution because it does not contain a
full microprocessor address. OFFSET- or WORD-based variables are addressed
relative to the current DS register.
•
The base cannot be subscripted. That is, it cannot be an array element.
The word BASED must immediately follow the name of the based variable in its
declaration, as in the following examples:
DECLARE
DECLARE
DECLARE
DECLARE
(AGE$PTR, INCOME$PTR, RATING$PTR, CATEGORY$PTR) POINTER;
AGE BASED AGE$PTR BYTE;
(INCOME BASED INCOME$PTR, RATING BASED RATING$PTR) WORD;
(CATEGORY BASED CATEGORY$PTR)(100) WORD;
In the first DECLARE statement, the POINTER variables AGE$PTR, INCOME$PTR,
RATING$PTR, and CATEGORY$PTR are declared. They are used as bases in the next
three DECLARE statements.
In the second DECLARE statement, a BYTE variable called AGE is declared. The
declaration implies that whenever AGE is referenced by the running program, its value
will be found at the location given by the current value of the POINTER variable
AGE$PTR.
The third DECLARE statement declares two based variables, both of type WORD.
The fourth DECLARE statement defines a 100-element WORD array called CATEGORY,
based on CATEGORY$PTR. When any element of CATEGORY is referenced at run
time, the current value of CATEGORY$PTR is the location of the array CATEGORY (i.e.,
its first element).
The other elements follow contiguously. The parentheses around the tokens
CATEGORY BASED CATEGORY$PTR make the statement more readable, but are not
required.
✏
Note
Debug information is available for only the first level of indirection
when using variables based on BASED variables.
PL/M-386 Programmer's Guide
Chapter 3
41
Location References and Based Variables
An important use of location references is to supply values for bases. Thus, the @
operator, together with the based variable concept, gives PL/M a very powerful
facility for manipulating pointers.
For example, to refer to the three different REAL variables NORTH$ERROR,
EAST$ERROR, and HEIGHT$ERROR at different times with the single identifier
ERROR, write:
DECLARE (NORTH$ERROR, EAST$ERROR, HEIGHT$ERROR) REAL;
DECLARE ERROR$PTR POINTER;
DECLARE ERROR BASED ERROR$PTR REAL;
. . .
ERROR$PTR = @NORTH$ERROR;
The value of ERROR$PTR is the location of NORTH$ERROR. A reference to ERROR is,
in effect, a reference to NORTH$ERROR. Later in the program, write:
ERROR$PTR = @HEIGHT$ERROR;
Now a reference to ERROR is, in effect, a reference to HEIGHT$ERROR. In the same
way, the value of the pointer can be made the location of EAST$ERROR, and a
reference to ERROR can be made a reference to EAST$ERROR.
This technique is useful for manipulating complicated data structures and for passing
locations to procedures as parameters. Examples are given in Chapter 8.
42
Chapter 3
Data Declarations, Types, and Based Variables
The AT Attribute
The AT attribute causes the address of a variable to be the specified location. The AT
attribute has the form:
AT (location)
Where:
location
must be a location reference formed with the @ operator.
AT must refer to a nonbased variable that has already been declared. If there is a
subscript expression, it must be a constant expression containing no operators except
+ and -.
The following are examples of valid AT attributes:
AT (@BUFFER)
AT (@BUFFER(128) )
AT (@NAMES(INDEX + 1) )
In the last example, INDEX represents a whole-number constant that has been
previously declared with a LITERALLY declaration. The compiler replaces this name
with the declared whole-number constant, thus satisfying the restrictions previously
mentioned.
The first nonbased variable in a factored declaration containing the AT attribute will
have the address specified by location. Other variables in the same declaration
will, in sequence, refer to successive locations thereafter.
For example, the declaration:
DECLARE (CHAR$A, CHAR$B, CHAR$C) BYTE AT (@BUFFER);
causes the BYTE variable CHAR$A to refer to the location of BUFFER. The variables
CHAR$B and CHAR$C are located in the next two bytes after CHAR$A.
The declaration:
DECLARE T(10) STRUCTURE
(X(3) BYTE,
Y(3) BYTE,
Z(3) BYTE) AT (@DATA$BUFFER);
sets up structure references to 90 bytes. They are organized so that each of the 10
members of T refers to nine bytes. The first three use the name X, the second three Y,
and the last three Z. Figure 3-1 illustrates this structure.
PL/M-386 Programmer's Guide
Chapter 3
43
T(0).X(0)
T(0).X(1)
T(0).X(2)
T(0).Y(0)
T(0).Y(1)
T(0).Y(2)
T(0).Z(0)
T(0).Z(1)
T(0).Z(2)
T(1).X(0)
T(1).X(1)
T(1).X(2)
T(1).Y(0)
T(1).Y(1)
And So On
OSD533
Figure 3-1. Successive Byte References of a Structure
44
Chapter 3
Data Declarations, Types, and Based Variables
The preceding declaration, using the AT attribute, causes the beginning of the
structure T, namely the scalar T(0).X(0), to be located at the same location as a
previously declared variable called DATA$BUFFER. The other scalars making up the
structure will follow this location in logical order: T(0).X(1), T(0).X(2), and so
on up to T(9).Z(2), which is the last scalar, located in the 89th byte after the
location of DATA$BUFFER.
However, no memory locations for these 90 scalars are allocated by this declaration.
You determine the contents of the memory space beginning at @DATA$BUFFER.
The following rules apply to the AT attribute:
•
AT cannot be used with variables that are based, EXTERNAL, or parameters.
•
AT can be used with the PUBLIC attribute, if it immediately follows the word
PUBLIC. However, the location cannot be a location reference to a variable that
is EXTERNAL.
The AT attribute can be used to make variables equivalent, providing more than one
way of referring to the same information. For example:
DECLARE DATUM HWORD;
DECLARE ITEM BYTE AT (@DATUM);
causes ITEM to be declared a BYTE variable at the same location that has just been
allocated for the HWORD variable DATUM. Thus, any reference to ITEM is, in effect, a
reference to the low-order byte of DATUM (because HWORD values are stored with the
low-order 8 bits preceding the high-order 8 bits).
The following is another example using the AT attribute:
DECLARE VECTOR (6) BYTE;
DECLARE SHORT$VECTOR STRUCTURE (FIRST (3) BYTE,
SECOND (3) BYTE)
AT (@VECTOR);
In this example, a six-element BYTE array called VECTOR is declared. Additionally, a
structure of two three-byte arrays, SHORT$VECTOR.FIRST and
SHORT$VECTOR.SECOND, is declared.
The first scalar of this structure, SHORT$VECTOR.FIRST(0), is located at the same
location as the first element of the array VECTOR.
Thus, there are two ways to refer to the same six bytes. For example, the fifth byte in
the group can be referenced as either VECTOR(4) or SHORT$VECTOR.SECOND(1).
When a variable is declared with the AT attribute, the compiler does not optimize the
machine code generated to access that variable.
PL/M-386 Programmer's Guide
Chapter 3
45
WORD32 | WORD16 Type Mapping
The PL/M-386 compiler supports two primary controls, WORD32 and WORD16, for
unsigned binary number and signed integer data types, which provide some basic data
type and language semantics compatibility for the Intel386 and Intel486 family of
microprocessors. These controls specify the basic WORD size and thus affect the
representation of certain data types. The default for PL/M-386 is WORD32. The
WORD16 control does not specify 16-bit code (a parameter pushed on the stack is still
four bytes), but maps the names of some data types into others. Internally, all
processing is the same (e.g., signed arithmetic is 32-bit for both WORD16 and
WORD32). To accommodate existing 16-bit code where data type representation is
critical, WORD16 can be used to map word size to the convention used in earlier
versions of the PL/M compiler. Table 3-3 lists the data type representation for
WORD32 and WORD16.
Table 3-3. WORD32 | WORD16 Data Type Mapping
✏
46
Unsigned Binary
Number Data Types
WORD32
(default)
WORD16
BYTE
8-bit
8-bit
HWORD
16-bit
8-bit
WORD
32-bit
16-bit
DWORD
64-bit
32-bit
QWORD
64-bit
64-bit
Signed Integer
Data Types
WORD32
WORD16
CHARINT
8-bit
8-bit
SHORTINT
16-bit
8-bit
INTEGER
32-bit
16-bit
LONGINT
32-bit
32-bit
Note
In PL/M-386, ADDRESS is equivalent to the OFFSET data type.
OFFSET is a 32-bit data type that represents the offset portion of a
POINTER. The size of OFFSET is not affected by the
WORD32|WORD16 compiler control.
Chapter 3
Data Declarations, Types, and Based Variables
When writing new PL/M code, or when updating existing PL/M code, it is best to
declare variables used for local addressing (i.e., those that are assigned from or
initialized to the dot operator location references, assigned from the OFFSET$OF
function, or used with the BUILD$PTR function or the STACK$PTR built-in) as
OFFSET (or ADDRESS).
In PL/M-386, WORD is the natural 32-bit data type of the language on which all
operations are available. However, in ASM386 a WORD is 16 bits and a DWORD is 32
bits.
Choosing WORD32 or WORD16
The WORD32|WORD16 compiler control determines how the data types in the source
code are interpreted by the PL/M-386 compiler. See Chapter 11 for a description of
the WORD32|WORD16 control and syntax.
When compiling new PL/M-386 source code, use WORD32 to take full advantage of
the Intel386 or Intel486 microprocessors' features.
When recompiling existing PL/M-86 or PL/M-286 code, consider the source code to
determine which compiler control to use. WORD32 is usually preferable. Use
WORD16 if one of the conditions listed below applies to the source code. Note that the
WORD16 control does not have any effect on the CMPB instruction. This always
remains as a 32-bit instruction.
•
Scalar types are mapped to external data, such as STRUCTUREs defined to
represent data records read from a peripheral device. The format of the data
from the peripheral device will not change regardless of the microprocessor
processing it.
•
Data is overlaid, for example:
DECLARE W HWORD (B1,B2) BYTE AT (@W);
DECLARE P POINTER, B BASED P (2) BYTE, WW BASED P WORD;
In this example, code may depend on the fact that two BYTES overlaying the
HWORD constitute both halves of the WORD completely. Similarly, code can
depend on the fact that the LOW or HIGH of an HWORD returns 8 bits.
•
Loops depend on the size of a WORD type. Operations dependent on a variable
overflow could produce unexpected results.
■■■
PL/M-386 Programmer's Guide
Chapter 3
47
48
Chapter 3
Data Declarations, Types, and Based Variables
Arrays and Structures
4
Arrays
For increased efficiency, it is often desirable to use a single identifier to refer to a
whole group of scalars, and to distinguish the individual scalars by means of a
subscript (i.e., a value enclosed in parentheses). Such a list, in which the scalars are
all the same type, is called an array.
An array is declared by using a dimension specifier. The dimension specifier is a
nonzero whole-number constant enclosed in parentheses. The value of the constant
specifies the number of array elements (individual scalar variables) making up the
array. For example:
DECLARE ITEMS (100) BYTE;
causes the identifier ITEMS to be associated with 100 array elements, each of type
BYTE. One byte of storage is allocated for each of these scalars.
The elements of an array are stored contiguously, with the first element in the lowest
location and the last element in the highest location. No storage is allocated for a
based array, but the elements are considered to be contiguous in memory.
The declaration:
DECLARE (WIDTH, LENGTH, HEIGHT) (100) REAL;
is similar to the following sequence:
DECLARE WIDTH (100) REAL;
DECLARE LENGTH (100) REAL;
DECLARE HEIGHT (100) REAL;
The difference between the two declarations is that contiguous storage is guaranteed
for variables declared in a single parenthesized list, whereas variables declared in
consecutive declarations are not necessarily stored contiguously.
This causes each of the three identifiers, WIDTH, LENGTH, and HEIGHT, to be
associated with 100 array elements of type REAL, so that 300 elements of type REAL
have been declared in all. For each of these scalars, four contiguous bytes of storage
are allocated.
PL/M-386 Programmer's Guide
Chapter 4
49
Subscripted Variables
To refer to a single element of a previously declared array, use the array name
followed by a subscript enclosed in parentheses. This construct is called a
subscripted variable.
For example, as a result of the following DECLARE statement:
DECLARE ITEMS (100) BYTE;
each byte can be referenced as an individual item using ITEMS(0), ITEMS(1),
ITEMS(2), and so on up to ITEMS(99).
Notice that the first element of an array has subscript 0, not 1. Thus, the subscript of
the last element is 1 less than the dimension specifier.
To add the third element of the array ITEMS to the fourth, and store the result in the
fifth, write the PL/M assignment statement as follows:
ITEMS(4) = ITEMS(2) + ITEMS(3);
The subscript of a subscripted variable need not be a whole-number constant. It can
be another variable, or any PL/M expression that yields a BYTE, HWORD, WORD,
OFFSET, SHORTINT, CHARINT, or INTEGER value.
Thus, the construction:
VECTOR(ITEMS(3) + 2)
refers to some element of the array VECTOR. Which element this construction refers
to depends on the expression ITEMS(3) + 2. This value, in turn, depends on the
value stored in ITEMS(3), the fourth element of array ITEMS, at the time when the
reference is processed by the running program. If ITEMS(3) contains the value 5,
then ITEMS(3) + 2 is equal to 7 and the reference is to VECTOR(7), the eighth
element of the array VECTOR.
The following sequence of statements will sum the elements of the 10-element array
NUMBERS by using an index variable named I, which takes values from 0 to 9:
DECLARE SUM BYTE;
DECLARE NUMBERS(10) BYTE;
DECLARE I BYTE;
/* To avoid overflow, */
/* SUM should add up */
/* to less than 255
*/
SUM = 0;
DO I = 0 TO 9;
SUM = SUM + NUMBERS(I);
END;
Subscripted array variables can be used anywhere a variable can be used, including
the left side of an assignment statement if the array elements are of a scalar type.
50
Chapter 4
Arrays and Structures
Structures
Just as an array enables one identifier to refer to a collection of elements of the same
type, a structure enables one identifier to refer to a collection of structure members
that may have different data types. Each member of a structure has a member
identifier.
A structure member can be another structure; these nested structures are described in
the section titled, Nested Structures.
The following is an example of a structure declaration:
DECLARE AIRPLANE STRUCTURE (
SPEED REAL,
ALTITUDE REAL);
This statement declares two REAL scalars, both associated with the identifier
AIRPLANE. Once this declaration has been made, the first scalar can be referred to as
AIRPLANE.SPEED and the second as AIRPLANE.ALTITUDE. These names are also
called the members of this structure.
A structure can have many members (see Appendix B for the correct limit). The
members of a structure are stored contiguously in the order in which they are
specified. (No storage is allocated for a based structure, but the members are
considered to be contiguous in memory.)
Individual structure members cannot be based and cannot have any attributes (see
Chapter 3).
Arrays of Structures
With PL/M, arrays of structures can be created. The following DECLARE statement
creates an array of structures that can be used to store SPEED and ALTITUDE for 20
AIRPLANEs instead of one:
DECLARE AIRPLANE (20) STRUCTURE (
SPEED REAL,
ALTITUDE REAL);
This statement declares 20 structures associated with the array identifier AIRPLANE,
each distinguished by subscripts from 0 to 19. Each of these structures consists of
two REAL scalar members. Thus, storage is allocated for 40 REAL scalars.
To refer to the ALTITUDE of the 17th AIRPLANE, write AIRPLANE(16).ALTITUDE.
Arrays Within Structures
An array can be used as a member of a structure, as follows:
PL/M-386 Programmer's Guide
Chapter 4
51
DECLARE PAYCHECK STRUCTURE (
LAST$NAME(15)BYTE,
FIRST$NAME(15)BYTE,
MI BYTE,
AMOUNT REAL);
This structure consists of two 15-element BYTE arrays, PAYCHECK.LAST$NAME and
PAYCHECK.FIRST$NAME, the BYTE scalar PAYCHECK.MI, and the REAL scalar
PAYCHECK.AMOUNT.
To refer to the fourth element of the array PAYCHECK.LAST$NAME, write
PAYCHECK.LAST$NAME(3).
Arrays of Structures With Arrays Inside the Structures
Given that an array can be made up of structures, and a structure can have arrays as
members, the two constructions can be combined to write:
DECLARE FLOOR (30) STRUCTURE (
OFFICE (55) BYTE);
The identifier FLOOR refers to an array of 30 structures, each of which contains one
array of 55 BYTE scalars. This could be thought of as a 30-by-55 matrix of BYTE
scalars. To reference a particular scalar value (for example, element 46 of structure
25) write FLOOR(24).OFFICE(45). Note that the scalar elements of each OFFICE
array are stored contiguously, and the OFFICE arrays are elements of the FLOOR array
and are stored contiguously.
Alter the preceding PAYCHECK structure declaration to make it an array of structures,
as follows:
DECLARE PAYROLL (100) STRUCTURE (
LAST$NAME(15)BYTE,
FIRST$NAME(15) BYTE,
MI BYTE,
AMOUNT REAL);
This is an array of 100 structures, each of which can be used during program
execution to store the last name, first name, middle initial, and amount of pay for one
employee. LAST$NAME and FIRST$NAME in each structure are 15-byte arrays for
storing the names as character strings.
52
Chapter 4
Arrays and Structures
To refer to the Kth character of the first name of the Nth employee, write:
PAYROLL(N-1).FIRST$NAME(K-1)
where N and K are previously declared variables to which appropriate values have
been assigned. This might be convenient in a routine for printing out payroll
information.
Nested Structures
A member of a structure can also be another structure; this is called a nested
structure.
Nested structures are subject to the same rules as all structures. They can contain
their own member identifiers, whether these are scalars, arrays, or structures.
The following example shows nested structures:
DECLARE EMPLOYEE (100) STRUCTURE (
ID WORD,
NAME STRUCTURE (
LAST$NAME (15) BYTE,
FIRST$NAME (15) BYTE,
MI BYTE),
AGE BYTE,
JOB WORD,
PAY STRUCTURE (
RATE REAL,
OTRATE REAL,
BENEFITS STRUCTURE (
OPTIONS REAL,
CHOSEN BYTE)
)
);
The preceding declaration statement is for an array (named EMPLOYEE) of 100
structures. Each of the 100 elements of EMPLOYEE is a structure with the following
members: a WORD scalar named ID, a nested structure called NAME, a BYTE scalar
named AGE, a WORD scalar named JOB, and a nested structure named PAY.
The NAME structure has two arrays (LAST$NAME and FIRST$NAME) of 15 bytes each
for members, as well as a BYTE scalar named MI.
The PAY structure has two REAL scalars (RATE and OTRATE) for members, as well as
a nested structure named BENEFITS. BENEFITS has the REAL scalar OPTIONS and
the BYTE scalar CHOSEN as members.
PL/M-386 Programmer's Guide
Chapter 4
53
The preceding example contains two levels of nested structures. The structures NAME
and PAY are at the first level of nesting; the structure BENEFITS is at the second level
of nesting. See Appendix B for the maximum limit on nested structures.
References to Arrays and Structures
A variable reference is the use, in program text, of the identifier of a variable that has
been declared. A variable reference can be fully qualified, partially qualified, or
unqualified.
Fully Qualified Variable References
A fully qualified variable reference specifies a single scalar. For example, given the
following declarations:
DECLARE AVERAGE REAL;
DECLARE ITEMS (100) BYTE;
.
.
.
DECLARE RECORD STRUCTURE (
KEY BYTE,
INFO WORD);
DECLARE NODE (25) STRUCTURE (
SUBLIST (100) BYTE,
RANK BYTE);
then AVERAGE, ITEMS(5), RECORD.INFO, and NODE(21).SUBLIST(32) are all
fully qualified variable references. Each refers unambiguously to a single scalar.
Note that qualification can only be applied to variables that have been appropriately
declared. A subscript can only be applied to an identifier that has been declared with
a dimension specifier. A member-identifier can be applied only to an identifier
declared as a structure identifier. The compiler flags violations of these rules as
errors.
54
Chapter 4
Arrays and Structures
Unqualified and Partially Qualified Variable References
Unqualified and partially qualified variable references can be used only in location
references (see Chapter 3) and in the built-in procedures LENGTH, LAST, and SIZE
(see Chapter 9).
An unqualified variable reference is the identifier of a structure or an array, without a
member-identifier or subscript. For example, with the declarations in the previous
section, ITEMS and RECORD are unqualified variable references. An unqualified
variable reference is a reference to the entire array or structure. @ITEMS is the
location of the entire array ITEMS (the location of its first byte). Similarly, @RECORD
is the location of the first byte of the structure RECORD.
A partially qualified variable reference does not refer to a single scalar even using a
subscript and/or member-identifier with an identifier.
For example, in the declaration in the previous section, NODE(15) and
NODE(12).SUBLIST are partially qualified variable references.
When used with the @ operator, partially qualified variable references are taken to
mean the first byte that fits the description. Thus, @NODE(15) is the location of the
first byte of the structure NODE(15), which is an element of the array NODE.
Similarly, @NODE(12).SUBLIST is the location of the first byte of the array
NODE(12).SUBLIST, which is a member of the structure NODE(12), which is an
element of the array NODE.
Because it is ambiguous, @NODE.SUBLIST cannot be used. In a location reference
referring to an array consisting of structures, a subscript must be given before a
member-identifier can be added to the reference. The rule is different for partially
qualified variable references in connection with the built-in procedures LENGTH,
LAST, and SIZE, as explained in Chapter 9.
■■■
PL/M-386 Programmer's Guide
Chapter 4
55
56
Chapter 4
Arrays and Structures
Expressions and Assignments
5
A PL/M expression consists of scalar operands (values) combined by arithmetic,
logical, and relational operators. For example:
A + B
A + B - C
A*B + C/D
A*(B + C) - (D - E)/F
where +, -, *, and / are arithmetic operators for addition, subtraction, multiplication,
and division, and A, B, C, D, E, and F represent operands. The parentheses group
operands and operators to control the order of evaluation.
This chapter describes the rules governing PL/M expressions. Although these rules
may appear complex, most of the expressions used in actual programs are simple. In
particular, when the operands of arithmetic and relational operators are all of the
same type, the resulting expression is easy to understand.
Operands
Operands are the building blocks of expressions. An operand is a quantity with a
value at run time on which an arithmetic, logical, or relational operation is performed
by an operator. In the preceding examples, A, B, C, etc., are identifiers of scalar
variables that have values at run time.
Operands in expressions can also be numeric constants and fully qualified variable
references. The following sections describe all of the types of operands that are
permitted.
PL/M-386 Programmer's Guide
Chapter 5
57
Constants
A numeric constant can be an operand in an expression. However, its type must be
appropriate, as discussed in the following paragraphs.
A numeric constant that contains a decimal point is of type REAL. A numeric
constant that does not contain a decimal point is a whole-number constant.
You can use a whole-number constant in either signed context or unsigned context.
In unsigned context, a whole-number constant is treated as an unsigned binary
number data type. In signed context, a whole-number constant is treated as a signed
integer data type (see Table 3-2).
Whole-number Constants in Unsigned Context
In PL/M-386, if the WORD32 control is in effect, a whole number constant in unsigned
context is treated as follows:
•
As a BYTE value if it ranges from 0 to 255
•
As a HWORD value if it ranges from 256 to 65,535
•
As a WORD value if it ranges from 65,536 to 4,294,967,295 (i.e., 2**32-1)
•
As a DWORD value if it ranges from 2**32 to 2**64-1
Whole-number Constants in Signed Context
In signed context, a whole-number constant is always treated as an INTEGER value.
In PL/M-386, the range is -2,147,483,648 to 2,147,483,647. Additionally, small
integer values are extended into 32-bit values with no change to the arithmetic value.
String Constants
A string constant containing not more than four characters can also be used as an
operand. If a string constant has only one character, it is treated as a BYTE constant
whose value is the 8-bit ASCII code for the character. If a string constant is a
two-character string, it is treated as an HWORD constant in PL/M-386. The value of
the two-character string is formed by stringing together the ASCII codes for the two
characters, with the code for the first character forming the most significant 8 bits of
the 16-bit number.
58
Chapter 5
Expressions and Assignments
In PL/M-386, if the WORD32 control is in effect, a three- or four-character string
constant is treated as a WORD constant whose value is formed by stringing together the
ASCII codes for all of the characters. The first character represents the high 8 bits,
the second character represents the second most significant 8 bits, and so on. If the
string has three characters, the ASCII NUL character is inserted in front of the first
character to form a four-character string.
Strings of more than four characters are illegal as operands in expressions, and can be
used in only two contexts: as initialization values for an array or as part of a location
reference that points to the location at which the string constant is stored (see Chapter
3).
PL/M-386 Programmer's Guide
Chapter 5
59
Variable and Location References
As described in Chapter 4, fully qualified variable references uniquely specify a
single scalar value. (Partially qualified references, also discussed in Chapter 4, have
very restricted uses.) Any fully qualified variable reference can be used as an
operand in an expression. When the expression is evaluated, the reference is replaced
by the value of the scalar.
A function reference is the name of a typed procedure that has been declared
previously, along with any parameters required by the procedure declaration. The
value of a function reference is the value returned by the procedure.
For example, in the statement:
I = J + ABS(L);
the absolute value of L will be returned by the function ABS and then added to the
value of J before being stored in I. If L is -27, the result will be the same as writing:
I = J + 27;
For a complete discussion of procedure and function references, see Chapter 8.
Location references are described in Chapter 3.
Subexpressions
A subexpression is an expression enclosed in parentheses, which can be used as an
operand in an expression. A subexpression can be used to group portions of an
expression together, just as in ordinary algebraic notation.
Compound Operands
All the operand types previously described are primary operands. An operand can
also be a value calculated by evaluating some portion of the total expression. For
example, in the expression:
A + B*C
(where A, B, and C are variable references), the operands of the * operator are B and
C. The operands of the + operator are A, and the result of the compound operand B *
C. Notice that this expression is evaluated as if it had been written as follows:
A + (B * C)
This analysis of an expression to determine which operands belong to which
operators, and which groups of operators and operands form compound operands, is
discussed in Expression Evaluation. Table 5-1 lists operator precedence.
60
Chapter 5
Expressions and Assignments
Table 5-1. Operator Precedence
Operator Class
Operator
Interpretation
Precedence
()
Controls order of evaluation: expressions within
parentheses are evaluated before the action of any
outside operator on the parenthesized items
Unary
+, -
Single positive operator, single negative operator
Arithmetic
*, /, MOD
+, -
Multiplication, division, modulo (remainder) division,
addition, subtraction
Relational
<, <= , <>
>=, >
Less than, less than or equal to, not equal to, =,
equals, greater than or equal to, greater than
Logical
NOT
AND
OR, XOR
Logical negation
Logical conjunction
Logical inclusion disjunction, logical exclusive
disjunction
Arithmetic Operators
PL/M has the following five principal arithmetic operators:
+ - * / MOD
These operators are used as in ordinary algebra to combine two operands. Each
operand can have an unsigned binary number data type value; a signed integer data
type value; or a REAL number data type value (except that REAL operands cannot be
used with the MOD operator).
Arithmetic operations cannot be used with POINTER and SELECTOR variables.
The +, -, *, and / Operators
The operators +, -, *, and / perform addition, subtraction, multiplication, and division
on operands of any data type except the POINTER and SELECTOR data types. The
following rules govern these operations.
•
Both operands must be of the same class (i.e., both operands must be unsigned,
signed, or real). Mixing operands of different classes is illegal. However, an
operand of one class can be converted, in an expression, to another class with the
use of a built-in conversion function (see Chapter 9).
PL/M-386 Programmer's Guide
Chapter 5
61
•
Unsigned Arithmetic
–
Unsigned Addition and Subtraction
If both operands are of the same data type, the result is of the same data type
(e.g., BYTE + or - BYTE produces a BYTE result).
If the operands are of different data types, the smaller operand is extended
with high-order 0 bits to the size of the larger operand; the addition or
subtraction is then performed as though both operands are of the same type.
For example, for BYTE + or - WORD, the BYTE operand is zero-extended by 8
bits to WORD size; then the operation is performed with the WORD operands to
produce a WORD result. For a BYTE + or - DWORD the BYTE operand is
zero-extended by 24 bits to DWORD size; then the operation is performed with
the DWORD operands to produce a DWORD result. For WORD + or - DWORD, the
WORD operand is zero-extended by 16 bits to DWORD size; then the operation
is performed with two DWORD operands producing a DWORD result.
–
Unsigned Multiplication and Division
Assuming WORD32, if both operands are of type BYTE, the * and / operations
produce an HWORD result; if both operands are of type HWORD, the * and /
operations produce a WORD result. If both operands are of type WORD, the *
and / operations produce a WORD result. If both operands are of type
OFFSET, the * and / operations produce an OFFSET result. If both operands
are of type DWORD, the * and / operations produce a DWORD result.
For mixed unsigned operands, the same rules as for addition and subtraction
apply. The smaller operand is zero-extended to the size of the larger
operand, then the multiplication or division is performed as though both
operands are of the same type. The results are as described in the preceding
paragraph.
If one operand is a whole-number constant or a string constant, it is treated
as a HWORD or WORD depending on its value (see Whole-number Constants in
Unsigned Context and String Constants).
All arithmetic for signed operands is 32-bit signed integer arithmetic. The
names of the storage type (e.g., CHARINT) do not imply what type of
arithmetic is performed, only the size of storage assigned for the variable.
62
Chapter 5
Expressions and Assignments
During signed arithmetic an expression can overflow only if it overflows the
machine word (32 bits). However, overflow is possible when the value is
assigned to a variable with SHORTINT or CHARINT data type. If the value is
assigned to CHARINT, 24 high-order bits are truncated to form the CHARINT
value. If the value is assigned to SHORTINT, 16 high-order bits are
truncated to form the SHORTINT value. If the value is assigned to INTEGER,
it is not changed.
Assignment overflow is detected using the OVERFLOW control (see
Chapter 11).
Constants are always represented as integer constants, regardless of their
value.
–
Real arithmetic
Both operands are always of type REAL. Thus, the +, -, *, and / operations
produce a result of type REAL.
If one operand is a constant, it must be typed as a floating-point constant,
that is, it must have a decimal point. Mixing REAL operands with
whole-number constants is not allowed. For example, if R is a REAL
variable, R+1.0 is a legal expression, but R+1 is illegal. Also, 1.0+1 is
illegal, because it mixes a REAL constant with a whole-number constant.
–
Arithmetic expressions containing operands of type SELECTOR or POINTER
are illegal.
–
If both operands are whole-number constants, the operation depends on the
context in which it occurs, as explained in Special Case: Constant
Expressions.
–
The result of division by 0 is undefined, except for REAL values (see
Appendix G).
A unary - operator is also defined in PL/M. It takes a single operand, to which it is
prefixed. A minus sign that has no operand to the left of it is taken to be a unary
minus.
A unary - operator makes (-A) equivalent to (0-A), where A is any operand. The 0 is
a BYTE value if A is an unsigned binary number data type. The 0 is an INTEGER
value if A is a signed integer data type; or a REAL value if A is a real number data
type. If A is a whole-number constant, its type and the unary - operation depend on
the context as explained in Special Case: Constant Expressions. In unsigned context,
(-1) is assigned a BYTE value (0-1) which is equivalent to 0FFH. In signed context,
(-1) is assigned an INTEGER value (0-1) which is equivalent to 0FFFFFFFFH for
PL/M-386.
Finally, a unary + has no effect; (+A) is equivalent to (A).
PL/M-386 Programmer's Guide
Chapter 5
63
The MOD Operator
MOD performs division, except the result is not the quotient, but rather the remainder
left after integer division. The result has the same sign as the operand on the left side
of the MOD operator.
REAL operands cannot be used with the MOD operator; only unsigned or signed
operands can be used.
For example, if A and B are INTEGER variables with values of 35 and 16,
respectively, then A MOD B yields an INTEGER result of 3, and -A MOD B yields -3.
Unlike the / operator, the MOD operator must be separated from surrounding letters
and digits by blanks or other separators.
64
Chapter 5
Expressions and Assignments
Relational Operators
The following relational operators are used to compare operands of the same type:
<
>
<=
>=
<>
=
less than
greater than
less than or equal to
greater than or equal to
not equal to
equal to
Relational operators are always binary operators, taking two operands, to yield a
BYTE result. Relational operators can be used with all types.
If both operands are unsigned, then unsigned arithmetic will be used to compare the
two values. As with the arithmetic operators, mixing unsigned data types are
allowed, with the smaller operand being zero-extended to the size of the larger
operand.
Whole-number and string constant operands are treated as BYTE, HWORD, WORD, or
OFFSET. Unsigned data types are primarily used to represent positive values.
Negative numbers are represented by two's complement in the smallest unsigned data
type that can hold the value. For example, -2 is represented as the BYTE value of
0FEH. If B is a BYTE variable, then the relational expression B>=-2 is TRUE only if
B has the value of 254 or 255, because the expression -2 (when evaluated unsigned)
has a BYTE value of 254.
In PL/M-386, if both operands are signed, then signed 32-bit integer arithmetic is
used to compare the two values. CHARINT or SHORTINT are sign extended to
INTEGER values. The calculated value is then assigned to the specified data type.
If both operands are real, floating point arithmetic will be used to compare the two
values. Only floating-point constants (i.e., constants containing a decimal point) can
be mixed with REAL operands.
Two POINTER operands can be compared for equality but greater than, less than, and
inequality operations cannot be used. In PL/M-386 only, two POINTERs are equal
only if they are bitwise equal (i.e., if both segment selector portions are equal and
both offset portions are equal).
Two SELECTOR operands can be compared for equality, inequality, less than, and
greater than.
Since constants cannot be typed as POINTER or SELECTOR, comparison between
POINTER or SELECTOR operands and constants is illegal.
PL/M-386 Programmer's Guide
Chapter 5
65
As with arithmetic operations, operands of different classes cannot be mixed together
in relational operations. An operand of one class can be converted, in an expression,
to another class using a built-in conversion function (see Chapter 9).
If the specified relation between the operands is true, a BYTE value of 0FFH (or
1111$1111B) is returned. Otherwise, the result is a BYTE value of 00H (or
0000$0000B). Thus, in all cases, the result is of type BYTE, with all 8 bits set to 1 for
a true condition, or to 0 for a false condition. For example:
(6>5)
(64<=4)
result is 0FFH (true)
result is 00H (false)
Values of true and false resulting from relational operations are useful in conjunction
with DO WHILE statements and IF statements, as described in Chapter 6. In the
context of a DO WHILE statement or IF statement, only the least significant bit of a
true or false value is used. Thus, each value with the least significant bit set
(including 0FFH) is considered true and each value with the least significant bit 0 is
considered false. A BYTE value is returned.
66
Chapter 5
Expressions and Assignments
Logical Operators
PL/M has the following four logical (Boolean) operators:
NOT AND OR XOR
These operators are used with the unsigned binary number data type, or
whole-number or string constant operands to perform logical operations on 8, 16, or
32 bits.
NOT is a unary operator, taking one operand only. It produces a result of the same
type as its operand: each bit of the result is the one's complement of the
corresponding bit of the original value.
The remaining operators (AND, OR, XOR) each take two operands, and perform bitwise
and, or, and exclusive or, respectively. The bits of an AND result are 1 only when the
corresponding bit in each operand is 1. The bits of an OR result are 1 when the
corresponding bit of either operand is 1, and 0 only when both operands are 0. The
bits of an XOR result are 0 only when the corresponding bits of the operand are the
same (i.e., both 1 or both 0); the result has a 1 when one operand is 1 and the
corresponding bit of the other operand is 0.
When both operands are of the same type, the result is the same type as the operands.
As with the arithmetic and relational operators, unsigned data types can be mixed in
any combination for logical operations. Whole-number operands are treated as BYTE,
HWORD, or WORD values in PL/M-386. The only exception is an expression composed
only of whole numbers within the cast parentheses; then the constants have integer
context and the numbers are extended to the 16-bit signed value. The usual bitwise
logical operation then takes place (as explained above for 16-bit numbers for bitwise
operations). Mixing OFFSET with WORD produces an OFFSET result.
The following are examples of logical operations:
NOT 11001100B
10101010B AND 11001100B
10101010B OR 11001100B
10101010B XOR 11001100B
PL/M-386 Programmer's Guide
result is 00110011B
result is 10001000B
result is 11101110B
result is 01100110B
Chapter 5
67
Note that true and false values resulting from relational operations can be combined
with logical operators:
NOT (6>5)
(6>5) AND (1>2)
(6>5) OR (1>2)
(LIM = Y) XOR (Z<2)
result is 00H (false)
result is 00H (false)
result is 0FFH (true)
result is 0FFH (true) if LIM = Y and Z<2 or if LIM<>Y and
Z<2, but result is 00H (false) if both relations are true or
both false
Note that in the statement:
A = (NOT B)
parentheses must be used as indicated. Failure to do so will result in a syntax error
because relational operators (=) have higher precedence than logical operators (NOT).
The following are examples of whole numbers. In this example the parentheses
enclose items to be converted (casted).
OFFSET(10101010B AND 11001100B) gives OFFSET(010001000B). This is the
16-bit result obtained with the simple logical operation written above, except that the
offset type is returned. PL/M-386 extends the result to 32 bits.
HWORD (-4 AND 7) gives HWORD (0FFFFFFFCH AND 00000007H) which gives
HWORD(4) or the unsigned 16-bit value of 4.
68
Chapter 5
Expressions and Assignments
Expression Evaluation
Precedence of Operators: Analyzing an Expression
In PL/M, operators have an implied order that determines how operands and
operators are grouped and analyzed during compilation.
The PL/M operators are listed in Table 5-1 (page 5-61) from highest to lowest
precedence; those that take effect first are listed first. Operators in the same line are
of equal precedence and are evaluated as they are encountered in a left to right
reading of an expression.
The order of evaluation in an expression is controlled first by parentheses, then by
operator precedence, and finally by left to right order.
The compiler first evaluates operands and operators enclosed in paired parentheses as
subexpressions, working from the innermost to the outermost pairs of parentheses.
The value of the subexpression is then used as an operand in the remainder of the
expression.
Parentheses are also used around both subscripts and the parameters of function or
procedure references. These are not subexpressions, but they too must be evaluated
before the remainder of the expressions or references can be evaluated at a higher
level.
When there is more than one operator in an expression, evaluate the results by
beginning with the operator with the highest precedence. If the operators are of equal
precedence, evaluate them left to right, as follows:
Example
(A + B) * C is not the same as A + B * C
A + B * C means the same as A + (B * C)
A/B * C means the same as (A/B) * C
PL/M-386 Programmer's Guide
Reason
Parentheses form subexpressions
Operator precedence
Left to right, equal precedence
Chapter 5
69
The following are examples of precedence ranking:
A+B*C
A+B-C*D
A+B+C+D
A/B*C/D
A > B AND NOT B<C-1
is equivalent to A + (B * C)
is equivalent to (A + B) - (C * D)
is equivalent to ( (A + B) + C) + D
is equivalent to ( (A / B) * C) / D
is equivalent to (A > B) AND
(NOT (B < (C < 1) ) )
In the last four examples, the application of the left-to-right rule for operators with
the same precedence is shown. In the second, third, and fifth examples, the
left-to-right rule for operators of equal precedence makes no difference in the value
of the expression. But in the fourth example, the left-to-right rule is critical.
The following example shows the action of the rules of precedence on a longer
expression:
(-B + SQRT ( B * B - 4.0 * A * C) )/(2.0 * A)
Assume A, B, and C are variables of type REAL, and SQRT is a procedure of type REAL
which returns the square root of the value passed to it as a parameter. In this case,
the parameter is the expression B * B -4.0 * A * C. Floating point constants
(4.0, 2.0) are used rather than whole-number constants (4,2) because it is invalid
to combine whole-number constants with REAL variables.
The compiler first analyzes the portions of the expressions within the innermost
parentheses, then the procedure parameter and the subexpression 2.0 * A. (The
subexpression is also called a compound operand because its result is used in
evaluating the whole expression.)
In a left-to-right scan, the two operands of the first * operator are both equal to the
value of B. The operands of the second * operator are 4.0 and the value of A. The
operands of the third * operator are the results of the second evaluation (i.e., the
compound operand 4.0 * A) and the value of C. The operands of the fourth *
operator are 2.0 and the value of A.
The subexpression 2.0 * A is now completely analyzed, but the parameter
expression still contains a minus (-) operator that has not been analyzed. The
operands of this operator are the result of evaluating B * B and the result of
evaluating 4.0 * A * C. Once the evaluations are done, the parameter expression
is analyzed and its value can be calculated.
This value does not become an operand in the overall expression. It is passed to the
procedure SQRT, which returns the square root of the parameter. This returned value
then becomes an operand in the remainder of the full original expression:
(-B + returned value) / (2.0 * A)
70
Chapter 5
Expressions and Assignments
Now that the innermost subexpressions have been analyzed and evaluated, a division
operator whose left operand must be evaluated further remains. This outer
subexpression is -B + the returned square root: there are two operators. The first is a
unary minus (-) and its operand is the value of B. The second is the binary plus (+)
operator, with two operands: the value of -B and the value of SQRT(B * B - 4.0
* A * C). -B has the same meaning as 0-B, which is to be added to the known
value of the square root indicated. The final operator is division (/), whose two
operands are known: the value of (-B + SQRT(B * B -4.0 * A * C) ) and the
value of (2.0 * A).
Three important points must be emphasized about expression evaluation, as discussed
in the next three sections.
Compound Operands Have Types
Compound operands have types as do primary operands. All of the primary operands
used in the preceding example were of type REAL, which results in compound
operands of type REAL. It is always valid for all the operands in an arithmetic
expression to be of the same type, and the result will be that type also. Combining
BYTE values can validly create a WORD or HWORD value. Combining a signed integer
data type value always creates an INTEGER value.
In an expression containing mixed data types, any combinations can be used as long
as the types belong to the same class (i.e., unsigned binary number, signed integer,
real, pointer, or selector). Data types (of the same class) can be mixed as operands in
expressions, whether they are constants or variables.
Mixing types of different classes in arithmetic, logical, or relational expressions is
invalid. For example, if F and G are INTEGER variables and H and K are REAL
variables, then the expressions F > K and H + G are invalid.
Due to operator precedence, some combinations can occur validly in the same
expression without being directly combined. In the following logical expression:
(F > G AND H < K)
the subexpression F > G yields a BYTE value, as does the subexpression H < K.
Then the BYTE values are ANDed together. This expression is legal despite an
apparent mixing of types. G and H could not be the operands for two reasons:
1.
The relational operators are of higher precedence than the AND operator.
2.
Only unsigned operands are legal with logical operators.
PL/M-386 Programmer's Guide
Chapter 5
71
Relational Operators Are Restricted
In the absence of parentheses denoting a subexpression, the result of a relational
operation (comparison) cannot become an operand in another relational operation.
Thus, the expression:
A <= X <= B
is invalid in PL/M because the second <= operator would have to use the result of the
first <= operator as one of its operands.
In PL/M the valid expression is as follows:
A <= X AND X <= B
Parentheses also could have created a valid expression; for example:
(A <= X) <= B
However, in this expression the result does not have the desired meaning: A <= X
becomes a byte of value 0 if A is greater than X, 0FFH if A is not greater than X.
Thus, if A is 0, X is 1, and B is 2:
(0 <= 1) <= 2
evaluates to:
(0FFH) <= 2
and yields a FALSE value. This is contrary to the original intention.
Order of Evaluation of Operands
Operators and operands are not bound in the same order as the order in which
operands are evaluated.
The rules of analysis specify which operands are bound to each operator. The
following example shows how operands are bound to operators:
A + B * C
B and C are the operands of the * operator, and A and the value of B * C are the
operands of the + operator. B and C must be evaluated before the * operation can be
performed, and the compound operand B * C must be evaluated before the +
operation is performed.
However, it is not obvious whether B will be evaluated before C or vice versa. A
could be evaluated before either B or C, and its value stored until the + operation is
performed.
72
Chapter 5
Expressions and Assignments
The rules of PL/M do not specify the order in which subexpressions or operands are
evaluated in each statement. This flexibility enables the compiler to optimize the
object code it produces, as described in Chapter 11. In most cases, the order of
evaluation makes no difference. However, certain embedded assignments (see
Assignment Statements) or function references (see Chapter 8) change the value of
an operand in the same expression.
Choice of Arithmetic: Summary of Rules
As described in Chapter 3, PL/M uses three distinct kinds of arithmetic: unsigned,
signed, and floating-point. Whenever an arithmetic or relational operation is carried
out, PL/M uses one of these types of arithmetic, depending on the types of the
operands.
Table 5-2 is a summary of the rules that determine which type of arithmetic is used in
each case. The table also lists the data type of the result for each kind of arithmetic
operation. The notes following the table provide additional information. (see
Relational Operators and Logical Operators for rules governing relational and logical
operations.)
In PL/M-386, OFFSET operands are always 32-bit unsigned operands.
In expressions, whole-number constants are always converted to the value of the
equivalent data type.
PL/M-386 Programmer's Guide
Chapter 5
73
Table 5-2. Summary of Expression Rules for PL/M-386
Variable
Type
Kind of
Arithmetic
Operand Type
Operation
Result
Notes
BYTE
HWORD
Unsigned
BYTE w/BYTE
+ or * / or MOD
BYTE
HWORD
range: 0 to 255
0 to 65,535
HWORD
w/HWORD
+ or * / or MOD
HWORD
WORD
range: 0-65,535
0 to 2**32-1
BYTE w/HWORD
becomes HWORD
w/HWORD
+ or * / or MOD
HWORD
WORD
BYTE is extended with
8 high-order zeros to
an HWORD value
WORD w/WORD
any
arithmetic
WORD
range:
0 to 2**32-1
BYTE w/WORD
becomes
WORD w/WORD
any
arithmetic
WORD
BYTE is extended with
24 high-order zeros
to a WORD value
HWORD w/WORD
becomes
WORD w/WORD
any
arithmetic
WORD
HWORD is extended
with 16 high-order
zeros to a WORD
value
DWORD
w/DWORD
any
arithmetic
DWORD
range:
0-2**63-1
BYTE w/DWORD
becomes DWORD
w/DWORD
any
arithmetic
DWORD
BYTE is extended
with 56 high-order
zeros to a DWORD
value
HWORD
w/DWORD
becomes DWORD
w/DWORD
any
arithmetic
DWORD
HWORD is extended
with 48 high-order
zeros to a DWORD
value
WORD w/DWORD
becomes DWORD
w/DWORD
any
arithmetic
DWORD
WORD is extended
with 32 high-order
zeros to a DWORD
value
WORD
DWORD
continued
74
Chapter 5
Expressions and Assignments
Table 5-2. Summary of Expression Rules for PL/M-386 (continued)
OFFSET
Unsigned
OFFSET
w/OFFSET
any
arithmetic
OFFSET
range:
0 to 2**32-1
BYTE w/OFFSET
becomes OFFSET
w/OFFSET
any
arithmetic
OFFSET
BYTE is extended with
24 high-order zeros to
an OFFSET value
HWORD
w/OFFSET
becomes OFFSET
w/OFFSET
any
arithmetic
OFFSET
HWORD is extended
with 16 high-order
zeros to an OFFSET
value
WORD w/OFFSET
becomes OFFSET
w/OFFSET
any
arithmetic
OFFSET
range: 0 to 2**32-1
OFFSET
w/DWORD
becomes DWORD
w/DWORD
any
arithmetic
DWORD
OFFSET is extended
with 32 high-order
zeros to a DWORD
value
-2**31 to
+2**31-1
CHARINT
SHORTINT
INTEGER
Signed
INTEGER
w/INTEGER
+ or * / or MOD
INTEGER
REAL
Floating
Point
REAL w/REAL
+ - * or /
REAL
POINTER
w/POINTER
=
BYTE
0 or 0FFH
SELECTOR
w/SELECTOR
=, <>,
<, or >
BYTE
0 or 0FFH
POINTER
SELECTOR
Unsigned
Note: CHARINT and SHORTINT are sign extended to INTEGER before expression evaluation.
The combinations of operands shown in Table 5-2 are the only usable combinations
of arithmetic operations and operands. For example, an operand of the signed integer
data type cannot be combined with an operand of the unsigned binary number data
type. However, explicit conversion can be coded in-line using the PL/M built-ins
described in Chapter 9.
PL/M-386 Programmer's Guide
Chapter 5
75
Special Case: Constant Expressions
The rules already given explain expressions like:
A + 3 * B
where there is a single whole-number constant. However, if there is an expression
like:
3 - 5 + A
then the kind of arithmetic that will be used to evaluate 3 - 5 must be considered,
because both operands are whole-number constants.
The answer, in this case, depends on the type of operand A. If A is an unsigned binary
number, then 3 - 5 is considered to be in unsigned context. Unsigned arithmetic is
used to evaluate 3 - 5, giving a BYTE result of 254. Unsigned arithmetic is then
used to add this result to A.
For PL/M-386, if A is a signed integer, then 3 - 5 is in signed context. Signed 32-bit
arithmetic is used to evaluate 3 - 5. Signed 32-bit arithmetic is then used to add
this result to A.
If A is of type REAL, POINTER, or SELECTOR, the expression is illegal.
Any compound operand, subexpression, or expression that contains only
whole-number constants as primary operands is called a constant expression.
Floating-point constants are of type REAL and are treated as the values of REAL
variables.
In this expression:
3 - 5 + 500 + A
3 - 5 is a constant expression that forms part of the larger constant expression
3 - 5 + 500.
If the constant expression is not the entire expression, its value is an operand in the
expression. The context is created by the other operand of the same operator.
In the preceding example, suppose the operand A has a BYTE value. Then the
constant expression 3 - 5 + 500 is in unsigned context. The constants 3 and 5 are
treated as BYTE values, and 500 is treated as a WORD or HWORD value. The operation
3 - 5 gives a BYTE result of 254, and this is extended to a WORD or HWORD value of
254 before adding 500. This results in a WORD or HWORD value of 754. It is exactly as
if the expression had been written as follows:
754 + A
76
Chapter 5
Expressions and Assignments
If A had a SHORTINT value, the constant 3 + 5 - 500 would be in signed context;
signed 32-bit arithmetic is used for the operation 3 - 5 + 500. The result (498) is
added to the value of A to form a 32-bit signed temporary result.
In summary, if the context is created by an unsigned binary number data type
operand, the constant expression is in unsigned context. If the context is created by a
signed integer data type operand, the constant expression is in signed context. Note
that if the context is created by a real number, pointer or selector data type operand,
the constant expression is illegal.
If the constant expression is the entire expression, then it belongs in one of the
categories listed below. For additional information, see Assignment Operators.
•
Constant expression as right-hand part of an assignment statement: context is
created by the variable to which the expression is being assigned.
•
Constant expression as subscript of an array variable: evaluated as if being
assigned to an INTEGER variable.
•
Constant expression in the IF part of an IF statement: evaluated as if being
assigned to a BYTE variable.
•
Constant expression in a DO WHILE statement: evaluated as if being assigned to a
BYTE variable.
•
Constant expression as start, step, or limit expression in an iterative DO
statement: evaluated as if being assigned to a variable of the same type as the
index variable in the same iterative DO statement.
•
Constant expression in a DO CASE statement: evaluated as if being assigned to a
WORD variable.
•
Constant expression as an actual parameter in a CALL statement or function
reference: evaluated as if being assigned to the corresponding formal parameter
in the procedure declaration.
•
Constant expression in a RETURN statement: evaluated as if being assigned to a
variable of the same type as the (typed) procedure that contains the RETURN
statement.
•
Constant expression inside an explicit type conversion (cast built-ins); evaluated
as if being assigned to an INTEGER variable, shorter values are extended to 16
bits or 32 bits. The only exception is that relational operators can be used and
are performed bitwise on 16-bit or 32-bit constant values.
PL/M-386 Programmer's Guide
Chapter 5
77
Assignment Statements
Results of computations can be stored as values of scalar variables. At any given
moment, a scalar variable has only one value; however, this value can change with
program execution. The PL/M assignment statement changes the value of a variable.
Its simplest form is:
variable =expression;
where expression is any PL/M expression, as described in the preceding sections.
This expression is evaluated, and the resulting value is assigned to (that is, stored in)
the variable. This variable can be any fully qualified variable reference except a
function reference. The old value of the variable is lost.
For example, following execution of the statement:
RESULT = A + B;
the variable RESULT will have a new value, calculated by evaluating the expression A
+ B.
Implicit Type Conversions
In an assignment statement, if the type of the value of the right-hand expression is not
the same as the type of the variable on the left side of the equal sign, then either the
assignment is illegal or an implicit type conversion occurs. For PL/M-386, all
unsigned binary number, signed integer and real data type values are converted
automatically. Chapter 9 includes a description of built-in functions that, when
invoked, perform explicit conversions for use in expressions or assignments.
For implicit type conversions, the data type of the value on the right-hand side of the
assignment statement is always forced to equal the data type of the value on the
left-hand side of the assignment statement. This is done either by extending the value
of the expression, or by truncating the value of the expression by the appropriate
number of high-order bits so that the data types of both sides of the assignment
statement are the same.
The implicit type conversions that occur for assignment statements are summarized
in Table 5-3.
78
Chapter 5
Expressions and Assignments
Table 5-3. Implicit Type Conversions in Assignment Statements for PL/M-386*
Expression
Result Type
Variable on Left of
Assignment Statement
BYTE
HWORD
BYTE value is extended by 8 high-order 0
bits to HWORD value
WORD
BYTE value is extended by 24 high-order 0
bits to WORD value
DWORD
BYTE value is extended by 56 high-order 0
bits to DWORD value
OFFSET
BYTE value is extended by 24 high-order 0
bits to OFFSET value
BYTE
8 high-order bits of HWORD value are
truncated to convert it to a BYTE value
WORD
HWORD value is extended by 16 high-order
0 bits to convert it to a WORD value
DWORD
HWORD value is extended by 48 high-order
0 bits to convert it to a DWORD value
OFFSET
HWORD value is extended by 16 high-order
0 bits to convert it to an OFFSET value
BYTE
24 high-order bits of WORD value are
truncated to convert it to a BYTE value
HWORD
16 high-order bits of WORD value are
truncated to convert it to a HWORD value
DWORD
WORD value is extended by 32 high-order 0
bits to convert it to a DWORD value
OFFSET
No conversion (both WORD and OFFSET
are 32-bits)
BYTE
56 high-order bits of DWORD value are
truncated to convert it to BYTE value
HWORD
48 high-order bits of DWORD value are
truncated to convert it to HWORD value
WORD
32 high-order bits of DWORD value are
truncated to convert it to WORD value
OFFSET
32 high-order bits of DWORD value are
truncated to convert it to OFFSET value
HWORD*
WORD*
DWORD*
*
Conversion
Assuming WORD32.
PL/M-386 Programmer's Guide
continued
Chapter 5
79
Table 5-3. Implicit Type Conversions in Assignment Statements for PL/M-386*
(continued)
Expression
Result Type
Variable on Left of
Assignment Statement
OFFSET**
BYTE
24 high-order bits of OFFSET value are
truncated to convert it to a BYTE value
HWORD
16 high-order bits of OFFSET value are
truncated to convert it to a HWORD value
WORD
No conversion is necessary (both WORD
and OFFSET are 32 bits)
DWORD
OFFSET value is extended by 32-high-order
0 bits to convert it to a DWORD value
CHARINT
24 high-order bits of INTEGER value are
truncated to convert it to CHARINT value
SHORTINT
16 high-order bits of INTEGER value are
truncated to convert it to SHORTINT value
REAL
Automatically converted to 32-bit value
INTEGER*
REAL
*
**
Conversion
Assuming WORD32.
A warning message is issued if OFFSET values are truncated.
Note that implicit conversion is not performed for POINTER or SELECTOR values.
For assignment statements with POINTER or SELECTOR expressions, the left side of
the assignment statement would be of the same type as the expression.
80
Chapter 5
Expressions and Assignments
Constant Expression
BYTE variable on the left: The constant expression is evaluated in unsigned context.
If the resulting value is equal to or greater than 0 and equal to or less than 255, it is
treated as a BYTE value and no conversion is necessary. If the resulting value is
greater than 255, it is truncated to type BYTE by dropping all except its 8 low-order
bits.
INTEGER variable on the left: The constant expression is evaluated in signed context.
No conversion is necessary.
REAL variable on the left: The assignment is illegal unless all values on the right are
floating-point constants. If the value of the constant expression is out of the range for
REAL variables, an overflow exception occurs (see Chapter 10 and Appendix G).
HWORD variable on the left: The constant expression is evaluated in unsigned context.
If the resulting value is equal to or greater than 0 and equal to or less than 65,535, it is
treated as an HWORD value, and no conversion is necessary. If the resulting value is
greater than 65,535, it is truncated to type HWORD by dropping all except its 16
low-order bits.
WORD variable on the left: The constant expression is evaluated in unsigned context.
No conversion is necessary.
DWORD variable on the left: The constant expression is evaluated in unsigned context
and is zero-extended to a DWORD value.
OFFSET variable on the left: The constant expression is evaluated in unsigned
context. No conversion is necessary.
CHARINT variable on the left: The constant expression is evaluated in 32-bit
INTEGER arithmetic. If the value is less than -128 or greater than +127, it is
truncated to 8 bits.
SHORTINT variable on the left: The constant expression is evaluated in 32-bit
INTEGER arithmetic. If the value is outside the given range for SHORTINT (-32,768
to +32,767), it is truncated to 16 bits.
Constants cannot be assigned to POINTER or SELECTOR variables.
Type conversion built-ins can be used to change the type of a constant expression to
the type required for assignment. The entire expression within the type conversion is
evaluated in signed context.
Multiple Assignment
It is often convenient to assign the same value to several variables at the same time.
