Free Pascal |
This is the programmer’s manual for Free Pascal .
It describes some of the peculiarities of the Free Pascal compiler, and provides a glimpse of how the compiler generates its code, and how you can change the generated code. It will not, however, provide a detailed account of the inner workings of the compiler, nor will it describe how to use the compiler (described in the Users guide ). It also will not describe the inner workings of the Run-Time Library (RTL). The best way to learn about the way the RTL is implemented is from the sources themselves.
The things described here are useful when things need to be done that require greater flexibility than the standard Pascal language constructs (described in the Reference guide ).
Since the compiler is continuously under development, this document may get out of date. Wherever possible, the information in this manual will be updated. If you find something which isn’t correct, or you think something is missing, feel free to contact me1.
Free Pascal supports compiler directives in the source file: Basically the same directives as in Turbo Pascal , Delphiand Mac OS pascal compilers