This is accomplished in PL/M by listing all the variables to the left of the equal sign,
PL/M-386 Programmer's Guide
Chapter 5
81
separated by commas. The variables LEFT, CENTER, and RIGHT can all be set to the
value of the expression INIT + CORR with the single assignment statement:
LEFT, CENTER, RIGHT = INIT + CORR;
The variables on the left-hand side of a multiple assignment must be all of the same
class, that is, all unsigned, all signed, all POINTER, all SELECTOR, or all REAL. Then
the conversion rules described previously in this chapter are applied separately to
each assignment.
✏
Note
The order in which the assignments are performed is not
guaranteed. Therefore, if a variable on the left side of a multiple
assignment also appears in the expression on the right side, the
results are undefined.
Embedded Assignments
A special form of assignment can be used within PL/M expressions. The form of this
embedded assignment is:
variable:=expression
and can appear anywhere an expression is allowed. The expression (everything to the
right of the := assignment symbol) is evaluated and stored in the variable on the left.
Parentheses are used to specify the limits of an embedded assignment within an
assignment statement. The value of the embedded assignment is the same as that of
its right half. For example, the expression:
ALT + (CORR := TCORR + PCORR) - (ELEV := HT/SCALE)
results in exactly the same value as:
ALT + (TCORR + PCORR) - (HT/SCALE)
except that the intermediate results TCORR + PCORR and HT/SCALE are stored in
CORR and ELEV, respectively. These names for intermediate results can then be used
at a later point in the program without recalculating their values. The names must
have been declared earlier.
82
Chapter 5
Expressions and Assignments
The rules of PL/M do not specify the order in which subexpressions or operands are
evaluated. When an embedded assignment changes the value of a variable that also
appears elsewhere in the same expression, the results cannot be guaranteed.
For example, the following expression:
A = (X:=X+4) + Y*Y + X;
could mean either of the following interpretations:
A1 = (X+4) + Y*Y + (X+4);
A2 = (X+4) + Y*Y + X;
Avoid this ambiguity by removing the embedded assignment from the expression and
using a separate assignment statement to achieve the desired effect as follows:
X = X + 4;
A1 = X + Y*Y + X;
X = X + 4;
A2 = X + Y*Y + X - 4;
A3 = X + 4 + Y*Y + X;
X = X + 4;
■■■
PL/M-386 Programmer's Guide
Chapter 5
83
84
Chapter 5
Expressions and Assignments
Flow Control Statements
6
This chapter describes statements that alter the sequence of PL/M statement
execution and that group statements into blocks.
DO and END Statements: DO Blocks
Procedures and DO blocks are the basic building units of modular programming in
PL/M. (Procedures are discussed in Chapter 8.)
This chapter discusses all four kinds of DO-blocks. Each DO block begins with a DO
statement and includes all subsequent statements through the closing END statement.
The four kinds of DO blocks are as follows:
•
Simple DO block
DO;
/* all statements executed, each in order */
statement-0;
statement-1;
statement-2;
.
.
END;
•
DO CASE block
DO CASE select_expression;
/* one statement executed */
case-0-statement; * executed if select_expression = 0 */
case-1-statement;/* executed if select_expression = 1 */
.
.
END;
•
DO WHILE block
DO WHILE expression_true;
statement-0;
/* all executed repeatedly if expression */
statement-1;
/* true, none executed if false. */
.
PL/M-386 Programmer's Guide
Chapter 6
85
.
END;
•
Iterative DO block
DO counter = start-expr TO limit-expr BY step-expr;
statement-0;
/* all statements executed a number */
statement-1;
/* of times depending on comparison */
/* of counter with limit expression */
.
.
END;
The last two blocks are also referred to as DO-loops because the executable
statements within them can be executed repeatedly (in sequence) depending on the
expressions in the DO statement.
Any DO statement can have multiple labels on it, and only the last of these can appear
between the word END and the next semicolon. For example:
A: B: C: D: EM: DO;
.
.
.
END EM ;
/* end of block EM; */
/* A, B, C, D also end here. */
As mentioned in Chapter 3, the placement of declarations is restricted. Except for
use in procedures, declarations are permitted only at the top of a simple DO block,
before any executable statements of the block. (This DO can, of course, be nested
within other DOs or procedures. Chapter 7 discusses the scope of declared names.)
Each DO block can contain any sequence of executable statements, including other DO
blocks. Each block is considered by the compiler as a unit, as if it were a single
executable statement. This fact is particularly useful in the DO CASE block and the IF
statement, both discussed in this chapter.
The discussions that follow describe the normal flow of control within each kind of
DO block. The normal exit from the block passes through the END statement to the
statement immediately following. These discussions assume that none of the
statements in the block causes control to bypass that process. A GOTO statement with
the target outside the block would be one such bypass. (GOTOs are discussed later in
this chapter.)
86
Chapter 6
Flow Control Statements
Simple DO Blocks
A simple DO block merely groups, as a unit, a set of statements that will be executed
sequentially (except for the effect of GOTOs or CALLs):
DO;
statement-0;
statement-1;
. . .
statement-n;
END;
For example:
DO;
NEW$VALUE = OLD$VALUE + TEMP;
COUNT = COUNT + 1;
END;
This simple DO block adds the value of TEMP to the value of OLD$VALUE and stores it
in NEW$VALUE. It then increments the value of COUNT by one.
DO blocks can be nested within each other as shown in the following example:
ABLE: DO;
statement-0;
statement-1;
BAKER:
DO;
statement-a;
statement-b;
statement-c;
END BAKER;
statement-2;
statement-3;
END ABLE;
The first DO statement and the second END statement bracket one simple DO block.
The second DO statement and the first END statement bracket a different DO block
inside the first one. Notice how indentation (using tabs or spaces) can be used to
make the sequence more readable, so that it can be seen at a glance that one DO block
is nested inside another. It is recommended that this practice be followed in writing
PL/M programs. See Appendix B for the number of DO blocks that can be nested.
A simple DO block can delimit the scope of variables, as discussed in Chapter 7.
PL/M-386 Programmer's Guide
Chapter 6
87
DO CASE Blocks
A DO CASE block begins with a DO CASE statement, and selectively executes one of
the statements in the block. The statement is selected by the value of an expression.
The maximum number of cases is given in Appendix B. The form of the DO CASE
block is:
DO CASE select_expression;
statement-0;
statement-1;
. . .
statement-n;
END;
In the DO CASE statement, select_expression must yield an unsigned binary
number (excluding DWORD) or a signed integer value. If the expression is a constant
expression, it is evaluated as if it were being assigned to a WORD variable. The value
(call the value K) must be between 0 and n, inclusive. K is used to select one of the
statements in the DO CASE block, which is then executed. The first case
(statement-0) corresponds to K = 0; the second (statement-1) corresponds to
K=1, and so forth. Only one statement from the block is selected. This statement is
then executed (only once). Control then passes to the statement following the END
statement of the DO CASE block.
✏
88
Note
If the run-time value of the expression in the DO CASE statement is
less than 0 or greater than n (where n + 1 is the number of
statements in the DO CASE block), the effect of the DO CASE
statement is undefined. This may have disastrous effects on
program execution. Therefore, if there is any possibility that this
out-of-range condition may occur, the DO CASE block should be
contained within an IF statement that tests the expression to make
sure that it has a value that will produce meaningful results.
Chapter 6
Flow Control Statements
An example of a DO CASE block is:
DO CASE SCORE;
;
CONVERSIONS=CONVERSIONS + 1;
SAFETIES = SAFETIES + 1;
FIELDGOALS = FIELDGOALS + 1;
;
;
TOUCHDOWNS=TOUCHDOWNS + 1;
END;
/*
/*
/*
/*
/*
/*
/*
case
case
case
case
case
case
case
0
1
2
3
4
5
6
*/
*/
*/
*/
*/
*/
*/
When execution of this CASE statement begins, the variable SCORE must be in the
range 0 to 6. If SCORE is 0, 4, or 5 then a null statement (consisting of only a
semicolon, and having no effect) is executed; otherwise the appropriate statement is
executed, causing the corresponding variable to be incremented.
A more complex DO CASE block is the following:
SELECT = COUNT - 5;
IF SELECT <= 2 AND SELECT >= 0 THEN
DO CASE SELECT;
X = X + 1;
/* Case 0 */
DO;
/* Begin Case 1 */
X = Y + 10;
Y = Y + 1;
END;
/* End Case 1 */
DO I = LAST$HI + 1 TO TOP - 6;
Z(I) = X * Y + 1;
W(I) = Z(I) * Z(I);
V(I) = W(I) - Z(I);
END;
END;
ELSE CALL ERROR;
/* Begin Case 2 */
/* End Case 2 */
/* End DO CASE block */
If SELECT and COUNT are INTEGER variables, negative values could occur. The DO
CASE block is placed within an IF statement to guarantee that execution of the DO
CASE block will not be attempted if the value of SELECT is less than 0 or greater than
2. Instead, a procedure called ERROR (declared previously) will be activated.
PL/M-386 Programmer's Guide
Chapter 6
89
The preceding example illustrates the use of a simple DO block as a single PL/M
statement. The DO CASE statement can select Case 1 or Case 2 and cause multiple
statements to be executed. This is only possible because they are grouped as a simple
DO block, which acts as a single statement.
DO WHILE Blocks
DO WHILE and IF statements examine only the least significant bit of the value of the
expression. If the value is an odd number (least significant bit = 1), it will be
considered true. If it is even (least significant bit = 0), it will be considered false. If
the expression is relational, e.g., A<B, the result will have a value of 00H or 0FFH,
but this is incidental; it may have any unsigned value.
A DO WHILE block begins with a DO WHILE statement, and has the following form:
DO WHILE expression;
statement-0;
statement-1;
. . .
statement-n;
END;
/* expression must yield */
/* an unsigned value */
The effect of this statement is as follows:
1.
First the unsigned expression following the reserved word WHILE is evaluated.
If the rightmost bit of the result is 1, then the sequence of statements up to the
END is executed.
2.
When the END is reached, the expression is evaluated again, and again the
sequence of statements is executed only if the value of the expression has a
rightmost bit of 1.
3.
The block is executed over and over until the expression has a value whose
rightmost bit is 0. Execution then skips the statements in the block and passes to
the statements following the END statement.
Consider the following example:
AMOUNT = 1;
DO WHILE AMOUNT <= 3;
AMOUNT = AMOUNT + 1;
END;
The statement AMOUNT = AMOUNT + 1 is executed exactly 3 times. The value of
AMOUNT when program control passes out of the block is 4.
90
Chapter 6
Flow Control Statements
Iterative DO Blocks
An iterative DO block begins with an iteration statement and executes each statement
in the block, in order, repeating the entire sequence. The form of the iterative DO
block is:
DO counter = start-expr TO limit-expr BY step-expr ;
statement-0 ;
statement-1 ;
.
.
.
END ;
The BY step-expr phrase is optional; if omitted, a step of 1 is the default.
For PL/M-386, the counter must be a non-subscripted variable of unsigned type:
BYTE, HWORD, WORD, or OFFSET, or a signed integer data type: INTEGER, CHARINT,
or SHORTINT.
An example of an iterative DO block is:
DO I = 1 TO 10;
CALL BELL;
END;
where BELL is the name of a procedure that causes a bell to ring. The bell will ring
ten times.
Another example shows how the index-variable can be used within the block:
AMOUNT = 0;
DO I = 1 TO 10;
AMOUNT = AMOUNT + I;
END;
The assignment statement is executed 10 times, each time with a new value for I.
The result is to sum the numbers from 1 to 10 (inclusive) and leave the sum (namely,
55) as the value of AMOUNT.
The next example uses step-expr:
/* Compute the product of the first N odd integers */
PROD = 1;
DO I = 1 TO (2*N-1) BY 2;
PROD = PROD*I;
END;
PL/M-386 Programmer's Guide
Chapter 6
91
The type of counter (signed or unsigned) affects the following factors in the
execution flow of iterative DOs:
•
When step-expr is evaluated.
•
What causes execution to exit the DO block.
The following steps constitute the general execution sequence of an iterative DO
block, with both signed and unsigned variables and expressions in the DO itself. Type
is mentioned only for steps in which actions or consequences vary according to type.
Where the signed case is different, it is described in parentheses. The discussion
following this description summarizes the rules and their results for signed and
unsigned data types.
1.
The start-expr is evaluated and assigned to counter.
2.
The limit-expr is evaluated and compared with counter. (If counter and
limit-expr are of signed type, then step-expr is also newly evaluated at this
time.)
a.
b.
If counter is greater than limit-expr, execution exits the DO and passes to
the statement following the next END (unless step-expr is a negative
signed value; if so, the exit occurs only if counter is less than limit-expr).
Otherwise, the statements within the DO block are executed in order until the
END statement is reached.
c.
3.
At the END, a step-expr of unsigned type (BYTE, HWORD, or WORD for
PL/M-386) is newly evaluated.
The counter is incremented by the value of step-expr. For unsigned counters,
if the new value is less than the old value (due to modulo arithmetic as explained
next), the loop is exited immediately. Otherwise, control returns to step 2.
An 8-bit BYTE can represent numbers no larger than 11111111B (255 decimal). The
largest number a 16-bit WORD (or HWORD) can represent is 1111111111111111B,
which is 65535 decimal. The largest number a 32-bit WORD can represent is
0FFFF$FFFFH, which is 4,294,967,295 decimal. Adding 1 to these values gives a
result of 0. Thus, the new counter can be less than the old.
92
Chapter 6
Flow Control Statements
These rules and their consequences can be summarized in two broad cases:
1.
2.
Starting with a non-negative step-expr, the loop is exited as soon as any one
of the following conditions become true:
a.
The new counter is greater the new limit-expr.
b.
A signed step-expr becomes negative and the new counter is still less
than the new limit-expr.
c.
An unsigned step-expr causes a lower counter than the one just used.
When starting with a negative and signed step-expr, then the loop is exited as
soon as either of the following two conditions occurs:
a.
The new counter is less than the new limit-expr.
b.
The new step-expr becomes non-negative and the new counter is greater
than the new limit-expr.
Upon exit from the iterative DO block:
1.
In all cases step-expr has been reevaluated.
2.
In all but one case limit-expr has been reevaluated. When an unsigned
counter has just gone over and become smaller, limit-expr is unchanged from
its value during the last loop.
3.
In all cases counter has been changed, but the step value that was added to it
varies. If signed, counter has been incremented by the former step value
before it was reevaluated. For unsigned counters, the newer step has been used.
The following distinctions are important:
•
In every case, start-expr is evaluated only once and limit-expr is
evaluated before any execution.
•
A signed step-expr is evaluated in step 2; other step-exprs are evaluated in
step 3.
•
With an unsigned counter, there cannot be a negative step. Furthermore,
stepping down to a limit-expr that is less than start-expr is not possible
because the loop will be exited immediately.
PL/M-386 Programmer's Guide
Chapter 6
93
END Statement
An END statement must terminate all DO blocks. An END statement has the following
syntax:
END [name];
Where:
name
is the optional name that (if present) should match the label of the
corresponding DO statement.
IF Statement
The IF statement provides conditional execution of statements. It takes the form:
IF expression THEN statement-a;
ELSE statement-b;
/*optional*/
The reserved word THEN and the statement following it are required. The reserved
word ELSE and the statement following it are optional.
The IF statement has the following effect: first expression is evaluated as if it
were being assigned to a variable of type BYTE. If the result is true (rightmost bit is
1) then statement-a is executed. If the result is false (rightmost bit is 0), then
statement-b is executed. Following execution of the chosen alternative, control
passes to the next statement following the IF statement. Thus, of the two statements
(statement-a and statement-b) only one is executed.
Consider the following program fragment:
IF NEW > OLD THEN RESULT = NEW;
ELSE RESULT = OLD;
Here, RESULT is assigned the value of NEW or the value of OLD, whichever is greater.
This code causes exactly one of the two assignment statements to be executed.
RESULT always gets assigned some value, but only one assignment to RESULT is
executed.
In the event that statement-b is not needed, the ELSE part may be omitted entirely.
Such an IF statement takes the form:
IF expression THEN statement-a;
94
Chapter 6
Flow Control Statements
Here, statement-a is executed if the value of expression has a rightmost bit of 1.
Otherwise, nothing happens, and control immediately passes on to the next statement
following the IF statement.
For example, the following sequence of PL/M statements will assign to INDEX either
the number 5, or the value of THRESHOLD, whichever is larger. The value of INIT
will change during execution of the IF statement only if THRESHOLD is greater than
5. The final value of INIT is copied to INDEX in any case:
INIT = 5;
IF THRESHOLD > INIT THEN INIT = THRESHOLD;
INDEX = INIT;
The power of the IF statement is enhanced by using DO blocks in the THEN and ELSE
parts. Since a DO block can be used wherever a single statement can be used, each of
the two statements in an IF statement may be a DO block. For example:
IF A = B THEN
DO;
EQUAL$EVENTS = EQUAL$EVENTS + 1;
PAIR$VALUE = A;
BOTTOM = B;
END;
ELSE
DO;
UNEQUAL$EVENTS = UNEQUAL$EVENTS + 1;
TOP = A;
BOTTOM = B;
END;
DO blocks nested within an IF statement can contain further nested DO blocks, IF
statements, variable and procedure declarations, and so on.
PL/M-386 Programmer's Guide
Chapter 6
95
Nested IF Statements
Any IF statement (including the ELSE part, if any) can be considered a single PL/M
statement (although it is not a block). Thus, the statement to be executed in a THEN
or an ELSE clause may in fact be another IF statement.
An IF statement inside a THEN clause is called a nested IF. Nesting may be carried
to several levels without needing to enclose any of the nested IF statements in DO
blocks, as in the following construction:
IF expression-1 THEN
IF expression-2 THEN
IF expression-3 THEN statement-a;
Here are three levels of nesting. Note that statement-a will be executed only if the
values of all three expressions are true. Thus, the preceding example is equivalent to:
IF expression-1 AND expression-2 AND expression-3
THEN statement-a;
Notice that the preceding example of nesting does not have an ELSE part. When
using nested IF statements, it is important to understand the following rule of PL/M:
•
A set of nested IF statements can have only one ELSE part, and it belongs to the
innermost (that is, the last) of the nested IF statements.
This rule could also be restated as follows:
•
When an IF statement is nested within the THEN part of an outer IF statement,
the outer IF statement may not have an ELSE part.
For example, the construction:
IF expression-1 THEN
IF expression-2 THEN statement-a
ELSE statement-b;
is legal and means that if the values of both expression-1 and expression-2 are
true, then statement-a will be executed. If the value of expression-1 is true
and the value of expression-2 is false, then statement-b will be executed. If
the value of expression-1 is false, neither statement-a nor statement-b will
be executed, regardless of the value of expression-2.
96
Chapter 6
Flow Control Statements
The preceding construction is equivalent to:
IF expression-1 THEN
DO;
IF expression-2 THEN statement-a;
ELSE statement-b;
END;
This construction is much more readable and offers less opportunity for error.
If the intention is for the ELSE part to belong to the outer IF statement, then the
nesting must be done by means of a DO block:
IF expression-1 THEN
DO;
IF expression-2 THEN statement-a;
END;
ELSE statement-b;
Note that the meaning of this construction differs completely from the previous one.
Finally, consider the following:
IF expression-1 THEN
IF expression-2 THEN
IF expression-3 THEN statement-a;
ELSE statement-b;
ELSE statement-c; /* illegal statement */
ELSE statement-d;
/* illegal statement */
This construction is illegal because only one ELSE part is allowed. If the intention is
for the ELSE parts to match the IF parts as indicated by the indenting, the nesting
must be done with DO blocks, as follows:
IF expression-1 THEN
DO;
IF expression-2 THEN
DO;
IF expression-3 THEN statement-a;
ELSE statement-b;
END;
ELSE statement-c;
END;
ELSE statement-d;
PL/M-386 Programmer's Guide
Chapter 6
97
Sequential IF Statements
Consider the following example. An ASCII-coded character is stored in a BYTE
variable named CHAR. If the character is an A, statement-a should be executed. If
the character is a B, statement-b should be executed. If the character is a C,
statement-c should be executed. If the character is not A, B, or C, statement-x
should be executed. The code for doing this could be written as follows, using IF
statements that are completely independent of one another:
IF
IF
IF
IF
CHAR
CHAR
CHAR
CHAR
THEN
= 'A' THEN statement-a;
= 'B' THEN statement-b;
= 'C' THEN statement-c;
<> 'A' AND CHAR <> 'B' and CHAR <> 'C'
statement-x;
This sequence is inefficient because all four IF statements (six tests in all) will be
carried out in every case, which is wasteful when one of the earlier tests succeeds.
A must be tested for in all cases. However, B needs to be tested only if the test for A
fails and C needs to be tested only if both previous tests fail. Finally, if the tests for
A, B, and C all fail, no further tests are needed and statement-x must be executed.
To improve the code, rewrite it as follows:
IF CHAR = 'A' THEN statement-a;
ELSE IF CHAR = 'B' THEN statement-b;
ELSE IF CHAR = 'C' THEN statement-c;
ELSE statement-x;
Notice that this sequence is not a case of nested IF statements as described in the
preceding section. IF statements are nested only when one IF statement is inside the
THEN part of another. In the next example, IF statements are inside the ELSE parts of
other IF statements. This construction is called sequential IF statements. It is
equivalent to the following:
IF CHAR = 'A' THEN statement-a;
ELSE DO;
IF CHAR = 'B' THEN statement-b;
ELSE DO;
IF CHAR = 'C' THEN statement-c;
ELSE statement-x;
END;
END;
Sequential IF statements are useful whenever a set of tests is to be made, but the
remaining tests should be skipped whenever one of the tests succeeds. This
construction works in such cases because all the remaining tests are in the ELSE part
of the current test.
98
Chapter 6
Flow Control Statements
GOTO Statements
A GOTO statement alters the sequential order of program execution by transferring
control directly to a labeled statement. Sequential execution then resumes, beginning
with the target statement. The GOTO statement has the following form:
GOTO label
For example:
GOTO ABORT;
The appearance of label in a GOTO statement is called a label reference, not a label
definition.
The reserved word GOTO can also be written GO TO, with an embedded blank.
For reasons discussed in Chapter 7, GOTO statements are restricted. The only possible
GOTO transfers are the following:
•
From a GOTO statement in the outer level of some block to a labeled statement in
the outer level of the same block.
•
From a GOTO statement in an inner block to a labeled statement in the outer level
of an enclosing block (not necessarily the smallest enclosing block). However, if
the inner block is a procedure block, the transfer can only be to a statement in the
outer level of the main program module.
•
From any point in one program module to a labeled statement in the outer level
of the main program module. To jump to such a label, the label must be declared
to have extended scope, (i.e., declare it PUBLIC in the main module and
EXTERNAL in the module containing the GOTO).
The use of GOTOs is necessary in some situations. However, in most situations where
control transfers are desired, the use of an iterative DO, DO WHILE, DO CASE, IF, or a
procedure activation (see Chapter 8) is preferable. Indiscriminate use of GOTOs will
result in a program that is difficult to understand, correct, and maintain.
PL/M-386 Programmer's Guide
Chapter 6
99
The CALL and RETURN Statements
The CALL and RETURN statements are mentioned here only for completeness, since
they control the flow of a program. However, they are discussed in detail in Chapter
8.
The CALL statement is used to activate an untyped procedure (one that does not return
a value).
The RETURN statement is used within a procedure body to cause a return of control
from the procedure to the point from which it was activated.
■■■
100
Chapter 6
Flow Control Statements
102
Chapter 6
Flow Control Statements
Block Structure and Scope
7
This chapter explains the meaning of outer level and the concept of scope, including
the use of the linkage attributes, PUBLIC and EXTERNAL.
The outer level of a block means statements (or labels) contained in the block but not
contained in any nested blocks. The term exclusive extent also has this meaning.
The inner level, or inclusive extent, includes this outer level and all nested blocks as
well.
A block at the same level as another block means that both blocks are contained by
exactly the same outer blocks.
The scope of an object means those parts of a program where its name, type, and
attributes are recognized (i.e., handled according to a given declaration). An object
means a variable, label, procedure, or symbolic (named) constant (i.e., a compilation
constant or execution constant as discussed in Chapter 3). A program is the complete
set of modules that are ultimately executed as a unit.
PL/M-386 Programmer's Guide
Chapter 7
103
Names Recognized Within Blocks
As shown throughout this manual, PL/M is a block-structured language that enables
design implementation for problem solving, data processing, and hardware control.
PL/M is used to create blocks of code containing declarations followed by executable
statements. These blocks are ordered and nested in such a way as to simplify and
clarify the flow of data and control. (See Appendix B for maximum block nesting.)
A collection of these blocks that performs a single function, or a small set of related
functions, is usually compiled as one module, as discussed in Chapter 1.
Beyond the advantages of modularity, simplicity, and clarity, the nesting of blocks
serves another very basic purpose: names declared at an outer level are known to all
statements of all nested blocks as well.
A new meaning can be declared for any such name within a nested simple DO or
procedure block, thereby cutting off its earlier meaning for this block. But if this
option is not chosen, its meaning is established by a single declaration at an outer
level. (The only objects that do not require declarations prior to use are labels and
reentrant procedures.)
In Figure 7-1, everything inside the figure (except the title) constitutes the inclusive
extent of block MMM (in this case, module MMM). KK is known throughout this block,
including all nested blocks.
Everything inside the large box constitutes the inclusive extent of block SORT. JJ
and II are known throughout this block, but not outside it. JJ and II are not known
before the label SORT or after the END SORT statement.
Everything inside the small box constitutes the inclusive extent of block FIND. Since
this is not a simple DO or procedure block, declarations are not allowed. All prior
declarations shown are available for use within FIND.
104
Chapter 7
Block Structure and Scope
MMM: DO;
/* Beginning of module */
DECLARE RECORD (128) STRUCTURE
(KEY BYTE,
INFO WORD);
DECLARE CURRENT STRUCTURE
(KEY BYTE,
INFO WORD);
DECLARE KK BYTE;
KK = 127;
/* Instructions here would read in data. */
SORT:
DO;
DECLARE (JJ,ii) INTEGER;
DO JJ = 1 TO 127;
CURRENT.KEY = RECORD(JJ).KEY;
CURRENT.INFO = RECORD(JJ).INFO;
II = JJ;
FIND:
DO WHILE II > 0 AND
RECORD(II-1).KEY > CURRENT.KEY;
RECORD(II).KEY = RECORD(II-1).KEY;
RECORD(II).INFO = RECORD(II-1).INFO;
II = II-1;
END FIND;
RECORD(II).KEY = CURRENT.KEY;
RECORD(II).INFO = CURRENT.INFO;
END;
END SORT;
/* Instructions here would write out data from the records. */
END MMM;
/* End of module */
Figure 7-1. Inclusive Extent of Blocks
PL/M-386 Programmer's Guide
Chapter 7
105
In Figure 7-1, the area within the large box and outside the small box is the exclusive
extent (the outer level) of block SORT. The area within the small box is the exclusive
(and inclusive) extent of block FIND. To the instructions within the FIND block,
SORT's exclusive extent is an outer level. The outermost level (or module level) is
the area outside the large box enclosing the SORT block.
Restrictions on Multiple Declarations
In any given block, a known name cannot be redeclared at the same level as its
original declaration. A new declaration is permitted inside a nested simple DO or
procedure block, where it automatically identifies a new object despite the existence
of the same name at a higher level. The new object will be the only one known by
this name within its block, and it will be unknown outside its block, where the prior
name maintains its meaning. These observations also apply when a name is
redeclared in another block at the same level as the block containing the original
declaration.
When a name is declared only in a separate block at the same level, there is no way to
access it except in that block where it is declared. The definition is not at an outer
level to the current block. Any local declaration that is supplied establishes a new
separate object whose values bear no relation to those of the other.
The reason for these rules, as for many in programming, is that there must be no
ambiguity about what address/location is meant by each name in the program. The
preceding declaration rules give freedom to choose names appropriate to a given
block, without interfering with exterior uses of them. But when a name is redeclared,
its outer-level meaning is inaccessible until execution exits the block containing the
new declaration. For example:
A: DO;
DECLARE X, Y, Z BYTE;
L1: X = 2;
Y = X;
Z = X;
B: DO;
DECLARE X, Y BYTE;
X = 3;
Y = X;
L2: Z = X;
END B;
L3:
/* At this point, X=2, Y=2, Z=3, because
/* the value of the redeclared X was used
/* to fill Z. If statement L2 were outside
/* the loop labeled B, then Z would be 2
/* because the outer X value would be used
106
Chapter 7
*/
*/
*/
*/
*/
Block Structure and Scope
Extended Scope: The PUBLIC and EXTERNAL
Attributes
The PUBLIC and EXTERNAL attributes permit the scope of names to be extended for
all objects except modules; a module name cannot be declared with either attribute.
To extend the scope means to make the names available for use in modules other than
the one where they are defined. (The names are already available to nested blocks in
this module.) Extended scope includes names for variables, labels, procedures, and
execution constants.
For example, the statement:
DECLARE FLAG BYTE PUBLIC;
causes a byte named FLAG to be allocated, and its address made known to any other
module where the following declaration occurs:
DECLARE FLAG BYTE EXTERNAL;
Similarly, if one module has a procedure declaration block that begins:
SUMMER: PROCEDURE (A,B) WORD PUBLIC;
DECLARE (A,B) BYTE;
.
/* other declarations can go here */
.
/* executable statements go here, */
.
/* defining the procedure
*/
.
END SUMMER;
then any other module may invoke SUMMER if it first declares:
SUMMER: PROCEDURE (A,B) WORD EXTERNAL;
/* A,B can be any
DECLARE (A,B) BYTE;
/* names but these names must
/* match them and each type must
END SUMMER;
/* match its public definition
PL/M-386 Programmer's Guide
Chapter 7
*/
*/
*/
*/
107
The use of PUBLIC and EXTERNAL must follow a strict set of rules to prevent
ambiguity of location or definition. These rules are as follows:
1.
These attributes can be used only in a declaration at the outermost level of a
module (i.e., never in a nested block).
2.
Only one can appear in any declaration, no more than once. Thus:
DECLARE ZETA BYTE PUBLIC EXTERNAL;
DECLARE RHO WORD PUBLIC PUBLIC;
/* error */
/* error */
and similar constructs are all invalid.
3.
Names can be declared PUBLIC no more than once. The PUBLIC declaration is
the defining declaration: the address it creates is used in each procedure or
module where the same name is declared EXTERNAL. Do not create more than
one PUBLIC address for any name.
4.
Names can be declared EXTERNAL only if they are also declared PUBLIC in a
different module of the program. The EXTERNAL attribute is essentially a
request to use a PUBLIC address. An EXTERNAL without a PUBLIC is a dead
letter. Lack of a definition elsewhere will result in a link-time error.
5.
Where the name is declared EXTERNAL, it must be given the same type as where
it is declared PUBLIC. Any contradiction of type would violate the intention to
use the location(s) and content(s) defined elsewhere. If the name is declared
PUBLIC and has the DATA attribute, all EXTERNAL declarations must also use
DATA, but cannot assign a value to the constant being declared.
6.
Similarly, names declared EXTERNAL must not be given a location (using the AT
clause), or an initialization (using DATA or INITIAL). Such usage would
contradict the fact that names are being defined in another module. However, in
the module where this name is declared PUBLIC, the use of AT, DATA (with
initialization values present), or INITIAL is allowed.
7.
Neither PUBLIC nor EXTERNAL can be applied to a name that is based. For
example:
DECLARE PTR1 POINTER;
DECLARE V1 BASED PTR1 PUBLIC;
is invalid. The reason: by definition, V1 has no home of its own; its location is
always determined by PTR1. Thus, to declare V1 PUBLIC or EXTERNAL does not
permit the correct assignment of addresses. PTR1, on the other hand, always
contains the current address of V1. Declaring the base, in this case PTR1, to be
PUBLIC or EXTERNAL is always permissible since it permits valid results.
108
Chapter 7
Block Structure and Scope
✏
8.
Note
The PL/M compiler will generate external records only for items
that are actually referenced in the program.
When extending the scope of a name with the PUBLIC attribute and DATA or
INITIAL, the placement in the DECLARE statement is critical. PUBLIC must be
placed after the type declaration and before the DATA or INITIAL attribute. For
example:
DECLARE a$p BYTE PUBLIC INITIAL(4);
(Additional restrictions on the use of PUBLIC and EXTERNAL procedures are
described in Chapter 8.)
Following these rules will enable consistent and reliable execution of programs using
names with extended scope. A PUBLIC definition occurring in one module will then
be used by all related references to that name in separate modules; that is, references
which declare the name EXTERNAL. The following diagram illustrates this:
MOD1: DO;
DECLARE V1 BYTE PUBLIC;
.
.
.
END MOD1;
MOD2: DO;
DECLARE V1 BYTE EXTERNAL;
QQ4: PROCEDURE PUBLIC;
.
.
.
END QQ4;
END MOD2;
PL/M-386 Programmer's Guide
Chapter 7
109
Both references to V1 will use the same definition (location) for V1, namely, the
definition in module MOD1. Similarly, if any module needed to call procedure QQ4, it
would first need a declaration like this:
QQ4:PROCEDURE EXTERNAL;
END QQ4;
so that a subsequent CALL QQ4 would correctly pass control to that procedure in
MOD2.
Scope of Labels and Restrictions on GOTOs
Labels are subject to exactly the same rules of scope previously discussed.
A label is unknown outside the block where it is declared. As discussed in Chapter 1,
a label is either declared explicitly at the beginning of a simple DO or procedure
block, or the compiler considers it to be declared there as soon as it is defined by use
anywhere in the block. Therefore, the discussion of what names are known in which
blocks applies directly to labels as well as to other names.
The label on a block is not part of the block it names. For example, the name on the
DO enclosing the module itself is not part of the DO; it merely names it. For nested
blocks, a label is again not part of the block it names, but belongs instead to the outer
level as part of that first enclosing block.
If a name used as a label on a block is defined inside that block, it will name a new
item, be it label, variable, or constant. There will be no confusion with the outer
label name. This fact leads to important restrictions on the use of the GOTO
statement:
1.
It is impossible for a GOTO to transfer control from an outer block to a labeled
statement inside a nested block.
2.
Moreover, a GOTO can transfer control from one block to another in the same
module only if the target block encloses the one containing the GOTO (and only if
the name of that target label is not declared in the nested block).
Furthermore, a label with the PUBLIC attribute is permitted only in the main module.
This has the interesting consequence of forcing all other transfers of control (i.e.,
those not involving a return to the main module) to use procedure calls. This favors
the development of orderly, modularized, traceable programs.
110
Chapter 7
Block Structure and Scope
Only four GOTO transfers are possible; these are as follows:
1.
From one point in a block to another statement also in the same level of the same
block.
2.
From an inner, nested DO block (not a nested procedure) to a statement in the
outer level of any enclosing block.
3.
From a procedure to a statement in the outer level of the main program in the
same module.
4.
To a main-program label that is declared PUBLIC, from any point in any module
that declares that label EXTERNAL.
Recall that only labels at the outer level of a main program can be declared PUBLIC.
Program structure and declarations are shown in Figure 7-2. Figure 7-3 illustrates the
only legal GOTO transfers that are permitted among the given labels in Figure 7-2. A
single-headed arrow means the transfer is valid only in the direction shown. A
double-headed arrow means that a GOTO can be used in either direction.
PL/M-386 Programmer's Guide
Chapter 7
111
MAIN: DO;
DECLARE (LAB33, LAB77) LABEL PUBLIC;
DECLARE IT BYTE;
. . .
LAB33: . . . ;
DO;
. . .
END;
. . .
LAB77: . . . ;
DO WHILE IT > 0;
. . .
END;
. . .
END MAIN;
MOD1: DO;
DECLARE (LAB33,LAB77) LABEL EXTERNAL;
P1: PROCEDURE;
L1: . . . ;
DO;
DECLARE KO BYTE;
P2: PROCEDURE;
. . .
L2; . . .;
. . .
END P2;
END;
L3: . . . ;
. . .
END P1;
END MOD1;
MOD2: DO;
. . .
DECLARE (LAB33,LAB77) LABEL EXTERNAL;
P4: PROCEDURE;
. . .
L4: . . . ;
. . .
L5: . . . ;
DO;
L6: . . . ;
. . .
END;
. . .
L7: . . . ;
. . .
END P4;
LB: . . .;
END MOD2;
Figure 7-2. Sample Program Modules Illustrating Valid GOTO Usage
112
Chapter 7
Block Structure and Scope
L5
L4
LAB33
L6
OR
L4
LAB77
L2
L7
L3
L1
OSD534
Figure 7-3. Sample Program Modules Illustrating Valid GOTO Transfers
■■■
PL/M-386 Programmer's Guide
Chapter 7
113
114
Chapter 7
Block Structure and Scope
Procedures
8
A procedure is a section of PL/M code that is declared without being executed, and
then activated from other parts of the program. A function reference or CALL
statement activates the procedure, even if it is physically located elsewhere. Program
control is transferred from the point of activation to the beginning of the procedure
code, and the code is executed. Upon exit from the procedure code, program control
is passed back to the statement immediately after the point of activation.
The use of procedures forms the basis of modular programming. It facilitates making
and using program libraries, eases programming and documentation, and reduces the
amount of object code generated by a program. The following sections review how
to declare and activate procedures.
Procedure Declarations
A procedure must be declared, just as variables must be declared. Thereafter, any
reference to a procedure must occur within the scope defined by the procedure
declaration. Also, a procedure cannot be used (called, or invoked in an expression)
until after the END statement of the procedure declaration unless it is reentrant.
A procedure declaration consists of three parts: a PROCEDURE statement, a sequence
of statements forming the procedure body, and an END statement.
The following is a simple example:
DOOR$CHECK: PROCEDURE;
IF FRONT$DOOR$LOCKED AND SIDE$DOOR$LOCKED THEN
CALL POWER$ON;
ELSE CALL DOOR$ALARM;
END DOOR$CHECK;
where POWER$ON and DOOR$ALARM are procedures declared previously in the same
program.
PL/M-386 Programmer's Guide
Chapter 8
115
✏
Note
The name DOOR$CHECK in a PROCEDURE statement has the same
appearance as a label definition, but it is not considered a label
definition, and a procedure name is not a label. PROCEDURE
statements cannot be labeled.
The name DOOR$CHECK is a PL/M identifier, which is associated with this procedure.
The scope of a procedure is governed by the placement of its declaration in the
program text, just as the scope of a variable is governed by the placement of its
DECLARE statement (see Chapter 7 for a detailed description). Within this scope, the
procedure can be activated by the name used in the PROCEDURE statement.
A procedure declaration, like a DO block, controls the scope of variables as described
in Chapter 7. Also, like a simple DO block, a procedure declaration can contain
DECLARE statements, which must precede the first executable statement in the
procedure body.
As in a DO block, the identifier in the END statement has no effect on the program, but
helps legibility and debugging. If used, it should be the same as the procedure name.
The parameter list and the type are discussed in the following two sections.
Parameters
Formal parameters are non-based scalar variables declared within a procedure
declaration; their identifiers appear in the parameter list in the PROCEDURE
statement. The identifiers in the list are separated by commas and the list is enclosed
in parentheses. No subscripts or member-identifiers can be used in the parameter list.
If the procedure has no formal parameters, the parameter list (including the
parentheses) is omitted from the PROCEDURE statement.
Each formal parameter must be declared as a non-based scalar variable in a DECLARE
statement preceding the first executable statement in the procedure body. However,
procedure parameters are not stored according to the same rules as other declared
variables. In particular, do not assume that a parameter is stored contiguously with
other variables declared in the same factored variable declaration.
When a procedure that has formal parameters is activated, the CALL statement or
function reference contains a list of actual parameters. Each actual parameter is an
expression whose value is assigned to the corresponding formal parameter in the
procedure before the procedure begins to execute.
For example, the following procedure takes four parameters, called PTR, N, LOWER,
and UPPER. It examines N contiguously stored BYTE variables. The parameter PTR is
the location of the first of these variables. If any of these variables is less than the
116
Chapter 8
Procedures
parameter LOWER or greater than the parameter UPPER, the ERRORSET procedure
(declared previously in the program) is activated:
RANGE$CHECK: PROCEDURE(PTR, N, LOWER, UPPER);
DECLARE PTR POINTER;
DECLARE (N, LOWER, UPPER, I) BYTE;
DECLARE ITEM BASED PTR(1) BYTE;
DO I = 0 TO N - 1;
IF (ITEM(I) < LOWER) OR (ITEM(I) > UPPER)
THEN CALL ERRORSET;
/* ERRORSET is a procedure declared previously */
END;
END RANGE$CHECK;
Notice that the array ITEM is declared to have only one element. Since it is a based
array, a reference to any element of ITEM is really a reference to some location
relative to the location represented by PTR. In writing the procedure RANGE$CHECK,
a dimension specifier that is any arbitrary number greater than zero must be supplied
for ITEM so that references to ITEM can be subscripted. But it does not matter what
the dimension specifier is (1 is arbitrarily used here).
Having made this declaration, suppose that 25 variables are stored contiguously in an
array called QUANTS. To check that all of these variables have values within the
range defined by the values of two other BYTE variables, SMALL and LARGE, write:
CALL RANGE$CHECK (@QUANTS, 25, SMALL, LARGE);
When this CALL statement is processed, the following sequence occurs:
•
The four actual parameters in the CALL statement (@QUANTS, 25, SMALL, and
LARGE) are assigned to the formal parameters PTR, N, LOWER, and UPPER, which
were declared within the procedure RANGE$CHECK. Since ITEM is based on PTR
and the value of PTR is @QUANTS, every reference to an element of ITEM
becomes a reference to the corresponding element of QUANTS.
•
The executable statements of the procedure RANGE$CHECK are executed. If any
of the values are less than the value of SMALL or greater than the value of LARGE,
the procedure ERRORSET is activated.
•
Finally, control returns to the statement following the CALL statement.
Notice how the use of a based variable, with the base passed as a parameter, allows
the procedure to have its own unchanging name (ITEM) for a set of variables which
may be a different set each time the procedure is activated.
Parameters are placed on the stack in left-to-right order. The stack grows from
higher locations to lower locations, so the first parameter occupies the highest
PL/M-386 Programmer's Guide
Chapter 8
117
position on the stack, and the last parameter occupies the lowest position. For more
information, see Appendix F.
✏
Note
PL/M does not guarantee the order in which multiple actual
parameters will be evaluated when the procedure is activated. If
one actual parameter changes another actual parameter, the results
are undefined. This can occur if an expression used as an actual
parameter contains an embedded assignment or function reference
that changes another actual parameter for the same procedure.
Typed Versus Untyped Procedures
The preceding RANGE$CHECK procedure is an untyped procedure. No type is given in
the PROCEDURE statement, and it does not return a value. An untyped procedure is
activated by using its name in a CALL statement.
A typed procedure, also called a function, has a type in its PROCEDURE statement: an
unsigned binary number, signed INTEGER, REAL number, POINTER or SELECTOR
data type. Such a procedure returns a value of this type, which is used in an
expression or stored as the value of a variable. The procedure is activated by using
its name as an operand in an expression; this special type of variable reference is
called a function reference.
When the expression is processed at run time, the function reference causes the
procedure to be executed. The function reference itself is then replaced by the value
returned by the procedure. The expression containing the function reference is then
evaluated, and program execution continues in normal sequence.
Like an untyped procedure, a typed procedure can have parameters. They are
handled as described in the previous section.
The body of a typed procedure can contain a RETURN statement with an expression,
as explained later in this chapter.
✏
118
Note
The body of a typed procedure can contain code (such as an
assignment statement) that changes the value of some variable
declared outside the procedure. This is called a side effect.
Chapter 8
Procedures
Recall that PL/M does not guarantee the order in which operands in an
expression are evaluated. Therefore, if a function used in an expression
changes the value of another variable in the same expression, the value
of the expression depends on whether the function reference or the
variable is evaluated first.
If the analysis of the expression does not force one of these operands to
be evaluated before the other, then the value of the expression is
undefined.
This situation can be avoided by using parentheses to segregate any
typed procedure that has a side effect, or by using this procedure in an
assignment statement first to create an unambiguous sequence.
Activating a Procedure: Function References and
CALL Statements
The two forms of procedure activation depend on whether the procedure is typed or
untyped. An untyped procedure is activated by means of a CALL statement, which
has the form:
CALL name;
or
CALL name (parameter list);
For example:
CALL REORDER (@RANK$TABLE,3);
(An alternate form of the CALL statement is discussed later.)
A typed procedure is activated by means of a function reference, which is an operand
in an expression. A function reference has the form:
name
or
name (parameter list)
PL/M-386 Programmer's Guide
Chapter 8
119
This occurs as an operand in an expression, as in the following example:
TOTAL = SUBTOTAL + SUM$ARRAY (@ITEMS,COUNT);
where SUM$ARRAY is a previously declared typed procedure. The value added to
SUBTOTAL will be the value returned by SUM$ARRAY using the actual parameters
(@ITEMS, COUNT).
In both forms of procedure activation, the elements of the parameter list are called
actual parameters to distinguish them from the formal parameters of the procedure
declaration. At the time of activation, each actual parameter is evaluated and the
result is assigned to the corresponding formal parameter in the procedure declaration.
Then, the procedure body is executed. Any PL/M expression may be an actual
parameter if its type is the same as that of the corresponding formal parameter.
The actual parameter list in a procedure activation must also match the formal
parameter list in the procedure declaration. That is, it must contain the same number
of parameters of the same type (except as described in the next paragraph) in the
same order. If the procedure is declared without a formal parameter list, then no
actual parameter list can be used in the activation.
As in expression evaluation and assignment statements (see Chapter 5), a few type
conversions are performed automatically when necessary in activating and returning
from a procedure. The built-in explicit type conversion procedures described in
Chapter 9 can also be used to force the value of an expression to a desired type.
Indirect Procedure Activation
The CALL statement, in the form shown in the preceding section, activates an untyped
procedure by its name. It is also possible to activate an untyped procedure by its
location. This is done by means of a CALL statement with the form:
CALL identifier[.member-identifier] [(parameter list)];
The identifier cannot be subscripted; however it can be a structure reference. The
identifier must be a fully qualified POINTER or WORD type variable reference for
PL/M-86 and PL/M-286, and a fully qualified POINTER, OFFSET, or WORD type
variable reference for PL/M-386. Its value is assumed to be the location of the
entry-point of the procedure being activated.
120
Chapter 8
Procedures
✏
Note
Calls through 48-bit POINTERs will be translated into long calls
whereas calls through 32-bit OFFSETs, WORDs, or POINTERs (in the
SMALL case) will be translated into short calls (relative to the
current code segment).
The identifier for the indirect procedure activation cannot be an
HWORD. Therefore, all variables used for indirect calling in
programs that are recompiled from PL/M-286 and use the WORD16
control should have DWORD, OFFSET (or ADDRESS) data types.
A normal CALL uses the name of the procedure; the compiler checks to make sure
that the correct number of parameters is supplied and performs automatic type
conversion on the actual parameters.
When the CALL statement uses a location, the compiler does not check the number of
parameters or perform type conversion. However, type conversion is performed if
the actual argument is a constant expression. The constant expression is evaluated in
unsigned context, as described in Chapter 5. If the number of parameters is wrong or
if an actual parameter is not of the same type as the corresponding formal parameter,
the results are unpredictable.
PL/M-386 Programmer's Guide
Chapter 8
121
Code Examples
The following code examples illustrate an indirect call for the COMPACT model.
The first example is a procedure which, when compiled, generates warnings.
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
$COMPACT
CALLF:DO;
DECLARE dummy word,
inner_p pointer,
main_p pointer;
funct1:PROCEDURE;
DECLARE i WORD;
i = 0;
RETURN;
END funct1;
funct:PROCEDURE;
DECLARE i WORD;
i = 0;
inner_p = @funct1;
call inner_p;
RETURN;
END funct;
dummy = .funct;
CALL dummy;
main_p = @funct;
CALL main_p;
END callf;
Warnings are generated at lines 16 and 23. The warnings occur because of conflicts
in FAR and NEAR calls. In most cases of using the COMPACT segmentation
model, indirect function calls are NEAR calls. The "@" operator causes FAR
function calls. Therefore, indirectly activating a function using the "@" operator in a
COMPACT model causes a FAR call, however, the function will execute a NEAR
return. This causes the compiler to generate a warning.
The warning is based on stack corruption. A long call pushes the segment selector
and offset addresses onto the stack. COMPACT functions do a NEAR RETURN
(unless they are on the EXPORT list). Therefore, only the OFFSET for the
RETURN address is popped. This leaves the previously pushed segment selector on
the stack.
122
Chapter 8
Procedures
The following example properly demonstrates indirect procedure calls in the
COMPACT model. This method uses the "." operator to generate a NEAR call. This
operator is similar to the "@" operator except it generates an address of the type
WORD.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
✏
$COMPACT
CALLF: DO;
DECLARE dummy WORD;
funct:PROCEDURE;
DECLARE i WORD;
i=0;
RETURN;
END funct;
dummy=funct;
CALL dummy;
END CALLF;
Note
Do not use the "." operator when using a pointer to a function as
required by certain iRMX system calls. These calls, such as
rq_create_task and rq_create_job, expect a pointer to a task address,
not just the offset. The interface to iRMX system libraries requires
a 32-bit pointer as a parameter. The "@" operator must be used
when the pointer to the start address of the task is passed to the
iRMX system call. No compiler warning is generated because the
task never returns, causing no stack corruption.
Exit from a Procedure: The RETURN Statement
The execution of a procedure is terminated in one of three ways:
•
By execution of a RETURN statement within the procedure body. A typed
procedure must terminate with a RETURN statement that has an expression.
•
By executing a GOTO to a statement outside the procedure body. The target of
the GOTO must be at the outer level of the main program (see Chapter 7).
•
By reaching the END statement that terminates the procedure declaration.
PL/M-386 Programmer's Guide
Chapter 8
123
The RETURN statement takes one of two forms:
RETURN;
or
RETURN expression;
The first form is used in an untyped procedure. The second form is used in a typed
procedure. The value of the expression becomes the value returned by the procedure.
It is evaluated as if it were being assigned to a variable of the same type as used on
the PROCEDURE statement.
124
Chapter 8
Procedures
The Procedure Body
The statements within the procedure body can be any valid PL/M statements,
including CALL statements as well as nested procedure declarations.
Examples
1.
The following is a typed procedure declaration:
AVG: PROCEDURE (X,Y) REAL;
DECLARE (X,Y) REAL;
RETURN (X + Y)/2.0;
END AVG;
This procedure could be used as follows:
SMALL = 3.0;
LARGE = 4.0;
MEAN = AVG (SMALL, LARGE);
The effect would be to assign the value 3.5 to MEAN.
2.
The following is an untyped procedure:
AOUT: PROCEDURE (ITEM);
DECLARE ITEM WORD;
IF ITEM >= OFFH THEN COUNTER = COUNTER + 1;
RETURN;
END AOUT;
Here COUNTER is some variable declared outside the procedure (i.e., it is a global
variable). This procedure could be activated as follows:
CALL AOUT (UNKNOWN);
If the value of the variable UNKNOWN is greater than or equal to 0FFH, the value
of COUNTER will be incremented.
PL/M-386 Programmer's Guide
Chapter 8
125
3.
This example demonstrates an important use of based variables:
SUM$ARRAY: PROCEDURE (PTR,N) BYTE;
DECLARE PTR POINTER,
ARRAY BASED PTR(1) BYTE,
(N,SUM,I)BYTE;
SUM = 0;
DO I = 0 TO N;
SUM = SUM + ARRAY(I);
END;
RETURN SUM;
END SUM$ARRAY;
This procedure returns the sum of the first N + 1 elements (from the zeroth to the
Nth) of a BYTE array pointed to by PTR. Notice that ARRAY is declared to have 1
element. Since it is a based variable, no space is allocated for it. It must be
declared as an array (with a non-zero dimension) so that it can be subscripted in
the iterative DO block. The choice of 1 as the constant in the dimension specifier
is arbitrary and does not restrict the value of N that may be supplied when the
procedure is activated.
The procedure could be used as follows to sum the elements of a 100-element
BYTE array named PRICE, and to assign the sum to the variable TOTAL:
TOTAL = SUM$ARRAY(@PRICE,99);
126
Chapter 8
Procedures
The Attributes: PUBLIC and EXTERNAL,
INTERRUPT, REENTRANT
The PUBLIC and EXTERNAL attributes can be included in PROCEDURE statements to
give procedures extended scope. Extended scope is discussed in Chapter 7.
A procedure declaration with the PUBLIC attribute is called a defining declaration. A
procedure declaration with the EXTERNAL attribute is called a usage declaration.
Most of the rules for PUBLIC and EXTERNAL appear in Chapter 7. The following
additional rules apply to the use of the EXTERNAL attribute in a procedure
declaration:
1.
The EXTERNAL attribute cannot be used in the same PROCEDURE statement as a
PUBLIC or REENTRANT attribute. Note, however, that the defining declaration of
a procedure may have the REENTRANT attribute.
2.
A usage (EXTERNAL) declaration of a procedure should have the same number of
parameters as the defining (PUBLIC) declaration. Variable types and dimension
specifiers should match up in the same sequence in both declarations. The
names of the parameters need not be the same. Note that a discrepancy between
the parameter lists in the defining declaration and in a usage declaration will not
be automatically detected (see Chapter 11 for a description of the TYPE control
to detect such an error at module linkage time).
3.
The procedure body of a usage declaration cannot contain anything except the
declarations of the formal parameters. The formal parameters must be declared
with the same types as in the defining declaration.
4.
No labels can appear in a usage declaration.
✏
Note
The PL/M compiler will generate external records only for items
that are actually referenced in the program.
For example, the procedure AVG (from example 1 in "The Procedure Body") can be
altered by giving it the PUBLIC attribute:
AVG: PROCEDURE (X,Y) REAL PUBLIC;
DECLARE (X,Y) REAL;
RETURN (X + Y)/2.0;
END AVG;
Another module would have a usage declaration, as follows:
AVG: PROCEDURE (X,Y) REAL EXTERNAL;
DECLARE (X,Y) REAL;
END AVG;
PL/M-386 Programmer's Guide
Chapter 8
127
Now, in the module with the usage declaration, AVG can be referenced in an
executable statement:
MIDDLE = AVG (FIRST, LATEST);
thereby activating the procedure AVG as declared in the first module.
Interrupts and the INTERRUPT Attribute
The INTERRUPT attribute enables definition of a procedure to handle some condition
signaled by a microprocessor interrupt (e.g., from a peripheral device). A procedure
with this attribute is activated when the corresponding interrupt signal is received in
the target system. The PL/M statement CAUSE$INTERRUPT (constant) can also be
used to initiate an interrupt signal (see Chapter 10).
Note that the following discussion applies only to interrupt procedures; interrupt tasks
are discussed in Appendix G.
The INTERRUPT attribute can be used only in declaring an untyped procedure with no
parameters at the outermost level of a program module. It must be declared PUBLIC
or EXTERNAL (and optionally REENTRANT). The form is:
INTERRUPT
At build time, an interrupt vector is assigned to each interrupt procedure.
The following discussion of the microprocessor interrupt mechanism clarifies how
interrupt procedures work. Additional information can be found in Appendix G.
The microprocessor interrupt mechanism has two states: enabled or disabled. With
the ENABLE statement, interrupts can take effect. The DISABLE statement prevents
interrupts from having any effect. The HALT statement also enables interrupts. (The
state of the microprocessor interrupt mechanism upon initialization is determined by
the operating system.)
128
Chapter 8
Procedures
When some peripheral device sends an interrupt to the CPU, it is ignored if the
interrupt mechanism is disabled. If interrupts are enabled, the interrupt is processed
as follows:
1.
The CPU completes any instruction currently in execution.
2.
The CPU sends an acknowledge interrupt signal, then the interrupting device
sends its interrupt number.
3.
The interrupt mechanism is disabled. This prevents any other device from
interfering.
4.
Control passes to the interrupt procedure whose number matches the number sent
by the peripheral device. If no such procedure has been established, the results
are undefined (since the vector that transfers control may be uninitialized).
5.
When the procedure is through (by executing a RETURN or reaching the END
of the procedure), the interrupt mechanism is enabled so other devices can be
serviced, and control returns to the point where the interrupt occurred.
It is possible (as with other untyped procedures) for the procedure to terminate by
executing a GOTO with a target outside the procedure in the outer level of the main
program module. In this case, control will never be returned to the point where the
program was interrupted, and interrupts will not be enabled automatically.
The following is an example of an interrupt procedure for a system where a
peripheral device initiates an interrupt whenever the temperature of a device exceeds
a certain threshold. The interrupt procedure turns on the annunciator light, updates a
status word, and returns control to the program:
HITEMP: PROCEDURE INTERRUPT 100 PUBLIC;
CALL ANNUNCIATOR(1);
/* This will result in an output from the microprocessor
to turn on annunciator light number 1, the
high-temperature warning. */
ALERT = ALERT OR 00000010B;
/* This puts a 1 in one of the bit positions
of ALERT, which contains a bit pattern
representing current alerts. */
END HITEMP;
Reentrancy and the REENTRANT Attribute
With the REENTRANT attribute, a procedure can suspend execution temporarily,
restart with new parameters, and then later complete the original execution
successfully as if there had been no interruption.
PL/M-386 Programmer's Guide
Chapter 8
129
This ability is desirable in two circumstances: (1) if the procedure (PROC1) activates
itself (called direct recursion), or (2) if the procedure activates another procedure
(PROC2) that will reactivate PROC1 before PROC1 has finished its original processing
(called indirect recursion).
Without the REENTRANT attribute, storage for procedure variables is allocated
statically, in fixed locations within the data segment of the object module.
Re-entering such a procedure would write over the earlier contents of such locations
making it impossible to complete the original suspended execution.
When the attribute REENTRANT is used in declaring a procedure, its variables are not
stored with other variables in the data section, but are stored on the stack. Thus
preserved, each set can be used independently by each invocation of the procedure.
Hence, multiple sets of variables might need to be stored on the stack during
recursive use of such procedures. A stack size must be specified (when binding the
program module) that is large enough for all such storage needed by all multiple
invocations that may be active at one time.
A procedure with the REENTRANT attribute may be activated before it is declared.
This permits direct recursion, where the procedure activates itself and permits
indirect recursion, where the procedure activates a second procedure and the second
procedure activates the first, or activates a third procedure, which activates a fourth,
and so forth, with the result that the first procedure is activated before it terminates.
The following rules summarize the use of the REENTRANT attribute:
•
Any procedure that can be interrupted and is also activated from within an
interrupt procedure should have the REENTRANT attribute.
Note that this may apply to an interrupt procedure that runs with interrupts enabled
because it contains an ENABLE statement. If there is any possibility that it will be
interrupted by its own interrupt, it should have the REENTRANT attribute. This
situation is equivalent to recursion.
130
•
Any procedure that is directly recursive (activates itself) should have the
REENTRANT attribute.
•
Any procedure that is indirectly recursive (activates another procedure and is
activated itself as a result) should have the REENTRANT attribute.
•
Any procedure that is activated by a reentrant procedure should also have the
REENTRANT attribute. In other words, if there is any possibility that a procedure
can be activated while it is already running, it should be REENTRANT.
•
The REENTRANT attribute cannot be used in the same declaration as the
EXTERNAL attribute. (It may be used with the PUBLIC attribute.)
Chapter 8
Procedures
•
The REENTRANT attribute can only be used in a PROCEDURE statement at the
outer level of a module.
•
A procedure declaration with the REENTRANT attribute cannot have a nested
procedure declaration.
■■■
PL/M-386 Programmer's Guide
Chapter 8
131
132
Chapter 8
Procedures
Built-in Procedures, Functions,
and Variables
9
Built-in procedures, functions, and variables are already declared in the PL/M code.
This makes it unnecessary to write code to perform the particular functions that
built-ins are designed to perform. The following built-in procedures, functions, and
variables are discussed in this chapter:
•
LENGTH, LAST, and SIZE functions – these functions return information
concerning variables. For example, the SIZE function returns the number of
bytes occupied by a scalar, array, or structure.
•
Explicit type and value conversion functions – these functions provide explicit
conversion for types and values.
•
Shift and rotate functions – these functions move bits using a pattern of 8, 16, or
32 bits.
•
String manipulation procedures and functions – these procedures and functions
move strings, compare strings, search strings for a match or a mismatch,
translate strings, and set strings to a specified value.
•
Bit manipulation procedures – these functions copy (and move) a bit string and
search bit strings for a set bit.
•
MOVE bytes – this procedure moves a specified number of bytes from one
location to another.
•
Time delay – this procedure causes a time delay.
•
Lock set – this function enables a software synchronization lock.
•
Lock bit – this function enables a memory location lock.
•
POINTER and SELECTOR functions – these functions enable the manipulation of
location addresses in the microprocessor's memory.
The identifiers for these built-ins are subject to the rules of scope (described in
Chapter 7). This means that the name of a built-in procedure or variable can be
declared to have a local meaning (scope) within the program. Within the scope of
such a declaration, the built-in is unavailable. This distinguishes these identifiers
from reserved words (listed in Appendix A), which cannot be used as identifiers in
declarations.
PL/M-386 Programmer's Guide
Chapter 9
133
No built-in procedure can be used within a location reference (e.g.,
@LENGTH(LIST) ). No built-in variable can be used within a location reference,
except as specifically noted in the following sections.
Obtaining Information About Variables
PL/M has three built-in procedures that take variable names as actual parameters and
return information based on the declarations of the variables: LENGTH, LAST, and
SIZE.
The LENGTH Function
LENGTH is a built-in WORD function that returns the number of elements in an array; it
is activated by a function reference with the form:
LENGTH (variable-ref)
Where:
variable-ref must be a non-subscripted reference to an array.
The array can be a member of a structure; it cannot be an EXTERNAL array using the
implicit dimension specifier (see Chapter 3).
The value returned is the number of elements assigned to the array in the declaration
statement (i.e., the value of the dimension specifier).
If the array is not a structure member, then the reference must be an unqualified
variable reference. If the array is a structure member, then the reference is a partially
qualified variable reference. For example, given the declaration:
DECLARE RECORD STRUCTURE (KEY BYTE,
INFO(3) WORD);
LENGTH(RECORD.INFO) is a valid function reference and returns a WORD value of 3.
If the array is a member of a structure, and that structure is an element of an array, a
special case arises. Given the declaration:
DECLARE LIST (4) STRUCTURE (KEY BYTE,
INFO (3) WORD);
then all of the following function references are correct and return the value 3:
LENGTH(LIST(0).INFO)
LENGTH(LIST(1).INFO)
LENGTH(LIST(2).INFO)
LENGTH(LIST(3).INFO)
134
Chapter 9
Built-in Procedures, Functions, and Variables
In other words, the subscript for the array LIST is irrelevant when a
member-identifier is supplied, because the arrays within the structures are all the
same length. PL/M has a shorthand form of partially qualified variable reference in
the LENGTH, LAST, and SIZE function references. For example:
LENGTH(LIST.INFO)
is a valid function and returns the value 3.
The LAST Function
LAST is a built-in WORD function that returns the subscript of the last element in an
array. It is activated by a function reference with the form:
LAST (variable)
Where:
variable
must be a non-subscripted reference to an array.
The array can be a member of a structure; it cannot be an EXTERNAL array using the
implicit dimension specifier (see Chapter 3).
The value returned is the subscript of the last element of the array. For a given array,
LAST will always be one less than LENGTH. When used with a based variable, LAST
returns the value assigned in the declaration statement. This is not necessarily the
actual value.
As in the LENGTH function, a shorthand form of partially qualified variable reference
is allowed in the case where the array is a member of a structure that is also an array
element.
The SIZE Function
SIZE is a built-in WORD function that returns the number of bytes occupied by a
scalar, array or structure. It is activated by a function reference with the form:
SIZE (variable)
Where:
variable
is a fully qualified, partially qualified, or unqualified reference to any
scalar, array, or structure. The variable cannot be an EXTERNAL
declaration that uses the implicit dimension specifier (see Chapter 3).
The value returned is the number of bytes required by the variable referenced. When
used with a based variable, SIZE returns the value assigned in the declaration
statement. This is not necessarily the actual (current) value.
PL/M-386 Programmer's Guide
Chapter 9
135
If the reference is fully qualified, it refers to a scalar, and the value is the number of
bytes required for the scalar. If the reference is unqualified, it refers to an entire
structure or array, and the value is the total number of bytes required for the structure
or array.
If the reference is partially qualified, it refers either to a structure member that is an
array or nested structure, or to an array element that is a structure. The value is the
number of bytes required for the array or structure.
As in the LENGTH function, a shorthand form of partially qualified variable reference
is allowed in the case where the array or scalar is a member of a structure and the
structure is an array element.
Explicit Type and Value Conversions
The functions in this section provide explicit conversion from one data type to
another and from signed values to or from absolute magnitudes.
Explicit type and value conversion functions are invoked as:
function-name (expression)
In Tables 9-1 and 9-2, each function name is followed by the expression type
expected, the purpose of the function, and the nature of the value it returns to the
expression that invoked it. For each function there is only one possible class of
expressions (e.g., HIGH accepts only unsigned values) that can be converted. For the
type conversions (BYTE, WORD, DWORD, INTEGER, REAL, POINTER, and SELECTOR,
OFFSET, HWORD, CHARINT, and SHORTINT), the context of the entire expression is
always a signed integer value. Table 9-1 gives the value and type conversions for
PL/M-386 when the WORD32 control is in effect.
136
Chapter 9
Built-in Procedures, Functions, and Variables
Table 9-1. Value and Type Conversions for PL/M-386
Procedure
Name
Parameter
Type
LOW
BYTE
HIGH
Function
Result Returned
HWORD
Converts HWORD value to
BYTE value
Low-order BYTE of HWORD
WORD or
OFFSET
Converts WORD or OFFSET
value to HWORD value
Low-order HWORD of WORD
or OFFSET
DWORD
Converts DWORD value to
WORD value
Low-order WORD of DWORD
BYTE value unchanged
BYTE
zero
HWORD
Converts HWORD value to
BYTE value
High-order BYTE of HWORD
WORD or
OFFSET
Converts WORD or OFFSET
value to HWORD value
High-order HWORD of WORD
or OFFSET
DWORD
Converts DWORD value to
WORD value
High-order WORD of DWORD
BYTE
Converts BYTE value to
HWORD value
HWORD, by appending 8
high-order zero bits
HWORD
Converts HWORD value to
WORD value
WORD, by appending 16
high-order zero bits
WORD or
OFFSET
DWORD
Converts WORD or OFFSET
value to DWORD value
DWORD value unchanged
DWORD, by appending 32
high-order zero bits
FLOAT
CHARINT
SHORTINT
INTEGER
Converts signed integer
value to REAL value
Same value of type REAL
FIX
REAL
Converts REAL value to
INTEGER value
Same value of type INTEGER if
within range -2**31 to
+(2**31)-1 otherwise undefined
INT
BYTE
HWORD
WORD
Converts unsigned binary
value to INTEGER value,
interprets parameter as
positive
Same value of type INTEGER if
within range -2**31 to
+(2**31)-1 otherwise
SIGNED
BYTE
Converts unsigned integer
value to INTEGER value
BYTE value is extended with
24 high-order zeros
HWORD value is extended with
16 high-order zeros
WORD value unchanged
DOUBLE
HWORD
WORD
PL/M-386 Programmer's Guide
Chapter 9
137
continued
138
Chapter 9
Built-in Procedures, Functions, and Variables
Table 9-1. Value and Type Conversions for PL/M-386 (continued)
Procedure
Name
Parameter
Type
UNSIGN
Function
Result Returned
CHARINT
SHORTINT
INTEGER
Converts INTEGER value to
WORD value
Signed INTEGER value is
interpreted as unsigned WORD
value
ABS
REAL
Converts negative real value
to positive real value
Absolute value of parameter:
value unchanged if positive
-(value) if negative. Result type
is same as parameter type.
IABS
CHARINT
SHORTINT
INTEGER
Converts negative integer to
positive integer
Absolute value of parameter:
value unchanged if positive (value) if negative. If -(value) is
out of range, result is
undefined. Result type is same
as parameter type.
BYTE
any
unsigned
type
Converts any unsigned type
to BYTE
BYTE value, by truncation
any signed
type
Converts any signed type to
BYTE
BYTE value, by truncation
REAL
Converts any REAL type to
BYTE
BYTE (CHARINT (real) )
SELECTOR
Converts SELECTOR to
BYTE
BYTE value, by truncation
POINTER
Converts offset portion of
POINTER to BYTE
BYTE (OFFSET$OF (pointer))
continued
PL/M-386 Programmer's Guide
Chapter 9
139
Table 9-1. Value and Type Conversions for PL/M-386 (continued)
Procedure
Name
Parameter
Type
Function
Result Returned
HWORD
any
unsigned
Converts any unsigned type
to HWORD
HWORD value, by truncation or
zero extension
any signed
type
Converts any signed type to
HWORD
HWORD value, by truncation or
sign extension
REAL
Converts any real type to
HWORD
HWORD (SHORTINT (real) )
SELECTOR
Converts SELECTOR to
HWORD
HWORD type, value
unchanged
POINTER
Converts offset portion of
POINTER to HWORD
HWORD (OFFSET$OF
(pointer) )
any
unsigned
type
Converts any unsigned type
to WORD
WORD value, by truncation or
zero extension
any signed
type
Converts any signed type to
WORD
WORD value, by sign
extension
REAL
Converts any real type to
WORD
WORD (INTEGER (real) )
SELECTOR
Converts SELECTOR to
WORD
WORD value, by zero
extension
POINTER
Converts offset portion of
POINTER to WORD
WORD (OFFSET$OF
(pointer) )
WORD
continued
140
Chapter 9
Built-in Procedures, Functions, and Variables
Table 9-1. Value and Type Conversions for PL/M-386 (continued)
Procedure
Name
Parameter
Type
DWORD
CHARINT
Function
Result Returned
any
unsigned
type
Converts any unsigned type
to DWORD
DWORD value, by zero
extension
any signed
type
Converts any signed type to
DWORD
DWORD value, by sign
extension
REAL
Converts any real type to
DWORD
DWORD (INTEGER (real) )
SELECTOR
Converts SELECTOR to
DWORD
DWORD value, by zero
extension
POINTER
Converts offset portion of
POINTER to DWORD
DWORD (OFFSET$OF
(pointer) )
any
unsigned
Converts any unsigned type
to CHARINT
CHARINT value, by truncation
any signed
type
Converts any signed type to
CHARINT
CHARINT value, by signextension
REAL
Converts any real type to
CHARINT
CHARINT (FIX(real) )
SELECTOR
Converts SELECTOR to
CHARINT
CHARINT value, by truncation
POINTER
Converts offset portion of
POINTER to CHARINT
CHARINT (OFFSET$OF
(pointer) )
continued
PL/M-386 Programmer's Guide
Chapter 9
141
Table 9-1. Value and Type Conversions for PL/M-386 (continued)
Procedure
Name
Parameter
Type
SHORTINT
INTEGER
REAL
Function
Result Returned
any
unsigned
type
Converts any unsigned type
to SHORTINT
SHORTINT value, by zero
extension or truncation
any signed
type
Converts any signed type to
SHORTINT
SHORTINT value, by sign
extension
REAL
Converts any real type to
SHORTINT
SHORTINT (FIX (real) )
SELECTOR
Converts SELECTOR to
SHORTINT
SHORTINT value
POINTER
Converts offset portion of
POINTER to SHORTINT
SHORTINT (OFFSET$OF
(pointer) )
any
unsigned
type
Converts any unsigned type
to INTEGER
INTEGER value, by zero
extension or truncation
any signed
type
Converts any signed type to
INTEGER
INTEGER value, by sign
extension
REAL
Converts any real type to
INTEGER
INTEGER (FIX (real) )
SELECTOR
Converts SELECTOR to
INTEGER
INTEGER value, by zero
extension
POINTER
Converts offset portion of
POINTER to INTEGER
INTEGER (OFFSET$OF
(pointer) )
any
unsigned
type (except
OFFSET)
Converts any unsigned type
to REAL
REAL (SIGNED (unsigned) )
any signed
type
Converts any signed type to
REAL
FLOAT (signed)
REAL
value unchanged
continued
142
Chapter 9
Built-in Procedures, Functions, and Variables
Table 9-1. Value and Type Conversions for PL/M-386 (continued)
Procedure
Name
Parameter
Type
SELECTOR
any
unsigned
binary type
Function
Result Returned
Converts any unsigned
binary type to SELECTOR
SELECTOR value, by zero
extension or truncation
OFFSET
any
unsigned
integer data
type
OFFSET
POINTER
Current data segment selector
Converts any signed integer
data type to SELECTOR
SELECTOR value by sign
extension or truncation
POINTER
Selector portion of the
POINTER
REAL
Cannot be used
any
unsigned
type
Converts any unsigned type
OFFSET
OFFSET, by zero extension or
truncation
any signed
type
Converts any signed type to
OFFSET
OFFSET, by sign extension
SELECTOR
zero (0)
POINTER
OFFSET$OF (pointer)
any
unsigned
type
Converts value of any
unsigned type to POINTER
BUILD$PTR (DS, OFFSET
(unsigned) ) (DS is selector of
current data segment)
any signed
type
Converts value of any signed
type to POINTER
BUILD$PTR (DS, OFFSET
(signed) ) (DS is selector of
current data segment)
SELECTOR
BUILD$PTR (SELECTOR, 0)
OFFSET
BUILD$PTR (DS, OFFSET)
(DS is selector of current data
segment)
Notes:
Conversions from REAL to OFFSET, or POINTER, and vice versa, are not allowed. Under WORD32 (the default),
LONGINT is equivalent to INTEGER. ADDRESS is equivalent to OFFSET.
PL/M-386 Programmer's Guide
Chapter 9
143
The PL/M-386 LOW, HIGH, and DOUBLE Functions
The PL/M-386 LOW built-in function converts DWORD values to WORD values, WORD or
OFFSET values to HWORD values, and HWORD values to BYTE values. LOW is activated
using the following form:
LOW (expression)
Where:
expression
has an unsigned binary number type.
If expression has a DWORD value, LOW returns the value of the low-order (least
significant) WORD of the expression value. If expression has a WORD or OFFSET
value, LOW returns the value of the low-order (least significant) HWORD of the
expression value. If expression has an HWORD value, LOW returns the value of
the low-order (least significant) BYTE of the expression value. If expression
has a BYTE value, LOW returns this value unchanged.
The PL/M-386 HIGH built-in function converts DWORD values to WORD values, WORD
or OFFSET values to HWORD values, and HWORD values to a BYTE values. HIGH is
activated using the following form:
HIGH (expression)
Where:
expression
has an unsigned binary number type.
If expression has a DWORD value, HIGH returns the value of the high-order (most
significant) WORD of the expression value. If expression has a WORD or OFFSET
value, HIGH returns the value of the high-order (most significant) HWORD of the
expression value. If expression has an HWORD value, HIGH returns the value of
the high-order (most significant) BYTE of the expression value. If expression
has a BYTE value, then HIGH will return a zero.
144
Chapter 9
Built-in Procedures, Functions, and Variables
The PL/M-386 DOUBLE built-in function converts BYTE values to HWORD values,
HWORD values to WORD values, and WORD or OFFSET values to DWORD values. DOUBLE
is activated using the following form:
DOUBLE (expression)
Where:
expression
has an unsigned binary number type.
If expression has a BYTE value, the DOUBLE function appends 8 high-order zero
bits to convert the expression to an HWORD value and returns this HWORD value. If
expression has an HWORD value, the DOUBLE function appends 16 high-order zero
bits to convert the expression to a WORD value and returns this WORD value. If
expression has a WORD or OFFSET value, the DOUBLE function appends 32
high-order bits to convert it to a DWORD value and returns this DWORD value. If
expression has a DWORD value, the DOUBLE function returns this DWORD value
unchanged.
The FLOAT Function
FLOAT is a built-in REAL function that converts a signed integer to the real number
data type. It is activated by a function reference with the following form:
FLOAT (expression)
Where:
expression
is a signed integer.
FLOAT converts the signed integer to the corresponding real number data type and
returns the real number. FLOAT can be replaced with
REAL (expression).
The FIX Function
FIX is a built-in INTEGER function that converts a REAL value to an INTEGER value.
It is activated by a function reference with the following form:
FIX (expression)
Where:
expression
has a REAL value.
FIX rounds the REAL value to the nearest INTEGER. If both INTEGER values are
equally near, FIX rounds to the even value. The resulting INTEGER value is then
returned.
PL/M-386 Programmer's Guide
Chapter 9
145
For example:
FIX(1.4)
FIX(-1.8)
FIX(3.5)
FIX(6.5)
/*
/*
/*
/*
would result in the INTEGER value 1, */
in -2, */
in 4, and */
in 6. */
If the result calculated by FIX is not within the implemented range of INTEGER
values, the result is undefined.
✏
Note
FIX is affected by the rounding mode; see Chapter 10. The default
mode (round to the nearest or even value) is used in the previous
examples.
FIX can be replaced with INTEGER (expression).
The INT Function
INT is a built-in INTEGER function that converts an unsigned binary value, excluding
DWORD values, to the signed integer data type. It is activated by a function reference
with the following form:
INT (expression)
Where:
expression
has an unsigned binary data type, excluding DWORD.
INT interprets the expression value as a positive number and returns the
corresponding INTEGER value.
If the result calculated by INT is not within the implemented range of INTEGER
values, the result is undefined (see Chapter 5 for ranges for INTEGER values).
The SIGNED Function
For PL/M-386, SIGNED is a built-in INTEGER function that converts a BYTE, HWORD,
or WORD value to an INTEGER value. SIGNED is activated by a function reference
with the following form:
SIGNED (expression)
Where:
expression
146
Chapter 9
has an unsigned binary number data type, excluding DWORD.
Built-in Procedures, Functions, and Variables
If expression has a BYTE or HWORD value, it will be extended by 24 or 16
high-order 0 bits, respectively, to produce a WORD value.
SIGNED interprets the WORD value as a 32-bit two's-complement number and returns
the corresponding integer value.
If the highest-order (most significant) bit of the WORD value is a 0, SIGNED interprets
the WORD value as a positive number and returns the corresponding INTEGER value.
For example:
SIGNED (0000$0000$0000$0100B)
returns an INTEGER value of 4.
If the highest-order bit of the WORD value is a 1, SIGNED returns a negative INTEGER
value whose absolute magnitude is the two's complement of the WORD value. For
example:
SIGNED(1111$1111$1111$1100B)
returns an INTEGER value of -4.
SIGNED can be replaced by INTEGER (expression).
The UNSIGN Function
The UNSIGN built-in function converts a signed integer to a WORD value. It is
activated by a function reference with the following form:
UNSIGN (expression)
Where:
expression
is a signed integer.
UNSIGN converts the INTEGER value to a WORD value.
If the INTEGER value is positive, the WORD value will be numerically the same as the
INTEGER value. However, if the INTEGER value is negative, the WORD value will be
the two's complement of the absolute magnitude of the INTEGER value. For
example:
UNSIGN(-4)
returns a WORD value of:
1111$1111$1111$1100B
UNSIGN can be replaced by WORD (expression).
PL/M-386 Programmer's Guide
Chapter 9
147
The Unsigned Binary Data Type Built-in Functions
The unsigned binary data type built-in functions convert any expression to the
specified unsigned binary data type. For example, the WORD and DWORD built-in
functions convert any expression to a WORD or DWORD value, respectively.
The built-in functions are activated with the form:
built-in (expression)
Where:
built-in
is the name of the data type to which the given expression is
converted (e.g., BYTE or WORD).
expression
has any value.
For example, WORD (INT1) converts the value of INT1 to a WORD value.
If expression is an unsigned binary number, it is converted by truncation or zero
extension, if necessary. If expression is a signed integer, it is converted by
truncation or sign extension, if necessary. If expression is a selector, it is
converted by truncation or zero extension. If expression is a pointer, the offset
portion of the pointer is converted by truncation or zero extension; the selector
portion of the pointer is discarded. If expression is a real number, it is first
converted to a signed integer using the numeric coprocessor's real to integer
conversion, then the resulting value is converted to the unsigned binary number data
type by truncation, if necessary.
Signed Integer Data Type Built-in Function
The signed integer data type built-in function converts any expression to a signed
integer data type. It has the form:
INTEGER (expression)
For example:
INTEGER (D)
converts the value of D to an INTEGER value within the INTEGER range.
If expression is an unsigned binary number or selector, it is converted by
truncation or zero extension. If expression is a pointer, the offset portion of the
pointer is converted by truncation or zero extension; the selector portion of the
pointer is discarded. If expression is a real number, it is converted using the
numeric coprocessor's real to integer conversion.
Specific to PL/M-386, if expression is a signed type, it is converted by sign
extension. Shorter data types are converted into longer data types by sign extending
148
Chapter 9
Built-in Procedures, Functions, and Variables
the shorter data type value. Longer data types are converted into shorter data types
by sign extension of the bits equivalent to the shorter data type. For example, if a
CHARINT built-in is used to convert an INTEGER value, the least significant 8 bits are
sign extended and the value returned is guaranteed to be in the CHARINT range.
REAL Built-in Functions
The REAL built-in function converts an expression to a REAL value. Expressions of
type SELECTOR, OFFSET, and POINTER cannot be converted. The conversion is
done using the numeric coprocessor's INTEGER to REAL conversion. If the
expression is an unsigned binary number it is zero extended, if necessary, and
interpreted as a signed value.
The SELECTOR Built-in Function
The SELECTOR built-in function converts any expression (except the real number
data type) to a SELECTOR value. If the expression is any unsigned binary number,
except OFFSET, it is truncated or zero extended to 16 bits. If the expression is a
signed integer, it is truncated or sign extended to 16 bits. If the expression is of type
POINTER, the selector portion of the pointer is returned. If the expression is of type
OFFSET, the current data segment selector is returned. Expressions of type REAL
cannot be converted.
The POINTER Built-in Function
The POINTER built-in function converts any expression (except real numbers) to a
POINTER value. If the expression is any unsigned binary number or signed integer, it
is converted to type OFFSET by truncation, zero, or sign extension, if necessary. This
OFFSET value is combined with the SELECTOR value of the current data segment to
create the POINTER value. If the expression is of type SELECTOR, it is combined
with an OFFSET value of zero to create the POINTER value. Expressions of type
REAL cannot be converted.
PL/M-386 Programmer's Guide
Chapter 9
149
The OFFSET Built-in Function
The OFFSET built-in function converts any expression (except real numbers) to an
OFFSET value. If the expression is any unsigned binary number or signed integer
data type, it is converted to type OFFSET by truncation, or by zero or sign extension.
If the expression is of type SELECTOR, an OFFSET value of zero is returned. If the
expression is of type POINTER, the offset portion of the pointer is returned. ADDRESS
values are equivalent to OFFSET. Expressions of type REAL cannot be converted.
The ABS and IABS Functions
The ABS built-in function returns the absolute value of a real number. It is activated
by a function reference with the following form:
ABS (expression)
Where:
expression
is a real number.
If the value of expression is positive, ABS returns it unchanged. If the value of
expression is negative, ABS returns -(expression).
The IABS built-in function returns the absolute value of a signed integer. It is
activated by a function reference with the following form:
IABS (expression)
Where:
expression
is a signed integer.
If the value of expression is positive, IABS returns it unchanged. If the value of
expression is negative, IABS returns -(expression).
150
Chapter 9
Built-in Procedures, Functions, and Variables
Shift and Rotate Functions
With the shift and rotate functions, bit patterns can be moved to the right and to the
left. In a shift, bits moved off one end of the pattern are lost, and zero bits move into
the pattern from the other end (except in the case of the algebraic shift right function,
SAR). In a rotate, bits moved off one end of the pattern are moved onto the other end
of the pattern. It is not possible to perform a rotate on a signed integer algebraic
pattern.
In PL/M-386, a value is handled as a pattern of 8 bits for a BYTE or CHARINT value,
16 bits for a HWORD or SHORTINT value, 32 bits for WORD, OFFSET, or INTEGER
values, or 64 bits for a DWORD value. The pattern is moved to the right or left by a
specified number of bits called the bit count.
Rotation Functions
The type of the rotate left (ROL) and rotate right (ROR) built-in functions depends on
the type of expression given as an actual parameter. These built-ins are activated by
function references with the following forms:
ROL (pattern, count)
ROR (pattern, count)
Where:
pattern and count are unsigned binary numbers.
If count is any unsigned binary number data type except BYTE, all but the low-order
bits will be dropped to produce a BYTE value. If the value of count is 0, no rotation
occurs.
The value of pattern is handled as an 8-bit, 16-bit, 32-bit, or 64-bit quantity. The
type of pattern determines which of the unsigned binary number data types is used.
This, in turn, determines the value of pattern. The number of bit positions by
which pattern is rotated is specified by count.
The following are examples of the action of these procedures:
ROR (10011101B, 1)
ROL (10011101B, 2)
ROR (1101011010011010B, 9)
PL/M-386 Programmer's Guide
returns a value of 11001110B
returns a value of 01110110B
returns a value of 0100110101101011B
Chapter 9
151
Logical-shift Functions
The type of the logical-shift left (SHL) and logical-shift right (SHR) built-in functions
depends on the type of the expression given as an actual parameter. SHL and SHR are
activated by function references with the forms:
SHL (pattern, count)
SHR (pattern, count)
Where:
pattern and count are expressions using an unsigned binary number data type.
If count is any unsigned binary number data type except BYTE, all but the 8
low-order bits will be dropped to produce a BYTE value. If the value of count is 0,
no shift occurs.
The value of pattern can be a BYTE, HWORD, WORD, or DWORD value and the value
will not be converted. If pattern is a BYTE value, the function will return a BYTE
value. If pattern is an HWORD value, the function will return an HWORD value. If
pattern is a WORD value, the function will return a WORD value; if pattern is a
DWORD value, the function will return a DWORD value.
The value of pattern is shifted left (by SHL) or right (by SHR), with the bit count
given by count.
A shift operation can force one bit out of the pattern. For example:
SHL(1000$0001B,1)
returns 0000$0010B, losing the former high-order bit, and:
SHR(1000$0001B,1)
becomes 0100$0000B, losing the former low-order bit.
If the specified pattern and count do not lose information, a shift of one bit
position has the effect of multiplication by two for a left shift, or division by two for a
right shift. For example, suppose that VAR is a BYTE variable with a value of eight.
This is represented as 0000$1000B. SHL(VAR,1) would return 0001$0000B, which
represents 16, and SHR(VAR,1) would return 0000$0100B, which represents four.
Casting can be used to ensure that no information is lost in a shift, as in the following
example:
SHL(WORD(LIT$MASK),3)
152
Chapter 9
Built-in Procedures, Functions, and Variables
Algebraic-shift Functions
The type of the algebraic-shift left (SAL) and algebraic-shift right (SAR) built-in
functions depends on the type of the expression given as an actual parameter. SAL
and SAR are activated by function references with the following forms:
SAL (pattern, count)
SAR (pattern, count)
Where:
pattern
is an expression using a signed integer data type.
count
is an expression using an unsigned binary data type.
If count is any unsigned binary data type except BYTE, all but the 8 low-order bits
will be dropped to produce a BYTE value. If the value of count is zero, no shift
occurs.
For PL/M-386, the type of pattern can be a CHARINT, SHORTINT, or INTEGER
value. All values are converted to INTEGER before the shift operations, and an
INTEGER value is returned.
In a left shift (SAL), zero-bits move into the pattern from the right (as in SHL and
SHR).
In a right shift (SAR), either zero-bits or one-bits move into the pattern from the left.
If the original value of pattern is positive, the sign bit (leftmost bit) is a 0, and zerobits move in from the left. If the original value is negative, the sign bit is a 1, and
one-bits move in from the left.
In some instances (as in logical shifts), an algebraic shift of one bit position can have
the effect of multiplication by two for a left shift or division by two for a right shift.
For example, suppose that VAL is an INTEGER variable with a value of -8. This value
is 1111$1111$1111$1000B. SAL(VAL,1) would return 1111$1111$1111$0000B,
which is -16, and SAR(VAL,1) would return 1111$1111$1111$1100B, which is -4.
PL/M-386 Programmer's Guide
Chapter 9
153
Concatenate Functions
The concatenate functions (SHLD and SHRD) are built-in WORD double-shift functions
that concatenate two WORD values to form a 64-bit string, shift the concatenated
pattern left (SHLD) or right (SHRD) by count bits, and return the destination WORD.
These built-ins are activated by function references with the following form:
keyword (high pattern, low pattern, count)
Where:
keyword
is SHLD or SHRD.
high pattern
is a WORD value.
low pattern
is a WORD value.
count
is a BYTE, HWORD, or WORD value that determines how many bits to shift
the concatenated pattern.
SHLD concatenates the bit pattern of the WORD value high pattern with the bit
pattern of the WORD value low pattern to form a 64-bit string. high pattern is
placed in the high 32 bits and low pattern is placed in the low 32 bits. The
concatenated pattern is shifted left by the number of bits given by count MODULO 32.
These operands are taken MODULO 32 to provide a number between 0 and 31 by
which to shift. This has the effect of shifting the high order bits of low pattern
into the low order bits of high pattern. SHLD returns the high 32 bits of the
shifted pattern.
SHRD concatenates the bit pattern of the WORD value high pattern with the bit
pattern of the WORD value low pattern to form a 64-bit string. high pattern is
placed in the high 32 bits and low pattern is placed in the low 32 bits. The
concatenated pattern is shifted right by the number of bits given by count MODULO
32. These operands are taken MODULO 32 to provide a number between 0 and 31 by
which to shift. This has the effect of shifting the low order bits of high pattern
into the high order bits of low pattern. SHRD returns the low 32 bits of the shifted
pattern.
154
Chapter 9
Built-in Procedures, Functions, and Variables
String Manipulation Procedures and Functions
The term string is used here in a broader sense than previously, in which string was
used to refer to a BYTE string. In this section, a string is any contiguously stored set
of unsigned binary number data type values (excluding DWORD and OFFSET). A
string can be regarded as if it were an unsigned binary number type (excluding
DWORD and OFFSET) array, and the array items can be referred to as elements.
The word index refers to the position of a given element within a string. The index is
similar to the subscript of an array reference. Thus, the index of the first element of a
string is 0, the index of the second element is 1, and so on.
In the following descriptions, the location of a string always means the location of its
first element. In each string manipulation procedure, the location of a string is
specified by a parameter called source or destination, which is an expression
with a POINTER value. The source points to the lowest element. For example, with
MOVB and MOVW, the lowest element (element 0) is the first element to be processed.
With MOVRB and MOVRW, the lowest element is the last element to be processed, as
discussed in the following sections.
The length of a string is the number of elements it contains. In each string
manipulation procedure, the number of elements to be processed is specified by a
parameter called count.
✏
Note
If the source or destination string address is in SELECTOR or
WORD form, use the @ operator of a variable based on the address.
Otherwise, the built-in function BUILD$PTR can be used to
construct the pointer-parameter for the string built-in.
In PL/M-386, each of the string-manipulation procedures described in the following
sections (except XLAT) is available for BYTE, HWORD, and WORD strings.
PL/M-386 Programmer's Guide
Chapter 9
155
The Copy String in Ascending Order Procedure
MOVxx is an untyped procedure that copies a string of length count from one
location to another. It is activated by a CALL statement with the following form:
CALL keyword (source, destination, count);
Where:
keyword
MOVB, MOVHW, MOVW
source and destination
expressions with POINTER values
expression with BYTE, HWORD, OFFSET, or WORD value
count
MOVB copies a BYTE string, MOVHW copies an HWORD string, and MOVW copies a WORD
string.
The string elements are copied in ascending order (i.e., element 0 is copied first, then
element 1, etc.). This order is significant if the source string and the destination
string overlap. If the value of destination is higher than the value of source, and
the two strings overlap, elements in the overlap area will be overwritten before they
are copied. To avoid the overwriting, use MOVRxx instead of MOVxx.
The Copy String in Descending Order Procedure
MOVRxx is an untyped procedure that copies a string of length count from one
location to another. It is activated by a call statement with the following form:
CALL keyword (source, destination, count);
Where:
keyword
MOVRB, MOVRHW, MOVRW
source and destination
expressions with POINTER values
count
156
expression with BYTE, HWORD, OFFSET, or WORD value
Chapter 9
Built-in Procedures, Functions, and Variables
The MOVRB built-in procedure is similar to the MOVB procedure except that the
elements in the MOVRB source string are copied to the destination string in
descending order (i.e., element (count-1) is copied first, then element (count-2), and
so on, with element 0 copied last). This order is significant when the two strings
overlap. If the value of destination is higher than the value of source, and an
overlap exists, elements in the overlap area will not be overwritten until they have
been copied. However, if the value of source is higher than the value of
destination, elements in the overlap area will be overwritten before they are
copied.
MOVHW performs the same function as MOVRB except that MOVHW copies an HWORD
string.
MOVRW performs the same function as MOVRB, except MOVRW copies a WORD string
instead of a BYTE string.
✏
Note
If two strings overlap, use a procedure such as the following to
make the correct choice between MOVB and MOVRB. This ensures
that elements in the overlap area will not be overwritten until after
they have been copied.
MOVBYTES: PROCEDURE (SRC, DST, CNT);
DECLARE (SRC, DST) POINTER, CNT HWORD;
IF (OFFSET(SRC)) > (OFFSET(DST)) THEN
CALL MOVB (SRC, DST, CNT);
ELSE CALL MOVRB (SRC, DST, CNT);
END MOVBYTES;
This procedure can be activated without the need to consider whether overlap may
occur or whether source or destination is higher.
The Compare String Function
CMPxx is a built-in WORD function that compares two strings of length count. It is
activated by a function reference with the following form:
keyword (source1, source2, count)
Where:
keyword
CMPB, CMPHW, CMPW
source1 source2
expressions with POINTER values
count
expression with BYTE, HWORD, OFFSET, or WORD value
PL/M-386 Programmer's Guide
Chapter 9
157
CMPB compares two BYTE strings of length count whose locations are source1 and
source2. It remains a 32-bit instruction even if the WORD16 control is in effect.
If every element in the string at source1 is equal to the corresponding element in the
string at source2, CMPxx returns a WORD value, 0FFFFFFFFH, for PL/M-386.
Otherwise, CMPxx returns the index (position within the strings) of the first pair of
elements found to be unequal.
CMPHW performs the same function as CMPB, except that CMPHW compares two HWORD
strings. CMPW performs the same as function as CMPB except that CMPW compares
two WORD strings instead of two BYTE strings.
The Find Element Functions
FIND is a built-in WORD function that searches a string to find an element that has a
specified value. It is activated by a function reference of the following form:
keyword (source, target, count)
Where:
keyword
FINDB, FINDHW, FINDW, FINDRB, FINDRHW, FINDRW
source
expression with POINTER value
target
expression with BYTE, HWORD, or WORD value
count
expression with BYTE, HWORD, OFFSET, or WORD value
FINDB examines each element of the source string (in ascending order) until it finds
an element whose value matches the BYTE value of target, or until count elements
have been searched, with none of them having matched the target. If the search is
successful, FINDB returns the index of the first element of the string that matches
target. If the search is unsuccessful, FINDB returns a WORD value.
FINDHW performs the same function as FINDB, except that FINDHW searches an
HWORD string. If target has a BYTE value, it is extended by 8 high-order, 0-bits to
produce an HWORD value. If target has a WORD value, it is truncated by 16
high-order bits to produce an HWORD value.
FINDW performs the same function as FINDB, except that FINDW searches a WORD
string. If target has a BYTE or HWORD value, target is extended appropriately to
produce a WORD value.
158
Chapter 9
Built-in Procedures, Functions, and Variables
FINDRB performs the same function as FINDB, except that FINDRB searches the
source string in descending order. Thus, if each search is successful, FINDRB
returns the index of the last (highest subscript) element that matches the BYTE value
of target. FINDRHW performs the same function as FINDRB, except that FINDRHW
searches an HWORD string (in descending order). FINDRW searches a WORD string (in
descending order).
The Find String Mismatch Function
SKIPB is a built-in WORD function that searches the BYTE string of length count at a
specified location (given by source) for the first BYTE value that does not match the
target BYTE. This search begins with the first BYTE value of the string. The result is
a WORD value, either 0FFFFFFFFH if the string contains only BYTE values equal to
the target BYTE, equal to the index of the first BYTE value not equal to the target
BYTE.
The function is activated by a function reference of the following form:
keyword (source, target, count)
Where:
keyword
SKIPB, SKIPHW, SKIPW, SKIPRB, SKIPRHW, SKIPRW
source
expression with POINTER value
target
expression with BYTE, WORD, or HWORD value
count
expression with BYTE, HWORD, OFFSET, or WORD value
SKIPW performs the same function as SKIPB, except that SKIPW searches a WORD
source string to find the first element that does not match the WORD value of target.
SKIPHW performs the same function as SKIPB, except that SKIPHW searches an
HWORD source string to find the first element that does not match the HWORD value of
target.
SKIPRB searches a BYTE string of the length specified by count, at the location
given by source, for the last BYTE value that does not match the target BYTE. This
search begins with the last BYTE value in the string. The result is a WORD value
(0FFFFFFFFH) if the string contains only BYTE values equal to the target BYTE, or
the index of the last BYTE value, if the last BYTE value in the string is not equal to the
target BYTE.
SKIPRW performs the same function as SKIPRB, except that SKIPRW searches for the
last element in the WORD source string that does not match the WORD value of the
target. SKIPRHW searches for the last element in the HWORD source string that does
not match the HWORD value of the target.
PL/M-386 Programmer's Guide
Chapter 9
159
The Translate String Procedure
XLAT is an untyped procedure that uses a translation table to translate a BYTE string
to produce another BYTE string. It is activated by a CALL statement of the form:
CALL XLAT (source, destination, count, table)
Where:
source, destination, table
expressions with POINTER values
count
expression with BYTE, HWORD, OFFSET, or WORD value
XLAT translates the count BYTE elements in the source string, placing the
translated elements in the destination string. The value of table is assumed to
be the location of a BYTE string of up to 256 elements. This string is used as a
translation table.
The value of an element in the source string is used as an index into the translation
table. The index selects one element from the translation table; this element is then
copied into the destination string.
For example, if the fifth element in the source string is 202, then 202 is used as an
index for the translation table. The 203rd element of the table is copied into the fifth
position in the destination string.
The elements of the source string are translated into the destination string in
ascending order.
160
Chapter 9
Built-in Procedures, Functions, and Variables
The Set String to Value Procedure
The SET built-in is an untyped procedure that sets each element of a BYTE string, the
length of which is specified by count, to a single specified value. SET is activated
by a CALL statement with the following form:
CALL keyword (newvalue, destination, count)
Where:
keyword
SETB, SETHW, SETW
newvalue
expression with BYTE, HWORD, OFFSET, or WORD value -- the
high-order bits are dropped to produce a BYTE WORD, or HWORD value
destination
expression with POINTER value
count
expression with BYTE, HWORD, OFFSET, or WORD value
SETB assigns the BYTE value of newvalue to each element of a BYTE string.
SETW performs the same function as SETB except that SETW assigns a single WORD
value to each element of a WORD string. If newvalue has a BYTE or an HWORD value,
it will be extended by 24 or 8 high-order 0 bits, respectively, to produce a WORD
value.
For information on WORD32|WORD16 mapping, see Tables 9-3, 10-1, and 11-3.
PL/M-386 Programmer's Guide
Chapter 9
161
PL/M-386 Bit Manipulation Built-ins
The Copy Bit String Procedure
MOVBIT is an untyped built-in procedure that copies a bit string of length count from
one location to another. MOVBIT is activated by a CALL statement with the following
form:
CALL MOVBIT (sbase, sbitoffset, dbase, dbitoffset, count);
Where:
sbase and dbase
are expressions with POINTER values.
sbitoffset
are expressions with BYTE, HWORD, OFFSET,
dbitoffset and count
are WORD values.
The MOVBIT built-in procedure moves the number of bits specified by count from
the bit location given by the base address sbase, and the bit offset sbitoffset, to
the location given by the base address dbase and the bit offset dbitoffset. These
bits are moved beginning with the low-order bit (least significant bit).
The MOVRBIT built-in procedure performs the same function as MOVBIT, except that
MOVRBIT moves bits in descending order, beginning with the high-order bit (most
significant bit).
The Find Set Bit Function
SCANBIT is a built-in WORD function that searches a bit string to find a set bit (i.e., a
bit with the value of 1). SCANBIT is activated by a function reference with the
following form:
SCANBIT (sbase, sbitoffset, count)
Where:
sbase
is an expression with a POINTER value.
sbitoffset and count
are expressions with BYTE, HWORD, OFFSET, or WORD values.
162
Chapter 9
Built-in Procedures, Functions, and Variables
The SCANBIT built-in function searches the bit string of length count at the bit
location given by the base address sbase and the bit offset sbitoffset for the first
set bit, beginning with the low-order bit (least significant bit) in the string. The result
of SCANBIT is either a WORD value of 0FFFFFFFFH if the string contains all 0 bits, or
the index of the first set bit.
SCANRBIT performs the same function as SCANBIT, except that SCANRBIT starts at
the high-order bit (most significant bit) in the string and searches for a set bit, in
descending order, and returns the location of the first set bit it encounters. The result
of SCANRBIT is either a WORD value of 0FFFFFFFFH if the string contains all 0-bits,
or the index of the first set bit encountered.
PL/M-386 Programmer's Guide
Chapter 9
163
Miscellaneous Built-ins
The Move Bytes Procedure
MOVE is an untyped procedure that moves the number of bytes specified by count to
the location given by the value of destination, starting at the location given by the
value of source. If the source and destination fields overlap, the result is
undefined. MOVE is provided for compatibility with PL/M-80 programs. MOVE is
activated by a CALL statement with the following form:
CALL MOVE (count, source, destination)
Where:
count
expression with BYTE, HWORD, OFFSET, or WORD value
source and destination
expressions with OFFSET values
If either source or destination is a value other than the value OFFSET, the value
will be extended by high-order 0 bits to produce the OFFSET value. The values of
source and destination are assumed to be the addresses of the source string
and the destination string.
The operation of the MOVE procedure differs from the MOVB procedure, as follows:
164
•
The source and destination parameters must be OFFSET values or they will
be converted. POINTER values cannot be used, nor can values be supplied with
the @ operator. Thus, MOVE can handle only strings whose locations can be
expressed as OFFSET addresses.
•
The parameter order is different from the one used by the other built-in string
functions.
•
The results are always undefined if the source and destination strings
overlap.
Chapter 9
Built-in Procedures, Functions, and Variables
The Time Delay Procedure
TIME is an untyped built-in procedure that causes a time delay specified by its actual
parameter. TIME is activated by a CALL statement with the following form:
CALL TIME (expression);
where the expression is converted, if necessary, to an HWORD quantity. The length
of time measured by the procedure is a multiple of 100 microseconds. If the actual
parameter evaluates to n, then the delay caused by the procedure is 100n
microseconds. For example, the statement:
CALL TIME (45);
causes a delay of 4.5 milliseconds. For PL/M-386, the maximum delay is 12 hours.
If required, longer delays can be obtained by repeated activations. The following
block takes one second to execute:
DO I = 1 TO 40;
CALL TIME (250);
END;
The TIME procedure is based on the microprocessor's CPU cycle times. The TIME
procedure assumes 16 MHz for Intel386 and Intel486 microprocessors.
Note that in generating code for a call to TIME, the computer generates a loop rather
than using interrupt processing. If a task containing a time delay is swapped out in a
multi-tasking environment, the time delay of that task stops executing.
The Lock Set Function
LOCKSET is a built-in BYTE function that enables implementation of a simple
software synchronization lock. It is called by a function reference with the following
form:
LOCKSET (lockptr, newvalue)
Where:
lockptr
expression with POINTER value
newvalue
expression with BYTE, HWORD, or WORD value -- the high-order bits are
dropped to produce a BYTE value
The action of LOCKSET is as follows: the lockptr parameter is used as a pointer to
a BYTE variable; the value of newvalue is assigned to this variable, and LOCKSET
returns the original value of the variable. During this transaction, the CPU prevents
any other process from accessing the same memory location.
PL/M-386 Programmer's Guide
Chapter 9
165
To see how this facility can be used, assume a system has more than one
microprocessor using the same memory, and has a program in one of these
microprocessors. This program uses memory locations that are also used by other
microprocessors in the system.
Within certain critical regions of the program, it is critical that no other
microprocessor can access the shared memory locations. To achieve this, declare a
global BYTE variable called LOCK, and establish a convention that if LOCK=0, any
microprocessor in the system can access the shared memory locations. However, if
LOCK=1, no microprocessor can access the shared memory locations except for the
microprocessor that set LOCK to 1.
Write the function reference LOCKSET(@LOCK,1). The value 1 will be assigned to
LOCK. If the value returned by LOCKSET is 0, then LOCK has not been set, and this
microprocessor is the one that set it. At the end of the critical region, the lock must
be released by writing LOCK=0.
If LOCKSET returns a value of 1, then LOCK has been set and this microprocessor was
not the one that set LOCK. Wait until a LOCKSET(@LOCK,1) function reference
returns a value of 0 before accessing the shared memory locations.
166
Chapter 9
Built-in Procedures, Functions, and Variables
Thus, the program could contain the following construction:
/*Begin critical region*/
DO WHILE LOCKSET(@LOCK,1);
/*Do nothing but repeat until LOCKSET returns 0*/
END;
/*Now LOCK has been set to 1 by this microprocessor*/
. . .
/*Critical region of program, where shared
memory locations are accessed*/
. . .
LOCK=0;
/*End critical region*/
In the simple case just described, only one software lock is used. It is represented by
the variable LOCK. If more than one set of memory locations need protection at
different times, it is possible to establish as many different software locks as
necessary, with each lock using a different BYTE variable.
Also, note that a software lock can be used for purposes other than protecting
memory locations. LOCKSET provides a mechanism that can be used to implement
various types of synchronization in a multiprocessor system.
The Lock Bit Functions
The BITLOCK functions are built-in BYTE functions similar to the LOCKSET built-in
described in the previous section. They are called by a function reference with the
form:
keyword (bbase, boffset)
Where:
keyword
is BITLOCKSET, BITLOCKRESET, or BITLOCKCOMPLEMENT.
bbase
is an expression with a POINTER value.
boffset
is an expression with a BYTE, HWORD, or WORD value.
The action of BITLOCKSET is as follows: the bbase and boffset parameters are
used as the base address and bit offset to point to a certain bit in memory. The value
1 is assigned to this variable, and BITLOCKSET returns a BYTE. The returned value is
TRUE (0FFH) if the original content of the bit was 1, otherwise it is FALSE. During
this transaction, the CPU prevents any other process from accessing the same
memory location. BITLOCKRESET performs the same function as BITLOCKSET,
except that BITLOCKRESET assigns the value 0 to the bit variable.
BITLOCKCOMPLEMENT performs the same function as BITLOCKSET, except that
BITLOCKCOMPLEMENT complements the BYTE variable; that is, if the value was
initially 0, it is set to 1 and vice versa.
PL/M-386 Programmer's Guide
Chapter 9
167
POINTER and SELECTOR-related Functions
With the following built-in functions, programs can manipulate POINTER and
SELECTOR values that serve as location addresses in the microprocessor's memory.
The Return POINTER Value Function
BUILD$PTR is a built-in POINTER function that takes the specified segment and
offset value and returns a POINTER value. It is activated by a function reference with
the following form:
BUILD$PTR (segment, offset)
Where:
segment
expression with SELECTOR value
offset
expression with OFFSET value
The Return Segment Portion of POINTER Function
SELECTOR$OF is a built-in SELECTOR function that returns the segment portion of a
POINTER. It is activated by a function reference with the following form:
SELECTOR$OF (pointer)
Where:
pointer
is an expression with a POINTER value.
The Return Offset Portion of POINTER Function
OFFSET$OF returns the offset portion of a POINTER. For PL/M-386, OFFSET$OF is
a built-in OFFSET function. It is activated by a function reference with the following
form:
OFFSET$OF (pointer)
Where:
pointer
168
is an expression with a POINTER value.
Chapter 9
Built-in Procedures, Functions, and Variables
The Set POINTER Bytes to Zero Variable
NIL is a built-in POINTER pseudo-variable that represents a pointer with all bytes set
to zero. NIL is activated by a function reference with the following form:
NIL
The pointer value NIL points to no object. The value NIL can be assigned to a
pointer to indicate, for instance, the end of a linked list.
Note that pointer values equal to NIL cannot be used to de-reference data values. For
example, if a program contains the following statements:
DECLARE P POINTER;
DECLARE B BASED P BYTE;
P = NIL;
any subsequent references to B are invalid and will cause a trap.
The NIL POINTER variable also has the property that @NIL is equal to NIL.
POINTER variables can be initialized to NIL by using @NIL with INITIAL. For
example:
DECLARE ENDOFLIST POINTER
INITIAL (@NIL);
initializes ENDOFLIST with the value of NIL (i.e., all zeros). OFFSET$OF(NIL) and
.NIL are also equal to zero.
WORD16 Built-in Mapping
The native machine word for Intel386 and Intel486 microprocessors is WORD32 (a
32-bit WORD). The WORD16 control affects the semantics of some data types and
built-ins as listed in Table 3-3. In PL/M-386, WORD16 keywords are mapped to the
equivalent WORD32 keyword. SELECTOR, POINTER, OFFSET (ADDRESS) are the
same under both WORD32 and WORD16. Table 11-5 in the discussion of the
WORD32|WORD16 controls shows the correspondence between default (WORD32)
built-ins and those available when WORD16 is in effect. For example, Table 11-5
shows that HWORD under WORD32 corresponds to WORD under WORD16.
■■■
PL/M-386 Programmer's Guide
Chapter 9
169
170
Chapter 9
Built-in Procedures, Functions, and Variables
Features Involving the Target CPU
and Numeric Coprocessor
10
The PL/M features described in this chapter make direct or indirect use of the target
microprocessor and numeric coprocessor hardware.
Microprocessor Hardware-dependent Statements
The ENABLE and DISABLE Statements
These statements enable and disable the microprocessor interrupt mechanism.
The ENABLE statement has the following form:
ENABLE;
ENABLE generates an STI instruction, causing the microprocessor to enable interrupts
after the next machine instruction is executed.
The DISABLE statement has the following form:
DISABLE;
DISABLE generates a CLI instruction, causing the microprocessor to disable
interrupts.
PL/M-386 Programmer's Guide
Chapter 10
171
The CAUSE$INTERRUPT Statement
The CAUSE$INTERRUPT statement causes a software interrupt to be generated. It has
the form:
CAUSE$INTERRUPT (constant);
Where:
constant
is a whole-number constant in the range 0 to 255.
CAUSE$INTERRUPT generates an INT instruction with the constant as the interrupt
type, causing the microprocessor to transfer control to the appropriate interrupt
vector. Appendix G contains more information on run-time interrupt processing.
The HALT Statement
The HALT statement causes a microprocessor halt. It has the form:
HALT;
HALT generates an STI instruction followed by an HLT instruction, causing the
microprocessor to halt with interrupts enabled.
172
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
Microprocessor Hardware Flags
Optimization and the Hardware Flags
To produce an efficient machine-code program from a PL/M source program, PL/M
compilers perform extensive optimizations of the machine code. This means that the
exact sequence of machine code produced to implement a given sequence of PL/M
source statements cannot be predicted.
Consequently, the state of the microprocessor hardware flags cannot be predicted for
any given point in the program. For example, suppose that a source program contains
the following fragment:
...
SUM = SUM + 250;
...
Where:
SUM
is a BYTE variable.
Now, if the value of SUM before this assignment statement is greater than five, the
addition will cause an overflow and the hardware CARRY flag will be set.
If there were no optimization of the machine code, this assignment statement could
be followed with one of the PL/M features described in the following sections. This
would ensure that the feature would operate in a certain fashion depending on
whether or not the addition caused the CARRY flag to be set. However, because of
the optimization, some machine code instructions could occur immediately after the
addition and change the CARRY flag. It cannot be safely predicted whether this will
happen or not.
✏
Note
Accordingly, any PL/M feature that is dependent on the CARRY
flag (or any of the other hardware flags) can cause the program to
run incorrectly. These features must therefore be used with
caution, and any program that uses them must be checked carefully
to make sure that it operates correctly.
PL/M-386 Programmer's Guide
Chapter 10
173
The CARRY, SIGN, ZERO, and PARITY Functions
These built-in BYTE functions return the logical values of the microprocessor
hardware flags. These functions take no parameters, and are activated by function
references with the following forms:
CARRY
ZERO
SIGN
PARITY
An occurrence of one of these activations (in an expression) generates a test of the
corresponding condition flag. If the flag is set (=1), a value of 0FFH is returned. If
the flag is clear (=0), a value of 0 is returned.
The PLUS and MINUS Operators
In addition to the arithmetic operators described in Chapter 5, PL/M has two more:
PLUS and MINUS.
PLUS and MINUS perform similarly to + and -, and have the same precedence.
However, PLUS sums two numbers and adds the CARRY bit to the result and MINUS
subtracts two numbers and subtracts the CARRY bit from the result.
Carry-rotation Functions
SCL and SCR are built-in rotation functions whose types depend on the type of the
expression given as an actual parameter. They are activated by function references
with the following forms:
keyword (pattern, count);
Where:
keyword
SCL, SCR
pattern and count
expressions with BYTE, HWORD, WORD, OFFSET, or DWORD value -for count the high-order bits are dropped to produce BYTE values
If the value of count is 0, no shift occurs.
For PL/M-386, the value of pattern is handled as an 8-bit, 16-bit, 32-bit, or 64-bit
binary quantity. This quantity is rotated to the left (by SCL) or to the right (by SCR).
This is similar to the ROL and ROR functions described in Chapter 9. The type of
pattern determines the type of rotate that is performed. The number of bit
positions by which the value of pattern is rotated is specified by count.
174
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
The bit rotated off one end of pattern is rotated into the CARRY flag, and the old
value of CARRY is rotated to the other end of pattern. In effect, SCL and SCR
perform 9-bit rotations on 8-bit values, 17-bit rotations on 16-bit values, and so on.
For example, if the value of CARRY is 0, then:
SCL(11001010B, 2)
returns a value of 00101001B and CARRY is set to 1
SCR(11001010B, 1)
returns a value of 01100101B and CARRY remains 0
The Decimal Adjust Function
DEC is a built-in BYTE function that performs a decimal adjust operation on the actual
parameter value and returns the result of this operation. For PL/M-386, DEC uses the
value of the hardware AUXILIARY CARRY flag internally. It is activated by a
function reference with the following form:
DEC (expression);
Where:
expression
is converted, if necessary, to a BYTE value.
Microprocessor Hardware Registers
The Flags Register Access Variable
FLAGS is a built-in WORD variable that provides access to the microprocessor's
hardware flags register (see Figure 10-1, which also has flags registers for the 8086
and 286 registers for comparison). The hardware flags register contains the hardware
flags that are altered by the execution of various instructions. The hardware flags
register for the Intel386 and Intel486 microprocessors are 32 bits long.
The FLAGS register is assigned to change the setting of the various flags. It can also
be read to determine the current flag settings.
For more information on setting the hardware register flags, see the appropriate
microprocessor programmer's reference manual.
PL/M-386 Programmer's Guide
Chapter 10
175
XX
...
X
VM RF
X
OF DF
IF
TF SF ZF
X
AF
X
PF
X
CF
8086
NT
IOPL
OF DF
IF
TF SF ZF
X
AF
X
PF
X
CF
286
NT
IOPL
OF DF
IF
TF SF ZF
X
AF
X
PF
X
CF
386
X
X
X
X
X
Carry Flag
Parity Flag
Auxiliary Carry Flag
Zero Flag
Sign Flag
Single-step Trap Flag
Interrupt Enable
Direction
Overflow
IOPL
Nested Task
Resume Flag
Virtual 8086 Mode
X Denotes Intel Reserved
OSD535
Figure 10-1. The Hardware Flags Register
The STACKPTR and STACKBASE Variables
For PL/M-386, STACKPTR is an OFFSET variable and STACKBASE is a SELECTOR
variable. They provide access to the microprocessor's hardware stack pointer and
stack base registers.
When setting these registers (that is, using STACKPTR or STACKBASE on the left side
of an assignment), care must be exercised because this takes control of the stack
away from the compiler. Thus, the compile-time checks on stack overflow and
assumptions by the compiler about the run-time status of the stack may be invalid.
176
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
Microprocessor Hardware I/O
Input from an I/O port of a single BYTE, HWORD, OFFSET, or WORD is performed by
the input built-ins as a function invocation in an expression on the right-hand side of
an assignment statement. Single BYTE, HWORD, OFFSET, or WORD output is achieved
by filling the appropriate element of the output array corresponding to the desired
output port of the target microprocessor.
Multiple BYTE, HWORD, OFFSET, or WORD input is performed as a procedure
invocation, reading in a string from the microprocessor's CPU port and storing it in a
user-specified memory location. Multiple BYTE, HWORD, OFFSET, or WORD output is
also performed as a procedure invocation, using a CALL statement to send a string
from memory into the target microprocessor port.
The Find Value in Input Port Function
The following built-in functions return the values in the specified input port. They
are activated by function references with the form:
keyword (expression);
Where:
keyword
INPUT, INHWORD INWORD
expression
expression with BYTE, HWORD or WORD value
The value of expression specifies one of the input ports of the target
microprocessor.
The value returned by keyword is the expression quantity found in the specified
input port.
PL/M-386 also has an INDWORD function when the WORD16 control is used.
The Access Output Port Array
For PL/M-386, OUTPUT, OUTHWORD, and OUTWORD are built-in BYTE, HWORD, and
WORD arrays, respectively. They are activated by a function reference with the
following form:
keyword (expression);
Where:
keyword
OUTPUT, OUTHWORD, OUTWORD
expression
expression with BYTE, HWORD, or WORD value
PL/M-386 Programmer's Guide
Chapter 10
177
These functions can access any port from 0 to 65,535, corresponding to the number
of output ports on the target CPU. References to these arrays cause the specified
expression quantity to be latched to the specified hardware output port.
A reference to keyword is legal only as the left part of an assignment statement or
embedded assignment. For PL/M-386, the right-hand side of the assignment must
have a BYTE, HWORD, or WORD value.
Specifying OUTPUT in the assignment statement places the BYTE value of the
expression on the right side of the assignment into the specified output port. (Since
OUTPUT is a BYTE built-in, the value of the expression is converted automatically to a
BYTE type if necessary.)
Specifying OUTWORD in the assignment statement places the WORD (or OFFSET)
value of the expression on the right side of the assignment into the specified output
port.
Similarly, of OUTHWORD places the HWORD value of the expression on the right side of
the assignment into the corresponding output port. PL/M-386 also has an OUTDWORD
built-in when the WORD16 control is used.
The Read and Store String Procedure
The read and store string procedures are built in. For PL/M-386, these built-ins read
the BYTE, HWORD, OFFSET, or WORD string values latched to the specified hardware
input port. The read values, of the length specified by count, are then stored at the
location specified by destination. These procedures are activated by a CALL
statement with the following form:
CALL keyword (port, destination, count);
Where:
keyword
BLOCKINPUT, BLOCKINHWORD, BLOCKINWORD
port
expression with BYTE or HWORD value
destination
expression with POINTER value
count
expression with BYTE, HWORD, OFFSET or WORD value
The keyword specifies the type of string found in the specified input port. The value
of port specifies one of the input ports of the CPU. The destination specifies the
location (in memory) at which to store the string. The value of count specifies the
length of the string.
PL/M-386 also has a BLOCKINDWORD procedure when the WORD16 control is used.
178
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
The Write String Procedure
The write string procedures are built-in procedures. For PL/M-386, these built-ins
write a BYTE, HWORD, OFFSET, or WORD string to the specified output hardware port.
These built-ins are activated by a CALL statement with the following form:
CALL keyword (port, source, count);
Where:
keyword
BLOCKOUTPUT, BLOCKOUTHWORD, BLOCKOUTWORD
port
expression with BYTE or HWORD value
source
expression with POINTER value
count
expression with BYTE, HWORD, OFFSET, or WORD value
The keyword specifies the type of string. The value of port specifies one of the
output ports of the microprocessor CPU. The source value specifies the location (in
memory) where the string is currently stored. The value of count specifies the string
length.
PL/M-386 also has a BLOCKOUTDWORD procedure when the WORD16 control is used
(see Chapter 10).
The Hardware Protection Model
The Intel386 and Intel486 microprocessors' protection mechanism provides up to four
privilege levels within each task. The highest privilege level (level 0) is reserved for
the operating system kernel. Below the kernel level, systems can be configured to
include a system service level (level 1), an applications service level (level 2), and an
application program level (level 3).
The following hardware protection built-in procedures and variables allow access to
the protection architecture of these microprocessors.
The Task Register
The TASK$REGISTER Variable
TASK$REGISTER is a built-in SELECTOR variable that provides access to the task
state register. This register points to a task state segment for the currently executing
task.
PL/M-386 Programmer's Guide
Chapter 10
179
The format of the task register for the Intel386 microprocessor is:
T
RPL
I
INDEX
31
16 15
24 23
8 7
2 1 0
OSD579
Values are assigned to TASK$REGISTER to reset the task state segment for the
current task or to enter the protected mode of the microprocessor. However, the
selector stored in TASK$REGISTER must point to a valid task state segment. Note
that values can be assigned to TASK$REGISTER only if the program is executed in
protection mode at level 0.
TASK$REGISTER can also be read to determine the task state segment of the
currently executing task.
The Global Descriptor Table Register
The global descriptor table register (GDTR) is a system-wide register used for
protected virtual address mode. The GDTR describes a memory area that contains an
array of descriptors for the global address space. The register occupies 6 bytes.
Its format for Intel386 and Intel486 microprocessors follows:
BASE
47
39 38
32 31
LIMIT
24 23
16 15
8 7
0
OSD580
180
LIMIT
size of the GDT segment (up to 64K bytes)
BASE
physical memory base address of the GDT segment
ACCESS
access control byte
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
The SAVE$GLOBAL$TABLE Procedure
SAVE$GLOBAL$TABLE is a built-in procedure. It is activated by a CALL statement
with the form:
CALL SAVE$GLOBAL$TABLE (location);
Where:
location
is an expression with a POINTER value.
SAVE$GLOBAL$TABLE saves the contents of the hardware global descriptor table
register in the 6-byte save area pointed to by location.
The RESTORE$GLOBAL$TABLE Procedure
RESTORE$GLOBAL$TABLE is a built-in procedure. It is activated by a CALL
statement with the form:
CALL RESTORE$GLOBAL$TABLE (location);
Where:
location
is an expression with a POINTER value.
RESTORE$GLOBAL$TABLE restores the contents of the hardware global descriptor
table register from the save area pointed to by location. This save area can be the
same area used in a preceding call to SAVE$GLOBAL$STATUS.
SAVE$GLOBAL$TABLE saves the value of the GDTR in a 6-byte memory area.
RESTORE$GLOBAL$TABLE restores the value of the GDTR.
PL/M-386 Programmer's Guide
Chapter 10
181
The Interrupt Descriptor Table Register
The interrupt descriptor table register (IDTR) is a system-wide register that is used
for interrupt processor management. The IDTR describes a segment that contains the
linear base address and the size of the interrupt descriptor table (IDT), and a segment
containing an array of gate descriptors for the interrupt handlers. The register
occupies 6 bytes.
Its format for Intel386 and Intel486 microprocessors follows:
BASE
47
39 38
32 31
LIMIT
24 23
16 15
8 7
0
OSD580
LIMIT
size of the segment (up to 64K bytes)
BASE
physical memory base address of the IDT segment
ACCESS
access control byte
The SAVE$INTERRUPT$TABLE Procedure
SAVE$INTERRUPT$TABLE is a built-in procedure that is activated by a CALL
statement with the following form:
CALL SAVE$INTERRUPT$TABLE (location);
Where:
location
is an expression with a POINTER value.
SAVE$INTERRUPT$TABLE saves the contents of the hardware interrupt descriptor
table register in the 6-byte save area pointed to by location.
182
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
The RESTORE$INTERRUPT$TABLE Procedure
RESTORE$INTERRUPT$TABLE is a built-in procedure that is activated by a CALL
statement with the following form:
CALL RESTORE$INTERRUPT$TABLE (location);
Where:
location
is an expression with a POINTER value.
RESTORE$INTERRUPT$TABLE restores the contents of the hardware interrupt
descriptor table register from the save area pointed to by location. This save area
can be the same area used in a preceding call to SAVE$INTERRUPT$TABLE.
A descriptor can be built that will initialize the interrupt processor operations.
RESTORE$GLOBAL$STATUS can then be called with a pointer to this descriptor.
The user must ensure that the save area contains a valid descriptor. Note that values
can be assigned to the IDTR only if the program is executed in protection mode at
level 0.
The Local Descriptor Table Register
The LOCAL$TABLE Variable
LOCAL$TABLE is a built-in SELECTOR variable that provides access to the local
descriptor table register (LDTR). The format of the register is a selector pointing to
an LDT in the GDT. The use of the local descriptor table is like the use of the
GDTR, except that it defines the local address space.
By assigning a value to LOCAL$TABLE, the local address space of the current task is
altered. If a task switch occurs, the new contents are not saved in the task state
segment. (To ensure proper operation, interrupts must be disabled.)
LOCAL$TABLE can be read to determine the current active descriptor array segment
for the current task.
The user must ensure that the selector in LOCAL$TABLE points to a valid descriptor
segment. Note that values can be assigned to the LDTR only when the program is
executed in protection mode at level 0.
PL/M-386 Programmer's Guide
Chapter 10
183
The Machine Status Register
The MACHINE$STATUS Variable
For PL/M-386, MACHINE$STATUS is a built-in HWORD variable. MACHINE$STATUS
provides access to the machine status word (MSW). The MSW register defines the
current status of the processor protection model and the real math unit support. The
format of MACHINE$STATUS is:
X
X
15
14
...
X
TS EM MP PE
4
3
2
1
0
Protection Enable
Real Math Unit (iAPX 287) Present
Emulation Mode
Task Switched
(Reserved)
(Reserved)
(Reserved)
OSD536
MACHINE$STATUS enables access to the protected mode of the microprocessor.
When a value is assigned to this register, the compiler generates a short jump to the
next instruction to clear the instruction queue. (Note, however, that values can be
assigned to MACHINE$STATUS only if the program is executed in protection mode at
level 0.)
The contents of MACHINE$STATUS can also be read to determine the current status of
various system components.
The CONTROL$REGISTER, DEBUG$REGISTER, and TEST$REGISTER
Built-in Arrays
The CONTROL$REGISTER is a built-in WORD array that provides access to the Intel386
and Intel486 microprocessors' 32-bit control registers that define the current status of
the processor and contain page table and page fault information.
184
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
The format of CONTROL$REGISTER (0) is:
P
G
E T E M P
T S M P E
Reserved
31
24 23
16 15
8 7
4 3 2 1 0
OSD575
where:
PG = paging enabled
ET = extension type
TS = task switched
EM = emulate coprocessor
MP = numeric coprocessor present
PE = protection enable
MSW is contained in the low-order 16 bits of CONTROL$REGISTER (0). However,
assigning a value to the MACHINE$STATUS built-in does not change the ET
(extension type) bit.
CONTROL$REGISTER (2) contains the 32-bit linear address that caused the last
detected page fault.
CONTROL$REGISTER (3) contains the physical page base address for the first level
of the page table structure. This address is in the high 20 bits (bits 12 to 31) of
CONTROL$REGISTER (3). The lower 12 bits are ignored when assigning to
CONTROL$REGISTER (3) and are undefined when reading CONTROL$REGISTER
(3). Note that the control registers are accessible only during execution at protected
mode level 0. Also note that CONTROL$REGISTER (1) is not accessible.
The DEBUG$REGISTER built-in WORD array provides access to six of the eight 32-bit
debug registers; DEBUG$REGISTER(4) and DEBUG$REGISTER(5) are not
accessible. The debug registers are accessible only during execution at protected
mode level 0.
TEST$REGISTER is a built-in WORD array that provides access to the 32-bit test
registers of the microprocessor. Of these test registers, only TEST$REGISTER (6)
and TEST$REGISTER (7) are accessible; these registers are accessible only when
executing in protection mode at level 0.
PL/M-386 Programmer's Guide
Chapter 10
185
The CLEAR$TASK$SWITCHED$FLAG Procedure
CLEAR$TASK$SWITCHED$FLAG is a built-in procedure that is activated by a CALL
statement with the form:
CALL CLEAR$TASK$SWITCHED$FLAG;
This procedure is used to clear the task switched flag in the machine status word.
The processor sets the task switched flag every time a task switch occurs. It can be
used to manage the sharing of the math coprocessor.
CLEAR$TASK$SWITCHED$FLAG can be called only when the program is executed in
protection mode at level 0.
Segment Information
The GET$ACCESS$RIGHTS Function
GET$ACCESS$RIGHTS is a built-in WORD function; it is activated by a function
reference with the form:
GET$ACCESS$RIGHTS (selector);
Where:
selector
is an expression with a SELECTOR value.
If the segment pointed to by selector is visible at the current privilege level, then
the hardware ZERO flag is set and a WORD value is returned. If the segment is not
visible, or if it is of the wrong type, the hardware ZERO flag is reset, and the returned
value is undefined.
✏
186
Note
The setting of the ZERO flag is guaranteed only if it is tested
immediately, before being altered by another operation. (For
example, if the value of the function is assigned to an array element
indexed by an expression, the value of the ZERO flag may be
incorrect.)
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
Specific to Intel386 and Intel486 microprocessors, the format of the return value is:
P
0
G
. . . . . . .
31
A
0 G 0 0 V X X X X
L
24 23
ACCESS
16 15
0
. . . . . . .
0
8 7
0
OSD576
X
reserved
G
granularity bit
AVL
available for software use
ACCESS
access rights byte
The following example illustrates how the GET$ACCESS$RIGHTS function can be
used:
DECLARE RIGHTS WORD;
DECLARE SEGMENT SELECTOR;
RIGHTS = GET$ACCESS$RIGHTS (SEGMENT);
IF ZERO THEN
/* The segment pointed to by SEGMENT is visible */
/* and RIGHTS contains the proper access
/* rights to it. */
ELSE
/* SEGMENT is not visible and the contents of */
/* RIGHTS is undefined. */
The GET$SEGMENT$LIMIT Function
For PL/M-386, GET$SEGMENT$LIMIT is a built-in OFFSET function.
GET$SEGMENT$LIMIT is activated by a function call of the form:
GET$SEGMENT$LIMIT (selector);
Where:
selector
is an expression with a SELECTOR value.
If the segment pointed to by selector is visible at the current protection level, then
the hardware ZERO flag is set and the value returned by GET$SEGMENT$LIMIT is
the size of the segment. If the segment is not visible, the ZERO flag is reset and the
returned value is undefined.
Set the ZERO flag with caution. See the note in section for the
GET$ACCESS$RIGHTS function.
PL/M-386 Programmer's Guide
Chapter 10
187
The following example illustrates how the GET$SEGMENT$LIMIT function can be
used:
DECLARE LIMITS OFFSET;
DECLARE SEGMENT SELECTOR;
LIMITS = GET$SEGMENT$LIMIT (SEGMENT);
IF ZERO THEN
/* The segment pointed to by SEGMENT is visible */
/* and LIMITS contains its proper size.*/
ELSE
/* SEGMENT is not visible and the contents of*/
/* LIMITS is undefined.*/
Segment Accessibility
It is sometimes helpful to know if the segment pointed to by a selector is readable or
writable from the current address space. This becomes particularly important when
the selector is a parameter that is passed to the current task.
If an attempt is made to access a segment that is inaccessible, an interrupt will occur.
To avoid this interrupt, segment readability and writability can be tested before the
segment is accessed.
The SEGMENT$READABLE Function
SEGMENT$READABLE is a built-in BYTE function. It is activated by a function
reference with the form:
SEGMENT$READABLE (selector);
Where:
selector
is an expression with a SELECTOR value.
SEGMENT$READABLE returns a value of TRUE (0FFH) if the segment pointed to by
selector is reachable and readable from the current privilege level; FALSE (0), if it is
not.
The SEGMENT$WRITABLE Function
SEGMENT$WRITABLE is a built-in BYTE function. It is activated by a function
reference with the form:
SEGMENT$WRITABLE (selector);
Where:
selector
188
is an expression with a SELECTOR value.
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
SEGMENT$WRITABLE returns a value of TRUE (0FFH) if the segment pointed to by
selector is reachable and writable from the current privilege level; FALSE (0), if it
is not.
Adjusting the Requested Privilege Level
The ADJUST$RPL Function
ADJUST$RPL is a built-in SELECTOR function that returns the argument of the
adjusted requested privilege level (RPL). It is activated by a function reference with
the form:
ADJUST$RPL (selector);
Where:
selector
is an expression with a SELECTOR value.
If the requested privilege level (RPL) field of the argument selector is less than the
RPL field of the code segment selector for the routine calling the procedure that
invoked ADJUST$RPL, then the hardware ZERO flag is set and the value returned is
the argument of an adjusted RPL field. Otherwise, the ZERO flag is reset, and the
value returned is the original value of the argument.
Setting the ZERO flag should be done cautiously; see the note in section on the
GET$ACCESS$RIGHTS function.
The following example illustrates how the ADJUST$RPL function can be used:
P: PROCEDURE (SEGMENT);
DECLARE SEGMENT SELECTOR;
SEGMENT = ADJUST$RPL (SEGMENT);
IF ZERO THEN
/* The RPL of SEGMENT was less than the RPL of*/
/* the routine that called P; SEGMENT now has*/
/* the RPL of the caller.*/
ELSE
/* The RPL of SEGMENT was not less than the RPL */
/* of the routine that called P; SEGMENT is unchanged */
END P;
PL/M-386 Programmer's Guide
Chapter 10
189
The REAL Math Facility
REAL math support for PL/M is provided by the numeric coprocessor. In relation to
the program, the REAL math facility consists of the following:
•
The REAL stack, used to hold operands and results during REAL operations.
•
The REAL error byte (see Figure 10-2), consisting of seven exception flags
initialized to all 0s. (The reserved bit is set to 1 by the numeric coprocessor.)
The first six bits in this byte correspond to the possible errors that can arise
during REAL operations (see Appendix G). When an error occurs, the facility
sets the corresponding bit to 1. A program can invoke a built-in procedure
(described in the next section) that reads and clears the REAL error byte.
The exception/error categories are discussed in Appendix G.
•
The REAL mode word (see Figure 10-3), consisting of 16 bits initialized to
03FFH (or 7FFH for the Intel387™ numeric coprocessor).
1.
7
IR
Bits 0-7 determine whether the corresponding error condition is to be
handled with the default recovery (described next) or with the
programmer-supplied exception procedure (see Appendix G for details on
writing these). When the bit is 1, the default is used; when it is 0, the user
routine is used. In either case, the facility records the error by setting the
corresponding bit of the REAL error byte. For most uses, the default
recovery is appropriate and less work.
0
PE UE OE ZE DE IE
Exception Flags (1 = Exception Has Occurred)
Invalid Operation
Denormalized Operand
Zerodivide
Overflow
Underflow
Precision
(Reserved)
Interrupt Request
OSD537
Figure 10-2. The REAL Error Byte
190
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
7
15
IC
RC
PC
IEM
0
PM UM OM ZM DM IM
Exception Masks (1 = Exception is Masked)
Invalid Operation
Denormalized Operand
Zerodivide
Overflow
Underflow
Precision
(Reserved)
1
Interrupt-Enable Mask (IEM)
Precision Control
Rounding Control
Infinity Control
2
3
4
(Reserved)
(1)
(2)
Interrupt - Enable Mask:
0 Interrupts Enabled
1 Interrupts Disabled (Masked)
Precision Control:
00 24 Bits
01 (Reserved)
10 53 Bits
11 64 Bits
(3)
Rounding Control:
00 Round To Nearest Or Even
01 Round Down (Toward
)
10 Round Up (Toward
)
11 Chop (Truncate Toward Zero)
(4)
Infinity Control:
0 Projective
1 Affine
OSD539
Figure 10-3. The REAL Mode Word
PL/M-386 Programmer's Guide
Chapter 10
191
This mode word is often called a mask; that is, it lets some signals through
(to interrupt processing), but not others. If one of the bits 0-5 is a 0, the
corresponding error is said to be unmasked (see the next section for setting
the mode word).
If the interrupt is enabled (IEM = 0), one of the masked bits is 0, and the
corresponding error occurs during floating point processing, then the REAL
math facility interrupts the host CPU. The numeric coprocessor's interrupt
number is dependent on the internal configuration. The exception condition
is thus reported and control is passed to the user-written error handling
routine. This situation is called an unmasked error. Chapter 8 and
Appendix G discuss aspects of interrupt procedures.
Conversely, a masked error means the mode bit corresponding to that error
is 1. Masked errors do not cause an interrupt, but are handled as described
in Appendix G.
2.
Bits 8 and 9 control precision. All intermediate results are held in an
internal format of 64-bit precision. The most-significant 24 bits of the final
result are returned (plus sign and 7-bit exponent) as the PL/M answer, and
rounded, if needed, according to the user-specified control. The default
precision setting preserves extended precision and operates slightly faster
than the other settings.
3.
Bits 10 and 11 control rounding. Rounding introduces an error of less than
one unit in the last place to which the result was rounded. Statistically, the
default provides the most accurate and unbiased estimate of the true result
(i.e., the 64-bit result). In all rounding modes except round down,
subtracting a number from itself yields +0; round down yields -0.
4.
Bit 12 controls how infinity is handled, as shown below.
0
Affine Closure
0
Projective Closure
OSD538
Bits 13, 14, and 15 are reserved and are not for PL/M use.
192
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
Built-ins Supporting the REAL Math Unit
The INIT$REAL$MATH$UNIT Procedure
INIT$REAL$MATH$UNIT is a built-in untyped procedure activated by a CALL
statement, as follows:
CALL INIT$REAL$MATH$UNIT;
This call is required as the first access to the math coprocessor.
This call initializes the REAL math unit for subsequent operations. This includes
setting a default value into the control (REAL mode) word, namely 03FFH or
0000001111111111B. This setting masks all exceptions and interrupts, sets precision
to 64 bits, and sets the rounding mode to nearest, with even preferred. This means no
interrupts will occur from the REAL math facility regardless of what errors are
detected.
Procedures activated after this call has taken effect do not need to do such
initialization.
The SET$REAL$MODE Procedure
This procedure should only be invoked to change the default mode word
(for example, to unmask the invalid exception).
SET$REAL$MODE is a built-in untyped procedure, activated by a CALL statement with
the following form:
CALL SET$REAL$MODE (modeword);
Where:
modeword
expression with HWORD value
The value of modeword becomes the new contents of the REAL mode word (see
Figure 10-3). The suggested value for modeword is 033EH, (0000001100111110B).
This value provides maximum precision, default rounding, and masked handling of
all exception conditions except an invalid operation, which can alert the user to errors
of initialization or stack usage (see Appendix G for facts and references on writing an
interrupt handling procedure).
PL/M-386 Programmer's Guide
Chapter 10
193
The GET$REAL$ERROR Function
GET$REAL$ERROR is a built-in BYTE function activated by a function reference with
the following form:
GET$REAL$ERROR
The BYTE value returned is the current contents of the REAL error byte (see Figure
10-2). This function also clears the error byte in the REAL math facility.
Saving and Restoring REAL Status
If an interrupt procedure performs any floating-point operation, it will change the
REAL status. If such an interrupt procedure is activated during a floating-point
operation, the program will be unable to continue the interrupted operation correctly
after returning from the interrupted procedure. Therefore, it is first necessary for any
interrupt procedure that performs a floating-point operation to save the REAL status
and subsequently restore it before returning. The built-in procedures
SAVE$REAL$STATUS and RESTORE$REAL$STATUS make this possible.
SAVE$REAL$STATUS also initializes the numeric coprocessor.
Additionally, these procedures can be used in a multi-tasking environment where a
running task using the numeric coprocessor can be preempted by another task that
also uses the numeric coprocessor. The preempting task must call
SAVE$REAL$STATUS before it executes any statements that affect the numeric
coprocessor, that is, before calling SET$REAL$MODE and before any arithmetic or
assignment of REALs (other than GET$REAL$ERROR, if needed).
New vectors will be required for the interrupt handlers appropriate to each new task
(e.g., to handle unmasked exception conditions). These vectors must be initialized by
the operating system.
After its processing is complete and it is ready to terminate, the preempting task must
call RESTORE$REAL$STATUS to reload the state information that applied at the time
the former running task was preempted. This enables that task to resume execution
from the point where it relinquished control.
✏
Note
REAL functions without REAL parameters should not call
GET$REAL$ ERRORS or SAVE$REAL$STATUS before executing at
least one floating-point instruction. To do so may result in loss of
processor synchronization.
194
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
The SAVE$REAL$STATUS Procedure
SAVE$REAL$STATUS is a built-in untyped procedure activated by a CALL statement
with the form:
CALL SAVE$REAL$STATUS (location);
Where:
location
is a pointer to a memory area 108 bytes long where the REAL status
information will be saved.
The REAL status is saved at the specified location, and the REAL stack and error bytes
are reinitialized.
If the state of the REAL math unit is unknown to this procedure when it is called, as in
the case previously mentioned for preempting tasks, then an initialization will destroy
existing error flags, masks, and control settings. To avoid this, the appropriate action
(except for error-recovery routines, discussed in Appendix G) is to issue:
CALL SAVE$REAL$STATUS (@location_1);
before any REAL math usage, and
CALL RESTORE$REAL$STATUS (@location_1);
prior to the procedure's return. The save automatically reinitializes the math unit and
the error byte.
This protects the status of preempted tasks or prior procedures and establishes a
known initialization state for the current procedure's actions. The microprocessor
interrupts are disabled during the save.
✏
Note
The microprocessor must be able to acknowledge numeric
coprocessor interrupts or loss of synchronization occurs.
PL/M-386 Programmer's Guide
Chapter 10
195
The RESTORE$REAL$STATUS Procedure
RESTORE$REAL$STATUS is a built-in untyped procedure activated by a CALL
statement with the form:
CALL RESTORE$REAL$STATUS (location);
Where:
location
is a pointer to a memory area where the REAL status information was
previously saved by a call to the SAVE$REAL$STATUS procedure.
This procedure should be called prior to returning from an interrupt procedure where
the real math unit's status was saved using SAVE$REAL$STATUS.
Interrupt Processing
The WAIT$FOR$INTERRUPT Procedure
WAIT$FOR$INTERRUPT is a built-in procedure that is activated by a CALL statement
with the form:
CALL WAIT$FOR$INTERRUPT;
This procedure is used to generate an IRET instruction in a nested interrupt task; if it
is used elsewhere, the results are undefined. The IRET instruction causes the
microprocessor to perform a task switch, saving the status of the outgoing task in its
TSS. The next time the interrupt task is activated, execution will begin at the
instruction immediately following the IRET, with all the registers unchanged.
The following example illustrates how the WAIT$FOR$INTERRUPT procedure can be
used:
NEW$INTERRUPT:
CALL INITIALIZE$INTERRUPT$LIST;
/* Start of a list of interrupts */
DO WHILE 1;
CALL WAIT$FOR$INTERRUPT;
/* Wait for next interrupt within list */
CALL PROCESS$INTERRUPT;
IF END$OF$INTERRUPT$LIST THEN DO;
CALL WAIT$FOR$INTERRUPT;
/* Wait for start of next interrupt sequence */
GOTO NEW$INTERRUPT;
END;
END;
196
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
WORD16 Mapping for Built-ins
Table 11-5, in the discussion of the WORD32|WORD16 control, shows the
correspondence between default (WORD32) machine built-ins and those available
when WORD16 is in effect. For example, Table 11-5 shows that HWORD (WORD32)
corresponds to WORD under WORD16.
Intel486 Processor Built-ins
The following are built-ins specific to the Intel486 processor. Specify the MOD486
control for the PL/M-386 compiler to use these functions:
•
BYTE$SWAP: This function generates an Intel486 processor instruction that
swaps bytes in a 32-bit expression to convert between big and little endian. The
BYTE$SWAP function takes a 32-bit expression and returns a value of the same
data type as the argument. An argument of less than 32 bits produces a semantic
error. To pass a pointer value, use a data type of WORD or OFFSET instead of
POINTER.
Invoke BYTE$SWAP as in the following example:
DECLARE (a, b) WORD;
a = BYTE$SWAP( b );
b = BYTE$SWAP( b + 10 );
TEST$REGISTER: This variable is an array of 8 elements. Each element is a 32bit unsigned scalar data type. The available registers of TEST$REGISTER
include elements (6) and (7). When the compiler control MOD486 is specified,
elements (3) through (5) are also available.
Use TEST$REGISTER as in the following example:
DECLARE a WORD;
a = TEST$REGISTER (4);
See also:
•
Test Registers, i486 Microprocessor Programmer's Reference Manual
INVALIDATE$DATA$CACHE: This function generates the Intel486 processor
instruction to clear the entire data cache.
Invoke INVALIDATE$DATA$CACHE as in the following example:
CALL INVALIDATE$DATA$CACHE;
PL/M-386 Programmer's Guide
Chapter 10
197
•
WB$INVALIDATE$DATA$CACHE: This function generates the Intel486 processor
instruction to first write out all changed lines to memory and then clear the entire
data cache.
Invoke WB$INVALIDATE$DATA$CACHE as in the following example:
CALL WB$INVALIDATE$DATA$CACHE;
•
INVALIDATE$TLB$ENTRY: This function generates the Intel486 processor
instruction to clear a specified entry in the paging cache (TLB). Specify the
entry to be cleared as an argument, preceded by an @ sign.
Invoke INVALIDATE$TLB$ENTRY as in the following example:
DECLARE a(10) BYTE;
DECLARE b WORD;
CALL INVALIDATE$TLB$ENTRY ( @a(5) );
CALL INVALIDATE$TLB$ENTRY ( @b );
■■■
198
Chapter 10
Features Involving the Target CPU and Numeric Coprocessor
11
Compiler Invocation and Controls
This chapter describes compiler controls, optimization, and invocation. There are
differences in invocation, depending on whether you are running on iRMX or DOS.
This chapter covers both operating systems.
Invocation Syntax on iRMX Systems
The general form of the invocation command is:
[:logical_name:]PLM386 filename [control]...
Where:
:logical_name:
is the optional logical name for the directory or device containing the
PL/M-386 compiler.
PLM386
is the name of the compiler.
filename
is the full filename (with directory path) of the file containing the
source code. The compiler accepts only one source file per invocation.
control
is zero or more of the compiler controls described later in this chapter.
Separate multiple controls with spaces to extend the invocation
command over multiple lines, use the ampersand (&) as a continuation
character.
The INCLUDE control must be the last control.
Errors detected in the invocation command cause the compiler to abort without
processing the source file.
The portion of the path set off with colons (:) is an iRMX logical name. A logical
name identifies the directory or device that contains the compiler files. In the
examples used here, the compiler resides in the :lang: directory. The subdirectory
mydir resides in the directory source. source resides in :home:. When you are
logged on as the user world, :home: is the logical name for the directory /user/world.
If the logical name is omitted from the invocation command, the operating system
automatically searches several directories for the invocation command. The
PL/M-386 Programmer's Guide
Chapter 11
199
directories searched and the order of the search are defined in the operating system
configuration.
Slashes (/) and carets (^), which are also called circumflexes, are used to move up or
down the directory tree. To identify a file, start with a logical name (or assume the
default). Continue through the directory tree using the slash to search down one level
or the caret to search up one level.
For example, if the source file textfile.plm is in directory source, and source is in the
directory identified by logical name :home:, use the following pathname:
- :HOME:SOURCE/TEXTFILE.PLM
If the default directory is :home:source/mydir, then the same source file can be
identified by specifying the path name as follows:
- ^TEXTFILE.PLM
The caret instructs the operating system to go up one level to find the file.
When you continue an invocation command over multiple lines by entering an
ampersand (&) before the line-feed character, the next line automatically appears with
the continuation prompt (**). The ampersand can also be used to insert comments.
The PL/M-386 compiler ignores characters that appear after an ampersand. For
example:
- PLM386 :HOME:SOURCE/TEXTFILE.PLM & Run compiler
** TITLE ("PROJECT SUPERVISOR") & for this file.
** OPTIMIZE(2) CODE XREF
200
Chapter 11
Compiler Invocation and Controls
Invocation Examples and Sign-on/Sign-off Messages under
the iRMX OS
The following example specifies compilation of a PL/M-386 source file named
myprog.src. The list file is sent to myprog.lst, with the heading TEST 24 on each
new page of output. Both the list and object files are written to the directory
/user/world/source.
- :LANG:PLM386 /USER/WORLD/SOURCE/MYPROG.SRC &
** TITLE("TEST 24")
The logical name :home: can be used in place of the directory pathname /user/world
if you are currently logged on as the user WORLD. One of these two specifications
must be used if your current directory is not /user/world. If your default, or current,
directory is /user/world/source only the actual file name, not including the directory
pathname, must be specified in the invocation command. To change the default
directory, use the ATTACHFILE command. Refer to the iRMX System Call Reference
for additional information on the ATTACHFILE command.
The :lang: logical name can be omitted if the default iRMX search path is used
(which automatically searches :lang: for commands).
When invoked, the compiler signs on with the following message:
host PL/M-386 COMPILER Vx.y
Copyright Intel Corporation, years
Where:
host
identifies the host system.
x.y
identifies the compiler version.
years
are the copyright years.
When compilation is complete, the compiler signs off with the following message:
PL/M-386 COMPLETE.
n WARNINGS,
m ERRORS.
where n and m are the numbers of warning and error messages generated during
compilation.
PL/M-386 Programmer's Guide
Chapter 11
201
Invocation Syntax on DOS Systems
The general form of the invocation command is:
PLM386 filename[control]...
Where:
is the name of the compiler. The directory containing the compiler
should be in your DOS PATH.
filename is the full filename (with directory path) of the file containing the
source code. The compiler accepts only one source file per invocation,
unless you use the INCLUDE control.
control
is zero or more of the compiler controls described later in this chapter.
Separate multiple controls with spaces.
Note that DOS limits the command line to 128 characters. You can extend the
command over multiple lines with the ampersand (&) continuation character.
The INCLUDE control (if used) must be the last control. Subsystem controls and
certain other controls, identified in this chapter, cannot be part of the invocation
command.
PLM386
Invocation Examples and Sign-on/Sign-off Messages under
DOS
The first example specifies compilation of a PL/M-386 source module called prog1.
The XREF control is used.
PLM386 PROG1.SRC XREF
The second example specifies compilation of a PL/M-386 source module called
myprog. The list file is sent to othrfile.lst, in which the heading TEST 24 appears on
each new page of output.
PLM386 MYPROG TITLE(`TEST 24') PRINT(OTHRFILE.LST)
When invoked, the compiler signs on with the following message:
host PL/M-386 COMPILER Vx.y
Copyright Intel Corporation, years
Where:
host
x.y
years
202
identifies the host system.
identifies the compiler version.
are the copyright years.
Chapter 11
Compiler Invocation and Controls
When compilation is complete, the compiler signs off with the following message:
PL/M-386 COMPLETE.
n WARNINGS,
m ERRORS.
where n and m are the numbers of warning and error messages generated during
compilation.
File Usage under DOS and the iRMX OS
The PL/M-386 compiler accepts a single source file as input. The compiler creates
and deletes work files, as further described below. By default, the compiler creates
two files: a print (or list) file and an object file.
Input Files
The pathname used in the invocation command identifies the source file to be
compiled. Other files containing source code can be included with the INCLUDE
control. The source file name and format must follow the file conventions of the OS.
Work Files
The PL/M-386 compiler uses temporary work files that are deleted after compilation.
All of these files are located on the device :WORK: under the iRMX OS. Under
DOS, use the set command for selecting an alternate drive for the work files. The
following example specifies that the work files be sent to directory d:.
SET :WORK: =
✏
d:\
Note
Using the set command to relocate work files is useful when the
DOS device driver has created a virtual disk. To change the default
location of work files, place the work files drive specification
command in the autoexec.bat file.
All work files have a .tmp extension. Avoid using .tmp as the extension on any
device used by the compiler. It is possible that an existing file with a .tmp extension
could be deleted or overwritten by the compiler.
The space required for work files is approximately equal to the space required for the
source file plus any included files. Be sure that the selected device provides adequate
disk space for the compiler work files.
PL/M-386 Programmer's Guide
Chapter 11
203
Print Files
The list file (also called the print file) contains a listing of the source program, the
messages collected during compilation, and other printed output specified by the
listing selection controls. By default, the list file has the same base name as the
source file and an .lst extension. Unless otherwise specified, the list file is located on
the same drive and in the same directory as the source file.
When the PRINT control is used, the compiler creates a list file with the same base
name as the source file. If another file exists with the same name, the existing file is
overwritten. To save the existing file use the PRINT control with a parameter; this
saves the new list file under another file name.
Object Files
The object file (also called the object code file or object module) contains the object
module format translation of the source code. By default, the object file has the same
base name as the source file and an .obj extension. Unless otherwise specified, the
object file is located on the same drive and in the directory as the source file.
The output of a PL/M-386 compiler is an object file containing a compiled module.
This object module may be linked with other object modules using the appropriate
linker or binder. A knowledge of the makeup of an object module is not necessary
for PL/M programming, but can aid in understanding the controls for program size
and linkage.
Object modules output by the PL/M-386 compilers contain three sections:
•
Code Section
•
Data Section
•
Stack Section
These sections can be combined in various ways into memory segments for
execution, depending on the size of the program.
204
Chapter 11
Compiler Invocation and Controls
Code Section
This section contains the instruction code generated for the source program. If either
the LARGE control or the ROM control is used, this section also contains all variables
initialized with the DATA attribute, all REAL constants, and all constant lists.
In addition, the code section for the main program module contains a main program
prologue generated by the compiler. This code precedes the code compiled from the
source program, and sets the microprocessor for program execution by initializing
various registers.
Data Section
All variables are allocated space in this section with the exception of parameters,
based variables, and variables located with an AT attribute or local to a REENTRANT
procedure. If the RAM control is used, this section also contains all variables
initialized with the DATA attribute, as well as all REAL constants and all constant lists.
If a nested procedure refers to any parameter of its calling procedure, then all
parameters of that calling procedure will be placed in the data section during
execution. The compiler reserves enough space during compilation to prepare for
this.
Stack Section
The stack section is used in executing procedures, as explained in Appendices F and
G. It is also used for any temporary storage used by the program but not explicitly
declared in the source module (such as temporary values generated by the compiler).
The exact size of the stack is automatically determined by the compiler except for
possible multiple invocations of reentrant procedures. You can override this
computation of stack size and explicitly state the stack requirement during the
binding (linking) process.
✏
Note
When using reentrant procedures or interrupt procedures, be sure to
allocate a stack section large enough to accommodate all possible
storage required by multiple invocations of such procedures. The
stack space requirement of each procedure is shown in the listing
produced by the SYMBOLS or XREF control. This information can
be used to compute the additional stack space required for reentrant
or interrupt procedures.
PL/M-386 Programmer's Guide
Chapter 11
205
Executable Programs
After the source file is compiled, related object modules must to be combined to form
executable modules. The libraries that provide the necessary run-time support for the
application must also be combined with the object modules. To do this you use the
BND386 utility, described in the Intel386 Family Utilities User's Guide.
DOS offers two ways of automatically invoking and executing multiple programs:
batch files and command files. For more information, refer to your DOS operating
system manuals.
206
Chapter 11
Compiler Invocation and Controls
Introduction to Compiler Controls
Use the compiler controls described in this chapter either in the command that
invokes the compiler or as control lines in the source input file.
A control line contains a dollar sign ($) in the left margin. Normally, the left margin
is set at column one, but you can change this with the LEFTMARGIN control. Control
lines allow selective control over sections of the program. For example, it may be
desirable to suppress the listing for certain sections of the program, or to cause page
ejects at certain places.
A line in a source file is considered to be a control line by the compiler if there is a
dollar sign in the left margin, even if the dollar sign appears to be part of a PL/M
comment or character string constant. Control lines within the source code must
begin with a dollar sign and can contain one or more controls, each separated by at
least one blank. Only the left margin column of a control line should contain a dollar
sign.
The following are examples of control lines:
$NOCODE XREF
$EJECT CODE
There are two types of controls: primary and general. Primary controls must occur
either in the invocation command or in a control line that precedes the first
noncontrol line of the source file. Primary controls cannot be changed within a
module. General controls can occur either in the invocation command or in a control
line anywhere in the source input, and can be changed freely within a module.
Certain controls can be negated by prefacing the control word with a NO. The control
descriptions in this chapter indicate that option by showing both options in the
headings.
Many controls are available, but a set of defaults is built into the compilers. The
controls are summarized in alphabetic order in Table 11-1.
A control consists of a control-name and, in some cases, a parameter. Parameters in
control lines must be enclosed in parentheses. Enclosing control parameters on the
invocation line in parentheses may be illegal, depending on the host operating system.
PL/M-386 Programmer's Guide
Chapter 11
207
The rest of this chapter is organized in the following manner:
208
•
Controls, default settings, abbreviations, and effects are listed in Table 11-1.
•
Compiler controls are categorized and an overview for each of the categories is
provided.
•
Compiler control descriptions are provided in alphabetical order, as listed in
Table 11-1. For example, the NOSYMBOLS description is located with the
SYMBOLS description.
•
A sample program listing is provided with a description of the listing.
Chapter 11
Compiler Invocation and Controls
Table 11-1. Compiler Controls
Control
Default
Abbreviation
Effect
CODE
NOCODE
NOCODE
CO
NOCO
Enables or disables listing of
pseudo-assembly code.
COND
NOCOND
COND
none
none
Determines whether text skipped
during compilation appears in the
listing.
*DEBUG
*NODEBUG
NODEBUG
N
DB
ODB
Generates debug records in the
object module.
EJECT
automatic paging
EJ
Forces a new print page.
IF
ELSEIF
ELSE
ENDIF
not applicable
none
Enables the conditional
compilation capability by testing
for conditions that use the value
of compile-time switches.
INCLUDE
not applicable
IC
Includes other source files as
input to the compiler.
*INTERFACE
none
ITF
Enables calls to other high-level
languages and to source code
translators.
LEFTMARGIN
LEFTMARGIN(1)
LM
Specifies that only input
beginning at position n should be
processed by the compiler.
LIST
NOLIST
LIST
LI
NOLI
Enables or disables listing of
source program.
*MOD486
none
none
Enables use of the Intel486
instruction set.
*OBJECT
*NOOBJECT
OBJECT
(source.obj)
OJ
NOOJ
Specifies a filename for an object
module, or prevents creation of
an object module.
*OPTIMIZE
OPTIMIZE(1)
OT
Determines the optimization level
during code generation.
* Denotes primary control
PL/M-386 Programmer's Guide
continued
Chapter 11
209
Table 11-1. Compiler Controls (continued)
Control
Default
Abbreviation
Effect
OVERFLOW
NOOVERFLOW
NOOVERFLOW
OV
NOOV
Enables or disables overflow
detection during signed
arithmetic.
*PAGELENGTH
PAGELENGTH(60)
PL
Specifies the maximum number
of lines per page.
*PAGEWIDTH
PAGEWIDTH(120)
PW
Specifies the maximum number
of characters per line.
PAGING
NOPAGING
PAGING
PI
NOPI
Specifies whether the program
listing should be page formatted
with a heading that identifies the
compiler and page number. A
user-specified title can also be
included (see TITLE).
PRINT
NOPRINT
PRINT
PR
NOPR
Enables or disables printed
output, or selects the device or
file to receive the printed output.
*RAM
*ROM
**RAM
none
Specifying RAM places the
CONSTANT section within the
DATA segment in all
segmentation. Specifying ROM
places constants in the CODE
segment.
SAVE
RESTORE
none
SA
RS
Enables the settings of certain
controls to be saved on the
stack and restores the control
settings after the included file.
SET
RESETaaa
RESET(0)
none
Controls the value of switches.
SET establishes a value.
RESET restores the value to 0.
* Denotes primary control
210
Chapter 11
continued
Compiler Invocation and Controls
Table 11-1. Compiler Controls (continued)
Control
Default
Abbreviation
Effect
*SMALL
*COMPACT
*FLAT
*MEDIUM
*LARGE
SMALL
SM
CP
MD
FL
LA
Determines the segmentation
model.
*SUBTITLE
no subtitle
ST
Puts a subtitle on each page of
printed output and causes a page
eject.
*SYMBOLS
*NOSYMBOLS
NOSYMBOLS
SB
NOSB
Specifies to the compiler whether
or not to produce a listing of
identifiers and attributes.
*TITLE
module name in
the source code
TT
Places a title on each page of the
printed output.
*TYPE
*NOTYPE
TYPE
TY
NOTY
Specifies whether or not to
include type records in the object
module.
*WORD16
*WORD32
WORD32
W16
W32
Defines the data type
terminology.
*XREF
*NOXREF
NOXREF
XR
NOXR
Enables or disables a
cross-reference listing of source
program identifiers.
* Denotes primary control
PL/M-386 Programmer's Guide
Chapter 11
211
Input Format Control
The LEFTMARGIN control specifies the left margin of the source file.
Code Generation and Object File Controls
These controls determine what type of object file is to be produced and in which
directory it is to appear. Object file controls include the following:
DEBUG|NODEBUG
INTERFACE
MOD486
OBJECT|NOOBJECT
OPTIMIZE
OVERFLOW|NOOVERFLOW
RAM|ROM
SMALL|COMPACT|FLAT|MEDIUM|LARGE
TYPE|NOTYPE
WORD32|WORD16
Segmentation Controls
For PL/M-386, the segmentation controls influence how locations are referenced in
the compiled program, which leads to certain programming restrictions for each of
the segmentation controls. These are primary controls. They have the following
form:
SMALL
COMPACT
MEDIUM
LARGE
FLAT
The segmentation controls SMALL and COMPACT determine the maximum allowable
size of the segments produced in the object program as well as the grouping of object
types (code, data, constants, and stack). These controls affect the operation of the
compiler in various ways and impose certain constraints on the source module being
compiled.
The MEDIUM control is equivalent to the SMALL control. The LARGE control is
equivalent to the COMPACT control except when LARGE is used to indicate a
subsystem whose name is unknown at compile time.
The FLAT control generates an object module containing separate code, data, and
stack segments, with constants in the code segment. You can use the BLD386 FLAT
control to link the segments together in a single segment up to 4 GB.
212
Chapter 11
Compiler Invocation and Controls
For maximum efficiency of the object code, the smallest possible size should be used
for any given program. Also, all modules of a program should be compiled with the
same segmentation control.
The segmentation controls are described later in this chapter; extensions to these
controls, i.e., the use of subsystems, are described in Chapter 13.
Listing Selection and Content Controls
These controls determine what types of listings are produced and where they appear.
The controls are as follows:
CODE|NOCODE
LIST|NOLIST
PRINT|NOPRINT
SYMBOLS|NOSYMBOLS
XREF|NOXREF
Listing Format Controls
Format controls determine the format of the listing output of the compiler. These
controls are as follows:
EJECT
PAGELENGTH
PAGEWIDTH
PAGING|NOPAGING
SUBTITLE
TITLE
Source Inclusion Controls
With these controls, the input source can be changed to a different file. The controls
are:
INCLUDE
SAVE|RESTORE
PL/M-386 Programmer's Guide
Chapter 11
213
Conditional Compilation Controls
These controls cause selected portions of the source file to be skipped by the
compiler if specified conditions are not met. Figure 11-1 shows an example program
using conditional compilation and Figure 11-2 shows the same example program
using the NOCOND control.
The conditional compilation controls are:
COND|NOCOND
IF|ELSEIF|ELSE|ENDIF
SET|RESET
214
Chapter 11
Compiler Invocation and Controls
PL/M-386 COMPILER
EXAMPLE
mm/dd/yy hh:mm:ss
PAGE 1
system-id PL/M-386 Vx.y COMPILATION OF MODULE EXAMPLE
OBJECT MODULE PLACED IN cex.obj
COMPILER INVOKED BY: plm386 cex.src PW(78) SET(DEBUG=3)
1
EXAMPLE: DO
2 1
DECLARE BOOLEAN LITERALLY 'BYTE',
TRUE LITERALLY '0FFH',
FALSE LITERALLY '0';
3 1
PRINT$DIAGNOSTICS: PROCEDURE (SWITCHES, TABLES) EXTERNAL;
4 2
DECLARE (SWITCHES, TABLES) BOOLEAN;
5 2
END PRINT$DIAGNOSTICS;
6
2
DISPLAY$PROMPT: PROCEDURE EXTERNAL; END DISPLAY$PROMPT;
8
2
AWAIT$CR: PROCEDURE EXTERNAL; END AWAIT$CR;
$IF DEBUG = 1
CALL PRINT$DIAGNOSTICS
$
RESET (TRAP)
$ELSEIF DEBUG = 2
CALL PRINT$DIAGNOSTICS
$
RESET (TRAP)
$ELSEIF DEBUG = 3
CALL PRINT$DIAGNOSTICS
CALL PRINT$DIAGNOSTICS
$
SET (TRAP)
$ENDIF
10
11
1
1
12
13
1
1
$IF TRAP
CALL DISPLAY$PROMPT;
CALL AWAIT$CR;
$ENDIF
14
1
END EXAMPLE;
(TRUE, FALSE);
(TRUE, TRUE);
(TRUE, TRUE);
(TRUE, TRUE);
Figure 11-1. Sample Program Using Conditional Compilation (SET control)
PL/M-386 Programmer's Guide
Chapter 11
215
PL/M-386 COMPILER
EXAMPLE
mm/dd/yy hh:mm:ss
PAGE 1
system-id PL/M-386 Vx.y COMPILATION OF MODULE EXAMPLE
OBJECT MODULE PLACED IN cex.obj
COMPILER INVOKED BY: plm386 cex.src PW(78) SET(DEBUG=3) NOCOND
1
EXAMPLE: DO;
2
1
3
4
5
1
2
2
DECLARE IS LITERALLY 'LITERALLY',
BOOLEAN IS 'BYTE',
TRUE IS '0FFH'
FALSE IS '0';
PRINT$DIAGNOSTICS: PROCEDURE (SWITCHES, TABLES) EXTERNAL;
DECLARE (SWITCHES, TABLES) BOOLEAN;
END PRINT$DIAGNOSTICS;
6
2
DISPLAY$PROMPT: PROCEDURE _EXTERNAL; END DISPLAY$PROMPT;
8
2
AWAIT$CR: PROCEDURE EXTERNAL; END AWAIT$CR;
10
11
1
1
$IF DEBUG = 1
$ELSEIF DEBUG = 3
CALL PRINT$DIAGNOSTICS (TRUE, TRUE);
CALL PRINT$DIAGNOSTICS (TRUE, TRUE);
$
SET (TRAP)
$ENDIF
12
13
1
1
14
1
$IF TRAP
CALL DISPLAY$PROMPT;
CALL AWAIT$CR;
$ENDIF
END EXAMPLE;
Figure 11-2. Sample Program Showing the NOCOND Control
216
Chapter 11
Compiler Invocation and Controls
Language Compatibility Control
The INTERFACE control enables PL/M to call procedures written in other languages
and vice versa. For PL/M-386, this control also enables the use of external
procedures compiled with PL/M-286 (or another OMF286 compiler).
Predefined Switches
If one of the switch names (in the following list) appears in an IF or ELSEIF
condition and has not been explicitly assigned a value using the SET or RESET
control, its default value is its primary control value.
SMALL
COMPACT
LARGE
MEDIUM
RAM
ROM
WORD16
WORD32
If a predefined switch is assigned a value using the SET or RESET control, it functions
from that point on like any other switch. A primary control value is not affected by
setting or resetting the predefined switch with the same name.
The four model switches are distinct. Even though the primary controls SMALL and
MEDIUM have the same control interpretation, specifying the MEDIUM control sets the
MEDIUM switch only, and specifying the SMALL control sets the SMALL switch only
(similarly for COMPACT and LARGE).
For example, given the following sequence of PL/M-386 control lines:
$RAM WORD16 MEDIUM
; line 1
$IF RAM
; line 2
.
$ELSEIF WORD32
.
$ELSEIF SMALL
.
$ENDIF
.
$SET (SMALL, WORD32) ; line x
At line 2, the switches RAM and WORD16 are true and their counterparts ROM and
WORD32 are false. The switch MEDIUM is true and the switches SMALL, COMPACT, and
LARGE are false. Therefore, the IF condition is true and the two ELSEIF conditions
are false. After line x, the switches RAM, WORD16, MEDIUM, SMALL, and WORD32 are
true; ROM, COMPACT, and LARGE remain false. The setting of SMALL and WORD32
compile time switches (whether set or reset) does not affect the existing segmentation
control or any of the other switches.
PL/M-386 Programmer's Guide
Chapter 11
217
Compiler Control Encyclopedia
The following sections present each of the PL/M-386 compiler controls. Note that
the segmentation controls are grouped under the SMALL control.
CODE | NOCODE
Form
CODE|NOCODE
Default
NOCODE
Type
General
The CODE control specifies that listing of the generated object code in
pseudo-assembly language format is to begin. This listing is placed at the end of the
program listing in the listing file. Note that the CODE control cannot override a
NOPRINT control.
The NOCODE control specifies that listing of the generated object code is to be
suppressed until the next occurrence, if any, of a CODE control.
COND | NOCOND
These controls determine whether text within an IF element will appear in the listing
if it is skipped during compilation.
Form
COND|NOCOND
Default
COND
Type
General
The COND control specifies that any text that is skipped is to be listed (without
statement or level numbers). Note that a COND control cannot override a NOLIST or
NOPRINT control, and that a COND control will not be processed if it is within text that
is skipped.
The NOCOND control specifies that text within an IF element that is skipped is not to
be listed; however, the controls that delimit the skipped text will be listed. This
provides an indication that something has been skipped. Note that a NOCOND control
will not be processed if it is within text that is skipped.
Figure 11-1 shows an example in which the program was compiled using the COND
(by default) and SET controls with the SET switch assignment DEBUG=3. Figure 11-2
is the same program, but it was compiled using the NOCOND control. These figures
demonstrate the use of conditional compilation. See also the description of
SET|RESET.
218
Chapter 11
Compiler Invocation and Controls
DEBUG | NODEBUG
Form
DEBUG|NODEBUG
Default
NODEBUG
Type
Primary
The DEBUG control specifies that the object module is to contain the statement
number and relative address of each source program statement, information about
each local symbol (including based symbols and procedure parameters), and block
information for each procedure. This information may be used later by a source level
debugging tool.
✏
Note
OPTIMIZE(0) is the only level of optimization that does not
optimize code between program lines. Thus, it is the only one that
gives guaranteed results when debugging programs.
EJECT
Form
EJECT
Default
None
Type
General
EJECT stops printing on the current page and starts a new page of printed output.
IF | ELSE | ELSEIF | ENDIF
These controls provide conditional compilation capability based on the values of
switches.
These controls cannot be used in the invocation of the compiler, and each must be the
only control on its control line. There are no default settings or abbreviations for
these controls.
PL/M-386 Programmer's Guide
Chapter 11
219
An IF control and an ENDIF control delimit an IF element, which can have several
different forms. The simplest form of an IF element is:
$IF condition
text
$ENDIF
Where:
condition is a limited form of a PL/M expression in which the only valid
operators are OR, XOR, NOT, AND, <, <=, =, < >, >=, and >, and the only
valid operands are switches and whole-number constants with a range
of 0 to 255. If the switch does not appear in a SET control, a value of
false (0) is assumed (except for predefined switches). Parenthesized
subexpressions cannot be used. Within these restrictions, condition
is evaluated according to the PL/M rules for expression evaluation.
Note that condition must be followed by an end-of-line.
text
is text that will be processed normally by the compiler if the least
significant bit of the value of condition is a 1, or skipped if the bit is
a 0. Note that text can contain any mixture of PL/M source and
compiler controls. If the text is skipped, any controls within it are not
processed.
The second form of the IF element contains an ELSE element:
$IF condition
text 1
$ELSE
text 2
$ENDIF
In this construction, text 1 will be processed if the least significant bit of the value
of condition is a 1, and text 2 will be skipped. If the bit is a 0, text 1 will be
skipped and text 2 will be processed.
Only one ELSE control can be used within an IF element.
220
Chapter 11
Compiler Invocation and Controls
With the most general form of the IF element, one or more ELSEIF controls can be
introduced before the ELSE (if any):
$IF condition 1
text 1
$ELSEIF condition 2
text 2
$ELSEIF condition 3
text 3
.
.
.
$ELSEIF condition n
text n
$ELSE
text n+1
$ENDIF
where any of the ELSEIF elements can be omitted, as can the ELSE element.
The conditions are tested in sequence. As soon as one of them yields a value with a 1
as its least significant bit, the associated text is processed. All other text in the IF
element is skipped. If none of the conditions yields a least significant bit of 1, the
text in the ELSE element (if any) is processed and all other text in the IF element is
skipped.
Parentheses cannot be used on a conditional control line. For example, the following
line is illegal:
$IF A+(B+C)
INCLUDE
Form
INCLUDE(pathname)
Default
None
Type
General
An INCLUDE control must be the right-most control in a control line or in the
invocation command.
The INCLUDE control causes the specified file to be included during compilation.
Input continues from this file until an end-of-file is detected, and then processing
resumes in the file containing the INCLUDE control.
An included file may also contain INCLUDE controls. Note that such nesting of
included files cannot exceed the depth given in Appendix B.
PL/M-386 Programmer's Guide
Chapter 11
221
INTERFACE
INTERFACE is a primary control that enhances the compatibility of PL/M with other
programming languages. The INTERFACE control enables PL/M programs to call
procedures written in other languages, such as iC-386, if those procedures use the
variable parameter list (VPL) calling convention. Additionally, with the INTERFACE
control, procedures written in PL/M can be called by procedures written in other
languages. The calling conventions for procedures written in Pascal, FORTRAN,
and PL/M are identical.
There are two types of calling conventions in iC-386. One is the FPL and the other is
VPL. The fixed parameter list (FPL) type is the default calling convention of the
iC-386 compiler. So whenever the C procedures are defined to be FPL, no special
designation is needed. But whenever the C procedure is defined to follow the VPL
convention, you must use the INTERFACE control. Note that INTERFACE cannot be
part of an invocation command.
For PL/M-386, INTERFACE is a primary control that enables PL/M-386 programs to
call or be called by procedures compiled with an Intel386 translator, such as iC-386
or ASM386. It can also be used to provide compatibility with procedures compiled
by a 286 translator, such as Fortran-286 or Pascal-286.
The INTERFACE control has the following form:
Form
INTERFACE(lang[/machine[/model[/ram|rom]]]
[=id[id]...])
Default
None
Type
Primary
Where:
222
lang
is the name of the language, eg. C, that requires a different calling
convention for VPL procedures.
machine
is Intel386 when calling VPL iC-386 procedures, and 286 when calling
VPL procedures compiled using a 286 translator. Only references to
286 ids from Intel386 modules are supported; Intel386 ids cannot be
referenced from 286 modules. Therefore, if machine is 286 then all
the identifiers in the id list must be declared EXTERNAL. If machine is
286 and an id is PUBLIC, it is an error.
Chapter 11
Compiler Invocation and Controls
model
is SMALL, COMPACT, MEDIUM, or LARGE and defines the model of
segmentation for the specified ids. model determines whether
Intel386 POINTER variables are offset-only or select-offset, as defined
by the PL/M-386 models of segmentation (see Chapter 13, Table 13-1).
If machine is 286, model defaults to LARGE. model should be
specified as the same model of segmentation used to compile the 286
code being referenced. If machine is Intel386, model is ignored.
ram|rom
is RAM or ROM and defines the placement of constant variables in either
the code or data segment. When used with the SMALL model, ram|rom
also defines whether POINTER variables are offset-only or
selector-offset. The default is RAM unless model is LARGE, in which
case the default is ROM. ram|rom is ignored if machine is Intel386.
id
specifies the procedures and variables that are implemented using the
specified language interface convention.
When the INTERFACE control is used to call procedures compiled with a 286
translator, the program switches from using 32-bit stack offsets to 16-bit offsets.
Therefore, the stack pointer for the called procedure must point within the lowest
64K of the stack segment, or else a gate must be used to switch to such a stack
segment. Parameters must fit within this boundary as well.
Except as noted above, the calling conventions for Intel386-based languages other
than VPL iC-386 procedures are identical to PL/M-386 and therefore do not require
the use of the INTERFACE control. Because the calling conventions differ for iC-386,
INTERFACE must be used to call or to be called for VPL iC-386 procedures. The C
(VPL) interface convention differs from the PL/M calling convention in the
following ways:
•
Parameters are evaluated and pushed onto the stack in reverse order.
•
A parameter whose size is less than two bytes (for Intel386 and Intel486
processors, four bytes) is zero-extended or sign-extended according to its type.
•
Real parameters for iC-386 are always 64-bit double floating-point numbers and
are passed on its stack.
•
The caller clears the parameters off the stack after return and the callee does not
pop parameters off the stack.
PL/M-386 Programmer's Guide
Chapter 11
223
If you define a function to be a C calling convention procedure, you can call it with
more arguments than the number of parameters you specify in the external
declaration. Thus, you can make variable parameter list (VPL) procedure calls to
functions such as the C function printf. This feature is similar to the ANSI C
prototyped function declarations using ellipses (,...). Type checking occurs for
arguments passed to the parameters you specify in the external declaration, not for
any additional arguments. For example, no type checking is done on a call to a
procedure declared with no parameters.
For example, the following is valid:
/* Define SAMP as a C Calling Convention procedure.
$INTERFACE (C=SAMP)
/* Declare SAMP; specify a parameter of type WORD.
SAMP: PROCEDURE (P) EXTERNAL;
DECLARE P WORD;
END SAMP;
/* Declare variables to pass to SAMP.
DECLARE (A, B, C) WORD;
/* Pass arguments to SAMP. Type checking occurs
/* for the argument A (parameter P) but not for
/* arguments B and C.
CALL SAMP (A, B, C);
224
Chapter 11
*/
*/
*/
*/
*/
*/
Compiler Invocation and Controls
Constant arguments are typed as in PL/M. This typing can affect the value of an
argument passed to a C routine, as demonstrated in the following example, where
SAMP2 and SAMP3 are C interface functions:
SAMP2: PROCEDURE (D) EXTERNAL;
DECLARE D INTEGER;
END SAMP2;
SAMP3: PROCEDURE EXTERNAL;
END SAMP3;
CALL SAMP2 ( 113 );
/* passes 113 as an INTEGER since D is
declared as type INTEGER. */
CALL SAMP3 ( 115 );
/* passes 115 as a BYTE
(since 0 < 115 < 255), but high
byte(s) are undefined since C
does integer promotion, even if
the first argument of SAMP3 is
an unsigned char argument stack. */
CALL SAMP3 ( INTEGER(115) );
/* uses an explicit cast to
ensure the constant is passed correctly. */
The following example demonstrates the use of the INTERFACE control to call a
PL/M-386 procedure:
$WORD32
$INTERFACE(PLM/386/FLAT/ROM=EXAMP)
EXAMP:PROCEDURE(A,B)EXTERNAL;
DECLARE A WORD, B POINTER;
END EXAMP;
DECLARE X WORD, Y POINTER;
CALL EXAMP(X,Y);
PL/M-386 Programmer's Guide
Chapter 11
225
In the preceding example, the INTERFACE control specifies the procedure EXAMP to
be defined as Intel386-compatible. The actual parameters X and Y will be
automatically converted to a 32-bit WORD and an Intel386 (48-bit) POINTER,
respectively.
Variables and formal parameters of Intel386-based procedures should be declared the
same as in the PL/M-386 code. The PL/M-386 compiler is also able to interpret the
terms in 286 context and perform the following mapping:
Term Used
BYTE
Maps to Data Type
HWORD
8-bit, unsigned
WORD
16-bit, unsigned
DWORD
32-bit, unsigned
QWORD
32-bit, unsigned
CHARINT
8-bit (interpretation dependent on 286 code)
SHORTINT
8 bit (interpretation dependent on 286 code)
INTEGER
16-bit, signed integer
LONGINT
32-bit(interpretation dependent on 286 code)
REAL
32-bit, real
SELECTOR
16-bit, selector
POINTER
see the following paragraphs
OFFSET
16-bit, unsigned
8-bit, unsigned
The PL/M-386 compiler converts Intel386-style 48-bit long POINTERs to 286
POINTERs by truncating the offset portion to 16 bits. In SMALL RAM, a POINTER is
the same as an OFFSET, and is treated as such by the compiler.
Note that this mapping is independent of WORD16|WORD32 (defined in Tables 9-3 and
11-4). This means that there is a third mapping of scalar terms to scalar data types.
226
Chapter 11
Compiler Invocation and Controls
LEFTMARGIN
This is the only control for specifying the format of the source input.
Form
LEFTMARGIN(n)
Default
LEFTMARGIN(1)
Type
General
All characters to the left of position n on subsequent input lines are not processed by
the compiler (but do appear on the listing). The first character on a line is in column
1.
The new setting of the left margin takes effect on the next input line. It remains in
effect for all input from this source file and any included files until it is reset by
another LEFTMARGIN control.
Note that a control line is one that contains a dollar sign in the column specified by
the most recent LEFTMARGIN control.
LIST | NOLIST
Form
LIST|NOLIST
Default
LIST
Type
General
The LIST control specifies that listing of the source program is to resume with the
next source line read. The PL/M-386 compiler numbers all source lines,
incrementing the line number for each new-line character. Note that the LIST
control cannot override a NOPRINT control. If NOPRINT is in effect, no listing is
produced.
The NOLIST control specifies that listing of the source program is to be suppressed
until the next occurrence, if any, of a LIST control.
When LIST is in effect, all input lines (from the source file or from an included file),
including control lines, are listed, provided there is not a NOPRINT control in effect.
When NOLIST is in effect, only source lines associated with error messages are
listed.
PL/M-386 Programmer's Guide
Chapter 11
227
MOD486
The MOD486 control, recognized by only the PL/M-386 compiler, is a switch
governing the instruction set available to the compiler. Use this control to compile
source text containing the following built-ins specific to the Intel486 processor:
NAME
BYTE$SWAP
USAGE
Byte swap function to convert between big and little
endian. The endian of a stored value indicates
whether the most-significant bit is in the highest (big
endian) or lowest (little endian) address of the
location.
TEST$REGISTER
Built-in variable extending the number of available
TEST$REGISTER elements
INVALIDATE$DATA$CACHE
Function to clear the entire data cache
WB$INVALIDATE$DATA$CACHE
Function to write-back changed lines to memory
and to clear the data cache
INVALIDATE$TLB$ENTRY
Function to invalidate a single entry in the paging
cache
OBJECT | NOOBJECT
Form
OBJECT(pathname)
NOOBJECT
Default
OBJECT(sourcefilename.OBJ)
Type
Primary
The OBJECT control specifies that an object module is to be created during
compilation. The pathname is a standard host operating system pathname that
specifies the file to receive the object module. If the control is absent or if an
OBJECT control appears without a pathname, the object module is directed to a file
that has the same name as the source input file, but with the extension .OBJ.
The NOOBJECT control specifies that no object module is to be produced.
228
Chapter 11
Compiler Invocation and Controls
OPTIMIZE
This control governs the level of optimization to be performed in generating object
code. The n parameter can be 0-3, representing the lowest to highest levels of
optimization. Figures 11-3 to 11-6 illustrate the different levels of optimization. The
same program was compiled for each level, but the source file was printed only for
OPTIMIZE(0).
Form
OPTIMIZE(n)
Where:
n = 0, 1, 2 or 3
Default
OPTIMIZE(1)
Type
Primary
OPTIMIZE(0) specifies only folding of constant expressions. Folding means
recognizing, during compilation, operations that are superfluous or combinable, and
removing or combining them so as to save memory space or execution time.
Examples include addition with a zero operand, multiplication by one, and logical
expressions with true or false constants.
OPTIMIZE(0) is the only level of optimization that is guaranteed to not optimize
code between lines. Figure 11-3 illustrates the OPTIMIZE(0) level of optimization.
PL/M-386 Programmer's Guide
Chapter 11
229
PL/M-386 COMPILER
EXAMPLES_OF_OPTIMIZATIONS
date time
PAGE 1
system-id PL/M-386 Vx.y COMPILATION OF MODULE
EXAMPLES_OF_OPTIMIZATIONS
OBJECT MODULE PLACED IN example.obj
COMPILER INVOKED BY: plm386 example.src PW(78) FLAT CODE OPTIMIZE(0)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
EXAMPLES_OF_OPTIMIZATIONS: DO;
DECLARE (A,B,C) WORD,
D(100) WORD,
(PTR_1, PTR_2) POINTER,
ABASED BASED PTR_1 (10) WORD;
DO WHILE D(A+B) < D(A+B+1);
IF (OFFSET(PTR_1) < (OFFSET(PTR_2)) THEN DO;
A = A * 2;
ABASED(A) = ABASED(B);
ABASED(B) = ABASED(C);
END;
ELSE A = A + 1;
END;
END EXAMPLES_OF_OPTIMIZATIONS;
1
1
1
1
1
2
3
3
3
3
2
2
1
PL/M-386 COMPILER
EXAMPLES_OF_OPTIMIZATIONS
date time
PAGE 2
ASSEMBLY LISTING OF OBJECT CODE
00000000
8BEC
MOV
EBP,ESP
00000002
00000008
0000000E
00000014
0000001A
0000001B
00000022
00000029
@1:
8B0500000000
030504000000
8B0D00000000
030D04000000
41
8B04850C000000
3B048D0C000000
0F8375000000
MOV
ADD
MOV
ADD
INC
MOV
CMP
JAE
0000002F
00000035
8B059C010000
8B0DA0010000
MOV
MOV
EAX,A
EAX,B
ECX,A
ECX,B
ECX
EAX,D[EAX*4]
EAX,D[ECX*4]
@2
; STATEMENT # 7
EAX,PTR_1
ECX,PTR_2
Figure 11-3. Sample Program Showing the OPTIMIZE(0) Control
230
Chapter 11
Compiler Invocation and Controls
0000003B
0000003D
3BC1
0F834F000000
CMP
JAE
EAX,ECX
@3
00000043
00000049
0000004B
8B0500000000
D1E0
890500000000
MOV
SHL
MOV
EAX,A
EAX,1
A,EAX
00000051
00000057
0000005D
00000060
00000066
0000006C
8B059C010000
8B0D04000000
8B0488
8B0D9C010000
8B1500000000
890491
MOV
MOV
MOV
MOV
MOV
MOV
0000006F
00000075
0000007B
0000007E
00000084
0000008A
0000008D
8B059C010000
8B0D08000000
8B0488
8B0D9C010000
8B1504000000
890491
E90D000000
MOV
MOV
MOV
MOV
MOV
MOV
JMP
00000092
00000098
00000099
@3:
8B0500000000
40
890500000000
MOV
INC
MOV
; STATEMENT # 8
; STATEMENT # 9
EAX,PTR_1
ECX,B
EAX,[EAX].ABASED[ECX*4]
ECX,PTR_1
EDX,A
[ECX].ABASED[EDX*4],EAX
; STATEMENT # 10
EAX,PTR_1
ECX,C
EAX,[EAX].ABASED[ECX*4]
ECX,PTR_1
EDX,B
[ECX].ABASED[EDX*4],EAX
@4
; STATEMENT # 12
EAX,A
EAX
A,EAX
; STATEMENT # 13
0000009F
@4:
E95EFFFFFF
@2:
JMP
@1
; STATEMENT # 15
Figure 11-3. Sample Program Showing the OPTIMIZE(0) Control (continued)
PL/M-386 Programmer's Guide
Chapter 11
231
OPTIMIZE(1) specifies strength reduction, elimination of common subexpressions
and short-circuit evaluation of some Boolean expressions, as well as the
optimizations of level (0).
Strength reduction means substituting quick operations (e.g., shifting by 1 instead of
multiplying by 2). This instruction requires less space and executes faster. Adding
identical subexpressions may also generate left shift instructions.
Elimination of common subexpressions means that if an expression reappears in the
same block, its value is re-used rather than recomputed. The compiler also
recognizes commutative forms of subexpressions (e.g., A + B and B + A are seen as
the same). Intermediate results during expression evaluation are saved in either
registers or on the stack for later use. For example:
A = B + C * D / 3;
C = E + D * C / 3;
The value of C*D/3 will not be recomputed for the second statement.
Optimizing the evaluation of Boolean expressions uses the fact that in certain cases
some of the terms are not needed to determine the value of the expression. For
example, in the expression:
( A > B AND I > J )
if the first term (A>B) is false, the entire expression is false, and it is not necessary to
evaluate the second term. The use of PL/M built-in procedures does not change this
optimization. However, if a user-written function or an embedded assignment is part
of the expression, this short evaluation is not done. For example:
(A > B AND ( UFUN (A) > J ) )
is evaluated in full.
Figure 11-4 illustrates the OPTIMIZE(1) level of optimization.
232
Chapter 11
Compiler Invocation and Controls
PL/M-386 COMPILER
EXAMPLES_OF_OPTIMIZATIONS
date time
PAGE 1
system-id PL/M-386 Vx.y COMPILATION OF MODULES
EXAMPLES_OF_OPTIMIZATIONS
OBJECT MODULE PLACED IN example.obj
COMPILER INVOKED BY: plm386 example.src PW(78) FLAT MODE
OPTIMIZE(1) NOLIST
PL/M-386 COMPILER
EXAMPLES_OF_OPTIMIZATIONS date time PAGE
ASSEMBLY LISTING OF OBJECT CODE
2
; STATEMENT # 6
00000000
8BEC
MOV
EBP,ESP
00000002
00000008
0000000E
00000010
00000011
00000012
00000013
0000001A
00000021
@1:
8B0500000000
8B0D04000000
03C1
50
40
5A
8B14950C000000
3B14850C000000
0F8356000000
MOV
MOV
ADD
PUSH
INC
POP
MOV
CMP
JAE
00000027
0000002D
00000033
00000035
8B059C010000
8B15A0010000
3BC2
0F8337000000
MOV
MOV
CMP
JAE
0000003B
00000041
00000043
8B0500000000
D1E0
890500000000
MOV
SHL
MOV
00000049
0000004F
00000052
8B159C010000
8B0C8A
890C82
MOV
MOV
MOV
00000055
0000005B
8B059C010000
8B0D08000000
MOV
MOV
EAX,A
ECX,B
EAX,ECX
EAX
; 1
EAX
EDX
; 1
EDX,D[EDX*4]
EDX,D[EAX*4]
@2
; STATEMENT # 7
EAX,PTR_1
EDX,PTR_2
EAX,EDX
@3
; STATEMENT # 8
EAX,A
EAX,1
A,EA
; STATEMENT # 9
EDX,PTR_1
ECX,[EDX].ABASED[ECX*4]
[EDX].ABASED[EAX*4],ECX
; STATEMENT # 10
EAX,PTR_1
ECX,C
Figure 11-4. Sample Program Showing the OPTIMIZE(1) Control
PL/M-386 Programmer's Guide
Chapter 11
233
00000061
00000064
0000006A
0000006D
8B0C88
8B1504000000
890C90
E906000000
MOV
MOV
MOV
JMP
ECX,[EAX].ABASED[ECX*4]
EDX,B
[EAX].ABASED[EDX*4],ECX
@4
; STATEMENT # 12
00000072
@3:
FF0500000000
INC
A
; STATEMENT # 13
00000078
@4:
E985FFFFFF
@2:
JMP
@1
; STATEMENT # 15
MODULE INFORMATION:
CODE AREA SIZE
CONSTANT AREA SIZE
VARIABLE AREA SIZE
MAXIMUM STACK SIZE
15 LINES READ
0 PROGRAM WARNINGS
0 PROGRAM ERRORS
=
=
=
=
0000007DH
00000000H
000001A4H
00000004H
125D
0D
420D
4D
DICTIONARY SUMMARY:
410KB MEMORY AVAILABLE
8KB MEMORY USED
(1%)
0KB DISK SPACE USED
END OF PL/M-386 COMPILATION
Figure 11-4. Sample Program Showing the OPTIMIZE(1) Control (continued)
234
Chapter 11
Compiler Invocation and Controls
OPTIMIZE(2) includes OPTIMIZE(0) and OPTIMIZE(1), plus the following:
•
Machine code optimizations (e.g., short jumps, moves)
•
Elimination of superfluous branches
•
Reuse of duplicate code
•
Removal of unreachable code and reversal of branch conditions
Optimizing machine code means saving space by using shorter forms for identical
machine instructions. This is possible because some instructions have multiple
forms. For example:
MOV RESLT1,AX;
/* move accumulator value to location RESLT1*/
can be generated by using three or four bytes for PL/M-86 and PL/M-286, and using
five or six bytes for PL/M-386. The former choice saves a byte of storage for the
program. Similarly, jumps that the compiler can recognize as within the same
segment or closer (within 127 bytes) permit the use of fewer byte instructions.
Elimination of superfluous branches means optimizing consecutive or multiple
branches into a single branch. For example:
JZ
JMP
LAB1:
LAB2:
LAB1;
LAB2;
.....
...
...
.....
/* jump on zero to LAB1 */
/* unconditional jump to LAB2 */
will be transformed into:
JNZ
LAB1
LAB2:
LAB2;
.....
...
...
.....
/* jump on non-zero to LAB2 */
Similarly, multiple branches like the following are eliminated:
LAB0:JMP
LAB1:JPM
LAB1
...
...
LAB2
...
...
LAB2:.....
PL/M-386 Programmer's Guide
Chapter 11
235
and transformed into:
LAB0:JMP
LAB1:JMP
LAB2
...
...
LAB2
...
...
LAB2:.....
Reuse of duplicate code can refer to identical code at the end of two converging
paths. In such a case, the code is inserted in only one path, and a jump to that path is
inserted in the other path. For example:
DECLARE A BYTE, SPOT POINTER;
DECLARE S BASED SPOT STRUCTURE (B BYTE, C BYTE);
IF A = 1 THEN
S.C = INPUT (0F7H) AND 07FH;
ELSE
S.C = INPUT (0F9H) AND 07FH;
Before
@1:
236
CMP
JZ
JMP
IN
AND
MOV
MOV
JMP
IN
AND
MOV
MOV
@2:
Chapter 11
After
A, 1H
& + 5H
@1
0F7H
AL, 7FH
BX, SPOT
S [BX+1H], AL
@2
0F9H
AL, 7FH
BX, SPOT
S [BX+1H], AL
CMP
JMP
A, 1H
@1
IN
JMP
0F7H
@2
@1: IN
@2: AND
MOV
MOV
0F9H
AL, 7FH
BX, SPOT
S [BX+1H],AL
Compiler Invocation and Controls
Reuse of duplicate code can also refer to machine instructions, immediately
preceding a loop, that are identical to those ending the loop. A branch can be
generated to reuse the code generated at the beginning of the loop. For example:
Before
ADD
MOV
LAB0: MOV
CMP
JNZ
...
...
ADD
MOV
JMP
LAB1:
After
AX, BX
ANS, AX
AL, DUM1
AL, DUM2
LAB1
AX, BX
ANS, AX
LAB0
...
LAB0: ADD AX, BX
MOV ANS, AX
MOV AL, DUM1
CMP AL, DUM2
JNZ LAB1
...
...
JMP LAB0
LAB1: ...
This is safe so long as LAB0 is not the target of a jump instruction. The compiler
normally handles a whole procedure at a time, and is aware of such a condition. This
optimization cannot be safely applied to labels in the outer level of the main program
module. This optimization will not change the program and will save code space.
Second level optimization removes unreachable code, takes a second look at the
generated object code, and finds areas that can never be reached due to the control
structures created earlier.
For example, if the following code were generated before optimization:
@1:
@2:
@3:
MOV
RCR
JB
JMP
AX, A
AL, 1
@1
@2
MOV
OUTW
JMP
MOV
ADD
JMP
....
....
....
....
.....
.....
AX, 0FFFFH
0F6H
@2
AX, B
A, AX
@3
PL/M-386 Programmer's Guide
Chapter 11
237
Then the removal of unreachable code would produce:
@1:
@2:
@3:
MOV
RCR
JB
JMP
AX, A
AL, 1
@1
@2
MOV
OUTW
JMP
...
...
.....
AX, 0FFFFH
0F6H
@2
This can be further optimized by reversing the branch condition in the third
instruction and removing the unnecessary JMP @2:
@1:
@2:
@3:
MOV
RCR
JNB
AX, A
AL, 1
@2
MOV
OUTW
...
...
.....
AX, 0FFFFH
0F6H
Figure 11-5 illustrates the OPTIMIZE(2) level of optimization.
238
Chapter 11
Compiler Invocation and Controls
PL/M-386 COMPILER
EXAMPLES_OF_OPTIMIZATIONS
date time
PAGE 1
system-id PL/M-386 Vx.y COMPILATION OF MODULES
EXAMPLES_OF_OPTIMIZATIONS
OBJECT MODULE PLACED IN example.obj
COMPILER INVOKED BY: plm386 example.src PW(78) FLAT MODE
OPTIMIZE(2) NOLIST
PL/M-386 COMPILER
EXAMPLES_OF_OPTIMIZATIONS date time PAGE
ASSEMBLY LISTING OF OBJECT CODE
2
; STATEMENT # 6
00000000
8BEC
MOV
EBP,ESP
00000002
00000007
0000000D
0000000F
00000010
00000011
00000012
00000019
00000020
@1:
A100000000
8B0D04000000
03C1
50
40
5A
8B14950C000000
3B14850C000000
7348
MOV
MOV
ADD
PUSH
INC
POP
MOV
CMP
JNB
00000022
00000027
0000002D
0000002F
A19C010000
8B15A0010000
3BC2
7331
MOV
MOV
CMP
JNB
00000031
00000036
00000038
A100000000
D1E0
A300000000
MOV
SHL
MOV
0000003D
00000043
00000046
8B159C010000
8B0C8A
890C82
MOV
MOV
MOV
00000049
0000004E
00000054
A19C010000
8B0D08000000
8B0C88
MOV
MOV
MOV
EAX,A
ECX,B
EAX,ECX
EAX
; 1
EAX
EDX
; 1
EDX,D[EDX*4]
EDX,D[EAX*4]
@2
; STATEMENT # 7
EAX,PTR_1
EDX,PTR_2
EAX,EDX
@3
; STATEMENT # 8
EAX,A
EAX,1
A,EAX
; STATEMENT # 9
EDX,PTR_1
ECX,[EDX].ABASED[ECX*4]
[EDX].ABASED[EAX*4],ECX
; STATEMENT # 10
EAX,PTR_1
ECX,C
ECX,[EAX].ABASED[ECX*4]
Figure 11-5. Sample Program Showing the OPTlMIZE(2) Control
PL/M-386 Programmer's Guide
Chapter 11
239
00000057
0000005D
00000060
8B1504000000
890C90
EBA0
MOV
MOV
JMP
EDX,B
[EAX].ABASED[EDX*4],ECX
@1
; STATEMENT # 12
00000062
@3:
FF0500000000
INC
A
00000068
EB98
JMP
@1
; STATEMENT # 13
@2:
; STATEMENT # 15
MODULE INFORMATION:
CODE AREA SIZE
CONSTANT AREA SIZE
VARIABLE AREA SIZE
MAXIMUM STACK SIZE
15 LINES READ
0 PROGRAM WARNINGS
0 PROGRAM ERRORS
=
=
=
=
0000006AH
00000000H
000001A4H
00000004H
106D
0D
420D
4D
DICTIONARY SUMMARY:
410KB MEMORY AVAILABLE
8KB MEMORY USED
(1%)
0KB DISK SPACE USED
END OF PL/M-386 COMPILATION
Figure 11-5. Sample Program Showing the OPTIMIZE(2) Control (continued)
240
Chapter 11
Compiler Invocation and Controls
OPTIMIZE(3) includes all of the preceding optimizations. It also optimizes
indeterminate storage operations (e. g., those using based variables or variables
declared with the AT attribute).
✏
Note
The assumption validating this new optimization is that based
variables (or AT variables) do not overlay other user-declared
variables.
On this optimization level, all Boolean expressions are short-circuited except those
containing embedded assignments. (For a description of how this optimization
occurs, see OPTIMIZE(1).)
The benefits of this optimization level include more efficient use of code space only
if needed values are not overlaid.
Caution in variable-declaration and usage is essential. For example, the sequence:
DECLARE (I, J) WORD;
DECLARE THETA (19) AT (@I);
DECLARE A BASED J (10);
STRUCTURE (Fl BYTE, F2 WORD);
..
J=.I;
....
..
A(I).F1 = 7;
A(I).F2 = 99;
THETA(I) = 31;
..
..
violates this caution because it causes the values being used as pointers and subscripts
to be overlaid. The compiler normally takes steps to avoid the difficulties implied
here. But, in OPTIMIZE(3), these steps are omitted due to the implicit requirement
that such situations must not be present at this level of optimization.
Figure 11-6 illustrates the OPTIMIZE(3) level of optimization.
PL/M-386 Programmer's Guide
Chapter 11
241
PL/M-386 COMPILER
EXAMPLES_OF_OPTIMIZATIONS
date time
PAGE 1
system-id PL/M-386 Vx.y COMPILATION OF MODULES
EXAMPLES_OF_OPTIMIZATIONS
OBJECT MODULE PLACED IN example.obj
COMPILER INVOKED BY: plm386 example.src PW(78) FLAT MODE
OPTIMIZE(3) NOLIST
PL/M-386 COMPILER EXAMPLES_OF_OPTIMIZATIONS date time PAGE 2
ASSEMBLY LISTING OF OBJECT CODE
; STATEMENT # 6
00000000
8BEC
MOV
EBP,ESP
00000002
00000007
0000000D
0000000F
00000010
00000011
00000012
00000019
00000020
@1:
A100000000
8B0D04000000
03C1
50
40
5A
8B14950C000000
3B14850C000000
733C
MOV
MOV
ADD
PUSH
INC
POP
MOV
CMP
JNB
00000022
00000027
0000002D
0000002F
A19C010000
8B15A0010000
3BC2
7325
MOV
MOV
CMP
JNB
00000031
00000036
00000038
A100000000
D1E0
A300000000
MOV
SHL
MOV
0000003D
00000043
00000046
8B159C010000
8B1C8A
891C82
MOV
MOV
MOV
00000049
0000004E
00000051
00000054
A108000000
8B0482
89048A
EBAC
MOV
MOV
MOV
JMP
EAX,A
ECX,B
EAX,ECX
EAX
;
EAX
EDX
; 1
EDX,D[EDX*4]
EDX,D[EAX*4]
@2
; STATEMENT # 7
EAX,PTR_1
EDX,PTR_2
EAX,EDX
@3
; STATEMENT # 8
EAX,A
EAX,1
A,EAX
; STATEMENT # 9
EDX,PTR_1
EBX,[EDX].ABASED[ECX*4]
[EDX].ABASED[EAX*4],EBX
; STATEMENT # 10
EAX,C
EAX,[EDX].ABASED[EAX*4]
[EDX].ABASED[ECX*4],EAX
@1
Figure 11-6. Sample Program Showing the OPTIMIZE(3) Control
242
Chapter 11
Compiler Invocation and Controls
; STATEMENT # 12
00000056
@3:
FF0500000000
INC
A
0000005C
EBA4
JMP
@1
; STATEMENT # 13
@2:
; STATEMENT # 15
MODULE INFORMATION:
CODE AREA SIZE
CONSTANT AREA SIZE
VARIABLE AREA SIZE
MAXIMUM STACK SIZE
15 LINES READ
0 PROGRAM WARNINGS
0 PROGRAM ERRORS
=
=
=
=
0000005EH
00000000H
000001A4H
00000004H
94D
0D
420D
4D
DICTIONARY SUMMARY:
410KB MEMORY AVAILABLE
8KB MEMORY USED
(1%)
0KB DISK SPACE USED
END OF PL/M-386 COMPILATION
Figure 11-6. Sample Program Showing the OPTIMIZE(3) Control (continued)
PL/M-386 Programmer's Guide
Chapter 11
243
OVERFLOW | NOOVERFLOW
Form
OVERFLOW|NOOVERFLOW
Default
NOOVERFLOW
Type
General
These controls specify whether to detect overflow when performing signed
arithmetic. If the NOOVERFLOW control is specified, no overflow detection is
implemented in the compiled module and the results of overflow in signed arithmetic
are undefined. If the OVERFLOW control is specified, overflow in signed arithmetic
results in a nonmaskable interrupt 4, and it is the programmer's responsibility to
provide an interrupt procedure to handle the interrupt. Failure to provide such a
procedure may result in unpredictable program behavior when overflow occurs.
If this control is nested within a program statement, overflow detection will begin
when the next complete statement is evaluated.
Note that the use of the OVERFLOW control results in some expansion of the object
code.
Specific to the Intel386 and Intel486 microprocessors, in-line checking code is
inserted for detecting machine overflow (32-bit arithmetic overflow) on signed
expressions, and value overflow on assignments to SHORTINT or CHARINT variables.
To save code space and execution time, avoid using SHORTINT and CHARINT when
compiling with the OVERFLOW control.
PAGELENGTH
Form
PAGELENGTH(n)
Default
PAGELENGTH(60)
Type
Primary
Pagelength is a non-zero, unsigned number specifying the maximum number of lines
to be printed per page of listing output. This number includes the page headings
printed on the page.
The minimum value for n is 5; the maximum value is 255.
244
Chapter 11
Compiler Invocation and Controls
PAGEWIDTH
Form
PAGEWIDTH(n)
Default
PAGEWIDTH(120)
Type
Primary
Pagewidth is a non-zero, unsigned number specifying the maximum line width, in
characters, to be used for listing output. The minimum value for n is 60; the
maximum value is 132.
PAGING | NOPAGING
Form
PAGING|NOPAGING
Default
PAGING
Type
Primary
The PAGING control specifies that the listed output is to be formatted onto pages.
Each page carries a heading identifying the compiler and a page number, and
possibly a user-specified title.
The NOPAGING control specifies that page ejecting, page heading, and page
numbering are not to be performed. Thus, the listing appears on one long page as
would be suitable for a slow serial output device. If NOPAGING is specified, a page
eject is not generated if an EJECT control is encountered.
PRINT | NOPRINT
Form
PRINT(pathname)
NOPRINT
Default
PRINT(sourcefilename.LST)
Type
Primary
The PRINT control specifies that printed output is to be produced. The parameter is a
standard host operating system pathname that specifies the file to receive the printed
output. Any output-type device, including a disk file, can also be given. If the
control is absent, or if a PRINT control appears without a pathname, printed output is
sent to a file that has the same name as the source input file but with the extension
.LST.
The NOPRINT control specifies that no printed output is to be produced, even if
implied by other listing controls such as LIST and CODE.
PL/M-386 Programmer's Guide
Chapter 11
245
RAM | ROM
Form
RAM|ROM
Default
RAM
Type
Primary
For PL/M-386, the RAM setting places the CONSTANT section within the DATA
segment in all segmentation models.
For all targets, the ROM setting places constants in the CODE segment. Under this
setting, the INITIAL attribute on a variable produces a warning message. Do not use
the dot operator for variable references under the ROM option because constants and
variables will be relative to different segment registers. If SMALL is specified with
the ROM control, then PL/M-386 pointers will be six bytes instead of four (see also
Appendix F).
If the keyword DATA is used in a PUBLIC declaration when compiling with the ROM
control, DATA must also be used in the EXTERNAL declaration of program modules
that reference it. However, no value list is then permitted because the data is defined
elsewhere.
SAVE | RESTORE
Form
SAVE|RESTORE
Default
None
Type
General
With these controls the settings of certain general controls can be saved on a stack
and then restored. The main usage of these controls is saving the controls before an
included file and restoring them after inclusion of that file is complete. The controls
whose settings are saved and restored are:
CODE|NOCODE
COND|NOCOND
LEFTMARGIN
LIST|NOLIST
OVERFLOW|NOOVERFLOW
The SAVE control saves all of these settings on a stack. The maximum capacity of
this stack corresponds to the maximum nesting depth for the INCLUDE control (the
maximum nesting depth is given in Appendix B).
The RESTORE control restores the most recently saved set of control settings from the
stack.
246
Chapter 11
Compiler Invocation and Controls
SET | RESET
These are general controls. The SET control has the following general form:
SET (switch_assignment_list)
Where:
switch_assignment_list
consists of one or more switch assignments separated by commas.
A switch assignment has the form:
switch[=value]
Where:
switch
is a name which is formed according to the PL/M rules for identifiers.
Note that a switch name exists only at the compiler control level, and
therefore a switch can have the same name as an identifier in the
program; no conflict is possible. Note however that no PL/M reserved
word other than a predefined switch can be used as a switch name.
value
is a whole-number constant in the range 0 to 255. This value is
assigned to the switch. If the value and the equal sign (=) are omitted
from the switch assignment, the default value 0FFH (true) is assigned to
a switch.
The following is an example of a SET control line:
$SET(TEST,ITERATION = 3)
This example sets the switch TEST to true (0FFH) and the switch ITERATION to 3.
Switches do not need to be declared.
Figure 11-1 and 11-2 are examples of a program that was compiled using the SET
control.
The RESET control has the form:
RESET (switch_list)
Where:
switch_list
consists of one or more switch names that have already occurred in SET
controls.
Each switch in the switch list is set to false (0).
PL/M-386 Programmer's Guide
Chapter 11
247
SMALL | COMPACT | MEDIUM | LARGE | FLAT
The following sections describe the SMALL, COMPACT, MEDIUM, LARGE and FLAT
controls (also called the segmentation controls). For application development under
the iRMX Operating System, see the note under the COMPACT model description.
SMALL
Form
SMALL
Default
SMALL
Type
Primary
Modules compiled with the SMALL control have three sections: code, data, and stack
(see the OBJECT control). When these modules are bound (linked), similar sections
from each module are combined to form two segments: code and data. For the
Intel386 and Intel486 microprocessors, the maximum size of each segment is 4G
bytes.
In the default SMALL case (RAM), the code sections from all modules are allocated
space within the code segment, which is addressed relative to the CS register.
Constants are combined with all the data and stack sections in the data segment. For
the Intel386 and Intel486 microprocessors, this segment is addressed relative to the
DS register, with an identical copy in the SS and ES registers. None of the segment
registers are changed during the course of program execution except ES, which is
used to perform string operations, and FS and GS, which are used to address data
exported by another subsystem. Subsystems are described in Chapter 13.
Therefore, the SMALL control can be used only if the total size of all code sections
does not exceed 4G bytes. The total size of the constants plus all data and stack
sections also cannot exceed 4G bytes.
If the ROM control is used, the constants from all the modules are placed with the code
in the code segment. The data segment then contains only the data and stack sections
from all the modules.
Because only one code segment exists, its segment selector (the CS register) is never
updated during program execution. (However, an interrupt will update the CS
register.) Likewise, when RAM is used, only one segment exists for all constant, data,
and stack sections. The segments' selectors (the DS and SS registers) are never
updated (except when an interrupt occurs, as explained in Appendix G). Therefore,
when any location is referenced, a 32-bit offset is calculated and used in conjunction
with the appropriate segment selector. POINTER values in the SMALL (RAM) case are
32-bit values for the Intel386 and Intel486 microprocessors.
The following restrictions must be observed:
248
Chapter 11
Compiler Invocation and Controls
1.
Do not use the @ and dot operations with variables based on SELECTOR. For
example:
DECLARE SEL SELECTOR;
DECLARE R BASED SEL BYTE;
DECLARE PO POINTER;
PO = @R;
/* invalid under SMALL RAM */
2.
Do not use the built-in function BUILD$PTR (see Chapter 9).
COMPACT
✏
Note
The iRMX Operating System supports only the COMPACT
model.
Form
COMPACT
Default
SMALL
Type
Primary
Modules compiled with the COMPACT control have three sections: code, data, and
stack (see the OBJECT control). When these modules are linked, similar sections
from each module are combined to form three segments: code, data, and stack. The
maximum size of each segment is 4G bytes for the Intel386 and Intel486
microprocessors.
In the default COMPACT case (RAM), the code sections from all modules are allocated
space within the code segment, which is addressed relative to the CS register.
Constants and all data sections are combined in the data segment, which is addressed
relative to the DS register and an identical copy is stored in the ES register. The
stack is addressed relative to SS. None of the segment registers are changed, except
ES, which is used to perform string operations, as well as FS and GS, which are used
to address data exported by another subsystem.
If the ROM control is used, the constants from all the modules are placed with the code
in the code segment. The data segment then contains only the data sections from all
the modules.
Since the code, data, and stack segments are fully defined by the time the program is
loaded, the segment selectors in the CS and SS registers are never changed.
PL/M-386 Programmer's Guide
Chapter 11
249
All six segment registers are initialized by the loader, with ES, FS, and GS initialized
to DS. The DS and ES registers are also saved and reinitialized in each interrupt
procedure prologue and epilogue to enable distinct interrupt environments. The FS
and GS registers are volatile after initialization. References to any location require
only a 32-bit offset against these segment selectors.
Observe the following restrictions when using COMPACT.
1.
When an exported procedure is indirectly activated, a POINTER variable must be
used in the CALL statement. For example:
$COMPACT(SUBSYS HAS MOD1, MOD2, MOD3; EXPORTS PROC)
MOD: DO
DECLARE P POINTER, W WORD;
PROC: PROCEDURE PUBLIC;
.
.
.
END PROC;
P = @PROC; CALL P;
/* POINTER must be used */
W = .PROC; CALL W;
/* not allowed */
END MOD1;
2.
When a procedure that is not exported is indirectly activated, an OFFSET variable
must be used. Note that OFFSET variables do not range over the entire
microprocessor address space, but are restricted to offsets within the current code
segment. For example:
DECLARE P POINTER, O OFFSET;
LPROC: PROCEDURE;
.
.
.
END LPROC;
P = @LPROC; CALL P;
O = .LPROC; CALL O;
250
Chapter 11
/* local */
/* not allowed */
/* OFFSET must be used */
Compiler Invocation and Controls
MEDIUM
Form
MEDIUM
Default
SMALL
Type
Primary
For PL/M-386, the MEDIUM control is provided for PL/M-86 and PL/M-286
compatibility. The MEDIUM control is interpreted exactly like the SMALL control. For
more information, refer to the SMALL control entry in this chapter.
LARGE
Form
LARGE
Default
SMALL
Type
Primary
The LARGE control is provided for PL/M-86 and PL/M-286 compatibility. The
LARGE control is interpreted exactly like the COMPACT control in most cases. For
more information, refer to the COMPACT control entry in this chapter. When the
LARGE control is used in a PL/M-386 subsystem definition, it behaves differently
from the COMPACT control. For more information about subsystems, see Chapter 13.
FLAT
The FLAT control is a member of the group of segmentation controls including
SMALL and COMPACT. Compiling with the FLAT control generates an object module
containing separate code, data, and stack segments, with constants included in the
code segment. The FLAT control overrides the RAM or ROM control. Using the
-CONST IN CODE- or -CONST IN DATA- attribute for extended segmentation
definition does not result in an error when you specify the FLAT control; however,
-CONST IN CODE- is redundant and -CONST IN DATA- is ignored when FLAT is in
effect.
Linking object modules compiled with the FLAT control produces the following
linked segments:
•
A single code segment (CODE32) containing all the code segments of the object
modules
•
A single data segment (DATA32) containing all the data segments of the object
modules
•
A single stack segment (STACK) containing all the stack segments of the object
modules
PL/M-386 Programmer's Guide
Chapter 11
251
Use the BLD386 FLAT control to map the three linked segments together to a single
segment of up to 4 Gigabytes.
Since only one segment exists during run-time, all pointers are short (a 32-bit offset
with no selector).. Also, compiling the following code with the FLAT control does
not result in the semantic error generated when compiling this code with the ROM
control and any other segmentation control:
DECLARE B WORD;
DECLARE A WORD AT (@B) DATA (10);
SUBTITLE
Form
SUBTITLE("subtitle")
Default
No subtitle
Type
General
The subtitle character sequence (truncated on the right to fit, if necessary) is printed
on the subtitle line of each page of listed output. Note that a subtitle specified on the
invocation line must be enclosed in quotation marks.
The maximum length for subtitle is 60 characters, but a narrow pagewidth may
restrict this number.
When a SUBTITLE control appears before the first noncontrol line in the source file,
it causes the specified subtitle to appear on the first page and all subsequent pages
until another SUBTITLE control appears.
A subsequent SUBTITLE control causes a page eject, and the new subtitle appears on
the next page and all subsequent pages until the next SUBTITLE control.
SYMBOLS | NOSYMBOLS
Form
SYMBOLS|NOSYMBOLS
Default
NOSYMBOLS
Type
Primary
The SYMBOLS control specifies that a listing of all identifiers in the PL/M source
program and their attributes is to be produced in the listing file.
The NOSYMBOLS control suppresses such a listing.
Note that the SYMBOLS control cannot override a NOPRINT control.
252
Chapter 11
Compiler Invocation and Controls
TITLE
Form
TITLE("title")
Default
TITLE ("modulename")
Type
Primary
The title character sequence, truncated on the right to fit, if necessary, is placed on
the title line of each page of listing output. Note that the character sequence for a
title must be enclosed in quotation marks when entered on the invocation line.
The maximum length for the title is 60 characters, but a narrow pagewidth may
restrict this number.
TYPE | NOTYPE
Form
TYPE|NOTYPE
Default
TYPE
Type
Primary
The TYPE control specifies that the object module is to contain information on the
variable types output in symbol records. TYPE records provide a mechanism for
promoting type compatibility between subprograms. This information may be used
later for type checking when the program modules are combined, or by a debugger.
The NOTYPE control specifies that such type definitions are not to be placed in the
object module.
WORD32 | WORD16
Form
WORD32|WORD16
Default
WORD32
Type
Primary
The WORD32|WORD16 control determines how the compiler interprets the unsigned
binary number and signed integer scalar types (as well as the built-ins that specify
these data types) in the code being compiled.
When compiling PL/M-286, PL/M-86, or PL/M-80 source code with the PL/M-386
compiler, there are several points to consider before accepting the default (WORD32)
or choosing WORD16. See Chapter 3 for a discussion of these points.
PL/M-386 Programmer's Guide
Chapter 11
253
Table 11-4 lists the data types as interpreted by the compiler under WORD32 and
WORD16. The WORD16 control does not mean creating PL/M-286 code, but rather that
PL/M-386 data types are mapped to the equivalent PL/M-286 data type. It affects
only the data types, it does not affect the operation of PL/M-386 functions.
Table 11-4. WORD32 | WORD16 Data Type Mapping
Unsigned Binary Number
Data Types
WORD32
(default)
WORD16
BYTE
8-bit
8-bit
HWORD
16-bit
8-bit
WORD
32-bit
16-bit
DWORD
64-bit
32-bit
QWORD
64-bit
64-bit
Signed Integer
Data Types
WORD32
WORD16
CHARINT
8-bit
8-bit
SHORTINT
16-bit
8-bit
INTEGER
32-bit
16-bit
LONGINT
32-bit
32-bit
Note that all built-ins that specify data types are different for WORD16. Table 11-5
lists the WORD32|WORD16 mapping for these built-ins. For example, the HWORD
built-in is a 16-bit, unsigned binary number under WORD32, whereas under WORD16,
the 16-bit, unsigned binary type is WORD.
254
Chapter 11
Compiler Invocation and Controls
Table 11-5. WORD32 | WORD16 Built-in Mapping
WORD32
WORD16
(type conversions)
BYTE
HWORD
WORD
DWORD, QWORD
CHARINT
SHORTINT
INTEGER
(type conversions)
BYTE, HWORD
WORD
DWORD
QWORD
SHORTINT, CHARINT
INTEGER
LONGINT
BLOCKINPUT
BLOCKOUTPUT
MOVB
MOVRB
FINDB
FINDRB
INPUT
OUTPUT
SKIPB
SKIPRB
CMPB
SETB
BLOCKINPUT
BLOCKOUTPUT
MOVB, MOVHW
MOVRB, MOVRHW
FINDB, FINDHW
FINDRB, FINDRHW
INPUT, INHWORD
OUTPUT, OUTHWORD
SKIPB, SKIPHW
SKIPRB, SKIPRWH
CMPB, CMPHW
SETB, SETHW
BLOCKINHWORD
BLOCKOUTHWORD
MOVHW
MOVRHW
FINDHW
FINDRHW
INHWORD
OUTHWORD
SKIPHW
SKIPRHW
CMPHW
SETHW
BLOCKINWORD
BLOCKOUTWORD
MOVW
MOVRW
FINDW
FINDRW
INWORD
OUTWORD
SKIPW
SKIPRW
CMPW
SETW
continued
PL/M-386 Programmer's Guide
Chapter 11
255
Table 11-5. WORD32 | WORD16 Built-in Mapping (continued)
WORD32
WORD16
(type conversions)
BLOCKINWORD
BLOCKOUTWORD
MOVW
MOVRW
FINDW
FINDRW
INWORD
OUTWORD
SKIPW
SKIPRW
CMPW
SETW
(type conversions)
BLOCKINDWORD
BLOCKOUTDWORD
MOVD
MOVRD
FINDD
FINDRD
INDWORD
OUTDWORD
SKIPD
SKIPRD
CMPD
SETD
XREF | NOXREF
Form
XREF|NOXREF
Default
NOXREF
Type
Primary
The XREF control specifies that a cross-reference listing of source program identifiers
is to be produced in the listing file.
The NOXREF control suppresses the cross-reference listing.
Note that the XREF control cannot override a NOPRINT control.
256
Chapter 11
Compiler Invocation and Controls
Program Listing
Sample Program Listing
During the compilation process, a listing of the source input is produced. Each page
of the listing carries a numbered page-header that identifies the compiler, prints a
time and date as designated by the host operating system, and optionally gives a title
and a subtitle, and/or a date (see Figure 11-7).
The first part of the listing contains a summary of the compilation, beginning with the
compiler identification and the name of the source module being compiled. The next
line names the file receiving the object code. The next line contains the command
used to invoke the compiler. The listing of the program itself is shown in Figure
11-7.
The listing contains a copy of the source input plus additional information. Two
columns of numbers appear to the left of the source image. The first column
provides a sequential numbering of PL/M statements. (Note that the PL/M-386
compiler treats each new-line character as a line terminator; therefore, blank lines are
counted.) Error messages, if any, refer to these statement numbers. The second
column gives the block nesting depth of the corresponding statement.
Lines included with the INCLUDE control are marked with an equal sign (=) just to
the left of the source image. If the included file contains another INCLUDE control,
lines included by this nested INCLUDE are marked with an =1. For yet another level
of nesting, =2 is used to mark each line, and so forth up to the compiler's limit of
nesting levels (see Appendix B). These markings make it easy to see where included
text begins and ends.
PL/M-386 Programmer's Guide
Chapter 11
257
PL/M-386 COMPILER Stack Module date time PAGE 1
system-id PL/M-386 Vx.y COMPILATION OF MODULE STACK
OBJECT MODULE PLACED IN stack.obj
COMPILER INVOKED BY: plm386 stack.src CODE XREF TITLE("Stack Module")
1
2
3
1
4
1
5
1
6
7
2
2
8
9
2
1
10
2
11
12
2
1
STACK: DO;
/* This module implements a BYTE stack with
push and pop */
DECLARE S(100) BYTE,
/* Stack Storage */
T BYTE PUBLIC INITIAL(-1);
/* Stack Index */
PPUSH: PROCEDURE (B) PUBLIC;
/* Pushes B onto the stack */
DECLARE B BYTE;
S(T:=T+1) = B;
/* Increment T and store B */
END PPUSH;
PPOP: PROCEDURE BYTE PUBLIC;
/* Returns value popped from stack */
RETURN S((T:=T-1)+1);
/* Decrement T, return S(T+1) */
END PPOP;
END STACK;
/* Module ends here */
Figure 11-7. Program Listing
258
Chapter 11
Compiler Invocation and Controls
Should a source line be too long to fit on the page in one line, it is continued on the
following line. Such continuation lines are marked with a hyphen (-) just to the left
of the source image.
The CODE control can be used to obtain the assembly code produced in the translation
of each PL/M statement. Figure 11-8 shows the assembly code listing for the
program given in Figure 11-7. This code listing appears in six columns of
information in a pseudo-assembly language format:
1.
Location counter (hexadecimal notation)
2.
Resultant binary code (hexadecimal notation)
3.
Label field
4.
Opcode mnemonic
5.
Symbolic arguments
6.
Comment field
PL/M-386 Programmer's Guide
Chapter 11
259
PL/M-386 COMPILER Stack Module date time PAGE 2
ASSEMBLY LISTING OF OBJECT CODE
; STATEMENT # 5
PPUSH
00000000
00000001
55
8BEC
PROC
PUSH
MOV
00000003
00000009
0000000B
00000011
00000014
00000017
8A0564000000
FEC0
880564000000
0FB6C0
8A4D08
888800000000
MOV
INC
MOV
MOVZX
MOV
MOV
0000001D
0000001E
5D
C20400
POP
RET
ENDP
PPUSH
NEAR
EBP
EBP,ESP
; STATEMENT # 7
AL,T
AL
T,AL
EAX,AL
CL,[EBP].B
[EAX].S,CL
; STATEMENT # 8
EBP
4H
; STATEMENT # 9
PPOP
00000024
00000025
55
8BEC
PROC
PUSH
MOV
00000027
0000002D
0000002F
00000035
00000037
0000003A
00000040
00000041
8A0564000000
FEC8
880564000000
FEC0
0FB6C0
8A8000000000
5D
C3
MOV
DEC
MOV
INC
MOVZX
MOV
POP
RET
NEAR
EBP
EBP,ESP
; STATEMENT # 10
AL,T
AL
T,AL
AL
EAX,AL
AL,[EAX].S
EBP
; STATEMENT # 11
PPOP
ENDP
; STATEMENT # 12
Figure 11-8. Code Listing (continued)
Not all six of the columns will appear on all lines of the code listing. Compiler
generated labels (e.g., those that mark the beginning and ending of a DO WHILE loop)
are preceded by an AT sign (@). The comments appearing on PUSH and POP
instructions indicate the stack depth associated with the stack instruction.
260
Chapter 11
Compiler Invocation and Controls
Symbol and Cross-reference Listing
Specifying the XREF or SYMBOLS control adds a summary of all identifier usage in
the program listing. Figure 11-9 shows the cross-reference listing of the program
given in Figure 11-7. The addresses in ADDR have four leading zeros.
PL/M-386 COMPILER Stack Module date time PAGE 3
CROSS-REFERENCE LISTING
DEFN ADDR
---- -----5
9
5
3
1
3
0008H
0024H
0000H
0000H
0000H
0064H
SIZE
---1
30
33
100
1
NAME, ATTRIBUTES, AND REFERENCES
-------------------------------B. . .
PPOP .
PPUSH.
S. . .
STACK.
T. . .
.
.
.
.
.
.
..
..
..
..
..
..
.
.
.
.
.
.
BYTE IN PROC(PPUSH) PARAMETER AUTOMATIC
PROCEDURE BYTE PUBLIC STACK=00000004H
PROCEDURE PUBLIC STACK=00000008H
BYTE ARRAY(100)
7* 10
MODULE STACK=00000000H
BYTE PUBLIC INITIAL
7 7* 10 10*
6 7
Figure 11-9. Cross-reference Listing
Depending on whether the SYMBOLS or XREF control was used to request the
identifier usage summary, five or seven types of information are provided in the
symbol or cross-reference listing. They are as follows:
1.
Statement number where the identifier was defined.
2.
Relative address associated with the identifier.
3.
Size of the object identified (in bytes).
4.
The identifier.
5.
Attributes of the identifier (including expansion for LITERALLYs and scoping
information for local variables and parameters). These attributes reflect the
WORD32|WORD16 terminology of the source file.
6.
Statement numbers where the identifier was referenced (XREF control only).
7.
Statement numbers where the identifier was assigned a value (XREF control
only).
PL/M-386 Programmer's Guide
Chapter 11
261
A single identifier can be declared more than once in a source module (i.e., an
identifier defined twice in different blocks). Each such unique object, even though
named by the same identifier, appears as a separate entry in the listing.
The address given for each object is the location of that object relative to the start of
its associated section. The object's attributes determine which section is applicable.
Identifiers in the SYMBOLS or XREF listing are given in alphabetical order with the
following exception: members of structures are listed, in order of declaration,
immediately following the entry for the structure itself. Indentation is used to
differentiate between these entries.
The XREF listing differentiates between items 6 and 7 by adding the asterisk (*)
character to statement numbers where a value is assigned. For example, if statement
17 reads as follows:
I = I + 1;
The list of statement numbers for I would include 17 and 17*, indicating a reference
and an assignment in statement 17.
The AUTOMATIC attribute indicates that the identifier was declared as a parameter or
as a local variable in a REENTRANT procedure and therefore is allocated dynamically
on the stack.
262
Chapter 11
Compiler Invocation and Controls
Compilation Summary
Following the listing (or appearing alone if NOLIST is in effect) is a compilation
summary. Eight pieces of information are provided:
•
Code area size gives the size in bytes of the code section of the output module
(not including constants, if any).
•
Constant area size gives the size in bytes of the constant area. The constant area
will be included with either the code or data section in the output module,
depending on the specified compiler controls.
•
Variable area size gives the size in bytes of the data section of the output module
(not including constants, if any).
•
Maximum stack size gives the size, in bytes, of the stack section allocated for the
output module.
•
Lines read gives the number of source lines processed during compilation.
•
Program warnings give the number of warning messages issued during
compilation.
•
Program errors give the number of error messages issued during compilation.
•
Dictionary summary gives the actual memory and disk space used by the
dictionary during compilation.
Figure 11-10 is an example of the compilation summary.
MODULE INFORMATION:
CODE AREA SIZE
CONSTANT AREA SIZE
VARIABLE AREA SIZE
MAXIMUM STACK SIZE
12 LINES READ
0 PROGRAM WARNINGS
0 PROGRAM ERRORS
=
=
=
=
00000042H
00000000H
00000065H
00000008H
66D
0D
101D
8D
DICTIONARY SUMMARY:
410KB MEMORY AVAILABLE
8KB MEMORY USED
(1%)
0KB DISK SPACE USED
END OF PL/M-386 COMPILATION
Figure 11-10. Compilation Summary
■■■
PL/M-386 Programmer's Guide
Chapter 11
263
264
Chapter 11
Compiler Invocation and Controls
Sample Program
12
Introduction
This chapter discusses a sample program consisting of three modules named FREQ,
OPEN, and PRINT. The purpose of this program is to illustrate the use of the PL/M
language. The program is written in PL/M-386 and compiled with the PL/M-386
compiler.
The program takes an input file, counts the uppercase and lowercase alphabetic
characters, and determines the percentage of use for each character. This is printed
either to the screen or, if one is specified, to an output file. The program's output lists
the number of times each character is used (for uppercase, for lowercase, and for both
uppercase and lowercase), and the percentage of use for each character. The source
program listings are shown in Figures 12-1 through 12-3.
In addition to the main program modules (FREQ, OPEN, and PRINT), this program
also has two include files. The include files, defns.inc and udi.inc (see Figures 12-4
and 12-5), contain definitions that are used in the program modules. The defns.inc
include file consists of global variable definitions. The udi.inc include file consists
of the universal development system interface (UDI) definitions. The UDI
definitions are used for operating system interfaces (e.g., file manipulation). Figure
12-6 is an example of the program output.
The following sections describe the source code in each of the program modules.
The line numbers in the figures are not part of the source code; they have been added
to simplify the discussion of the source code.
FREQ Module
FREQ is the main module. The source code is shown in Figure 12-1. As indicated,
the line numbers in the figure have been added to simplify the discussion of the
source code.
The program lines that begin with a dollar sign ($) are compiler control lines. Lines
that begin with a dollar sign instruct the compiler and are not part of the source
program. In any position other than the first character (or the position specified with
PL/M-386 Programmer's Guide
Chapter 12
265
the LEFTMARGIN control), the dollar sign is an insignificant character and can be
used as a separator to simplify the reading of variable names.
1
2
$DEBUG PW(75)
freq:DO;
3
$INCLUDE (defns.inc)
4
5
6
7
$NOLIST
8
9
open$files:PROCEDURE EXTERNAL;
END open$files;
10
11
12
13
14
15
16
17
18
19
20
21
22
23
/*** LIST of UDI procedures is in OPEN.PLM ***/
$INCLUDE (udi.inc)
$LIST
print$stats:PROCEDURE(arr$ptr, arr$len) EXTERNAL;
DECLARE arr$ptr POINTER;
DECLARE arr$len WORD;
END print$stats;
DECLARE
DECLARE
DECLARE
DECLARE
DECLARE
DECLARE
DECLARE
DECLARE
DECLARE
DECLARE
buf(80) BYTE;
console CONNECTION EXTERNAL;
i BYTE;
infile CONNECTION EXTERNAL;
lfreq(26) Freq_Struc;
num$read BYTE;
outfile CONNECTION EXTERNAL;
quit$time BYTE INITIAL(False);
status WORD;
total WORD PUBLIC INITIAL (0);
Figure 12-1. Source Code for FREQ Module
266
Chapter 12
Sample Program
24
$EJECT
25
main:
26
CALL open$files;
27
CALL init$real$math$unit;
28
29
30
31
32
33
34
35
36
DO i = 0 to LENGTH(lfreq);
lfreq(i).let.low = 0;
lfreq(i).let.up = 0;
lfreq(i).percent = 0.0;
END;
37
38
39
40
41
42
43
44
/*** Now, read the files ***/
read$file:DO WHILE (NOT quit$time);
num$read = dq$read(infile,@buf,LENGTH(buf),@status);
IF num$read <> LENGTH(buf) THEN quit$time = True;
DO i = 0 to num$read;
total = total + 1;
/*** Total keeps track of ALL
characters ***/
/*** Read, not just the letters. ***/
sh_which_letter:IF (buf(i) >= 'A' AND buf(i) <= 'Z') THEN
lfreq(buf(i)-'A').let.up = lfreq(buf(i)-'A').let.up + 1;
ELSE IF (buf(i) >= 'a' AND buf(i) <= 'z') THEN
lfreq(buf(i)-'a').let.low = lfreq(buf(i)-'a').let.low +
1;
END;
/*** Loop i = 0 to num$read ***/
45
read$file:END;
46
47
48
stats:
CALL print$stats(@lfreq,LENGTH(lfreq) );
CALL dq$exit(0);
49
END freq;
Figure 12-1. Source Code for FREQ Module (continued)
PL/M-386 Programmer's Guide
Chapter 12
267
Line 1 specifies the DEBUG control and the pagewidth. The DEBUG control instructs
the compiler to collect debug information such as the statement number and relative
address of each source program module. PW(75) specifies an output page 75
characters wide.
Line 2 names the module and establishes the beginning of the module's DO block. As
stated in Chapter 1, a module must begin with a labeled DO statement and end with an
END statement.
Lines 3 through 6 specify the include files to be used in the program module. Line 4
indicates to the compiler to not list anything until the LIST control is encountered,
which happens at line 7.
Line 5 is a user comment and will not be interpreted by the compiler. User comment
lines begin with a slash/asterisk (/*) combination and end with an asterisk/slash (*/)
combination.
Lines 8 through 23 are the procedure and variable declarations used in the FREQ
module. Note the EXTERNAL declarations in lines 8 through 13. These procedures
are declared EXTERNAL, which means that the procedure is defined in another
module. The calling module must declare the procedure as EXTERNAL. The module
in which these procedures are defined must declare the procedures as PUBLIC.
The variable declarations (see lines 15, 17, and 20) are also EXTERNAL. The same
rules apply for variables as for procedures. The calling module must declare the
variable as EXTERNAL and the defining module must declare the variable as PUBLIC.
If the variable definition is included in the calling module, the definition must be
identical to the definition in the declaring module.
Line 18 declares the lfreq structure, which is declared in the defns.inc file (see
Figure 12-4). Line 21 declares quit$time as a variable (with the INITIAL
attribute) of type BYTE. In an initialization, the initialization attribute must be placed
after the variable attributes. In line 23, total is declared as a variable of type BYTE.
Note also the PUBLIC declaration. This indicates that this variable can be used by
other modules within the program (if it is declared EXTERNAL within the module
which uses it).
Line 24 specifies the beginning of a new page (used when the program listing is
printed).
The program begins at line 25. Line 26 calls the open$files procedure (declared
as EXTERNAL in line 8). This procedure opens the input file, and if one is specified,
the output file. Line 27 calls the compiler built-in procedure,
init$real$math$unit. This call is required to initialize the REAL math facility
for subsequent operations.
Lines 28 through 32 consist of more initializations. These lines set (or reset) the
values of the structure variable used in the module. Freq_struc is an array of
268
Chapter 12
Sample Program
nested structures (see Chapter 4). Freq_struc is a 26 element array (one element
for each letter in the alphabet). Each element of the freq_struc array contains the
let structure, which consists of a letter and a percent. Nested within the let
structure is another structure (low and up). This structure holds the count of
uppercase and lowercase characters. To see how freq_struc is declared, refer to
Figure 12-4.
Lines 34 through 45 show an example of a nested DO block. With PL/M, DO blocks
can be nested up to 18 levels. Line 37 begins a second DO block within the DO block
that begins at line 34. The DO block nested within the first DO block ends at line 44.
The first DO block ends at line 45.
Lines 34 through 36 use the UDI function, dq$read, to read from a file (infile).
A specified number of characters are read from the file into an array. The array is
buf and the number of characters read is LENGTH(buf). The value of buf was set
in line 14. LENGTH is a built-in function (see Chapter 11) that returns the number of
elements in an array. The UDI function, dq$read, returns the number of characters
read (num$read) and an error code (status).
The nested loop (lines 37 through 44) keeps totals for all the characters read, the
uppercase letters read, and the lowercase characters read. This entire loop repeats
until the number of characters read in from the input file is less than 80 (this indicates
that the input file is empty).
Line 47 calls the external procedure print$stats. This procedure is defined in the
PRINT module. Line 48 calls a UDI procedure, dq$exit. Finally, line 49 ends the
FREQ module.
OPEN Module
The OPEN module takes care of the majority of the file-handling procedures for the
program. This module makes extensive use of the UDI procedures provided by the
run-time support library. The source code is shown in Figure 12-2. Note that the line
numbers in the figure are not part of the source code, nor are they the line numbers
that the compiler would assign. The line numbers have been added to simplify the
discussion of the source code.
PL/M-386 Programmer's Guide
Chapter 12
269
1
$DEBUG PW(75)
2
open:DO;
3
4
5
$NOLIST
$INCLUDE (defns.inc)
$LIST
6
7
$EJECT
$INCLUDE(udi.inc)
8
9
10
11
$EJECT
DECLARE console CONNECTION PUBLIC;
DECLARE infile CONNECTION PUBLIC;
DECLARE outfile CONNECTION PUBLIC;
12
13
DECLARE NeedFile(*) BYTE INITIAL('Enter input file name: ');
DECLARE OpenError(*) BYTE INITIAL ('Error opening input
file',CR,LF);
14
15
16
17
18
19
20
21
open$files:PROCEDURE PUBLIC;
DECLARE delim BYTE;
DECLARE console$in CONNECTION;
DECLARE buffer(80) BYTE;
DECLARE status WORD;
DECLARE in$buf(81) BYTE;
DECLARE i BYTE;
DECLARE num$read BYTE;
22
23
console = dq$create(@(4,':CO:'),@status);
CALL dq$open(console,WriteOnly,0,@status);
24
25
26
27
28
29
/*** Process the command line. It consists of three parts,
1) the program name (lf.exe)
2) the input file name, if this is not present then
ask for it
3) the output file name, if this is not present then
the output goes to the console ***/
Figure 12-2. Source Code for OPEN Module
270
Chapter 12
Sample Program
30
31
/*** Read past the program name ***/
delim = dq$get$argument(@buffer,@status);
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
/*** Find out name of the input file ***/
IF delim = CR THEN
DO;
/*** No input file specified, ask for it ***/
CALL dq$write(console,@NeedFile,LENGTH(NeedFile),@status);
console$in = dq$attach(@(4,':CI:'),@status);
CALL dq$open(console$in,ReadOnly,0,@status);
sch001:num$read =
dq$read(console$in,@in$buf,LENGTH(in$buf),@status);
CALL dq$close(console$in,@status);
48
49
50
51
/*** Convert the read in buffer to the infile buffer ***/
sh_infile:buffer(0) = num$read;
DO i = 0 to num$read;
IF (in$buf(i) <> CR) AND (in$buf(i) <> LF)
THEN buffer(i+1) = in$buf(i);
ELSE
buffer(0) = buffer(0) - 1;
/*** Adjust count for
CR/LF ***/
END;
/*** End of DO loop to Convert buffer ***/
END;
ELSE
delim = dq$get$argument(@buffer,@status);
52
/*** END; get file name to process ***/
53
54
55
56
57
/*** Open input file ***/
infile = dq$attach(@buffer,@status);
CALL dq$open(infile,ReadOnly,2,@status);
IF status <> E$OK THEN DO;
CALL dq$write(console,@OpenError,LENGTH(OpenError),
@status);
CALL dq$exit(1);
END;
/** Status is not ok **/
58
59
Figure 12-2. Source Code for OPEN Module (continued)
PL/M-386 Programmer's Guide
Chapter 12
271
60
61
62
63
64
65
66
67
68
/*** Find out if an output file was specified. If so, ***/
/*** open it, if not use the console output ***/
IF delim = CR THEN
outfile = console;
ELSE DO;
delim = dq$get$argument(@buffer,@status);
outfile = dq$create(@buffer,@status);
CALL dq$open(outfile,WriteOnly,2,@status);
END;
69
END open$files;
70
END open;
Figure 12-2. Source Code for OPEN Module (continued)
272
Chapter 12
Sample Program
Line 1 instructs the compiler to collect debug information and sets the page width for
printed output. Line 2 names the module and establishes the beginning of the
module's DO block. Lines 3 through 8 specify the inclusion of the program's include
files, turn the listing function on and off, and specify a few new pages for printed
output ($EJECT).
Lines 9 through 11 define and declare some PUBLIC variables. Because these
variables are declared PUBLIC, they can be used in another module. The calling
module must declare the variable as EXTERNAL. The variable definition is included
in the calling module, and it is the same as the definition in the defining module.
Lines 12 and 13 are error messages to be used by the OPEN module if the necessary
information is not included in the invocation line (which causes an error). Note the
use of the asterisk in each of these lines. The asterisk is used as an implicit
dimension specifier. The implicit dimension specifier can be used when the size of
the array is either unknown or insignificant. In this instance, the size of the array is
unknown. The implicit dimension specifier in lines 12 and 13 specifies that the
NeedFile array and the OpenError array will have the same number of elements as
the value list (the number of characters in the message).
Line 14 begins the open$files procedure. This procedure is declared as PUBLIC
(it is called by the FREQ module) and continues until the end of the module (line 69).
Lines 22 and 23 get and open a connection with the console using predefined UDI
procedures. Note the use of the @ operator in these two lines. The first @ operator in
line 22 allocates storage for the constants 4 and :CO:. The other @ operators are for
location references. This means that the value of the reference (e.g., the value of
@status) is the actual run-time location of the variable.
Lines 31 through 51 use the UDI procedure, dq$get$argument, to parse the input
line. Line 31 gets the first part of the command line, as well as the delimiter used to
separate this part of the command line from the next part (if there is any). Line 33
tests the delimiter. If the delimiter is a carriage return then lines 34 through 49 are
processed. Lines 34 through 49 request a file name. If the delimiter is not a carriage
return then dq$get$argument is called again. This routine is also called by lines 62
through 68 to determine whether the program output should go to a file or to the
console.
Line 31 passes the invocation line to the following IF/THEN/ELSE construct (lines
33 through 51). The IF/THEN/ELSE construct checks for an input file name. If no
input file is specified, line 36 uses the NeedFile string declared in line 12. This
prompts the user to enter an input file name. If no input file name is specified in
response to the prompt, the program aborts. Otherwise, the string is converted as
discussed in the preceding paragraph.
PL/M-386 Programmer's Guide
Chapter 12
273
Lines 43 through 48 convert the file name to a UDI call.
Lines 50 and 51 are the ELSE clause of IF delim = CR.
Lines 53 through 59 open the input file. Lines 62 through 68 open an output file, if
one is specified. Otherwise, the program data is sent to the console.
Line 69 is the END statement for the open$files procedure and line 70 is the END
statement for the OPEN module.
PRINT Module
The PRINT module performs the program calculations and prints the information
(either to the console or to the specified output file). The source code is shown in
Figure 12-3.
274
Chapter 12
Sample Program
1
2
$DEBUG PW(75)
print:DO;
3
4
5
6
$NOLIST
$INCLUDE (defns.inc)
$INCLUDE (udi.inc)
$LIST
7
8
9
10
11
12
13
DECLARE BLANK$OUT$LINE LITERALLY
'DO j = 0 TO LENGTH(line);line(j) = SPACE;END';
DECLARE LETTER LITERALLY '3';
DECLARE LOWER LITERALLY '24';
DECLARE PCT LITERALLY '33';
DECLARE SUM
LITERALLY '8';
DECLARE UPPER LITERALLY '16';
14
15
16
17
18
19
20
DECLARE outfile CONNECTION EXTERNAL;
DECLARE topline(*) BYTE INITIAL
('LETTER TOTAL UPPER LOWER % ',CR,LF);
/** (
A
00000
00000
00000
000.0
***/
/** ( 123456789 123456789 123456789 123456789 123456789 ***/
DECLARE total WORD EXTERNAL;
DECLARE total$str (5) BYTE INITIAL ('TOTAL');
21
22
23
24
int2asc:PROCEDURE(number,stg$ptr,count) BYTE;
DECLARE number WORD;
DECLARE stg$ptr POINTER;
DECLARE count BYTE;
25
26
27
28
DECLARE
DECLARE
DECLARE
DECLARE
i BYTE, j BYTE;
max DWORD;
string BASED stg$ptr(1) BYTE;
tmpstg(10) BYTE;
29
30
31
32
33
34
35
36
max = 1;
DO i = 1 TO count;
max = 10 * max;
END;
max = max - 1;
DO i = 0 TO LAST(tmpstg);
tmpstg(i) = SPACE;
END;
Figure 12-3. Source Code for PRINT Module
PL/M-386 Programmer's Guide
Chapter 12
275
37
38
39
40
41
42
43
IF number <= max THEN DO;
i = 0;
loop:
tmpstg(i) = (number MOD 10) " '0';
i = i " 1;
number = number/10;
IF number 0 THEN GOTO loop;
44
45
46
47
48
49
50
51
52
DO j = 0 TO count;
string(count-j) = tmpstg(j);
END;
END;
ELSE DO;
DO i = 0 to count;
string(i) = '*';
END;
END;
53
54
RETURN(i);
END int2asc;
55
56
57
58
59
60
61
62
63
real2asc:PROCEDURE(number,stg$ptr,count);
DECLARE number REAL;
DECLARE stg$ptr POINTER;
DECLARE count WORD;
DECLARE
DECLARE
DECLARE
DECLARE
DECLARE
i BYTE, j BYTE;
int$len BYTE;
string BASED stg$ptr(1) BYTE;
tmpnum DWORD;
tmpstg(10) BYTE;
64
/*** Convert the number to an INTEGER to convert
it, assume one ***/
decimal place ***/
65
66
tmpnum = DWORD(number*10.0);
Figure 12-3. Source Code for PRINT Module (continued)
276
Chapter 12
Sample Program
67
68
69
70
71
72
73
74
75
int$len = int2asc(tmpnum,@tmpstg,LAST(tmpstg) );
IF int$len = 1 THEN DO;
/*** Handle the case where
the number ***/
/*** is less than 1.0 ***/
int$len = 2;
tmpstg(LAST(tmpstg)-1) = '0';
END;
DO i = 0 TO int$len-2;
string(count-i) = tmpstg(LAST(tmpstg)-i);
END;
76
77
string(count-int$len) = '.';
string(count-int$len-1) = tmpstg(LAST(tmpstg)-int$len+1);
78
79
80
81
82
83
84
85
86
87
88
END real2asc;
$EJECT
print$stats:PROCEDURE (arr$ptr, arr$len) PUBLIC;
DECLARE arr$ptr POINTER;
DECLARE arr$len WORD;
DECLARE array BASED arr$ptr(1) Freq_Struc;
DECLARE i BYTE, j BYTE;
DECLARE line(50) BYTE;
DECLARE status WORD;
DECLARE tmp BYTE;
DECLARE ii BYTE;
89
call dq$write(outfile,@topline,LENGTH(topline),@status);
90
printlines:DO ii = 0 TO arr$len-1;
91
92
93
94
95
96
97
98
99
BLANK$OUT$LINE;
line(LETTER) = ii + 'A';
/*** Get the total and convert number to ascii ***/
tmp = int2asc (array(ii).let.low + array(ii).let.up),
@line(SUM),5);
tmp = int2asc (array(ii).let.low, @line(LOWER),5);
tmp = int2asc (array(ii).let.up, @line(UPPER),5);
array(ii).percent = REAL((array(ii).let.low) +
(array(ii).let.up)) /
REAL(total) * 100.0;
CALL real2asc (array(ii).percent, @line(PCT),5);
Figure 12-3. Source Code for PRINT Module (continued)
PL/M-386 Programmer's Guide
Chapter 12
277
100
101
line(LAST(line)-1) = CR;
line(LAST(line)) = LF;
102
CALL dq$write(outfile,@line,LENGTH(line),@status);
103
END printlines;
104
105
106
107
BLANK$OUT$LINE;
DO i = 0 TO LAST(total$str);
line(LETTER-2"i) = total$str(i);
END;
108
tmp = int2asc( total, @line(SUM),5);
109
call dq$write(outfile,@line,LENGTH(line),@status);
110
END print$stats;
111
END print;
/*** print loop ***/
Figure 12-3. Source Code for PRINT Module (continued)
Line 1 instructs the compiler to collect debug information and sets the page width for
printed output. Line 2 names the module and establishes the beginning of the
module's DO block. Lines 3 through 6 specify the inclusion of the program's include
files and turn the listing function on and off.
Lines 7 through 13 are a group of literally definitions; each one creates an
alternate name for a sequence of characters. Lines 7 and 8 declare
BLANK$OUT$LINE as the alternate name for the DO loop used to blank out the output
line buffer. Additionally, after line 13, the number 16 will reference UPPER (for
uppercase character). This is a useful function to eliminate keystrokes, to make the
program more readable, and to declare quantities that may be fixed in one module,
but subject to change in another module.
Lines 14 through 20 contain more declarations, as well as the header string for the
output (line 16).
Lines 21 through 54 perform an integer-to-ASCII translation. Lines 55 through 78
convert real numbers to ASCII characters.
278
Chapter 12
Sample Program
Line 80 is the beginning of the print$stats procedure. The print$stats
procedure is called by the FREQ module, therefore it is declared PUBLIC in this
module. Note the based variable in line 83. In this instance, the location of array is
based on the address of arr$ptr, which is passed into the print$stats procedure.
The size of the array is unknown (except through the parameter). The 1 enclosed in
parentheses enables the use of arr$ptr as an array (any number can be used).
Line 89 calls a UDI procedure that writes to an external connection declared in the
OPEN module. Note the use of BLANK$OUT$LINE in line 91.
Lines 90 through 103 are a DO loop that is repeated for each letter in the alphabet.
For each character, the line(LETTER) array is filled with the letter, the total, the
total uppercase, the total lowercase, and the percent. This information is then sent to
the specified output device (the console or a file).
Lines 93 through 96 call the procedure to convert the total into ASCII characters.
Lines 97 and 98 figure the percentage of use for each character. Line 99 calls the
procedure to convert the percentage to ASCII characters. Lines 100 and 101 insert a
carriage return and a line feed in the console display or in the output file.
Line 110 ends the print$stats procedure and line 111 ends the PRINT module.
PL/M-386 Programmer's Guide
Chapter 12
279
Include Files
As stated earlier, there are two include files with this program (see Figures 12-4 and
12-5).
DECLARE DCL LITERALLY 'DECLARE';
DCL LIT
LITERALLY 'LITERALLY';
DCL CR
DCL LF
LITERALLY '0DH';
LITERALLY '0AH';
DCL True
DCL False
LITERALLY '0FFH';
LITERALLY '000H';
DCL Freq_Struc LITERALLY 'STRUCTURE (let STRUCTURE
(low WORD, up WORD),
percent REAL)';
DCL SPACE
LITERALLY '020H';
Figure 12-4. Include File -- defns.inc
Figure 12-4 is the defns.inc file. It contains definitions for terms used in common by
all of the modules in the program (excluding the UDI definitions). Note the
declaration of a structure in this include file (Freq_Struc). This structure is used in
the PRINT module and the FREQ module. This structure declaration illustrates
several levels of nesting. Structures can be nested up to 32 levels.
Figure 12-5 is the udi.inc file. It contains UDI definitions that are used throughout
the modules. The UDI is a predefined set of procedure calls that enables use of
operating system functions.
280
Chapter 12
Sample Program
DECLARE
DECLARE
DECLARE
DECLARE
CONNECTION literally 'WORD';
ReadOnly LITERALLY '1';
WriteOnly LITERALLY '2';
E$OK LITERALLY '0H';
dq$attach:procedure (path$p,except$p) CONNECTION external;
declare path$p pointer; declare except$p pointer;
end dq$attach;
dq$close:procedure (aftn,exception$ptr) external;
declare aftn CONNECTION, exception$ptr pointer;
end dq$close;
dq$create:procedure (path$p,exception$ptr) CONNECTION external;
declare (path$p,exception$ptr) pointer;
end dq$create;
dq$exit:procedure (completion$code) external;
declare completion$code word;
end dq$exit;
dq$get$argument:PROCEDURE (arg$ptr, ex$ptr) BYTE EXTERNAL;
declare arg$ptr POINTER, ex$ptr POINTER;
END dq$get$argument;
dq$open:procedure (aftn,mode,num$buf,exception$ptr) external;
declare aftn CONNECTION, exception$ptr pointer;
declare (mode,num$buf) byte;
end dq$open;
dq$read:PROCEDURE(aftn,buf$ptr,count,ex$ptr) WORD EXTERNAL;
declare aftn CONNECTION;
declare buf$ptr POINTER;
declare count WORD;
declare ex$ptr POINTER;
END dq$read;
dq$write:procedure (aftn,buffer,count,exception$ptr) external;
declare aftn CONNECTION;
declare count word;
declare (buffer,exception$ptr) pointer;
end dq$write;
Figure 12-5. Include File -- udi.inc
■■■
PL/M-386 Programmer's Guide
Chapter 12
281
282
Chapter 12
Sample Program
Extended Segmentation Models
13
Overview
Program segmentation is the division of a program into memory segments. It is a
technique used to optimize the code produced by the compiler. The segmentation
controls (COMPACT, LARGE, MEDIUM, SMALL, and FLAT) manage program
segmentation by defining the physical relationship in memory of a program's code,
data, constants, and stack. They determine which (if any) segments get combined.
For example, specifying the SMALL segmentation control for a program module
locates all of the module's code, data, constants, and stack in two segments, CODE and
DATA. When the program's modules are combined, sections from the separately
compiled modules are combined into segments according to the specified
segmentation controls. This optimizes code because references to locations in the
same memory segment are more efficient.
Extended segmentation models are a super-set of the segmentation controls. The
extended segmentation models (which consist of the SMALL, COMPACT, and LARGE
subsystems) provide enhanced program speed and aid in the construction of large
programs. An extended segmentation model consists of a number of subsystems. A
subsystem is a collection of program modules that use the same segmentation
controls. A program is made up of one or more subsystems. With subsystems,
program modules that are compiled with different segmentation controls can be
combined.
This chapter defines the use of extended segmentation models, and contains the
following sections:
•
Introduction
•
Segmentation controls architecture overview
•
Using subsystems
•
Syntax
•
Exporting procedures
•
Large matrix example
PL/M-386 Programmer's Guide
Chapter 13
283
Introduction
Extended segmentation models provide the following programming advantages:
•
Efficient use of memory.
•
Access to the microprocessor's segmented architecture.
•
Storage reduction for external references to pointers and code.
•
Increased program execution speed for intersegment calls and data access.
Additionally, to simplify the development of large programs, the segmentation
controls can be used to partition the program into a collection of related subsystems.
Partitioning a large program into a series of subsystems isolates code references
within the same segment. The compiler processes each program module
individually, assigning code, data and stack segments for each module (according to
the specified segmentation control). As a source file is translated, the compiler
generates a STACK segment for the program stack, as well as a DATA segment for the
program data and a CODE segment for the program's executable code. When the
program modules are combined, the CODE, DATA and STACK segments from all of the
individual program modules are combined. Use of the segmentation controls ensures
that the segment names generated by the compiler are combined according to the
overall structure of the program.
A subsystem is either open or closed. An extended segmentation model can have
only one open subsystem, but any number of closed subsystems.
An open subsystem does not have a name and claims the program modules that are
not claimed by another subsystem. Effectively, a program that uses only the
segmentation controls is an open subsystem. Modules can be added to the open
system without having to change the subsystem definition.
A closed subsystem has a name and, optionally, a list of program modules used in the
subsystem. To add a module to a closed subsystem, the subsystem definition must be
changed.
284
Chapter 13
Extended Segmentation Models
Segmentation Controls Architecture Overview
The segmentation controls described in Chapter 11 define the physical relationship in
memory of program code, data, constants, and stack during program execution.
When a PL/M source file is compiled, an object module conforms to a particular
extended segmentation model.
There are three extended segmentation models: SMALL, COMPACT, and LARGE. For
Intel386 and Intel486 microprocessors, each segment can be as large as 4G bytes.
There are two submodels within each model: RAM and ROM. Specifying RAM places
the program constants in the DATA segment. Specifying ROM places the program
constants in the CODE segment.
Tables 13-1 and 13-2 define the memory partitions and the placement of pointers in
the various architectural models available with the segmentation controls. Table 13-1
shows how memory is partitioned. Table 13-2 defines the register addresses and the
pointer values. Table 13-3 defines the register addresses and the pointer values for
the Intel386 and Intel486 microprocessor-specific ES register. Note that the
POINTER variable value for these microprocessors, when using the SMALL ROM
extended segmentation controls, is 6 bytes.
PL/M-386 Programmer's Guide
Chapter 13
285
Table 13-1. Segmentation Controls and Memory Partitions
CODE
SMALL RAM
code
data
constants
stack
SMALL ROM
constants
data
code
stack
COMPACT RAM
code
constants
data
stack
COMPACT ROM
constants
code
data
stack
MEDIUM RAM*
separate CODE
segment for each
module's code
data
constants
stack
MEDIUM ROM*
separate CODE
segment for each
module's code
and constants
data
stack
LARGE RAM*
separate CODE
segment for each
module's code
separate DATA
segment for
each module's
data and
constants
stack
LARGE ROM*
separate CODE
segment for each
module's code
and constants
separate DATA
segment for each
module's data
stack
*
286
Segment Name
DATA
Control
STACK
The Intel386 and Intel486 microprocessors use only the SMALL and COMPACT
segmentation controls. For the segmentation controls (not subsystems), MEDIUM is
equivalent to SMALL and LARGE is equivalent to COMPACT.
Chapter 13
Extended Segmentation Models
Table 13-2. Segmentation Controls, Register Addresses and Pointer Values
Register Address
Pointer
Variable Value
Control
CS
DS
SS
SMALL
RAM
CODE seg.
Offset-reference
relative to DS
DATA seg.
Offset-reference
relative to DS
DATA seg.
Has same value
as DS
Offset-reference
4-byte offset
only
SMALL
ROM
CODE seg.
Constant reference
requires selectoroffset containing CS
value and offset
within CODE
segment Code
reference requires
offset-reference
relative to DS
DATA seg.
Offset reference
DATA seg.
Has same value
as DS
Offset-reference
6-byte selectoroffset
COMPACT
RAM
CODE seg.
Selector-offset
reference
DATA seg.
Selector-offset
reference
STACK seg.
Selector-offset
reference
6-byte
selector-offset
COMPACT
ROM
CODE seg.
Selector-offset
reference
CODE seg.
Selector-offset
reference
STACK seg.
Selector-offset
reference
6-byte
Selector-offset
MEDIUM
RAM
Current CODE seg.
Selector-offset
reference
Updated when
PUBLIC or
EXTERNAL
procedure is
activated
DATA seg.
Selector-offset
reference
DATA seg.
Selector-offset
reference
6-byte
Selector-offset
continued
PL/M-386 Programmer's Guide
Chapter 13
287
Table 13-2. Segmentation Controls, Register Addresses and Pointer Values
(continued)
Register Address
Pointer
Variable Value
Control
CS
DS
SS
MEDIUM
ROM
Current CODE seg.
Selector-offset
reference Updated
when PUBLIC or
EXTERNAL
procedure is
activated
DATA seg.
Selector-offset
reference
DATA seg.
Selector-offset
reference
6-byte
Selector-offset
LARGE
RAM
Current CODE seg.
Selector-offset
reference Updated
when PUBLIC or
EXTERNAL
procedure is
activated
Current DATA
seg. Selectoroffset reference
Updated when
PUBLIC or
EXTERNAL
procedure is
activated
STACK seg.
Selector-offset
reference
6-byte
Selector-offset
LARGE
ROM
Current CODE seg.
Selector-offset
reference Updated
when PUBLIC or
EXTERNAL
procedure is
activated
Current CODE
seg. Selectoroffset reference
Updated when
PUBLIC or
EXTERNAL
procedure is
activated
STACK seg.
Selector-offset
reference
6-byte
Selector-offset
The values given in Tables 13-1 and 13-2 are identical for Intel386 and Intel486
microprocessors. Additionally, these microprocessors have the ES register address.
Table 13-3 states the values for the ES register.
288
Chapter 13
Extended Segmentation Models
Table 13-3. Intel386 and Intel486 Microprocessor-specific ES Register Segmentation
Controls, Register Addresses and Pointer Values
Control
ES Register Address
POINTER Variable Value
SMALL RAM
DATA seg.
Offset reference
4-byte offset only
SMALL ROM
DATA seg.
Offset reference
6-byte selector offset
COMPACT RAM
DATA seg.
Selector-offset reference
6-byte selector-offset
COMPACT ROM
DATA seg.
Selector-offset reference
6-byte selector-offset
The SMALL RAM segmentation control is the most efficient. Because all of the code
resides in one segment, jumps and calls are always within the same segment
(intrasegment). However, the SMALL RAM segmentation control provides less
protection and cannot be used to pass pointers to library procedures unless the library
procedure is also a SMALL RAM model.
Use the COMPACT segmentation controls (COMPACT RAM and COMPACT ROM) for
separate management of the code, data, and stack, or to improve segment-limit
protection. To reference stack-based variables, the COMPACT segmentation controls
use selector-offset references. This is less efficient than using offset-only references.
However, data and constant references within a COMPACT segmentation control
module are within the same segment (intrasegment). Note that if a COMPACT
program must pass a data address to a procedure in a different subsystem, it must use
a selector-offset reference.
Using Subsystems
Subsystems offer an efficient way to manage programs with large amounts of data, to
share data between program modules, and to communicate with other programs.
For example, subsystems are useful when several programmers are each writing a
separate module for a highly structured program in which sharing data between
modules is accomplished with parameter passing, by value only. To maintain the
integrity of each section's data requires that each section have its own DATA segment.
In this way, code in one module of the program cannot mistakenly destroy data
belonging to another section of the program. In this instance, each module could be a
COMPACT subsystem, with its own CODE and DATA segments.
PL/M-386 Programmer's Guide
Chapter 13
289
As another example, a program performing I/O usually requires operating system
support routines. In many cases, the operating system will operate at a higher
protection level than the application program. Thus, operating system procedure calls
are intersegment calls. The application program views the operating system as a
separate subsystem. Usually, operating system interface libraries are supplied to
application programmers; these libraries perform the inter-subsystem communication
details. If a program needs to make a direct operating system call without using a
presupplied library, the program itself must define the necessary subsystem
environments at compile time.
It is usually more efficient to structure a large program with subsystems. With
subsystems the code and data can be partitioned into manageable pieces bigger than
one module. Within each subsystem, calls and jumps are near (4 byte offset),
references can be offset only, and the data of each subsystem is protected from being
overwritten by other subsystems. Calls and jumps between subsystems are still far,
and references between subsystems need to be selector-offset. In general, a
program's structure is such that it is possible to break the program into pieces with a
minimum number of intersegment calls, jumps, and references.
For example, consider a program consisting of 10 modules, mod_1 through mod_10.
Modules 1 through 3 deal with input and initial processing. Modules 4 through 8 do
the main data processing. Modules 9 and 10 output the data. The following figure
illustrates the structure of the program:
input «»
INPUT
(mod_1
mod_2
mod_3)
290
Chapter 13
data
flow
«»
PROCESS
data
flow
«»
OUTPUT
(mod_4
mod_5
mod_6
mod_7
mod_8)
(mod_9
mod_10)
Extended Segmentation Models
The total code space required by this program exceeds 64K bytes, and the total data
space also exceeds 64K bytes. The LARGE segmentation control can be used. This
control provides each module with its own CODE and DATA segment. For this
example, this results in a total of 21 segments (10 CODE, 10 DATA, and 1 STACK). For
the LARGE segmentation control, all calls and jumps are far, and all intermodule
references must be through selector-offset POINTERs.
If, for example, COMPACT subsystems are used instead of the LARGE segmentation
control, modules 1 through 3 can form one subsystem, which you could call
SUB_INPUT. Modules 4 through 8 can form subsystem SUB_PROCESS. Finally,
modules 9 and 10 can form subsystem SUB_OUTPUT. The number of segments has
been reduced to seven: 3 CODE, 3 DATA, and 1 STACK. Since most of the calls,
jumps, and references now take place within only one of the subsystems, the program
is much more efficient. The only far calls and jumps, and the only selector-offset
references needed are those in the interfaces between the subsystems.
A typical program does not require subsystems. The code space of 4 Gigabytes and
the data space of 4 Gigabytes is quite sufficient for most programs. However,
consider a program that processes a large amount of data such as a 10x1,000,000,000
REAL matrix. A REAL scalar consists of 4 bytes, so the total memory needed is 40
billion bytes. Rows could be used to partition the matrix. Each row would be 4
billion bytes, which would fit into a single DATA segment.
Ten COMPACT subsystems (named ROW1, ROW2, etc.) could be created, each
containing a 1-billion element REAL array. Procedures to store and retrieve particular
matrix elements can be written and called from the normal matrix processing code.
An example of such a program is shown later in this chapter.
It is not just dividing a program into subsystems that increases its efficiency. If all
the even numbered modules had been placed in one subsystem, for instance, and all
the odd numbered ones into another, the efficiency of the program would not have
improved as it did when the modules were grouped into subsystems according to the
logical structure of the program.
PL/M-386 Programmer's Guide
Chapter 13
291
Note also the following points:
1.
Not all subsystems must use the same segmentation control. For instance, if
SUB_PROCESS in the preceding example is small enough, it could be a SMALL
subsystem.
2.
If a SMALL subsystem is mixed with subsystems using other segmentation
controls, the main program must be in SMALL. This is because anything
compiled in SMALL assumes that DS and SS are identical. This will be so only if
the main program is SMALL. Notice that in this case, the STACK segment
resulting from the COMPACT and LARGE subsystems will not be used, since the
stack of the main program is in the combined DATA-STACK segment of the
SMALL model.
3.
SMALL RAM subsystems have the limitation that the SMALL segmentation control
uses short (offset only) pointers. A SMALL RAM subsystem cannot receive a
pointer from another subsystem, because it cannot save the selector portion. A
SMALL RAM subsystem can, however, pass a pointer to a subsystem that is not
SMALL RAM, because its own DS is known to it. However, a SMALL RAM
subsystem cannot pass a pointer (which points to a procedure), since DS is
assumed as the selector to all pointers.
4.
MEDIUM is a segmentation control only, not an extended segmentation model.
Later sections describe the memory layouts of programs using the standard
segmentation controls: FLAT|COMPACT|LARGE|MEDIUM|SMALL. To understand the
memory layouts of programs structured with subsystems, it is necessary to make the
distinction between compiling modules and combining modules into a program.
The compiler compiles only one module at a time. When modules are combined into
a program, many CODE, DATA and STACK segments, which were generated during
separate compilations, are combined. When combining program modules, all
segments with the same name are combined. The segmentation controls work by
controlling the names of the segments generated by the compiler. This ensures that
the segment names will be combined as desired when the modules are combined into
a program.
The standard SMALL segmentation control causes the compiler to name the CODE
segment CODE, and the DATA-STACK segment DATA. Since under the standard
SMALL model all CODE segments have the same name, and all DATA-STACK segments
have the same name, they are combined when the modules are combined.
292
Chapter 13
Extended Segmentation Models
A module belonging to a SMALL subsystem, on the other hand, takes the name of its
CODE segment from the name of the subsystem. The name of its DATA-STACK
segment is still DATA. Thus, a SMALL subsystem named SUB1 contains one CODE
segment named SUB1_CODE, and one DATA-STACK segment named DATA. A SMALL
subsystem named SUB2 contains one CODE segment named SUB2_CODE, and one
DATA-STACK segment named DATA. When the program modules are combined, all
segments with the same name are combined.
The memory layout of the loaded program containing the two subsystems SUB1 and
SUB2 is as follows (it is assumed that both subsystems are SMALL RAM):
HIGH
CODE
SUB1_CODE
CODE
SUB2_CODE
SUB1 & SUB2
PROGRAM
STACK
PROGRAM
CODE
OF
SUBSYSTEM
SUB1
PROGRAM
CODE
OF
SUBSYSTEM
SUB2
DATA
SEGMENT
SUB1 & SUB2
PROGRAM
DATA &
CONSTANTS
… CS
SS
… DS
LOW
Note that a program using the MEDIUM segmentation control is equivalent to a
program in which each module is declared to be in a unique SMALL subsystem.
A module belonging to a COMPACT subsystem takes the name of its CODE segment
and the name of its DATA segment from the subsystem name. So a COMPACT
subsystem named SUB1 contains one CODE segment named SUB1_CODE, one DATA
segment named SUB1_DATA, and one STACK segment named STACK. A COMPACT
subsystem named SUB2 contains one CODE segment named SUB2_CODE, one DATA
segment named SUB2_DATA, and one STACK segment named STACK. The loaded
program will contain five segments, two CODE segments, two DATA segments, and
one STACK segment. Note that a program using the LARGE segmentation control is
equivalent to a program in which each module is declared to be in a unique COMPACT
subsystem.
PL/M-386 Programmer's Guide
Chapter 13
293
A LARGE subsystem can be simulated by a COMPACT subsystem containing only one
module. However, LARGE subsystems are useful for the following reason. A LARGE
subsystem named SUB1, which contains the modules MOD1, MOD2, and MOD3, has
three CODE segments named MOD1_CODE, MOD2_CODE, and MOD3_CODE, and three
DATA segments named MOD1_DATA, MOD2_DATA, and MOD3_DATA. As usual, it
contains one STACK segment named STACK. It is possible to use a LARGE subsystem
instead of inventing names for many COMPACT subsystems, each containing only one
module. Note that the segment name in the LARGE subsystem is derived from the
module names and not from the subsystem name.
The LARGE segmentation control is identical to the COMPACT segmentation control.
However, there is a difference between LARGE and COMPACT subsystems. In a
LARGE subsystem, the external definition of all symbols in the EXPORTS list have
their segment field set to an unknown value. This enables the creation of external far
objects with public locations that are unknown at compile time. In all other respects,
a LARGE subsystem is identical to a COMPACT subsystem.
Open Subsystems
Compiling files using only the segmentation controls and using no other subsystem
controls produces open subsystems. When object modules are combined, all modules
created from compilations specifying a particular segmentation control are
automatically combined. Segments are created according to the rules for the
segmentation control. A list of modules belonging to an open subsystem is therefore
not needed at compile time. Modules can be freely added to or deleted from an open
subsystem at any time during program development.
Note that both RAM and ROM modules are combined into the single open subsystem.
For a SMALL subsystem, be careful when combining RAM and ROM modules,
particularly concerning the passing of pointer parameters and the accessing of
constants not in the current module.
It is not possible to pass pointer parameters between SMALL RAM and SMALL ROM
modules, because pointers are defined differently in each submodel. Also, it is not
possible to directly reference constants defined in a ROM module from a RAM module,
and vice versa, because RAM modules define constants to be in the data segment, and
ROM modules define constants to be in the code segment.
In the COMPACT model, passing pointer parameters between RAM and ROM modules is
not a problem, because pointers are always long. As in SMALL, the restriction on
direct reference to constants applies.
294
Chapter 13
Extended Segmentation Models
The names of the segments in both SMALL and COMPACT models are identical:
CODE32 for the code segment, DATA for the data segment. This means that if SMALL
and COMPACT modules are combined, they will also be combined to form a single
open subsystem consisting of the CODE32, DATA, and STACK segments. Care must be
taken regarding stack references, because COMPACT defines a separate stack segment
and SMALL does not. For more information on Intel386 microprocessor segment
combining, see the binder chapter in the Intel386 Family Utilities User's Guide.
Closed Subsystems
A closed subsystem differs from an open subsystem in two ways: it has a name and it
consists of a specific list of modules. The compiler must know the name of the
subsystem and the modules belonging to the subsystem in order to create a closed
subsystem.
The need for a closed subsystem name is simply to differentiate a particular closed
subsystem from another closed subsystem or from the open subsystem. This is done
as follows: the name of the subsystem is added to the beginning of the segment
names to create unique code and data segments.
For example, if a subsystem is named PHASE1, then the code sections from all
modules belonging to the PHASE1 subsystem are combined into a single code
segment called PHASE1_CODE32; similarly for COMPACT subsystems the data
sections are combined into a single data segment called PHASE1_DATA. When using
COMPACT, however, the stack sections are still combined into a segment called STACK
because only one execution-time stack is usually necessary. Using SMALL all data
and stack segments are combined in one segment called DATA, as usual.
A closed subsystem module list is needed for differentiation. For instance, if the
compiler is not informed that module SCANNER belongs to subsystem PHASE1, then
the compiler has no choice but to assume that module SCANNER belongs to the open
subsystem.
Thus, every module in a program either is specified as part of a closed subsystem or,
by default, becomes part of the open subsystem. A program can consist of only
closed subsystems, or of both closed subsystems and the open subsystem, or of only
the open subsystem (by default). There is only one open subsystem per program; all
open subsystems are treated as one subsystem by the utility used to combine the
program modules.
Communication Between Subsystems
Within a subsystem there can be code and/or data items (procedures and variables)
that must be known by other subsystems; that is, they are meant to be referenced
from other subsystems. Such items are said to be exported. The export of a symbol
PL/M-386 Programmer's Guide
Chapter 13
295
is not directed at any one particular subsystem; it is directed at all subsystems in the
program, including its own subsystem.
It is important to realize that the subsystem definitions are additions to normal
intermodule PUBLIC/EXTERNAL definitions, not replacements.
For instance, module MOD1 belongs to subsystem SUB1 and makes a reference to
symbol SYM2; SYM2 belongs to subsystem SUB2. SYM2 must be declared as
EXTERNAL in MOD1, as usual, and must also be declared as PUBLIC and exported
from SUB2. Using this information, the compiler generates an intersegment reference
to SYM2.
Syntax
Defining subsystems means telling the compiler what extended segmentation model
each subsystem uses, and which modules belong to each subsystem. In addition, it
means telling the compiler which procedures and data are accessible from outside the
subsystem.
Making everything available to all subsystems defeats the purpose of subsystems.
For example, if a procedure is declared to be accessible from outside the subsystem,
it is a far procedure. This means that all calls are far calls, even if the procedure is
never actually accessed from outside its subsystem.
Each subsystem in a PL/M program has one extended segmentation model definition,
which takes one of the following forms:
1. $ model (subsystem-id [submodel] [x])
2. $ model ([submodel] [x])
3. $ model (submodel [x])
where [x] is of the form;
[HAS module-list]
or
[HAS module-list; EXPORTS public-list]
or
[EXPORTS public-list]
296
Chapter 13
Extended Segmentation Models
Where:
model
is SMALL, COMPACT, or LARGE and specifies the extended segmentation
model for the subsystem. All modules in the subsystem must be
compiled with the same extended segmentation model.
submodel
is -CONST IN CODE- or -CONST IN DATA- and specifies the
placement of constants. -CONST IN CODE- corresponds to the ROM
submodel; -CONST IN DATA- corresponds to the RAM submodel.
The default depends on the segmentation control and corresponds to the
defaults of RAM|ROM for each model. The use of the RAM and ROM
controls (see Chapter 11) can create conflicts when subsystems are
defined. RAM is specified by -CONST IN DATA-; ROM is specified by
-CONST IN CODE-.
subsystem-id
is any PL/M identifier that can be used as a module name, and specifies
the name of the subsystem. This ID does not conflict with any IDs used
within the program. A subsystem control without subsystem-id
defines the open subsystem.
HAS module-list
is a list of module names, separated by commas, specifying the modules
belonging to the subsystem. These module names must exactly match
the module names from each source file comprising the subsystem. (A
module name is the name of the outermost DO block of a source file.) A
particular module name can appear in only one module-list. There
are no default modules in the module-list. Any module for which a
name does not appear in a module-list becomes part of the open
subsystem.
EXPORTS public-list
is a list of procedure, variable, and constant IDs, specifying the code
and data objects exported by the subsystem (i.e., accessible outside of
the subsystem). Using a dollar sign ($) in a procedure name within a
subsystem definition will cause an error. Any symbol in the exports list
may be declared PUBLIC in at most one of the modules belonging to the
subsystem, and should be declared EXTERNAL in all modules in and out
of the subsystem that access the symbol.
A particular exported symbol can appear in only one public-list.
The public-list is exhaustive. Only the symbols in the
public-list can be referenced from other subsystems. Symbols in
the subsystem declared PUBLIC but not appearing in the public-list
are accessible only from within the subsystem itself. Conversely,
PUBLIC symbols that are not intended to be referenced from outside the
subsystem should not appear in the public-list. These symbols are
called domestic symbols.
PL/M-386 Programmer's Guide
Chapter 13
297
In most applications of the subsystem controls, the HAS and EXPORTS lists will have
several dozen entries apiece. To accommodate lists of this length, a subsystem
control may be continued over more than one control line. (The continuation lines
must be contiguous, and each must begin with a dollar sign ($) in the first column.)
Keep in mind that using a dollar sign in a procedure name within a subsystem
definition will cause an error. Also, note that any number of HAS and EXPORTS lists
can appear in a control, in any order. This enables formatting of the subsystem
specification so it can be easily read and maintained.
Consider the following subsystem definition:
$COMPACT(SUB_INPUT -CONST IN CODE- HAS mod_1, mod_2, mod_3; EXPORTS
input)
$SMALL(SUB_PROCESS HAS mod_4, mod_5, mod_6, mod_7, mod_8)
$COMPACT(SUB_OUTPUT HAS mod_9, mod_10; EXPORTS format, output)
This sample program contains three subsystems: SUB_INPUT, SUB_PROCESS, and
SUB_OUTPUT. SUB_INPUT and SUB_OUTPUT use the COMPACT extended
segmentation model. SUB_PROCESS uses the SMALL extended segmentation model.
Constants are stored with the code in SUB_INPUT. The SUB_INPUT subsystem
contains the modules mod_1, mod_2, and mod_3, and exports one symbol, input.
SUB_PROCESS contains modules 4 through 8. SUB_PROCESS contains the main
program, as it must, since it is the only SMALL subsystem in the program. (Recall
that when mixing SMALL with other models, the main program must be SMALL.) For
this reason it does not need to export any symbols. A subsystem containing the main
program can export symbols (for instance, global variables). But other subsystems
must export at least one symbol, otherwise they are totally unaccessible to the main
program, and therefore useless to the program of which they are a part.)
SUB_OUTPUT supplies two symbols called format and output.
The preceding subsystem definition should appear in all 10 modules (mod_1 through
mod_10), even though not all the exported symbols are used by all subsystems. It is
recommended that the subsystem definition be kept in an include file, then included
in each module compiled. This avoids any problems in maintaining consistency
between the subsystem definitions of all source modules.
Consider another example, this time containing an open subsystem. Start from an
existing COMPACT program that does not use extended segmentation models, but
whose code has grown too large. Assume that the following modules from the
original program (ATTACH, OPEN, CLOSE, ERRORS, ALLOCATE, FREE) were compiled
with the following segmentation control:
$COMPACT
298
Chapter 13
Extended Segmentation Models
If the modules ALLOCATE and FREE are factored out from the original program,
creating SUBSYS1, the subsystem definition is as follows:
$COMPACT(SUBSYS1 HAS ALLOCATE, FREE)
Now, suppose that the modules remaining in the open subsystem reference entry
points AllocBuff and FreeBuff in SUBSYS1. These must be exported from
SUBSYS1 as follows:
$COMPACT(SUBSYS1 HAS ALLOCATE, FREE;
$
EXPORTS AllocBuff, FreeBuff)
or
$COMPACT(SUBSYS1 HAS ALLOCATE; EXPORTS AllocBuff:
$
HAS FREE; EXPORTS FreeBuff)
The second form illustrates how multiple HAS and EXPORTS lists can be used to
document the items exported from each module.
If a routine in SUBSYS1 references the procedure FatalError in the module
ERRORS, the definition of the open subsystem is as follows:
$COMPACT (EXPORTS FatalError)
No data structures need to be changed, because data reference values can be two
bytes. All procedures except AllocBuff and FreeBuff use the short call and return
mechanism.
Placement of Segmentation Controls
The segmentation controls have special restrictions associated with their placement.
These rules are as follows:
•
The segmentation controls are primary controls. They must appear before the DO
statement of the module name.
•
Only the definition of the open subsystem (with no submodel and no EXPORTS
list) can be placed on the invocation line; definitions of all other subsystems
must occur inside the source program.
The subsystem definitions for the entire program can be included in the compilation
of each module using the INCLUDE control. The compiler extracts the information
needed to correctly and efficiently compile each module's intrasubsystem and
inter-subsystem references.
PL/M-386 Programmer's Guide
Chapter 13
299
Exporting Procedures
A symbol included in a subsystem's EXPORTS list must be declared PUBLIC in one of
the modules in that subsystem. The symbol, called an exported symbol, can be
referenced by modules in other subsystems. A PUBLIC symbol defined within a
subsystem but not listed in its EXPORTS list is called a domestic symbol. It should be
referenced only by modules within the same subsystem.
A procedure should be exported only if it must be referenced outside the defining
subsystem, because accessing exported procedures will, in general, require more code
and time than is required for domestic procedures.
Exported procedures have the following characteristics:
•
The long form of call and return is used.
•
The caller's DS and ES registers are saved and restored upon entry and exit.
•
The DS and ES registers are loaded with the associated data segment upon entry.
Note that if a SMALL or MEDIUM module calls a procedure that is exported from a
COMPACT or LARGE subsystem, the stack sections of the two will not be combined
when the modules are combined because the segments containing them have different
names (see Chapter 11). To get the proper stack size, the SEGSIZE control on the
utility used to combine the program modules must be used to increase the size of the
DATA segment. This segment must be increased by the sum of the stack requirements
for both the SMALL or MEDIUM module and the subsystem.
300
Chapter 13
Extended Segmentation Models
The SMALL RAM segmentation control uses short pointers. Therefore, care must be
taken when calling procedures that have pointer parameters and are exported from a
SMALL subsystem. In these cases, the compiler always uses the value of the current
DS register as the selector portion of the long pointer. This means that passing a
pointer to any data items declared in the SMALL module will produce the proper
result, but the following restrictions must be observed for the special cases:
1.
If the actual parameter is the NIL pointer, DS:0 will be passed to the exported
procedure. Consequently, the procedure executes differently if it is called from a
SMALL module than if it had been called from a COMPACT, MEDIUM, or LARGE
module. For example:]
$COMPACT (FOO HAS N; EXPORTS FOO)
$SMALL
M: DO;
DECLARE PTR POINTER;
FOO: PROCEDURE (P) EXTERNAL;
DECLARE P POINTER;
END FOO;
CALL FOO (NIL);
/* Wrong, will pass DS:0
*/
PTR = NIL;
CALL FOO (PTR);
/* Wrong, will pass DS:0
*/
END M;
$COMPACT (FOO HAS N; EXPORTS FOO)
N: DO;
FOO: PROCEDURE (P) PUBLIC;
DECLARE P POINTER;
DECLARE B BYTE;
B = (P=NIL);
/* Will assign FALSE (000H) to B
*/
/* if FOO is called from SMALL; Will
*/
/* assign TRUE (0FFH) to B otherwise
*/
END FOO;
CALL FOO (NIL);
/* Right, will pass 0:0
*/
END N;
PL/M-386 Programmer's Guide
Chapter 13
301
2.
If the actual parameter is a pointer to a procedure, the compiler extends the short
pointer with DS and then passes the value of DS:(offset of procedure) to the
exported procedure. This situation should be avoided because the result of any
reference through such a pointer is undefined. For example:
$COMPACT (FOO HAS N; EXPORTS FOO)
$SMALL
M: DO;
DECLARE PTR POINTER;
DECLARE TABLE(10) BYTE;
FOO: PROCEDURE (P) EXTERNAL;
DECLARE P POINTER;
END FOO;
BAZ: PROCEDURE;
...
END BAZ;
CALL FOO (@BAZ);
PTR = @BAZ;
CALL FOO (PTR);
CALL FOO (@TABLE);
END M;
/* Wrong, will pass */
/* DS:offset-of-BAZ */
/* Wrong, will pass DS:PTR
/* Right, will pass pointer
/* to TABLE
*/
*/
*/
Large Matrix Example
The large REAL matrix example can now be fully developed (see Using Subsystems).
Recall that one module for each row is needed, with each module containing a
1-billion element REAL array. Running such an application is possible only on
systems having virtual memory management for supporting such large data. The first
module could be:
ROW0_MOD: DO;
/* ROW0_MOD is the module name */
DECLARE ROW0 (1000000000) REAL PUBLIC;
END ROW0_MOD;
302
Chapter 13
Extended Segmentation Models
The modules for ROW1 through ROW9 are similar. The subsystem definition at this
point is:
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
(
(
(
(
(
(
(
(
(
(
ROW0_SYS
ROW1_SYS
ROW2_SYS
ROW3_SYS
ROW4_SYS
ROW5_SYS
ROW6_SYS
ROW7_SYS
ROW8_SYS
ROW9_SYS
HAS
HAS
HAS
HAS
HAS
HAS
HAS
HAS
HAS
HAS
ROW0_MOD;
ROW1_MOD;
ROW2_MOD;
ROW3_MOD;
ROW4_MOD;
ROW5_MOD;
ROW6_MOD;
ROW7_MOD;
ROW8_MOD;
ROW9_MOD;
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
ROW0
ROW1
ROW2
ROW3
ROW4
ROW5
ROW6
ROW7
ROW8
ROW9
)
)
)
)
)
)
)
)
)
)
Now define the program:
MATRIX_MOD: DO;
DECLARE ROW0 (1000000000)
DECLARE ROW1 (1000000000)
DECLARE ROW2 (1000000000)
DECLARE ROW3 (1000000000)
DECLARE ROW4 (1000000000)
DECLARE ROW5 (1000000000)
DECLARE ROW6 (1000000000)
DECLARE ROW7 (1000000000)
DECLARE ROW8 (1000000000)
DECLARE ROW9 (1000000000)
PL/M-386 Programmer's Guide
REAL
REAL
REAL
REAL
REAL
REAL
REAL
REAL
REAL
REAL
EXTERNAL;
EXTERNAL;
EXTERNAL;
EXTERNAL;
EXTERNAL;
EXTERNAL;
EXTERNAL;
EXTERNAL;
EXTERNAL;
EXTERNAL;
Chapter 13
303
DECLARE ROW_POINTERS (10) POINTER INITIAL (
@ROW0, @ROW1, @ROW2, @ROW3, @ROW4,
@ROW5, @ROW6, @ROW7, @ROW8, @ROW9 );
RETRIEVE_ELEMENT: PROCEDURE (ROW,COL) REAL PUBLIC;
DECLARE (ROW,COL) WORD;
DECLARE ROW_PTR POINTER,
ROW_ARRAY BASED ROW_PTR (1) REAL;
ROW_PTR = ROW_POINTERS (ROW);
RETURN ROW_ARRAY (COL);
END RETRIEVE_ELEMENT;
STORE_ELEMENT: PROCEDURE (ROW,COL,VAL) PUBLIC;
DECLARE (ROW,COL) WORD;
DECLARE VAL REAL;
DECLARE ROW_PTR POINTER,
ROW_ARRAY BASED ROW_PTR (1) REAL;
ROW_PTR = ROW_POINTERS (ROW);
ROW_ARRAY (COL) = VAL;
END STORE_ELEMENT;
/* the matrix processing code inserted here */
END MATRIX_MOD;
Now assume that other modules will be added to this program later. In this case, it is
better to put MATRIX_MOD and these other modules in the COMPACT OPEN subsystem.
This way modules can freely be added or deleted without having to redefine the
overall subsystem structure. Also assume the need to calculate sines and cosines of
various matrix elements. The functions SINE and COSINE are supplied in an external
math package. The only thing known about this package is that all its routines
require long calls.
The final subsystem definition is now:
$LARGE (
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
$COMPACT
304
Chapter 13
EXPORTS SINE, COSINE )
( ROW0_SYS HAS ROW0_MOD;
( ROW1_SYS HAS ROW1_MOD;
( ROW2_SYS HAS ROW2_MOD;
( ROW3_SYS HAS ROW3_MOD;
( ROW4_SYS HAS ROW4_MOD;
( ROW5_SYS HAS ROW5_MOD;
( ROW6_SYS HAS ROW6_MOD;
( ROW7_SYS HAS ROW7_MOD;
( ROW8_SYS HAS ROW8_MOD;
( ROW9_SYS HAS ROW9_MOD;
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
EXPORTS
ROW0
ROW1
ROW2
ROW3
ROW4
ROW5
ROW6
ROW7
ROW8
ROW9
)
)
)
)
)
)
)
)
)
)
Extended Segmentation Models
The COMPACT control should appear in the invocation line. The first control line
indicates that the symbols SINE and COSINE require long references and belong to
some unknown subsystem. The next ten lines define the ten closed subsystems, each
containing a row of the matrix. The COMPACT control is specified on the invocation
line when compiling MATRIX_MOD (and when compiling any other module in the
program except the ROW modules).
Every subsystem definition should be consistent. For example, ROW0_MOD must
reside in the same subsystem in each definition. It is convenient to put control lines,
such as those shown above, in an include file. If any changes to the subsystem
definitions are made later, only one file needs to be updated.
■■■
PL/M-386 Programmer's Guide
Chapter 13
305
306
Chapter 13
Extended Segmentation Models
Error and Warning Messages
14
The compiler may issue these kinds of error and warning messages:
•
PL/M program error messages
•
Fatal command tail and control error messages
•
Fatal input/output error messages
•
Fatal insufficient memory error messages
•
Fatal compiler failure error messages
•
Insufficient memory warning messages
The source errors are reported in the program listing; the fatal errors are reported on
the console device.
PL/M Program Error and Warning Messages
Nearly all of the source PL/M program error messages are interspersed in the listing
at the point of error and follow the general format:
*** ERROR mmm IN ppp (LINE ppp), NEAR 'aaa', message
or:
*** WARNING mmm IN ppp (LINE ppp), NEAR 'aaa', message
Where:
mmm
is an error number from the following list.
ppp
is the actual source line number where the error occurs.
aaa
is the source text near where the error is detected.
message
is a message from the following list.
The following source error messages may be encountered.
*** ERROR 1 INVALID CONTROL
An unrecognized control in the control line; for example:
$NXCODE;
PL/M-386 Programmer's Guide
/* probably intended NOCODE */
Chapter 14
307
*** ERROR 2 PRIMARY CONTROL FOLLOWS NON-CONTROL LINE
Primary controls can be control lines in the source program, but they must come first.
No other statements can precede them.
*** ERROR 3 MISSING CONTROL PARAMETER
Certain controls (e.g., INCLUDE), require a parameter.
*** ERROR 4 INVALID CONTROL PARAMETER
Examples are an illegal pathname for a control such as OBJECT or a string where a
number is required.
*** ERROR 5 INVALID CONTROL FORMAT
See Chapter 11 for correct formatting of control lines.
*** ERROR 7 INVALID PATHNAME
The pathname for a file is incorrectly specified; see the host-system operating
instructions.
*** WARNING 8 ILLEGAL PAGELENGTH, IGNORED
The pagelength specified is less than 5 or greater than 255; the default is 60.
*** ERROR 9 ILLEGAL PAGEWIDTH, IGNORED
The pagewidth specified is less than 60 or more than 132; the default is 120.
*** WARNING 10 RESPECIFIED PRIMARY CONTROL, IGNORED
Primary controls can be specified only once and cannot alter a previous setting.
*** ERROR 11 MISPLACED ELSE OR ELSEIF CONTROL
ELSE or ELSEIF control occurred without a corresponding IF control.
*** ERROR 12 MISPLACED ENDIF CONTROL
ENDIF control occurred without a corresponding IF control.
*** ERROR 13 MISSING ENDIF CONTROL
End of source file without an ENDIF control to match a previous IF.
*** ERROR 14 NAME TOO LONG(31), TRUNCATED
Switch variable name in IF, ELSE, SET, or RESET statement is too long.
*** ERROR 15 MISSING OPERATOR
Two operands in an expression must be separated by an arithmetic, logical, or
relational operator.
*** WARNING 16 INVALID CONSTANT, ZERO ASSUMED
The constant specified by SET, IF, or ELSEIF is invalid.
*** ERROR 17 INVALID OPERAND
SET, RESET, IF, or ELSEIF is used in an invalid position.
*** WARNING 18 PARENTHESES IGNORED WITHIN CONDITIONAL COMPILATION
CONDITION
Parentheses within conditional compilation conditions are ignored and the expression
is evaluated according to the regular precedence rules.
*** ERROR 19 LIMIT EXCEEDED: SAVE NESTING
See Appendix B for the correct limit.
308
Chapter 14
Error and Warning Messages
*** ERROR 20 LIMIT EXCEEDED: INCLUDE NESTING
See Appendix B for the correct limit.
*** ERROR 21 MISPLACED RESTORE CONTROL
RESTORE works only if there is a preceding SAVE.
*** ERROR 22 UNEXPECTED END OF CONTROL
A segmentation control requires a continuation line or a right parenthesis.
*** ERROR 23 SYMBOL EXISTS IN MORE THAN ONE HAS LIST
A module name can occur in only one HAS list.
*** ERROR 24 SUBSYSTEM ALREADY DEFINED
The subsystem name has already been defined.
*** ERROR 25 CONFLICTING SEGMENTATION CONTROLS
More than one segmentation control affecting the module being compiled was
encountered. One common cause is specifying both -CONST IN CODE- and ROM in
a module with a subsystem definition.
*** ERROR 26 ILLEGAL PL/M IDENTIFIER
Identifiers can be up to 31 alphanumeric characters or the underscore; the first
character must be alphabetic or the underscore.
*** ERROR 27 PREDEFINED SWITCHES ARE NOT VALID BEFORE MODULE NAME
Predefined switches can be used only after the first DO statement.
*** WARNING 28 INVALID PL/M CHARACTER, IGNORED
Look near the text flagged for an invalid character, or one that is inappropriate in
context. Delete it or retype the statement.
*** WARNING 29 UNPRINTABLE CHARACTER, IGNORED
Retype the line in question using valid characters.
*** ERROR 30 STRING TOO LONG, TRUNCATED
See Appendix B for the correct limit.
*** ERROR 31 ILLEGAL CONSTANT TYPE
A constant contains illegal characters. This might reflect missing operators (e.g.,
A=4T instead of A=4+T).
*** ERROR 32 INVALID CHARACTER IN CONSTANT
For example, 107B and 0ABCD will cause this error because neither can be valid in
any PL/M interpretation; 7 is not a binary numeral, B cannot occur in decimal or
octal, and neither string ends in H.
PL/M-386 Programmer's Guide
Chapter 14
309
*** ERROR 33 RECURSIVE MACRO EXPANSION
Following is an example causing this error:
DECLARE A LITERALLY 'B';
DECLARE B LITERALLY 'A';
.
.
.
B=4;/* error discovered here */
LITERALLYs cannot be declared circularly (i.e., solely in terms of each other).
*** ERROR 34 LIMIT EXCEEDED: MACRO NESTING (5)
This error occurs when too many DECLARE statements refer back through each other
to the one that actually supplies a type. See Appendix B for the correct limit. For
example:
DECLARE A
DECLARE B
. .
. .
DECLARE Y
DECLARE Z
.
.
.
A=7;
LITERALLY 'B';
LITERALLY 'C';
. .
. .
LITERALLY 'Z';
BYTE INITIAL (77);
/* error discovered here */
*** ERROR 35 LIMIT EXCEEDED: SOURCE LINE LENGTH (128)
See Appendix B for the correct limit.
*** ERROR 37 INVALID REAL CONSTANT
*** WARNING 38 REAL CONSTANT UNDERFLOW
An underflow occurred when conversion into floating-point was attempted.
*** WARNING 39 REAL CONSTANT OVERFLOW
An overflow occurred when conversion into floating-point was attempted.
*** ERROR 40 NULL STRING NOT ALLOWED
Strings of length zero are not supported.
*** ERROR 41 DELETED: "tokens"
The compiler deleted tokens while attempting to recover from a syntax error.
*** ERROR 42 NEAR "syntax" INSERTED: "tokens"
The compiler inserted tokens while attempting to recover from a syntax error.
*** ERROR 43 STATEMENTS FOLLOW MODULE END
Statements follow the logical end-of-module.
*** ERROR 44 CONSTANT TOO LARGE
A constant value (e.g., 999,999,999,999) is too large for the compiler.
310
Chapter 14
Error and Warning Messages
*** WARNING 45 MISMATCHED BLOCK IDENTIFIER
If a label is supplied in an END statement, the label must match the first unmatched
DO statement above the END. Sometimes the error involves a module name confused
with a procedure name.
*** ERROR 46 DUPLICATE PROCEDURE NAME
Procedure names must be unique.
*** ERROR 47 LIMIT EXCEEDED: PROCEDURES
Too many procedures in this module. Break it into smaller modules. See Appendix
B for the correct limit.
*** ERROR 48 DUPLICATE PARAMETER NAME
A parameter must be declared exactly once. This message indicates that the flagged
parameter already has a definition at this block level, as in:
YAR: PROCEDURE (YAR77, YAR78);
DECLARE YAR77 BYTE;
DECLARE YAR77 BYTE;
Perhaps a different spelling was intended.
*** ERROR 49 NOT AT MODULE LEVEL
The flagged attribute or initialization can be valid only at the module level, not in a
procedure.
*** ERROR 50 DUPLICATE ATTRIBUTE
Attributes should be specified only once. This message means the compiler has
found a declaration like:
DECLARE B BYTE EXTERNAL EXTERNAL;
*** ERROR 51 MISSING OR ILLEGAL INTERRUPT VALUE
Interrupt numbers must be whole-number constants between 0 and 255. Thus -7 or
272 would be invalid.
*** ERROR 52 INTERRUPT WITH PARAMETERS
No parameters can be used in interrupt procedures.
*** ERROR 53 INTERRUPT WITH TYPED PROCEDURE
Interrupt procedures must be untyped.
*** ERROR 54 INVALID DIMENSION
*** ERROR 55 LIMIT EXCEEDED: NESTED STRUCTURES
See Appendix B for the correct limit.
*** ERROR 56 STAR DIMENSION WITH STRUCTURE MEMBER
Star dimension (*) must not be used with structures. The dimensions for an array
that is a structure member must be specified explicitly.
*** ERROR 57 CONFLICT WITH PARAMETER
Object cannot be a parameter.
*** ERROR 58 DUPLICATE DECLARATION
The flagged item already has a definition declared at this block level.
PL/M-386 Programmer's Guide
Chapter 14
311
*** ERROR 59 ILLEGAL PARAMETER TYPE
Parameters cannot be declared of type structure or array.
*** ERROR 60 DUPLICATE LABEL
Each label must be unique within its block or scope. Otherwise, GOTOs and CALLs
would have ambiguous targets.
*** ERROR 61 DUPLICATE MEMBER NAME
Member has been declared twice in the same structure. For example, in:
DECLARE AIR STRUCTURE (F4 BYTE, F4 BYTE);
subsequent references to AIR.F4 would be ambiguous.
*** ERROR 62 UNDECLARED PARAMETER
A parameter named in the procedure statement was not defined in the body of the
procedure.
*** ERROR 63 CONFLICTING ATTRIBUTES
A variable has been declared with inconsistent attributes (e.g., PUBLIC or EXTERNAL,
DATA or INITIAL, AT or BASED).
*** ERROR 64 LIMIT EXCEEDED: DO BLOCKS
See Appendix B for the correct limit.
*** ERROR 65 ILLEGAL PARAMETER ATTRIBUTE
Certain attributes cannot be used to declare a parameter (e.g., PUBLIC, EXTERNAL,
DATA, INITIAL, AT, or BASED).
*** ERROR 66 UNDEFINED BASE
A variable was declared BASED using an undeclared identifier.
*** ERROR 67 INVALID TYPE OR ATTRIBUTE FOR BASE
A base must be a non-subscripted scalar of type ADDRESS, POINTER, WORD,
SELECTOR, or OFFSET.
*** ERROR 68 MISPLACED DECLARATION
Declarations and procedures can be interspersed, but not declarations and executable
statements.
*** ERROR 69 INVALID BASE WITH LABEL OR MACRO
BASED cannot be used with LABEL or LITERALLY types.
*** ERROR 70 INVALID DIMENSION WITH LABEL OR MACRO
LABEL or LITERALLY cannot be dimensioned.
*** ERROR 71 INITIALIZATION LIST REQUIRED
A list of initial values is required if the INITIAL attribute, the non-external *
dimension form, or the non-external DATA attribute is used.
*** ERROR 72 BASED CONFLICTS WITH ATTRIBUTES
Examples of attributes conflicting with base include AT, DATA, INITIAL, PUBLIC,
and EXTERNAL.
312
Chapter 14
Error and Warning Messages
*** ERROR 73 DATA OR EXECUTABLE STATEMENTS IN EXTERNAL
An EXTERNAL procedure, being defined elsewhere, cannot contain executable
statements or data declarations for variables that are not formal parameters.
*** ERROR 74 MISSING RETURN FOR TYPED PROCEDURE
A typed procedure must return a value; thus, it must include a RETURN statement.
*** ERROR 75 INVALID NESTED REENTRANT PROCEDURE
Reentrant procedures cannot contain procedures.
*** ERROR 76 LIMIT EXCEEDED: FACTORED LIST
Too many variables were named in a factored declaration. Break it into several
declarations. See Appendix B for the correct value.
*** ERROR 77 LIMIT EXCEEDED: STRUCTURE MEMBERS
See Appendix B for the correct value.
*** ERROR 78 MISSING PROCEDURE NAME
Every procedure must have a name.
*** ERROR 79 MULTIPLE PROCEDURE LABELS
Procedures must have only one name.
*** ERROR 80 DECLARATIONS MAY NOT BE LABELED
Labels cannot be used on declaration statements.
*** ERROR 81 STAR DIM WITH FACTORED LIST NOT ALLOWED
Separate the array declarations giving the data initializations for each array
separately, or explicitly state the dimensions of the factored array declarations as in
the following examples:
DECLARE (A,B) (*) BYTE DATA ('abcd', 'xywz'); /* illegal */
DECLARE (A) (*) BYTE DATA ('abcd');
/* legal */
DECLARE (B)(*) BYTE DATA ('xywz');
/* legal */
or
DECLARE (A,B) (4) BYTE DATA ('abcd', 'xywz');
/* legal */
*** ERROR 82 SIZE EXCEEDS nn BYTES
Storage for the declared item exceeds the maximum storage for the microprocessor.
For the Intel386 and Intel486 microprocessor nn is 4G bytes.
*** WARNING 83 PROCEDURE CONTAINS NO EXECUTABLE STATEMENTS
This procedure does nothing, but executes successfully.
*** ERROR 85 INITIAL USED WITH ROM OPTION
Variables declared with INITIAL are not initialized until load-time. Thus, if the
program is in ROM, these initializations will never occur.
*** ERROR 86 LIMIT EXCEEDED: NUMBER OF PARAMETERS
The procedure declaration includes too many parameters. See Appendix B for the
correct limit.
PL/M-386 Programmer's Guide
Chapter 14
313
*** ERROR 88 LIMIT EXCEEDED: PROGRAM TOO COMPLEX
The program has too many complex expressions, cases, or procedures. Break it into
smaller procedures.
*** ERROR 89 COMPILER ERROR: BAD ERROR RECOVERY
An unrecoverable error occurred. Trying a different copy of the compiler on a
different drive might reveal that the first copy has been damaged. Contact your
RadiSys representative.
*** ERROR 90 COMPILER ERROR: MULTIPLE PARSE ARGS
See source error message number 89.
*** ERROR 91 LIMIT EXCEEDED: PROGRAM TOO COMPLEX
The program has too many complex expressions, cases, or procedures. Break it into
smaller procedures.
*** ERROR 92 COMPILER ERROR: PARSE ARG STACK UNDERFLOW
See source error message number 89.
*** ERROR 93 LIMIT EXCEEDED: PROGRAM TOO COMPLEX
The program has too many complex expressions, cases, or procedures. Break it into
smaller modules.
*** ERROR 94 COMPILER ERROR: PARSE STACK UNDERFLOW
See source error message number 89.
*** ERROR 95 COMPILER ERROR: PARSE BUFFER OVERFLOW
See source error message number 89.
*** ERROR 96 LIMIT EXCEEDED: BLOCK NESTING
The program has too many nested DO blocks. Break it into smaller procedures. See
Appendix B for the correct limit.
*** ERROR 97 COMPILER ERROR: SCOPE STACK UNDERFLOW
See source error message number 89.
*** ERROR 98 LIMIT EXCEEDED: STATEMENT TOO COMPLEX
The statement is too large for the compiler. Break it into several smaller statements.
*** ERROR 99 COMPILER ERROR: SEMANTIC UNDERFLOW
See source error message number 89.
*** ERROR 100 STRING CONSTANT TOO LONG
String constants used as scalars have a maximum of four characters.
*** ERROR 101 UNSUBSCRIPTED ARRAY
The array reference requires a subscript.
*** WARNING 102 UNQUALIFIED STRUCTURE
This statement is ambiguous as to which structure or member it references.
*** ERROR 103 NOT AN ARRAY
Subscripts are permitted only on identifiers declared as arrays. Check spelling
consistency.
314
Chapter 14
Error and Warning Messages
*** ERROR 104 MULTIPLE SUBSCRIPTS
PL/M has only single dimension arrays. Therefore, only one subscript is permitted in
an array reference. For example, for any array TING references of the form
TING(2,4) or TING(3,7,9,6) are invalid.
*** ERROR 105 NOT A STRUCTURE
For example, a reference of the form GNU.F1, where GNU was not declared a
structure.
*** ERROR 106 UNDEFINED IDENTIFIER
Every identifier must be declared.
*** ERROR 107 UNDEFINED MEMBER
For example, KAPI.HORN, where KAPI is a valid, declared structure but HORN is an
undeclared member of the structure.
*** ERROR 108 ILLEGAL ITERATIVE DO INDEX TYPE
Only expressions of type BYTE, WORD, and INTEGER can be used.
*** ERROR 109 UNDEFINED OR NOT A LABEL
The identifier following GOTO must be a label; the flagged item was declared
otherwise, or the identifier was declared as a label but was not defined.
*** ERROR 110 MISSING RETURN VALUE
A typed procedure must return a value that is specified by its RETURN statement.
*** ERROR 111 INVALID RETURN WITH UNTYPED PROCEDURE
An untyped procedure does not return a value; thus, its RETURN statement cannot
specify one.
*** ERROR 112 INVALID INDIRECT TYPE
Only WORD or POINTER scalars can be used for indirect calls. This excludes WORD or
POINTER expressions; BYTE, DWORD, INTEGER, or REAL scalars; all structures; and
all arrays.
*** ERROR 113 INVALID PARAMETER COUNT
The number of actual parameters supplied in a CALL must be equal to the number of
formal parameters declared in the procedure.
*** ERROR 114 QUALIFIED PROCEDURE NAME
Procedure names cannot be qualified.
*** ERROR 115 INVALID FUNCTION REFERENCE
Typed procedures can be invoked only by use in an expression, not by a CALL.
*** ERROR 116 INVALID CASE EXPRESSION TYPE
Case expressions must be of type BYTE, WORD, or INTEGER.
*** ERROR 117 LIMIT EXCEEDED: NUMBER OF ACTIVE CASES
Reduce the number of cases in this case statement; the maximum number has been
exceeded.
*** ERROR 118 TYPE CONFLICT
An example of type conflict is WORD and REAL mixed in a reference.
PL/M-386 Programmer's Guide
Chapter 14
315
*** ERROR 119 INVALID BUILT-IN REFERENCE
Built-in reference was qualified with a member name, or OUTPUT/OUTWORD did not
appear on the left side of an assignment.
*** ERROR 120 INVALID PROCEDURE REFERENCE
Untyped procedures must be invoked by a CALL statement; references to such
procedures are not permitted in expressions.
*** ERROR 121 INVALID LEFT-HAND SIDE OF ASSIGNMENT
The left-hand side of the assignment must be a scalar variable. For example,
PROCEDURE=4 or INWORD(7)=9.
*** ERROR 122 INVALID REFERENCE
Invalid label reference.
*** ERROR 123 USE OF "." MAY BE UNSAFE
The "dot" operator does not always produce correct results in a PL/M program that
contains more than one data segment or more than one code segment.
*** ERROR 124 PROCEDURE NAME REQUIRED
Procedure name is required for SET$INTERRUPT and INTERRUPT$PTR built-ins.
*** ERROR 125 PROCEDURE NAME ONLY
Parameters are not allowed on the procedure name in SET$INTERRUPT and
INTERRUPT$PTR.
*** ERROR 126 BAD INTERRUPT NUMBER
Interrupt numbers in a CAUSE$INTERRUPT statement must be whole-number
constants in the range (0 - 255).
*** ERROR 127 CONSTANT ONLY
In this instance, a constant is required.
*** ERROR 128 ARRAY REQUIRED
Some built-ins need an array name as a parameter.
*** ERROR 129 INTERRUPT PROCEDURE REQUIRED
The name declared in a SET$INTERRUPT procedure or INTERRUPT$PTR function
must be a previously declared procedure.
*** ERROR 130 INVALID RESTRICTED OPERAND
Illegal use of a dot operator.
*** ERROR 131 INVALID RESTRICTED OPERATOR
Only + and – can be used in restricted expressions.
*** ERROR 133 REFERENCE REQUIRED
A variable reference is required for LENGTH, LAST, and SIZE.
*** ERROR 134 VARIABLE REQUIRED
The operand to LENGTH, LAST, and SIZE must be a variable.
*** ERROR 135 VALUE TOO LARGE
A value is too large for its contextually determined type.
316
Chapter 14
Error and Warning Messages
*** ERROR 136 ABSOLUTE POINTER WITH SHORT POINTERS
Two possible causes in the SMALL (RAM) case: pointer variables cannot be initialized
with or assigned whole number constants; or the @ operator cannot be used with a
variable that was located at an absolute address that was specified by a whole number
constant.
*** ERROR 137 INVALID RESTRICTED EXPRESSION
Only addresses or constant types can be used in restricted expressions.
*** ERROR 138 PUBLIC AT EXTERNAL
PUBLIC declarations must be fully defined within the procedure. For example:
DECLARE KUN BYTE EXTERNAL;
DECLARE JAN BYTE PUBLIC AT (.KUN);
is illegal.
*** ERROR 139 PUBLIC AT ABSOLUTE
Absolute locations for PUBLICs are supported only in the LARGE model.
*** ERROR 140 PUBLIC AT MEMORY
PUBLIC at @ MEMORY is not supported by COMPACT.
*** ERROR 141 AT BASED VARIABLE
Based variables cannot be used in AT clauses.
*** ERROR 142 ILLEGAL FORWARD REFERENCE
An AT expression cannot have a forward reference. Any location reference in the AT
expression must refer to previously declared variables.
*** ERROR 143 VARIABLE TYPE REQUIRED IN AN AT EXPRESSION
The AT expression must be a variable name. For example:
DECLARE B BYTE AT (.proc_name);
is illegal.
*** ERROR 144 LIMIT EXCEEDED: DATA OR STACK SEGMENT TOO LARGE
*** ERROR 145 LIMIT EXCEEDED: CODE OR CONST SEGMENT TOO LARGE
*** ERROR 146 LIMIT EXCEEDED: NUMBER OF EXTERNALS
See Appendix B for the correct limit.
*** ERROR 147 LABEL NOT AT LOCAL OR MODULE LEVEL
Label was not used correctly.
*** ERROR 148 INITIALIZING MORE SPACE THAN DECLARED
The number of initialization values exceeds the number of declared elements.
*** ERROR 149 ILLEGAL MODULE NAME REFERENCE
Module names cannot be referenced.
*** WARNING 150 USE OF "." WITH FAR PROCEDURE
A subsequent indirect call made through the respective address/pointer generates the
wrong type of call.
PL/M-386 Programmer's Guide
Chapter 14
317
*** WARNING 151 USE OF "@" WITH NEAR PROCEDURE
See source error message number 150.
*** ERROR 152 INVALID "." OR "@" OPERAND
Must be used with a variable, procedure, or constant list.
*** ERROR 153 INVALID RETURN IN MAIN PROGRAM
A main program must have no returns.
*** ERROR 154 STAR DIMENSIONED VARIABLE WITH LENGTH, SIZE OR LAST
The LENGTH, LAST, and SIZE built-in functions cannot be used with variables
declared with the implicit dimension specifier (*) and the EXTERNAL attribute.
*** ERROR 155 SYMBOL EXPORTED FROM ANOTHER SUBSYSTEM
A PUBLIC symbol in this module is also exported by another subsystem.
*** ERROR 156 LONG POINTER REQUIRED FOR THIS CONSTRUCT
A model with long pointers is required.
*** ERROR 158 INITIALIZATION CONFLICTS WITH ATTRIBUTES
An external variable cannot be initialized.
*** ERROR 159 ILLEGAL INTERRUPT PROCEDURE REFERENCE
An interrupt procedure cannot be invoked with the CALL statement.
*** ERROR 160 INTERRUPT PROCEDURES MUST BE PUBLIC
An interrupt procedure must also be given the PUBLIC attribute.
*** ERROR 161 ILLEGAL ABSOLUTE POINTER OR SELECTOR
Constants cannot be assigned to POINTERs or SELECTORs, nor used to initialize
them. POINTERs and SELECTORs also cannot be passed as actual parameters.
*** ERROR 162 LIMIT EXCEEDED: STATEMENT TOO COMPLEX
The statement is too large for the compiler. Break it into several smaller statements.
*** WARNING 162 LIMIT EXCEEDED: PROGRAM COMPLEXITY
Too many complex expressions, cases, etc. Break it into smaller procedures.
*** ERROR 163 COMPILER ERROR: SEMANTIC UNDERFLOW
See source error message number 89.
*** ERROR 164 COMPILER ERROR: INVALID NODE
See source error message number 89.
*** ERROR 165: 286 INTERFACE OBJECT NOT EXTERNAL
If the machine parameter is 286, all identifiers in the id list must be declared
EXTERNAL.
*** ERROR 166 COMPILER ERROR: INVALID TREE
See source error message number 89.
*** ERROR 167 COMPILER ERROR: SCOPE STACK UNDERFLOW
See source error message number 89.
318
Chapter 14
Error and Warning Messages
*** ERROR 168 LIMIT EXCEEDED: PROGRAM COMPLEXITY
The program has too many complex expressions, cases, or procedures. Break it into
smaller procedures.
*** ERROR 169 COMPILER ERROR: INVALID RECORD
See source error message number 89.
*** ERROR 170 INVALID DO CASE BLOCK, AT LEAST ONE CASE REQUIRED
The DO CASE block is described in Chapter 6.
*** ERROR 171 LIMIT EXCEEDED: NUMBER OF CASES
*** ERROR 172 LIMIT EXCEEDED: NESTING OF TYPED PROCEDURE CALLS
*** ERROR 173 LIMIT EXCEEDED: NUMBER OF ACTIVE PROCEDURES AND DO CASE
GROUPS
See Appendix B for the correct limit.
*** ERROR 174 ILLEGAL NESTING OF BLOCKS, ENDS NOT BALANCED
For every DO, an END is needed.
*** ERROR 175 COMPILER ERROR: INVALID OPERATION
See source error message number 89.
*** ERROR 176 LIMIT EXCEEDED: REAL EXPRESSION COMPLEXITY
The REAL stack has eight registers. Heavily nested use of REAL functions with REAL
expressions as parameters can get excessively complex. See Appendix F.
*** ERROR 177 COMPILER ERROR: REAL STACK UNDERFLOW
See source error message numbers 89 and 176.
*** ERROR 178 LIMIT EXCEEDED: BASIC BLOCK COMPLEXITY
There is a very long list of statements without labels, CASEs, IFs, GOTOs, and/or
RETURNs. Either break the procedure into several smaller procedures, or add labels
to some of the statements.
*** ERROR 179 LIMIT EXCEEDED: STATEMENT SIZE
The statement is too large for the compiler. Break it into several smaller statements.
*** ERROR 199 LIMIT EXCEEDED: PROCEDURE COMPLEXITY FOR OPTIMIZE (2)
The combined complexity of statements, user labels, and compiler-generated labels is
too great. Simplify as much as possible, perhaps breaking the procedure into several
smaller procedures.
*** ERROR 200 ILLEGAL INITIALIZATION OF MORE SPACE THAN DECLARED
The number of initialization values exceeds the number of declared elements.
*** ERROR 201 INVALID LABEL: UNDEFINED
No definition for this label was found.
*** ERROR 202 LIMIT EXCEEDED: NUMBER OF EXTERNAL ITEMS
See Appendix B for the correct limit.
*** ERROR 203 COMPILER ERROR: BAD LABEL ADDRESS
See source error message number 89.
PL/M-386 Programmer's Guide
Chapter 14
319
*** ERROR 204 LIMIT EXCEEDED: CODE SEGMENT SIZE
See Appendix B for the correct limit.
*** ERROR 205 COMPILER ERROR: BAD CODE GENERATED
See source error message number 89.
*** ERROR 206 LIMIT EXCEEDED: DATA SEGMENT SIZE
See Appendix B for the correct limit.
*** ERROR 207 ATTEMPT TO USE 0 AS DIVISOR IN DIVISION/MODULO
Zero cannot be used as a divisor in division/modulo; use 1. This error appears at the
end as a semantic error.
***
***
***
***
ERROR
ERROR
ERROR
ERROR
210
211
212
230
COMPILER
COMPILER
COMPILER
COMPILER
ERROR:
ERROR:
ERROR:
ERROR:
OBJECT MODULE GENERATION ERROR
DEBUG SEGMENT SIZE OVERFLOW
ILLEGAL FIXUP
INVALID INTERNAL TYPE
See source error message number 89.
*** ERROR 241 ILLEGAL TYPE CASTING
For example:
pt2=pointer (real_value)
is illegal.
*** ERROR 242 TRUNCATION OF n BIT OFFSET
OFFSET was assigned to a variable with a size less than 32 bits; the assigned value
may not be a valid OFFSET. For the 8086 and 286 microprocessors, n is 16. For the
Intel386 and Intel486 microprocessors, n is 32.
*** ERROR 243 286 INTERFACE OBJECT NOT EXTERNAL
If the machine parameter is 286, all identifiers in the id list must be declared
EXTERNAL.
*** ERROR 244 SYMBOL REPEATED IN INTERFACE SPECIFICATION
Symbols can be used only once in an INTERFACE control (i.e., a symbol cannot be
repeated in the INTERFACE control).
*** ERROR 245 AT VARIABLE IN DIFFERENT SEGMENT
A variable cannot be declared using both the DATA attribute and the AT attribute when
using the ROM option. DATA should be in CODE segments and INITIAL should be in
DATA segments.
*** WARNING 247 INDIRECT CALL THROUGH 16 BIT VARIABLE
An indirect call through a 16-bit variable is not recommended because a 16-bit
variable can address only the first 64K of a segment.
*** WARNING 248 BASE TYPE HAS ONLY 16 BITS OFFSET
Use of a 16-bit base specifier is not recommended because it can address only the
first 64K of a segment.
*** ERROR 251 COMPILER ERROR: INVALID OBJECT
*** ERROR 252 COMPILER ERROR: SELF NAME LINK
320
Chapter 14
Error and Warning Messages
*** ERROR 253 COMPILER ERROR: SELF ATTR LINK
See source error message number 89.
*** ERROR 254 LIMIT EXCEEDED: PROGRAM COMPLEXITY
The program has too many complex expressions, cases, or procedures. Break it into
smaller modules.
*** ERROR 255 LIMIT EXCEEDED: SYMBOLS
See Appendix B for the correct limit.
✏
Note
If a terminal error is encountered, program text beyond the point of
error is not compiled. A terminal error message will appear at the
point of error in the program listing.
Fatal Command Tail and Control Error Messages
Fatal command tail errors are caused by an improperly specified compiler invocation
command or an improper control. The errors that can occur are as follows:
COMMAND TAIL TOO LONG
COMMAND TAIL BUFFER LIMIT EXCEEDED AT OR NEAR: xxx
ILLEGAL COMMAND TAIL SYNTAX OR VALUE
UNABLE TO PARSE COMMAND TAIL AT OR NEAR: xxx
ILLEGAL COMMAND TAIL SYNTAX OR VALUE
UNRECOGNIZED CONTROL IN COMMAND TAIL
INVOCATION COMMAND DOES NOT END WITH <CR><LF>
ILLEGAL COMMAND TAIL SYNTAX
PL/M-386 Programmer's Guide
Chapter 14
321
Fatal Input/Output Error Messages
Fatal input/output errors occur when the user specifies an incorrect pathname for
compiler input or output. These error messages are of the form:
PL/M-386 xxx ERROR -FILE:
NAME:
ERROR:
COMPILATION TERMINATED
These errors also occur when the device runs out of space (e.g., the list file is larger
than the available memory).
Fatal Insufficient Memory Error Messages
The fatal insufficient memory errors are caused by a system configuration with
insufficient RAM memory to support the compiler.
The errors that can occur due to insufficient memory are as follows:
NOT ENOUGH MEMORY FOR COMPILATION
DYNAMIC STORAGE OVERFLOW
NOT ENOUGH MEMORY FOR CODE GENERATION
322
Chapter 14
Error and Warning Messages
Fatal Compiler Failure Error Messages
The fatal compiler failure errors are internal errors that should never occur. If you
encounter such an error, please contact your RadiSys representative. The errors
falling into this class are as follows:
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***
***
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
ERROR
89 COMPILER ERROR: BAD ERROR RECOVERY
90 COMPILER ERROR: MULTIPLE PARSE ARGS
92 COMPILER ERROR: PARSE ARG STACK UNDERFLOW
94 COMPILER ERROR: PARSE STACK UNDERFLOW
95 COMPILER ERROR: PARSE BUFFER OVERFLOW
97 COMPILER ERROR: SCOPE STACK UNDERFLOW
99 COMPILER ERROR: SEMANTIC UNDERFLOW
163 COMPILER ERROR: SEMANTIC UNDERFLOW
164 COMPILER ERROR: INVALID NODE
166 COMPILER ERROR: INVALID TREE
167 COMPILER ERROR: SCOPE STACK UNDERFLOW
175 COMPILER ERROR: INVALID OPERATION
177 COMPILER ERROR: REAL STACK UNDERFLOW
203 COMPILER ERROR: BAD LABEL ADDRESS
205 COMPILER ERROR: BAD CODE GENERATED
210 COMPILER ERROR: OBJECT MODULE GENERATION
211 COMPILER ERROR: DEBUG SEGMENT SIZE OVERFLOW
212 COMPILER ERROR: ILLEGAL FIXUP
230 COMPILER ERROR: INVALID INTERNAL TYPE
251 COMPILER ERROR: INVALID OBJECT
252 COMPILER ERROR: SELF NAME LINK
253 COMPILER ERROR: SELF ATTR LINK
It is also possible to receive an UNKNOWN FATAL ERROR message.
Insufficient Memory Warning Messages
The following warnings may occur if there are too many symbols for symbol
processing:
NOT ENOUGH MEMORY FOR FULL DICTIONARY LISTING
NOT ENOUGH MEMORY FOR ANY XREF PROCESSING
NOT ENOUGH MEMORY FOR FULL XREF PROCESSING
■■■
PL/M-386 Programmer's Guide
Chapter 14
323
324
Chapter 14
Error and Warning Messages
PL/M Reserved Words and
Predeclared Identifiers
A
Introduction
These are reserved words in PL/M-386. They cannot be used as identifiers.
ADDRESS
AND
AT
BASED
BY
BYTE
CALL
CASE
CHARINT
DATA
DECLARE
DISABLE
DO
DWORD
ELSE
ENABLE
END
EOF
EXTERNAL
GO
GOTO
HALT
HWORD
IF
INITIAL
PL/M-386 Programmer's Guide
INTEGER
INTERRUPT
LABEL
LITERALLY
LONGINT
MINUS
MOD
NOT
OFFSET
OR
PLUS
POINTER
PROCEDURE
PUBLIC
REAL
REENTRANT
RETURN
SELECTOR
SHORTINT
STRUCTURE
THEN
TO
WHILE
WORD
QWORD
XOR
Appendix A
325
The following are PL/M-386 identifiers, built-in procedures and predeclared
variables. If one of these identifiers is declared in a DECLARE statement, the
corresponding built-in procedure or predeclared variable becomes unavailable within
the scope of the declaration.
ABS
326
NIL
ADJUSTRPL
OFFSETOF
BLOCKINPUT
OUTHWORD
BLOCKINWORD
OUTPUT
BLOCKOUTPUT
OUTWORD
BLOCKOUTWORD
PARITY
BUILDPTR
RESTOREGLOBALTABLE
CARRY
RESTOREINTERRUPTABLE
CAUSEINTERRUPT
RESTOREREALSTATUS
CLEARTASKSWITCHEDFLAG
ROL
CONTROLREGISTER
ROR
CMPB
SAL
CMPW
SAR
DEBUGREGISTER
SAVEGLOBALTABLE
DEC
SAVEINTERRUPTTABLE
DOUBLE
SAVEREALSTATUS
FINDB
SCANBIT
FINDHW
SCANRBIT
FINDRB
SCL
FINDRHW
SCR
FINDRW
SEGMENTREADABLE
FINDW
SEGMENTWRITABLE
FIX
SELECTOROF
FLAGS
SETB
FLOAT
SETHW
GETACCESSRIGHTS
SETREALMODE
GETREALERROR
SETW
GETSEGMENTLIMIT
SHL
HIGH
SHLD
IABS
SHR
Appendix A
PL/M Reserved Words and Predeclared Identifiers
INHWORD
SHRD
INITREALMATHUNITSKIPRB
SIGN
INPUT
SIGNED
INT
SIZE
INWORD
SIZE
LAST
SKIPB
LENGTH
SKIPHW
LOCALTABLE
SKIPRHW
LOCKSET
SKIPRW
LOW
SKIPW
MACHINESTATUS
STACKBASE
MOVB
STACKPTR
MOVBIT
TASKREGISTER
MOVE
TESTREGISTER
MOVHW
TIME
MOVRB
UNSIGN
MOVRBIT
WAITFORINTERRUPT
MOVRHW
XLAT
MOVRW
ZERO
MOVW
Identifiers with WORD16 Control
The following identifiers are specific to PL/M-386 when using the WORD16 control.
BLOCKINDWORD
BLOCKOUTDWORD
CMPD
FINDD
FINDRD
INDWORD
MOVD
MOVRD
OUTDWORD
SETD
SKIPD
SKIPRD
PL/M-386 Programmer's Guide
Appendix A
327
Identifiers with MOD486 Control
The following identifiers are specific to PL/M-386 when using the MOD486 control.
BYTESWAP
TESTREGISTER
INVALIDATEDATACACHE
WBINVALIDATEDATACACHE
INVALIDATETLBENTRY
■■■
328
Appendix A
PL/M Reserved Words and Predeclared Identifiers
330
Appendix A
PL/M Reserved Words and Predeclared Identifiers
PL/M Program Limits
Feature
Indirection level (A BASED on B,
B BASED on C)
Length of a string constant
Nesting of blocks
unlimited***
255
18
5
Nesting of LITERALLY invocations
5
32
Number of active cases
255
Number of cases in a DO CASE block
255
Number of DO blocks in a procedure
Number of declared EXTERNAL items
Number of elements in a factored list
Number of EXTERNAL items used
Number of labels on a statement
Number of nested procedures
and DO cases
Number of nested typed procedures
Number of procedures in a module
Numbers of characters in a line
Segment size
Size of LITERALLY string
65536
**
64
**
unlimited*
255
18
1016
128
4G
unlimited*
Structure size
4G-1
Symbol capacity
2500
Total number of members in
a structure (at all levels)
*
PL/M-386
Nesting of INCLUDE controls
Nesting of structures
B
128
Limited by the total size of the symbol table.
PL/M-386 Programmer's Guide
Appendix B
331
** Limited by either the number of procedures or the number of symbols, or both.
*** Unlimited means limited only by the amount of free memory allocated by the compiler.
■■■
332
PL/M Program Limits
334
Appendix B
PL/M Program Limits
Grammar of the PL/M Language
C
This appendix lists the entire syntax of the PL/M language in Backus-Naur Form
(BNF) notation. Since the semantic rules are not included here, this syntax permits
certain constructions that are not actually allowed. The terminology used in the BNF
syntax has been designed for convenience in constructing concise and rigorous
definitions. Its appearance differs substantially from the main body of the manual.
The notations used here are slightly extended from standard BNF notations. An
ellipsis (...) indicates that the syntactic element preceding it can be repeated
indefinitely. The vertical bar (|) separates alternatives. Braces ({ }) enclose required
alternatives and brackets ([ ]) enclose optional alternatives. The vertical bar within
braces and brackets is also a separator of alternatives.
When items are stacked vertically within brackets, only one of the items can be used.
PL/M-386 Programmer's Guide
Appendix C
335
Lexical Elements
Character Sets
<character>::= <apostrophe> | <non-quote character>
<apostrophe>::= '
<non-quote character>::= <letter> | <decimal digit> | $ |
<special character> | blank
<letter>::= <uppercase letter> | <lowercase letter>
<uppercase letter>::= A | B | C | D | E | F | G | H | I | J | K | L |
M | N | O | P | Q | R | S | T | U | V | W | X |
Y | Z
<lowercase letter>::= a | b | c | d | e | f | g | h | i | j | k | l |
m | n | o | p | q | r | s | t | u | v | w | x |
y | z
<decimal digit>::= 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9
<special character>::= + | - | * | / | < | > | = | : | ; | . | , |
( | ) | @ | _
Tokens
<token>::= <delimiter> | <identifier> | <reserved word> |
<numeric constant> | <string>
Delimiters
<delimiter>::= <simple delimiter> | <compound delimiter>
<simple delimiter>::= + | - | * | / | < | > | = | : | ; | . | , | ( |
) | @
<compound delimiter>::= <> | <= | >= | :=
Identifiers
<first id character>::= <letter> | _
<identifier>::= <first id character> [<letter> |
<decimal digit> | $ | _ ]...
<reserved word>
336
(For a list of reserved words, see Appendix A.)
Appendix C
Grammar of the PL/M Language
Numeric Constants
<numeric constant>::= <binary number> | <octal number> |
<decimal number> | <hexadecimal number> |
<floating point number>
<binary number>::= <binary digit> [<binary digit> | $]... B | Q
<octal number>::= <octal digit> [<octal digit> | $]... { O | Q }
<decimal number>::= <decimal digit> [<decimal digit> | $]... [D]
<hexadecimal number>::= <decimal digit> [<hexadecimal digit>
| $]... H
<floating point number>::= <digit string> <fractional part>
[<exponent part>]
<fractional part>::= [<.digit string>]
<exponent part>::= E [+ | -] <digit string>
<digit string>::= <decimal digit> [<decimal digit> | $]...
<binary digit>::= 0 | 1
<octal digit>::= <binary digit> | 2 | 3 | 4 | 5 | 6 | 7
<decimal digit>::= <octal digit> | 8 | 9
<hexadecimal digit>::= <decimal digit> | A | B | C | D | E | F
Strings
<string>::= '<string body element> [<string body element>]... '
<string body element>::= <non-quote character> |"
PL/M Text Structure: Tokens, Blanks, and Comments
<pl/m text>::= <token> | <separator> [<token> | <separator>]...
<separator>::= blank | <comment>
<comment>::= /* [<character>]... */
PL/M-386 Programmer's Guide
Appendix C
337
Modules and the Main Program
<compilation>::= <module> [EOF]
<module>::= <module name>:<simple do block>
<module name>::= <identifier>
338
Appendix C
Grammar of the PL/M Language
Declarations
<declaration>::= <declare statement> | <procedure definition>
DECLARE Statement
<declare statement>::= DECLARE <declare element list>;
<declare element list>::= <declare element>[,<declare element>]...
<declare element>::= <factored element> | <unfactored element>
<unfactored element>::= <variable element> | <literal element> |
<label element>
<factored element>::= <factored variable element> |
<factored label element>
Variable Elements
<variable element>::= <variable name specifier> [<array specifier>]
<variable type> | [<variable attributes>]
<variable name specifier>::= <non-based name> |
<based name> BASED <base specifier>
<non-based name>::= <variable name>
<based name>::= <variable name>
<variable name>::= <identifier>
<base specifier>::= <identifier>[.<identifier>]
<variable attributes>::= [PUBLIC] [<locator>][<initialization>] |
[EXTERNAL] [<constant attribute>]
<locator>::= AT(<expression>)
<constant attribute>::= DATA
<array specifier>::= <explicit dimension> | <implicit dimension>
<explicit dimension>::= (<numeric constant>)
<implicit dimension>::= (*)
PL/M-386 Programmer's Guide
Appendix C
339
<variable type>::= <basic type> | <structure type>
<basic type>::= Address | BYTE | HWORD | DWORD | QWORD | CHARINT |
OFFSET | SHORTINT | INTEGER | REAL | SELECTOR |
POINTER | OFFSET
Label Element
<label element>::= <identifier> LABEL [PUBLIC | EXTERNAL]
Literal Elements
<literal element>::= <identifier> LITERALLY <string>
Factored Variable Element
<factored variable element>::= ( <variable name specifier>
[,<variable name specifier>]...)
[<explicit dimension>] <variable type>
[<variable attributes>]
Factored Label Element
<factored label element>::= (<identifier> [,<identifier>]...)
LABEL [PUBLIC | EXTERNAL]
The Structure Type
<structure type>::= STRUCTURE (<member element>
[,<member element>]...)
<member element>::= <unfactored member> | <factored member>
<unfactored member>::= <member name> [<explicit dimension>]
<variable type>
<member name>::= <identifier>
<factored member>::= (<member name>(,<member name>)...)
[<explicit dimension>] <variable type>
340
Appendix C
Grammar of the PL/M Language
Procedure Definition
<procedure definition>::= <procedure statement> [<declaration>...]
[<unit>...] <ending>
<procedure statement>::= <procedure name> : PROCEDURE
[<formal parameter list>] [<procedure type>]
[<procedure attributes>];
<procedure name>::= <identifier>
<procedure type>::= <basic type>
<formal parameter list>::= (<formal parameter>
[,<formal parameter>]...)
<formal parameter>::= <identifier>
<procedure attributes>::= {EXTERNAL | PUBLIC | <interrupt> |
REENTRANT}...
Attributes
AT
<locator>::= AT (<expression>)
INTERRUPT
<interrupt>::= INTERRUPT
Initialization
<initialization>::= {INITIAL | DATA} (<initial
[,<initial value>]...)
value>
<initial value>::= <expression> | <string>
PL/M-386 Programmer's Guide
Appendix C
341
Units
<unit>::= <conditional clause> | <do block> | <basic statement> |
<label definition><unit>
<basic statement>::= <assignment statement> | <call statement> |
<goto statement> | <null statement> |
<return statement> |
<microprocessor dependent statement>
<scoping statement>::= <simple do statement> | <do-case statement> |
<do-while statement> |
<iterative do statement> | <end statement> |
<procedure statement>
<label definition>::= <identifier>:
Basic Statements
Assignment Statement
<assignment statement>::= <left part >=<expression>;
<left part>::= <variable reference> [,<variable reference>]...
CALL Statement
<call statement>::= CALL <simple variable>[<parameter list>];
<parameter list>::= (<expression>[,<expression>]...)
<simple variable>::= <identifier> | <identifier>.<identifier>
GOTO Statement
<goto statement>::= {GOTO | GO TO} <identifier>
Null Statement
<null statement>::= ;
RETURN Statement
<return statement>::= <typed return> | <untyped return>
<typed return>::= RETURN <expression>;
<untyped return>::= RETURN;
342
Appendix C
Grammar of the PL/M Language
Microprocessor-dependent Statements
<microprocessor dependent statement>::= <disable statement> |
<enable statement> |
<halt statement> |
<cause interrupt statement>
<disable statement>::= DISABLE;
<enable statement>::= ENABLE;
<halt statement>::= HALT;
<cause interrupt statement>::= CAUSE$INTERRUPT (numeric constant);
PL/M-386 Programmer's Guide
Appendix C
343
Scoping Statements
Simple DO Statement
<simple do statement>::= DO;
DO-CASE Statement
<do-case statement>::= DO CASE <expression>;
DO-WHILE Statement
<do-while statement>::= DO WHILE <expression>;
Iterative DO Statement
<iterative do statement>::= DO <index part> <to part> [<by part>];
<index part>::= <index variable>=<start expression>
<to part>::= TO <bound expression>
<by part>::= BY <step expression>
<index variable>::= <simple variable>
<start expression>::= <expression>
<bound expression>::= <expression>
<step expression>::= <expression>
END Statement
<end statement>::= END [<identifier>];
Procedure Statement
<procedure statement>::= <procedure name> : PROCEDURE
[<formal parameter list>] [<procedure type>]
[<procedure attributes>];
344
Appendix C
Grammar of the PL/M Language
Conditional Clause
<conditional clause>::= <if condition><true unit> |
<if condition><true element> ELSE
<false element>
<if condition>::= IF <expression> THEN <true unit>
<true element>::= [<label definition>...] <do block> |
[<label definition>...] <basic statement>
<false element>::= <unit>
<true unit>::= <unit>
DO Blocks
<do block>::= <simple do block> | <do-case block> | <do-while block> |
<iterative do block>
Simple DO Blocks
<simple do block>::= <simple do statement>[<declaration>...]
[<unit>...]<ending>
<ending>::= [<label definition>...]<end statement>
DO-CASE Blocks
<do-case block>::= <do-case statement> [<unit>...] <ending>
DO-WHILE Blocks
<do-while block>::= <do-while statement> [<unit>...] <ending>
Iterative DO Blocks
<iterative do block>::= <iterative do statement> [<unit>...] <ending>
PL/M-386 Programmer's Guide
Appendix C
345
Expressions
Primaries
<primary>::= <constant> | <variable reference> | <location reference>
| <subexpression>
<subexpression>::= (<expression>)
Constants
<constant>::= <numeric constant> | <string>
Variable References
<variable reference>::= <data reference> | <function reference>
<data reference>::= <name>[<subscript>] [<member specifier>]
<subscript>::= (<expression>)
<member specifier>::= .<member name>[<subscript>]
<function reference>::= <name>[<actual parameters>]
<actual parameters>::= (<expression>[,<expression>]...)
<member name>::= <identifier>
<name>::= <identifier>
Location References
<location reference>::= @<constant list> | @<variable reference>
<constant list>::= (<constant>[,<constant>]...)
Operators
<operator>::= <logical operator> | <relational operator> |
<arithmetic operator>
<logical operator>::= AND | OR | NOT | XOR
<relational operator>::= < | > | <= | >= | <> | =
<arithmetic operator>::=
346
Appendix C
+ | - | PLUS | MINUS | * | / | MOD
Grammar of the PL/M Language
Structure of Expressions
<expression>::= <logical expression> | <embedded assignment>
<embedded assignment>::= <variable reference> := <logical expression>
<logical expression>::= <logical factor> | <logical expression>
<or operator> <logical factor>
<or operator>::= OR | XOR
<logical factor>::= <logical secondary> | <logical factor>
<and operator> <logical secondary>
<and operator>::= AND
<logical secondary>::= [<not operator>] <logical primary>
<not operator>::= NOT
<logical primary>::= <arithmetic expression> [<relational operator>
<arithmetic expression>]
<relational operator>::= < | > | <= | >= | <> | =
<arithmetic expression>::= <term> | <arithmetic expression>
<adding operator> <term>
<adding operator>::= + | - | PLUS | MINUS
<term>::= <secondary> | <term> <multiplying operator> <secondary>
<multiplying operator>::= * | / | MOD
<secondary>::= [<unary minus> | <unary plus>] <primary>
<unary minus>::= <unary plus>::= +
■■■
PL/M-386 Programmer's Guide
Appendix C
347
348
Appendix C
Grammar of the PL/M Language
Differences Between PL/M Compilers
D
Differences between PL/M-86 and PL/M-80
PL/M-86 differs from PL/M-80 in the following respects:
•
Support for floating-point arithmetic
•
Support for signed arithmetic
•
Addition of REAL, INTEGER, POINTER, and SELECTOR data types
•
Addition of the @ location reference operator
•
Support for nested structures
•
Expanded set of built-in procedures
In addition, the PL/M-80 reserved word ADDRESS is replaced by the PL/M-86
reserved word WORD. PL/M-80 has only the BYTE and ADDRESS data types.
However, PL/M-86 has the following data types: BYTE, WORD, DWORD, INTEGER,
REAL, POINTER, and SELECTOR.
The PL/M-86 rules for expression evaluation are more complete than those of
PL/M-80. Other differences stem from the ones noted here. For example, an
iterative DO block operates differently if its index variable is an INTEGER variable.
PL/M-386 Programmer's Guide
Appendix D
349
Compatibility of PL/M-80 Programs and the PL/M-86
Compiler
PL/M-80 programs that operate correctly on an 8080 microprocessor can be
recompiled with the PL/M-86 compiler to produce code that will run on an 8086
microprocessor. You may need to edit the PL/M-80 source code to change identifiers
that are PL/M-86 reserved words. (It is not necessary to change ADDRESS to WORD;
ADDRESS is a PL/M-86 reserved word with the same meaning as WORD.)
Note that where PL/M-86 programs would normally have POINTER variables and
location references formed with the @ operator, PL/M-80 programs have ADDRESS
(WORD) variables and location references formed with the dot operator. PL/M-80
usage is less restricted than PL/M-86 usage, because arithmetic operations can be
used on WORD values. In general, the PL/M-86 compiler supports PL/M-80 usage to
provide upward compatibility. Some restrictions affect the types of expressions that
can be used in the AT attribute, the INITIAL and DATA initializations, and location
references. See also the discussions of size controls and the dot and @ operators in
this manual.
Differences between PL/M-286 and PL/M-86
PL/M-286 differs from PL/M-86 in the following respects:
350
•
POINTER and SELECTOR variables cannot be assigned absolute (i.e., constant)
values. Only the equals operator (=) can be used with POINTER variables. For
SELECTOR variables the logical (AND, OR, NOT, XOR) and relational (<, >, <=, >=,
<>, =) operators can be used.
•
Access to the hardware flag register is provided with the built-in variable FLAGS.
•
Four built-in functions have been added to support multiple byte and word input:
BLOCKINPUT, BLOCKINWORD, BLOCKOUTPUT, and BLOCKOUTWORD (available to
PL/M-86 via the MOD86|MOD186 control).
•
The type of the STACKBASE variable has been changed from WORD to SELECTOR.
•
New built-in procedures and functions have been added to support the 286
hardware protection model.
•
Interrupt procedures are no longer assigned numbers in the source program.
(This is done by the 286 system builder.) Interrupt procedures also cannot be
called directly, and the SET$INTERRUPT and INTERRUPT$PTR built-ins have
been removed.
•
The memory array has been removed.
Appendix D
Differences Between PL/M Compilers
Compatibility of PL/M-86 Programs and the PL/M-286
Compiler
PL/M-86 programs that operate correctly on an 8086 microprocessor can be
recompiled with the PL/M-286 compiler to produce code that will run on an 286
microprocessor. The PL/M-86 source code must be edited as follows:
•
Assignments to the STACKBASE built-in variable must be changed from WORD to
SELECTOR.
•
All absolute pointer and selector assignments must be changed. (Pointers can be
assigned a zero value using the new built-in function NIL.) Also, relational
operations on pointer and selector values for any operation other than equality
and inequality must be changed.
•
The interrupt numbers on all interrupt procedures must be deleted. Interrupt
vectors will be assigned to these procedures by the 286 system builder. Direct
calls to interrupt procedures must also be changed.
•
References to the SET$INTERRUPT, INTERRUPT$PTR, and MEMORY built-ins
must be removed.
Differences between PL/M-386 and PL/M-286
PL/M-386 differs from PL/M-286 in the following respects:
•
The string built-ins FIND, CMP, and SKIP return a value of 0FFFFFFFFH for the
not found and string equal results.
•
Support for 64-bit unsigned scalars.
•
Support for 8-bit and 32-bit signed scalars.
•
Addition of HWORD and QWORD unsigned integers, and the CHARINT, SHORTINT,
and LONGINT signed integer data types.
•
ADDRESS is the same as OFFSET (and not as WORD as in PL/M-286).
•
Support for casting functions.
•
Support for WORD32 and WORD16 mapping for data type identifiers.
•
Addition of the WORD32|WORD16 primary compiler controls, which ensure PL/M
data type and language compatibility.
•
MEDIUM and LARGE segmentation controls no longer indicate unique meaning to
the compiler; MEDIUM is interpreted as SMALL and LARGE is interpreted as
COMPACT except when LARGE is used to indicate a subsystem whose name is
unknown at compile time.
PL/M-386 Programmer's Guide
Appendix D
351
•
Several new built-in procedures and functions have been added to support the
new data types (for example, CMPHW, BLOCKINHWORD; see Chapters 9 and 10),
and some bit-string operations (for example, SCANBIT, MOVBIT).
•
The built-ins CONTROL$REGISTER, DEBUG$REGISTER, and TEST$REGISTER
have been added to support the Intel386 microprocessor.
•
The following built-ins have been added to support the Intel486 microprocessor:
BYTE$SWAP, TEST$REGISTER, INVALIDATE$DATA$CACHE,
WB$INVALIDATE$DATA$CACHE, and INVALIDATE$TLB$ENTRY.
•
The FLAT and MOD486 compiler controls have been added.
Compatibility of PL/M-286 Programs and the
PL/M-386 Compiler
PL/M-286 programs can be compiled with the PL/M-286 compiler to produce code
that will run on Intel386 and Intel486 microprocessors in 286 microprocessor mode
and interface with PL/M-386 code through INTERFACE(/286). PL/M-286
programs can be recompiled with the PL/M-386 compiler to produce code that will
run on an Intel386 and Intel486 microprocessors in their native microprocessor mode.
■■■
352
Appendix D
Differences Between PL/M Compilers
Character Set
E
This appendix lists the ASCII character set and indicates whether the characters are
part of the PL/M source character set. Table E-1 is a list of codes.
Table E-1. Character Set
Dec
Hex
PL/M
Character
0
00
NO
NULL
1
01
NO
SOH
2
02
NO
STX
3
03
NO
ETX
4
04
NO
EOT
5
05
NO
ENQ
6
06
NO
ACK
7
07
NO
BEL
8
08
NO
BS
9
09
YES
HT
10
0A
YES
LF
11
0B
NO
VT
12
0C
NO
FF
13
0D
YES
CR
14
0E
NO
SO
15
0F
NO
SI
16
10
NO
DLE
17
11
NO
DC1
18
12
NO
DC2
19
13
NO
DC3
20
14
NO
DC4
21
15
NO
NAK
PL/M-386 Programmer's Guide
Appendix E
353
22
16
NO
SYN
Table E-1. Character Set (continued)
354
Dec
Hex
PL/M
Character
23
17
NO
ETB
24
18
NO
CAN
25
19
NO
EM
26
1A
NO
SUB
27
1B
NO
ESC
28
1C
NO
FS
29
1D
NO
GS
30
1E
NO
RS
31
1F
NO
US
32
20
YES
SP
33
21
NO
!
34
22
NO
"
35
23
NO
#
36
24
YES
$
37
25
NO
%
38
26
NO
&
39
27
YES
'
40
28
YES
(
41
29
YES
)
42
2A
YES
*
43
2B
YES
+
44
2C
YES
,
45
2D
YES
-
46
2E
YES
.
47
2F
YES
/
48
30
YES
0
49
31
YES
1
50
32
YES
2
51
33
YES
3
52
34
YES
4
Appendix E
Character Set
53
35
YES
5
Table E-1. Character Set (continued)
Dec
Hex
PL/M
Character
54
36
YES
6
55
37
YES
7
56
38
YES
8
57
39
YES
9
58
3A
YES
:
59
3B
YES
;
60
3C
YES
<
61
3D
YES
=
62
3E
YES
>
63
3F
YES
?
64
40
YES
@
65
41
YES
A
66
42
YES
B
67
43
YES
C
68
44
YES
D
69
45
YES
E
70
46
YES
F
71
47
YES
G
72
48
YES
H
73
49
YES
I
74
4A
YES
J
75
4B
YES
K
76
4C
YES
L
77
4D
YES
M
78
4E
YES
N
79
4F
YES
O
80
50
YES
P
81
51
YES
Q
82
52
YES
R
83
53
YES
S
PL/M-386 Programmer's Guide
Appendix E
355
84
54
YES
T
Table E-1. Character Set (continued)
356
Dec
Hex
PL/M
Character
85
55
YES
U
86
56
YES
V
87
57
YES
W
88
58
YES
X
89
59
YES
Y
90
5A
YES
Z
91
5B
NO
[
92
5C
NO
\
93
5D
NO
]
94
5E
NO
^
95
5F
YES
_
96
60
NO
`
97
61
YES
a
98
62
YES
b
99
63
YES
c
100
64
YES
d
101
65
YES
e
102
66
YES
f
103
67
YES
g
104
68
YES
h
105
69
YES
i
106
6A
YES
j
107
6B
YES
k
108
6C
YES
l
109
6D
YES
m
110
6E
YES
n
111
6F
YES
o
112
70
YES
p
113
71
YES
q
114
72
YES
r
Appendix E
Character Set
115
73
PL/M-386 Programmer's Guide
YES
s
Appendix E
357
Table E-1. Character Set (continued)
Dec
Hex
PL/M
Character
116
74
YES
t
117
75
YES
u
118
76
YES
v
119
77
YES
w
120
78
YES
x
121
79
YES
y
122
7A
YES
z
123
7B
NO
{
124
7C
NO
|
125
7D
NO
}
126
7E
NO
~
127
7F
NO
DEL
■■■
358
Appendix E
Character Set
360
Appendix E
Character Set
Linking to Modules
Written in Other Languages
F
Introduction
This appendix describes the calling conventions used by the [x]86 family of
languages. These calling conventions are standardized so that a module written in
PL/M can freely call procedures, subroutines, and subprograms in other modules
written in other [x]86 languages.
The information in this appendix is not necessary to call PL/M procedures and
functions from PL/M. See Chapter 8 for information about parameters and
arguments.
The calling conventions and stack and register usage described in this appendix are
needed to call ASM subroutines. Also, the corresponding data types listed at the end
of this appendix are needed to write a subroutine that can pick up the data in the
PL/M program. Refer to the ASM macro assembler operating instructions for more
information about combining PL/M programs with ASM programs and for examples.
The easiest way to ensure compatibility between assembly-language subroutines that
are combined with PL/M programs or procedures is to write a dummy procedure in
PL/M. This procedure would have the same argument list and the same attributes as
the assembly language subroutine. Then compile the PL/M procedure with the
correct segmentation control and the CODE control. This will produce a
pseudo-assembly listing of the generated microprocessor code, which can then be
copied to the prologue and epilogue of the assembly language subroutine.
PL/M-386 Programmer's Guide
Appendix F
361
With PL/M, separate modules can be written and compiled, and combined at a later
time. This allows you to create separately tested modules that are combined after
they are internally bug-free. Not all modules have to be in PL/M: you can choose the
appropriate language for each module. Be sure to combine the modules properly
with a binder or a linker in order to satisfy references to externals. Because the [x]86
languages (excluding C) follow the same calling sequence, control will pass to a
called module correctly. The standard calling sequence is described in the following
section.) However, the called module might not be able to deal intelligently with the
data passed to it since languages treat some data structures differently.
By specifying arguments in a reference to an external procedure, data is passed to the
external procedure. The number of arguments and the order in which they are
specified must match the number and order of the corresponding parameters in the
external procedures declaration (see Chapter 8).
All arguments for parameters are passed on the microprocessor's stack, or the
numeric coprocessor's register stack, in the order in which they were specified. For
Intel386 and Intel486 microprocessors, the space occupied by a parameter pushed on
the microprocessor's stack is always a multiple of four bytes. Functions return
non-real values in a register, and REAL values on the top of the numeric coprocessor's
register stack.
362
Appendix F
Linking to Modules Written in Other Languages
Calling Sequence
The calling sequence for each procedure activation places the procedure's actual
parameters (if any) on the stack and then activates the procedure with a CALL
instruction.
Parameters are placed on the microprocessor's stack or the numeric coprocessor's
register stack in left-to-right order. Because the stack grows from higher locations to
lower locations, the first parameter occupies the highest position on the stack, and the
last parameter occupies the lowest position. Stack representation for the different
PL/M parameters is described in Table F-1.
Table F-1. Stack Representation for PL/M Parameters
Parameter
Intel386 and Intel486 CPU
Stack Representation
BYTE
Four bytes, with the higher three bytes undefined.
CHARINT
Four bytes, with the higher three bytes undefined.
HWORD
Four bytes, with the high two bytes undefined.
SELECTOR
Four bytes, with the high two bytes undefined.
SHORTINT
Four bytes, with the high two bytes undefined.
WORD
Four bytes, with no undefined bytes.
OFFSET
Four bytes, with no undefined bytes.
INTEGER
Four bytes, with no undefined bytes.
REAL
Four bytes, with no undefined bytes.
DWORD
Eight bytes with the high 32 bits pushed first and
the low 32 bits 16 bits pushed second.
For Intel386 and Intel486 microprocessors, a POINTER parameter in the
SMALL(ROM) and COMPACT cases consists of a selector and an offset. The 16-bit
selector is pushed first, followed by the 32-bit offset.
The left-most seven REAL parameters are passed on the numeric coprocessor's stack.
If more than seven REAL parameters are present, the rest (after the left-most seven)
are passed on the microprocessor's stack and are intermixed with the other non-real
parameters in the order in which all parameters were declared.
After the parameters are passed, the CALL instruction places the return address on the
stack. In the SMALL and COMPACT cases with local (or non-exported) procedures, this
address is a 32-bit offset (the contents of the EIP register) and occupies four bytes on
the stack.
PL/M-386 Programmer's Guide
Appendix F
363
For procedures exported from a subsystem, the return address is a POINTER value
consisting of a selector and offset; the return address is placed on the stack in the
same way a POINTER parameter is passed. The 16-bit segment selector (contents of
the CS register) is pushed first, then the 32-bit offset (EIP register contents) is
pushed.
For all of the microprocessors, control is passed to the code of the procedure by
updating the EIP register. For procedures exported from a subsystem, the CS register
is also updated.
Figure F-1 shows the stack layout at the point where the procedure gains control.
Stack Marker (BP Reg. Contents)
Stack
Counter
Higher
Locations
Parameter 0
Parameter
1
.
.
.
Parameter N
Return Segment Selector
Lower
Locations
Return Offset
Each Parameter Occupies 2 or 4
Bytes - See Text
Absent In Small Or Compact
Program Or Local Procedure
Stack Pointer ( SP Reg. Contents)
OSD540
Figure F-1. Stack Layout at Point Where
a Non-interrupt Procedure is Activated
364
Appendix F
Linking to Modules Written in Other Languages
Procedure Prologue
In compiling the procedure itself, the compiler inserts a sequence of instructions
called the procedure prologue. The procedure prologue varies depending on the type
of procedure being compiled as follows:
•
If the procedure has the PUBLIC attribute and the program size is LARGE, or if it
is exported from a subsystem, the content of the DS register is placed on the
stack and is then updated to the data segment of the procedure. ES is set to DS.
•
If any parameter of the procedure is referenced by a nested procedure, all
parameters are copied from the stack to space reserved for them in the data
segment.
•
The stack marker offset (EBP register contents) is placed on the stack, and the
current stack pointer (ESP register contents) is used to update the BP or EBP
register.
•
If the procedure has the REENTRANT attribute, space is reserved on the stack for
any variables declared within the procedure (this does not include based
variables, variables with the DATA attribute, or variables with the AT attribute).
Control then passes to the code compiled from the executable statements in the
procedure body. Figure F-2 shows the stack layout at this point.
Higher
Locations
Parameter 0
Parameter 1
.
.
.
Parameter N
Stack Counter
Return Segment Selector
Return Offset
Old Data Segment Selector
Old Stack Marker
Absent if Any Parameter is Referenced
Within A Nested Procedure
Absent In Small Or Compact
Program Or Local Procedure
Only In Public Procedure In
Large Program
New Stack Marker (BP Reg. Contents)
Only In Reentrant Procedure
Lower
Locations
This Space May Be Used
During Procedure Execution
Stack Pointer May Change
During Procedure Execution
OSD541
Figure F-2. Stack Layout During Execution of a Non-interrupt Procedure Body
PL/M-386 Programmer's Guide
Appendix F
365
366
Appendix F
Linking to Modules Written in Other Languages
Procedure Epilogue
To return from the procedure, the compiler inserts an instruction sequence called the
epilogue. This accomplishes the following:
•
If the compiler has used stack locations for temporary storage or local variables
during procedure execution, the stack pointer (the ESP register) is loaded with
the stack marker (the EBP register), discarding the temporary storage.
•
The old stack marker is restored by popping the stored value from the stack into
the EBP register.
•
If the procedure has the PUBLIC attribute and the program size is LARGE or it is
exported from a subsystem, the old data segment selector is restored by popping
the stored value from the stack into the DS register. Additionally, ES is set to
DS.
•
The stored return address (a 32-bit offset) is popped into the EIP register. If the
procedure is exported, the stored return address selector is also popped into the
CS register. Any parameters stored on the stack are discarded.
PL/M-386 Programmer's Guide
Appendix F
367
Register Usage
Table F-2 provides a summary register usage.
Table F-2. Summary of the Intel386 Microprocessor Register Usage
Register
Must Preserve
Usage
EAX
No
Return BYTE (AL), HWORD (AX), WORD,
DWORD, CHARINT (AL), SHORTINT
(AX), INTEGER, SELECTOR (AX),
POINTER offset portion, and OFFSET.
EBX
No** (Yes, when using
C language interface.)
--
ECX
No
--
EDX
No
Return upper half of DWORD values,
POINTER segment selector.
ESP
Yes*
Stack pointer
EBP
Yes
Stack marker
ESI
No (Yes when using
--
C language interface.)
EDI
No (Yes when using
C language interface.)
--
FLAGS
No
--
CS
Yes
Called procedure's code segment.
DS
Yes
Caller's data segment.
SS
Yes
Caller's stack segment.
ES
Yes
Caller's data segment.
FS, GS
No
--
* ESP must be adjusted so that all arguments are removed from the stack on return (except when using
C language interface.
** The C language interface referred to in the table is the variable parameter list format.
368
Appendix F
Linking to Modules Written in Other Languages
The numeric coprocessor's stack contains the first seven REAL arguments passed by
the calling program. The numeric coprocessor's status word is unknown and does not
need to be saved. If the status word is changed, the numeric coprocessor's mode
word must be saved on entry and restored before exit.
If an assembly language subroutine alters the DS or SS registers, and expects to be
called by a PL/M program, the subroutine must save the contents of these registers
upon entry and restore them before returning to the PL/M program. Additionally, the
CS and ES registers must be preserved by the called procedure.
PL/M uses the ESP and EBP registers to address the stack. If a called assembly
language subroutine uses the stack register, the subroutine must save the contents of
the register on entry and restore the register's contents before returning control to the
PL/M program. Before returning, the called subroutine must also adjust the ESP
register to remove all parameters from the microprocessor's stack. Additionally, the
CS and ES registers must be preserved by the called procedure.
The EAX, EBX, ECX, EDX, ESI, EDI, FS, and GS registers do not need to be
preserved. A called subroutine can freely use these registers.
An assembly language program calling a PL/M procedure cannot expect the contents
of the general-purpose registers (EBP and ESP) to be preserved. If the contents of
these registers are needed, they must be saved prior to calling the PL/M procedure.
Table F-3 summarizes the microprocessor registers used to hold simple data types
that are returned by typed procedures.
PL/M-386 Programmer's Guide
Appendix F
369
Table F-3. Registers Used to Hold Simple Data Types
Intel386=Microprocessor
Procedure Type
Register
BYTE
AL
CHARINT
HWORD
AX
SHORTINT
DWORD
DX:AX
INTEGER
AX
WORD
EAX
OFFSET
INTEGER
DWORD
EDX:EAX
POINTER (SHORT, SMALL RAM)
EAX
POINTER (LONG,
EDX:EAX
COMPACT,
SMALL ROM)
370
SELECTOR
AX
REAL
Top of the numeric coprocessor's stack.
Appendix F
Linking to Modules Written in Other Languages
Segment Name Conventions
Tables F-4 summarizes the segmentation of a subsystem under the SMALL and
COMPACT program controls. The table shows the name of the segment in which each
type of program section is stored for each control and for subsystems.
Table F-4. Summary of PL/M-386 Segment Names
Model
SubModel
Code
Data
Const
Stack
SMALL
IN DATA
IN CODE
CODE32
CODE32
DATA
DATA
DATA
CODE32
DATA
DATA
SMALL
IN DATA
(subsystem) IN CODE
S_CODE32
S_CODE32
DATA
DATA
DATA
S_CODE32
DATA
DATA
COMPACT
IN DATA
IN CODE
S_CODE32
S_CODE32
S_DATA S_CODE32
S_DATA S_CODE32
STACK
STACK
COMPACT IN DATA
(subsystem) IN CODE
S_CODE32
S_CODE32
DATA
DATA
STACK
STACK
S_DATA
S_CODE32
Notes:
CODE32 denotes a segment name composed of CODE32.
DATA denotes a segment name composed of DATA.
S_CODE32 denotes a segment name composed of the subsystem name and CODE32.
S_DATA denotes a segment name composed of the subsystem name and DATA.
PL/M-386 Programmer's Guide
Appendix F
371
C Language Compatibility
The iC-n86 calling conventions, procedure prologue and epilogue, and register usage
differ from other Intel n86 languages. However, the INTERFACE control, described
in Chapter 11, allows C procedures to call procedures written in PL/M and vice versa.
The procedure prologue and epilogue and register usage for the VPL (variable
parameter list) calling convention for iC-n86 differ from other n86 languages. These
differences are as follows:
372
•
All parameters (real and non-real), are passed on the microprocessor's stack.
The last parameter is pushed first and the first parameter is pushed last so that
the first parameter is in the lowest memory location.
•
An integral parameter that is four bytes must be zero or sign-extended, as
required by the C language.
•
The space occupied by a parameter pushed on the microprocessor stack is always
a multiple of four bytes for Intel386 and Intel486 microprocessors.
•
Both short (floating-point) and long (double) real parameters are pushed as long
real parameters, as required by the C language. Therefore, all real parameters
passed from or to iC-386 procedures must be typed as 64-bit REAL in the
PL/M-386 code.
•
The calling procedure pops the parameters from the microprocessor stack after
the called procedure has returned. Except when the called procedure is a
function returning real results, the called procedure must not leave any entries in
the numeric coprocessor stack.
•
The ESP, EBP, CS, DS, ES, and SS registers should be preserved by the called
procedure. (They are used for global storage). The EBX, ESI, and EDI registers
should also be preserved by the called procedure. These registers can be used by
the caller for local data storage.
•
The EAX, ECX, EDX, FS and GS registers do not need to be preserved by the
called procedure.
Appendix F
Linking to Modules Written in Other Languages
Design Guidelines
The following guidelines should be followed when combining C and PL/M modules.
These guidelines are demonstrated in the code example which follows afterwards.
1.
Identify all C functions which use the VPL calling convention. Library function
calling conventions are found by checking the .h include files.
2.
Use the PL/M-386 INTERFACE compiler control to allow the PL/M compiler
control to generate VPL code.
3.
All PL/M functions should be in a "#pragma fixed-params("plmf,...") list. This
will guarantee that any call to a PL/M function will use the FPL calling
convention.
4.
Compile all files and link them in the same way C files are linked.
Code Example
This code example, run under the iRMX Operating System, discuses how a PL/M
application makes C function calls. An example of this is when a large PL/M
application is being converted to C. Mixing C and PL/M modules allows the
converted C modules to be debugged one at a time after conversion. Another
example is a PL/M application which needs access to the extensive I/O routines
available in C libraries.
The PL/M example, named ptest.plm, shows how a PL/M function calls a C function
that uses the VPL calling convention. It also includes a C procedure called by the C
example named candplm.c, which the FPL calling convention.
The code example uses the $INTERFACE control to signal the compiler that printf is
a C function that uses the VPL convention. Any function that has the "varparams"
attribute should be included in the $INTERFACE list, such as the printf function. C
functions compiled under the iC-386 C compiler use the FPL convention by default
and should not be included in the "varparams" list.
PL/M-386 Programmer's Guide
Appendix F
373
/* PL/M module - ptest.plm */
/* Only printf uses the VPL convention */
$INTERFACE(C=printf)
ptest: DO;
/* Uses VPL convention */
printf:PROCEDURE EXTERNAL;
END printf;
/* Uses FPL convention */
c_call_plm_funct:PROCEDURE (i) WORD EXTERNAL;
DECLARE i WORD;
END c_call_plm_funct;
/* Uses FPL convention */
c_fpl_funct:PROCEDURE (i) WORD EXTERNAL;
DECLARE i WORD;
END c_fpl_funct;
/* This procedure is called by a C function which uses FPL */
plmproc:PROCEDURE (i) WORD PUBLIC;
DECLARE i WORD;
i = i-1;
/* The string in the PL/M call to printf terminates with 0DH, 0AH, 00H
so it conforms to C string conventions. These symbols cause a <CR>,
<LF>, and C end of string. */
/* Call to VPL C function. */
CALL printf(@(0Dh,'In plmproc, i = %d',0Dh,0aH,00H),i);
/* Call to FPL C function */
i = c_fpl_funct(i);
RETURN(i);
END plmproc;
374
Appendix F
Linking to Modules Written in Other Languages
/* This main function is written in PL/M and may be used as a
substitute for a C main modules. The function calls printf to
demonstrate how PL/M calls a C function which uses VPL calling
conventions. It also calls a C function which uses FPL calling
conventions. */
main:PROCEDURE PUBLIC;
DECLARE i WORD, j WORD, k DWORD;
i = 5;
k = 12345678H;
/* Call to VPL C function */
CALL printf(@('The value of i = %d, and k =
%xH',0Dh,0aH,00H),i,k);
/* Call to FPL C function */
j = c_call_plm_funct(i);
CALL printf(@(0Dh,'The value of j = %d',0aH,00H)j);
END main;
END ptest;
PL/M-386 Programmer's Guide
Appendix F
375
The following C example, named candplm.c, demonstrates how a PL/M call is made
from a C application.
/* C module - candplm.c */
#include <stdio.h>
#include <reent.h>
#include <locale.h>
/* Sets FPL for PL/M functions */
#pragma fixedparams("plmproc")
extern unsigned int plmproc(unsigned int);
/* The C function c_fpl_funct uses the FPL calling conventions. The
PL/M function "plmproc" calls this function to demonstrate how a PL/M
procedure calls a C function which uses the FPL calling convention. */
unsigned int
c_fpl_funct(unsigned int i)
{
i -= 1;
printf("c_fpl_funct, i = %d\n",i);
return (i);
}
/* The C function call_plm_funct uses the FPL calling conventions.
This function calls "plmproc" to demonstrate how a C function calls a
PL/M function. */
unsigned int
c_call_plm_funct(unsigned int i)
{
i -= 1;
printf("c_call_plm_funct, i = %d\n",i);
/* Call to PL/M function */
i = plmproc(i);
return(i);
}
376
Appendix F
Linking to Modules Written in Other Languages
Compiling C and PL/M Modules
The submit file, named plmsub.csd, contains the following command syntax to
compile and bind ptest.plm and candplm.c.
ic386 candplm.c debug code compact
plm386 ptest.plm debug code compact
bnd386 /intel/obj/cstart32.obj, &
ptest.obj, candplm.obj, &
/intel/lib/clibxf32.lib &
renameseg (code32 to code) &
segsize (stack(2400H)) &
debug object(plmsub) &
rc(dm(4000h,0FFFFFh))
To invoke the submit files, enter the following command at the iRMX "-" prompt:
- submit plmsub over plmsub.out echo
When the processing stops and the prompt has returned, enter the following to run the
example:
- plmsub
The output of the code example follows:
The value of i = 5, and k is 12345678H
c_call_plmfunct, i = 4
In plmproc, i = 3
c_fpl_funct, i = 2
The value of j = 2
Note that the variable "i" changes as it is passed as a parameter. The value is initially
set to 5. As it passes through each function, it is decremented and its new value is
displayed.
■■■
PL/M-386 Programmer's Guide
Appendix F
377
378
Appendix F
Linking to Modules Written in Other Languages
Run-time Interrupt Processing
G
General Information
Interrupts can be initialized when the CPU receives a signal on its maskable interrupt
pin from a peripheral device, or when control is transferred to an interrupt vector by
the CAUSE$INTERRUPT statement. If the program runs under an operating system
that traps interrupts, the information in this appendix may not be applicable.
Note that the CPU does not respond to the interrupt signal unless interrupts are
enabled. The PL/M-386 compiler do not generate any code to enable or disable
interrupts at the start of the main program.
If interrupts are enabled and vectored through an interrupt gate, the following actions
take place:
1.
The CPU completes any instruction currently in execution.
2.
The CPU issues an acknowledge interrupt signal and waits for the interrupting
device to send an interrupt number.
3.
The CPU flag register is placed on the stack (occupying two bytes of stack
storage).
4.
Interrupts are disabled by clearing the IF flag.
5.
Single stepping is disabled by clearing the TF flag.
6.
The CPU activates the interrupt procedure corresponding to the interrupt number
sent by the interrupting device.
7.
When that procedure terminates, the stack is automatically restored to the state it
was in when the interrupt was received, and control returns to the point where it
was interrupted.
The mechanism for this activation and restoration are described in the following
sections. If interrupts are vectored through a trap gate, the fourth step is not
performed; if they are vectored through a task gate, all seven steps are replaced by a
task switch.
See also:
interrupt processing, System Concepts
PL/M-386 Programmer's Guide
Appendix G
379
The Interrupt Descriptor Table
The interrupt descriptor table (IDT) contains descriptors that vector interrupts, traps,
and protection exceptions to their respective handling routines.
These descriptors are called gates; they can be either interrupt gates, trap gates, or
task gates. Interrupt gates and trap gates point to a particular entry point in the
address space of the interrupted user (i.e., to an interrupt procedure). Task gates
point to an interrupt processing task state segment (TSS).
BLD386 sets up the IDT and to assign numbers to vector the individual gates to the
appropriate interrupt procedure or task. For more information, see the Intel386
Family Utilities User's Guide.
The IDT can hold up to 256 gates. Gates 0 through 31 are reserved for internal use.
Procedures and Tasks
For Intel386 and Intel486 microprocessors, when an interrupt is vectored through an
interrupt gate, all registers must be pushed onto the stack, and interrupts are
automatically disabled. (Interrupts must be explicitly enabled.) The interrupt
procedure then begins execution. The interrupt procedure ends with an IRET
instruction that acts as a normal return. Hence, execution starts at the beginning of a
procedure each time it is entered.
The interrupt process differs for an interrupt vectored through a task gate. The
registers for the interrupted task are saved in the TSS, and the microprocessor's
registers are loaded from the TSS of the interrupt task. Thus, no explicit register
saving is necessary. Interrupts are enabled or disabled depending on the flag settings
in the interrupt task's TSS during execution of the interrupt task (unless explicitly
changed). This enables interruption of the interrupt task. However, a protection
violation occurs if an interrupt task is busy and an attempt is made to vector through
the busy interrupt task.
The interrupt task also ends with an IRET instruction, but in this case it acts as a task
switch, saving the status of the outgoing interrupt task in memory. When the task is
re-entered, execution continues at the first instruction after the IRET instruction.
380
Appendix G
Run-time Interrupt Processing
Interrupt Procedure Prologue and Epilogue
An interrupt procedure begins by declaring its name and its PUBLIC or EXTERNAL
attribute. The following interrupt procedure declaration is the correct form for
PL/M-386:
HANDLER: PROCEDURE INTERRUPT PUBLIC;
This alerts the compiler to create a code prologue appropriate to a routine that will, in
general, be invoked by interrupts.
At the beginning of each interrupt procedure, the interrupt procedure prologue
inserted by the compiler accomplishes the following tasks:
1.
Pushes the CPU registers onto the stack in the following order: EAX, ECX,
EDX, EBX, ESP, EBP, ESI, EDI.
2.
Pushes the ES, FS, GS, and DS register content on the stack.
3.
If the interrupt procedure has the PUBLIC attribute, and if it is exported from a
subsystem, the contents of the DS register is placed on the stack and is then
updated to the data segment of the procedure. In addition, ES is set to DS.
4.
The stack marker offset (EBP register contents) is placed on the stack, and the
current stack pointer (ESP register contents) is used to update the EBP register.
5.
If the procedure has the REENTRANT attribute, space is reserved on the stack for
any variables declared within the procedure. (This does not include based
variables, variables with the DATA attribute, or variables with the AT attribute.)
✏
Note
The compiler may temporarily use the DS register and the ES
register in some cases (e.g., string built-ins), but always restores it.
Take care to note this possibility when writing an interrupt
procedure in assembly language.
PL/M-386 Programmer's Guide
Appendix G
381
At the point where the interrupt procedure prologue gains control, the stack layout is
as shown in Figure G-1.
Higher
Locations
Stack
Counter
Flag Reg. Contents
Return Segment Selector
2 Bytes
Present Regardless Of
Program Size
Return Offset
Stack Pointer
Lower
Locations
OSD542
Figure G-1. Stack Layout at Point Where an Interrupt Procedure Gains Control
382
Appendix G
Run-time Interrupt Processing
After the interrupt procedure prologue is executed (at the point where the code
compiled from the procedure body gains control), the stack layout is as shown in
Figure G-2.
Higher
Locations
Flag Reg. Contents
Return Segment Selector
(In Interrupted Program)
2 Bytes
Present Regardless of
Program Size
Return Offset
Stack Counter
EAX Reg. Contents
ECX Reg. Contents
EDX Reg. Contents
EBX Reg. Contents
ESP Reg. Contents
EBP Reg. Contents
ESI Reg. Contents
EDI Reg. Contents
ES Reg. Contents
DS Reg. Contents
Old Stack Marker
Local Variables
CPU Status Information
New Stack Marker
(BP Reg. Contents)
Only In Reentrant Procedure
This Space May Be Used During
Procedure Execution
Lower
Locations
Stack Pointer May Change
During Procedure Execution
OM02063
Figure G-2. Stack Layout during Execution of Interrupt Procedure Body
PL/M-386 Programmer's Guide
Appendix G
383
The return from the procedure body is called the interrupt procedure epilogue; it
restores the stack to the state it was in before the interrupt occurred. The interrupt
procedure epilogue contains the following steps:
1.
If the compiler has used stack locations for temporary storage or local variables
during procedure execution, the stack pointer (the ESP register) is loaded with
the current stack marker (the EBP register) discarding the temporary storage.
2.
The old stack marker is restored by popping the stored value from the stack into
the EBP register.
3.
The old data segment is restored by popping the stored value from the stack into
the DS register. This step will occur only if the procedure has a PUBLIC
attribute and it is exported from a subsystem.
4.
The stack is popped into the CPU registers in the following order: EDI, ESI,
EBP, ESP, EBX, EDX, ECX, EAX. Note that the ESP register value is
discarded.
5.
An IRET instruction is executed to return from the interrupt procedure restoring
the IP or EIP, CS, and the flag register contents from the stack.
At this point, the stack has been restored to the state it was in before the interrupt
occurred, and processing continues normally.
Interrupt Tasks
A task on the microprocessor is a single thread of execution; that is, a stream of
instructions and data with a task state image. The task state image is made up of the
contents of the task registers, the task's status word, and the virtual locations of the
task's instructions and data segments.
Tasks are initiated with a task switch operation. The CPU stores the task state image
of the outgoing task (held in the processor registers) in memory, and loads the task
state image of the incoming task into task registers. Because all the registers are
reloaded and a new address space is entered, it is impossible to jump directly from
one task to another.
Interrupt tasks are frequently written as one loop. At the beginning is the code
needed to initialize the task, followed by the steps needed to handle the interrupt.
Call the WAIT$FOR$INTERRUPT built-in procedure (see Chapter 10) to generate an
IRET instruction. When the task is activated again, execution continues at the
instruction following the IRET, with all the registers unchanged. At the end of the
interrupt task, use a GOTO statement to loop back to the top of the interrupt task.
Thus, an interrupt task never terminates, unless an operating system function removes
the task.
384
Appendix G
Run-time Interrupt Processing
Use of the WAIT$FOR$INTERRUPT procedure is demonstrated in the following
example. This task is designed to handle messages that arrive in pieces, each one
being preceded by an interrupt.
TASK: DO
DECLARE local variables;
local procedures;
NEW$MESSAGE:
CALL INITIALIZE$MESSAGE$PROCESSING;
DO FOREVER;
CALL WAIT$FOR$INTERRUPT;
/* IRET to wait for next interrupt, which continues here */
CALL PROCESS$PIECE$OF$MESSAGE;
IF LAST$PIECE$OF$MESSAGE$ THEN DO;
CALL TERMINATE$MESSAGE$PROCESSING;
CALL WAIT$FOR$INTERRUPT;
/* IRET to wait for start of next message */
GOTO NEW$MESSAGE
END;
END;
END TASK;
PL/M-386 Programmer's Guide
Appendix G
385
Exception Conditions in REAL Arithmetic
Six exception conditions can occur during normal numeric operations:
•
Invalid operation
•
Denormalized operand
•
Zero divide
•
Overflow
•
Underflow
•
Precision
These exceptions are discussed in the following sections. In each case, only a few of
the possible causes are described because most are not likely to occur with PL/M
usage. To perform sophisticated numeric processing of extreme precision and
flexibility, refer to the microprocessor-specific programmer's reference manual.
The six exceptions and their default responses are summarized in Table G-1.
As the following sections indicate, the masked, default response to most exceptions
will provide the least abrupt, most appropriate action for PL/M applications. Many
real math exceptions that occur in other processors will not occur with the numeric
coprocessor because of the extended range of intermediate results it holds. The soft
recovery of gradual underflow (described in the denormalized exception section) also
extends the range of permissible execution rather than reporting a hard-failure
condition.
386
Appendix G
Run-time Interrupt Processing
Table G-1. Exception and Response Summary
Exception
Masked Response
Unmasked Response
Invalid
Operation
In one operand is NAN, return it;
if both are NAN's return NAN with
larger absolute value; if neither is
NAN, return indefinite NAN.
Request interrupt. (Numeric
coprocessor stack unchanged.)
Zero divide
Return infinity signed with
"exclusive or" of operand signs.
Request interrupt. (Numeric
coprocessor stack unchanged.)
Denormalized
Memory operand: proceed as
usual. Register operand: convert
to valid unnormal, then
reevaluate for then reevaluate for
exceptions.
Request interrupt. (Numeric
coprocessor stack unchanged.)
Overflow
Return properly signed infinity.
Register destination: adjust
exponent, store result (see note),
request interrupt. Memory
destination: request interrupt.
Underflow
Denormalize result.
Register destination: adjust
exponent, store result (see note),
request interrupt. Memory
destination: request interrupt.
Precision
Return rounded result.
Return rounded result, request
interrupt.
Note:
On overflow, 24,576 decimal is subtracted from the true result's exponent. This forces the exponent back into
range and enables a user exception handler to ascertain the true result from the adjusted result that is returned.
On underflow, the same constant is added to the true result's exponent.
Programmers who use the recommended setting of the REAL mode word (see Chapter
10) need to handle only the invalid exception. Study of the other exception
conditions is advised, however, to gain a general understanding of their use.
PL/M-386 Programmer's Guide
Appendix G
387
Invalid Operation Exception
This exception generally indicates a program error. It could be caused by referencing
an uninitialized REAL variable or by referencing a location that does not contain a
REAL value (as might occur with an out-of-range subscript for a REAL array).
Attempting to take the square root of a negative number or to store a number too
large for integer format would also generate this exception.
Another interpretation of this exception is stack error. This may be caused by failing
to restore the math unit status before returning from an interrupt routine that had
saved the status. Another cause is the generation of more than eight intermediate
results during REAL arithmetic, which can arise if REAL procedure function calls are
nested too deeply. The compiler ensures that no single procedure does this, but
cannot check what may occur only at run time. This exception can also occur when
REAL functions (typed procedures) are used as operands within longer REAL
expressions. For example:
DELTA$1 = ALPHA * (BETA/GAMMA) + CHI (PSI, RHO, PI)
where all these names are typed REAL and CHI is some previously declared REAL
function having three parameters.
The following is less likely to cause an exception condition:
EPS = CHI (PSI, RHO, PI)
DELTA$1 = ALPHA * (BETA/GAMMA) + EPS
because all REAL arithmetic is performed using the numeric coprocessor's stack,
which has eight registers. The first seven REAL parameters supplied in procedure
calls are placed on this stack. If the procedure is typed (i.e., invoked as a function), it
can be embedded as one operand within a longer REAL expression.
Because the evaluation of such an expression also involves the use of this stack for
prior and subsequent arithmetic operations, stack overflow may occur. This overflow
amounts to unpredictable destruction of original parameters or intermediate results.
It becomes more likely as the complexity of REAL expressions containing REAL
functions is increased. Thus, it is safer to use an assignment statement first to store
the function's value in a real variable; then use that variable in the larger expression.
If stack error might apply, modify the code for the effected procedures to call the
built-in procedures SAVE$REAL$STATUS and RESTORE$REAL$STATUS as their first
and last operations, respectively.
388
Appendix G
Run-time Interrupt Processing
The masked (default) response is to set the result to one of the special bit patterns
called Not-A-Number (NANs), usually the indefinite value, the smallest NAN
representable in the specified precision. It also sets bit 0 of the REAL error byte.
If bit 0 of the REAL mode word is 0 (invalid exception unmasked), an interrupt
occurs, transferring control to the user-written interrupt handler.
Denormal Operand Exception
This condition arises when numeric operations have resulted in a number whose
exponent is literally zero and whose significand is non-zero, or have resulted in a
number whose significand does not begin with a one. Denormals usually arise in
response to masked underflow. Gradual underflow is the masked, default response to
underflow. A small denormal added to a large normal REAL number can give an
acceptable, in-range answer if the denormal exception is masked. In practice,
denormals are very rare since intermediate results are kept in temporary real format
(15-bit exponent).
This condition causes bit 1 of the REAL error byte to be set to 1. If bit 1 of the REAL
mode word is 1, the response described previously occurs. If bit 1 is 0, an interrupt
occurs, transferring control to the user-written interrupt handler.
Zero Divide Exception
This condition arises when in the course of some REAL computation a divisor turns
out to be zero. The masked response, when bit 2 of the REAL mode word is 1, is to
return infinity appropriately signed. If bit 1 is 0, an interrupt occurs, giving control to
the user-written interrupt handler. In either case, bit 2 of the REAL error byte is set to
1.
Overflow Exception
This error occurs when a real result is too large for the format in use. For assigning
to REAL scalar types, it occurs if the result is greater than about 3.37 x 10**38. For
intermediate REAL computations, it occurs if the result is greater than about
10**4932. The overflow exception can arise during assignment, addition,
subtraction, multiplication, division, or conversion to integer.
The masked, default response (bit 3 of REAL mode word = 1) is to return infinity
(signed if affine mode is set) and set bit 3 of the REAL error byte to 1. Unmasked
overflow must go through a user-written interrupt handler.
PL/M-386 Programmer's Guide
Appendix G
389
Underflow Exception
Underflow exception is caused by an exponent too small for the format in use. For
REAL assignments, it occurs if the exponent is less than -127; and for intermediate
results if the exponent is less than -16383. Underflow can be caused by the same
type of REAL operations as overflow.
The masked, default response (bit 4 of REAL mode word = 1) is to use the denormal
number created by adjusting the very small result. It adjusts the significand, moving
significant digits off to the right and raising the exponent until the latter becomes
non-zero. For example, with single precision values, a 24-bit significand of .01 with
an exponent of zero implies the number 1 x 2**-129 because a zero exponent in this
format means -127. If the denormal exception is masked, this number would be
adjusted into a significand of .001 with an exponent of 1 (i.e., 0.001 x 2**-126), prior
to operation. This number would then be available for use in subsequent REAL
operations that might yield valid results. Zero is returned if it is the rounded result.
Bit 4 of the REAL error byte is set to 1. Unmasked underflow must go through a
user-written interrupt handler.
Precision Exception
This error occurs when the result of an operation is inexact (i.e., rounded) or when an
overflow exception occurs. No special action is performed by a masked response (bit
5 of REAL mode word = 1) other than setting bit 5 of the REAL error byte. Unmasked
response is as chosen by the user.
390
Appendix G
Run-time Interrupt Processing
Writing a Procedure to Handle REAL Interrupts
This section partially summarizes the information pertaining to interrupts,
floating-point usage, and procedures. (Additional facilities for handling REAL
interrupts may be provided by the operating system, or can be performed with the
system builder.)
An interrupt-handling procedure may, for example, begin as follows:
HANDLER: PROCEDURE INTERRUPT PUBLIC;
If HANDLER will do any REAL arithmetic or assignments, its first executable
statements should be of the form:
ERR$INFO = GET$REAL$ERROR;
/* must declare ERR$INFO$ BYTE earlier */
or:
CALL SAVE$REAL$STATUS (@Local_Save_Area);
/* also declare earlier */
Each procedure clears the error byte. The latter procedure also clears out the REAL
stack. Thus, after either procedure is used, the REAL error byte no longer contains the
flagged cause of the exception condition that invoked HANDLER.
Using SAVE$REAL$STATUS is a way of avoiding possible stack errors from
cumulative usage. This enables errors in HANDLER to be detected independently of
the originating exception condition. It also enables HANDLER to restore the state of
the interrupted procedure despite HANDLER's own use of the REAL facility.
SAVE$REAL$STATUS also makes available all the information regarding the state of
the numeric coprocessor exceptions, stack, and operations, as shown in the following
paragraph.
PL/M-386 Programmer's Guide
Appendix G
391
Thus, the beginning of a typical routine to handle REAL exception conditions could
look like this:
HANDLER: PROCEDURE INTERRUPT PUBLIC;
DECLARE ERR$INFO BYTE;
DECLARE LOCAL$SAVE$AREA (94) BYTE;
ERR$INFO = GET$REAL$ERROR;
or, to perform extensive manipulations on the save area, declare a structure
permitting access to the save area's component parts by name and/or byte, as follows:
HANDLER: PROCEDURE INTERRUPT PUBLIC;
DECLARE ERR$INFO BYTE;
DECLARE SAVE$AREA STRUCTURE (
CONTROL(2)
BYTE,
STATUS(2)
BYTE,
TAG
HWORD,
INSTR_OFF
WORD,
INSTR_SEL
SELECTOR,
OPERAND_OFF
WORD,
OPERAND_SEL
SELECTOR,
STACK_TOP(5) WORD,
STACK_ONE(5) WORD,
STACK_TWO(5) WORD,
STACK_3 (5)
WORD,
STACK_4 (5)
WORD,
STACK_5 (5)
WORD,
STACK_6 (5)
WORD,
STACK_7 (5)
WORD);
CALL SAVE$REAL$STATUS (@SAVE_AREA);
ERR$INFO = SAVE_AREA.STATUS(0);
✏
392
Note
To make use of the TAG word, use the masks and shifts to access
the individual fields shown in Figure G-3.
Appendix G
Run-time Interrupt Processing
Call either the SAVE$REAL$STATUS procedure or the GET$REAL$ERROR function,
but not both. If the extra information gained by the save is not needed (i.e., only the
exceptions are needed), use the GET$REAL$ERROR function. If both are called, the
second call will produce incorrect results.
7
15
0
TAG(7) TAG(6) TAG(5) TAG(4) TAG(3) TAG(2) TAG(1) TAG(0)
Tag Values:
00 = Valid (Normal or Unnormal)
01 = Zero (True)
10 = Special (Not-A-Number, , or Denormal)
11 = Empty
OSD544
Figure G-3. Tag Word Format
PL/M-386 Programmer's Guide
Appendix G
393
The rest of HANDLER can perform any appropriate actions. This is an application
dependent decision. Among the possibilities:
•
Incrementing an exception counter for later display
•
Printing diagnostic data (e.g., the contents of SAVE$AREA)
•
Aborting further execution of the calculation causing exception
•
Aborting all further execution
The format of the LOCAL_SAVE_AREA as it is filled by the save procedure is shown
in Figure G-4
The final action prior to returning (if desired) to the interrupted procedure is to
restore the status of the REAL math unit:
CALL RESTORE$REAL$STATUS (@LOCAL_SAVE_AREA);
However, if GET$REAL$ERROR is not used prior to the SAVE$REAL$STATUS call, the
local save area will contain the original contents of the error byte. Under these
circumstances, first clear the lower byte of the saved status word before the RESTORE
statement to avoid retriggering the same exception that invoked HANDLER in the
beginning.
To do so, use a command of the form:
LOCAL_SAVE_AREA (2) = 0;
/* should precede restore */
or:
SAVE_AREA.STATUS (0) = 0;
394
Appendix G
Run-time Interrupt Processing
16 15
31
Next to
Last
Element :
ST(6)
Last
Stack
Element :
ST(7)
Control Word
+0
Reserved
Status Word
+4
Reserved
Tag Word
+8
+ 12
Instruction Pointer Offset
Instruction
Pointer
Operand
Pointer
Top
Stack
Element :
ST
Next
Stack
Element :
ST(1)
0
Reserved
Instr. PTR Selector
Reserved
+ 20
Operand Offset
Reserved
OP Selector
ST(0) Significand 31-0
+ 32
ST(0) Exponent
S
ST(1) Significand 63-48
ST(1) Exponent
+ 44
ST(6) Significand 31-0
+ 88
ST(6) Significand 63-32
+ 92
ST(7) 15-0
S
ST(6) Exponent
ST(7) Exp
+ 96
+ 100
ST(7) 47-16
S
+ 36
+ 40
ST(1) Significand 47-16
S
+ 24
+ 28
ST(0) Significand 63-32
ST(1) Significand 15-0
+ 16
ST(7) 63-48
+ 104
OSD546
Figure G-4. Memory Layout of the REAL Save Area in Protected Mode for the
386 Microprocessor
■■■
PL/M-386 Programmer's Guide
Appendix G
395
396
Appendix G
Run-time Interrupt Processing
Run-time Support for
PL/M Applications
H
In addition to tools that support the software development process, RadiSys provides
run-time support for application programs.
Numeric Coprocessor Support Libraries
Three specific libraries contained in the Intel387 numeric coprocessor support
directory are of use to PL/M-386 programmers. The directory, /intel/ndp387,
contains these libraries:
dc387n.lib/dc387f.lib (near and far) converts floating-point
representations from ASCII decimal format to internal binary format,
and vice versa.
cl387n.lib/cl387f.lib (near and far) is a common elementary function
library that provides an assortment of common elementary functions,
i.e., logarithmic, exponential, trigonometric, and hyperbolic, involving
floating-point numbers, such as rounding.
eh387n.lib/eh387f.lib (near and far) includes floating-point
exception-handling procedures.
For additional information on the libraries contained with Intel387 numeric
coprocessors, see the numeric coprocessor reference manual.
PL/M-386 Programmer's Guide
Appendix H
397
PL/M Support Libraries
The PL/M support libraries contain connection procedures and complex built-ins
written in assembly language. The following support library modules are provided:
Interface to 286 CPU code:
INTERFACE286_FAR
INTERFACE286_NEAR
Math function for double words:
LQ_DWORD_DIVIDE
LQ_DWORD_MULTIPLY
Bit manipulation functions:
MOVBIT
MOVRBIT
SCANBIT
SCANRBIT
■■■
398
Appendix H
Run-time Support for PL/M Applications
Index
@ operator, 39
A
ABS function, 151
ADJUST$RPL function, 189
Algebraic-shift functions, 154
Apostrophe in string, to include, 15
Arithmetic operators, 61, 73
Arrays, 49
ASM interface, 223
assignment, 78
Assignment, 57
AT attribute, 25, 43
Attributes
EXTERNAL, 127
INTERRUPT, 127
PUBLIC, 127
REENTRANT, 127
B
based variables, 36
Based variables, 17, 40
Binary number variables, 32
BITLOCK functions, 166
blanks, 11
Block structure, 103
BLOCKINDWORD, BLOCKINHWORD,
BLOCKINPUT, BLOCKINWORD
procedures, 178
BLOCKOUTDWORD, BLOCKOUTHWORD,
BLOCKOUTPUT, BLOCKOUTWORD
procedures, 179
blocks, 5
BUILD$PTR function, 169
Built-in arrays, 181, 182
PL/M-386 Programmer’s Guide
Built-in procedures and variables, 7
Built-ins, 134
BYTE$SWAP built-in function, 197
C
C language compatibility, 372
cache, clearing, 197
CALL statement, 100, 119
Calling sequence, 363
CARRY flag, 173, 175
CAUSE$INTERRUPT statement, 172
Character set, 9
character strings, 15
Character strings, 15
CLEAR$TASK$SWITCHED$FLAG built-in
procedure, 186
Closed subsystems, 295
CODE control, 218
Comments, 15
Communication between subsystems, 295
COMPACT control, 207, 217, 249
Compilation summary listing example, 257
Compound operands, 60
Concatenate functions, 155
COND control, 218
Constants, 12, 25, 58
CONTROL$REGISTER built-in array, 184
Cross-reference listing example, 256
D
Data attribute declaration, 17
DATA keyword, 21, 25
data types, 30
Data types, 17
DEBUG control, 219
DEBUG$REGISTER built-in array, 184
Index
399
DEC built-in function, 175
Decimal adjust, 175
Declaration statements, 5, 17, 18, 27, 106
Denormal operand exception, 389
Dimension specifier, 23
DISABLE statement, 171
DO block and statement, 5
DO statement, 85
dollar sign, 12
Global descriptor table register, 180
GOTO restrictions, 110
GOTO statement, 99
E
I/O hardware, 178
IABS function, 151
Identifiers, 12
IDTR register, 182, 183
IF control, 220
IF statement, 94
IF|ELSE|ELSEIF|ENDIF controls, 219
Implicit dimension specifier, 23
INCLUDE control, 221
INIT$REAL$MATH$UNIT built-in procedure,
193
INITIAL keyword, 20
Initialization, 21
Input files, 203
INPUT, INHWORD, INWORD functions, 177
Input/Output support, 8
INT function, 147
INTEGER keyword and variables, 33
INTERFACE control, 222
intermodule references, 29
Interrupt
Mechanism, to enable or disable, 171
Processing, 196
Software, to generate, 172
Interrupt descriptor table, 380
Interrupt descriptor table register, 182
Interrupt procedures, 381
Interrupt processing, 379
Invalid operation exception, 388
INVALIDATE$DATA$CACHE built-in
function, 197
INVALIDATE$TLB$ENTRY built-in function,
198
IRET instruction, to generate, 196
EJECT control, 219
ENABLE statement, 171
END statement, 5, 85, 94
Evaluation of expressions, 69
Example of subsystem, 302
Example program, 265
executable statements, 6
Exporting procedures, 300
expressions, 7, 57
Extended segmentation model syntax, 296
EXTERNAL attribute, 107
F
factored declaration, 18
File inclusion with compiler controls, 213
File usage, 203
FIND element functions, 159
Find string mismatch function, 160
Find value in input port function, 177
FIX function, 146
FLAGS function, 175
Flags, hardware, 173, 174
FLAT control, 251
FLOAT function, 146
Floating-point arithmetic, 32
Flow of control, 85
Function references, 60, 119
G
GDT register, 180
GET$ACCESS$RIGHTS function, 186
GET$REAL$ERROR function, 194
GET$SEGMENT$LIMIT function, 187
400
Index
H
HALT statement, 172
I
L
O
label declarations, 28
Label declarations, 17
Languages interface, 217, 222
LAST function, 136
LDT register, 183
LEFTMARGIN control, 207, 212, 227
LENGTH function, 135
Line numbers, 227
Linkage attributes, 103
Linking to modules in other languages, 361
LIST control, 227
Listing example, 257
LITERALLY declarations, 26, 27
local descriptor table register, 183
LOCAL$TABLE variable, 183
Location references, 36, 38, 42, 60
LOCKSET function, 166
Logical operators, 67
Logical-shift functions, 153
Object files, 204
OFFSET function, 151
OFFSET type, 39
OFFSET$OF function, 169
Open subsystems, 294
operands, 57
Operator precedence, 69
OPTIMIZE control, 229
OUTPUT, OUTDWORD, OUTHWORD,
OUTWORD functions, 177
OVERFLOW control, 244
Overflow exception, 389
M
Machine overflow, 244
Machine status register, 184
MACHINE$STATUS built-in variable, 184
Math facility, 190
MEDIUM control, 217, 251
Messages, 307
MINUS operator, 174
MOD486 control, 228
module, 5
MOVB, MOVHW, MOVW procedures, 157
MOVBIT procedure, 163
Move bit patterns right or left, 152
MOVE procedure, 165
MOVRB, MOVRHW, MOVRW procedures,
157
MSW register, 184
N
NIL function, 170
null statement, 89
Number base (binary, decimal, hexadecimal, and
octal), 13
PL/M-386 Programmer’s Guide
P
PAGELENGTH control, 244
PAGEWIDTH control, 245
PAGING control, 245
Parameters, actual and formal, 116
PARITY flag, 174
PLUS operator, 174
POINTER function, 150
POINTER keyword and type, 30, 36
Precision exception, 390
PRINT control, 245
Print files, 204
Privilege level, to adjust, 186, 188
Procedure declarations, 29
Procedure epilogue, 367
procedures, 115
Procedures, 5
Activation, 119
Declaration, 115
Definition, 116
Exit from, 123
Parameters, 116
Scope, 116
Typed, 118
Untyped, 118
Procedures and tasks, 380
Protection architecture of the microprocessor, to
access, 179
PUBLIC attribute, 107
Index
401
R
RAM control, 205, 217, 246
Read string procedure, 178, 180
REAL functions, 146
REAL interrupts, 391
REAL keyword and variables, 33
REAL math facility, 190
Recursion, direct and indirect, 130
Register usage, 368
Registers, 175
Relational operators, 65
Requested privilege level, to adjust, 189
Reserved words, 325
RESET control, 214, 217, 218, 247
RESTORE control, 213, 246
RESTORE$GLOBAL$TABLE built-in
procedure, 181
RESTORE$INTERRUPT$TABLE built-in
procedure, 183
RESTORE$REAL$STATUS built-in procedure,
194
RETURN statement, 123
ROL function, 152
ROM control, 217
ROR function, 152
Rotation functions, 152, 174
RPL, to adjust, 189
Run-time support, 397
S
SAL function, 154
Sample program, 265
SAR function, 154
SAVE control, 246
SAVE$GLOBAL$TABLE built-in procedure,
181
SAVE$INTERRUPT$TABLE built-in
procedure, 182
SAVE$REAL$STATUS built-in procedure, 194
SCANBIT function, 163
scientific notation, 34
SCL built-in function, 174
scope, 103
Scope, 107, 110
SCR built-in function, 174
402
Index
Segment information and accessibility functions,
186, 188
Segment name conventions, 371
SEGMENT$READABLE function, 188
SEGMENT$WRITABLE function, 188
Segmentation controls, 285, 299
SELECTOR function, 150
SELECTOR keyword and type, 36, 39
SELECTOR$OF function, 169
Separators, 11
set command, 203
SET control, 247
SET procedures, 162
SET$REAL$MODE procedure, 193
SHL function, 153
SHLD function, 155
SHR function, 153
SHRD function, 155
SIGN flag, 174
Signed arithmetic, 33
SIGNED function, 147
Signed integer data type built-in function, 146
SIZE function, 136
SKIP functions, 160
SMALL control, 248
Source code, to insert compiler control line, 207
special characters, 11
Stack layout, 384
Stack representation, 363
STACKBASE variable, 176
STACKPTR variable, 176
Statements
CALL, 178, 179, 181, 182, 186, 193, 195,
196
CAUSE$INTERRUPT, 172
DISABLE, 171
ENABLE, 171
HALT, 172
String manipulation procedures and functions,
156
Strings, 15
Structures, 51
Subexpressions, 60
Subscripted variables, 50
Substitution (characters/values/quantities), 26,
27
Subsystems, 283
SUBTITLE control, 252
Support libraries, 397
SYMBOLS control, 206, 252
T
TASK$REGISTER variable, 179
Tasks, 384
temporary-real format, 34
TEST$REGISTER built-in array, 184, 197
TIME procedure, 166
TITLE control, 253
Tokens, 11
Translate string procedure, 161
TYPE control, 253
Type conversion, 78, 137
Typed procedures, 118
U
Underflow exception, 390
underscore, 12
UNSIGN function, 148
Unsigned arithmetic, 32
Unsigned binary data type built-in functions,
149
Untyped procedures, 116
Using subsystems, 289
PL/M-386 Programmer’s Guide
V
Value conversion, 137
Variable declarations, 18
Variable references, 60
W
WAIT$FOR$INTERRUPT built-in procedure,
196
WB$INVALIDATE$DATA$CACHE built-in
function, 198
WMOVB function, 157
WORD16 mapping for built-ins, 197
WORD32|WORD16 control, 253
WORD32|WORD16 type mapping, 46
Work files, 203
Write string procedure, 179
X
XLAT procedure, 161
XREF control, 206, 207, 256
Z
Zero divide exception, 389
ZERO flag, 186
Index
403
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement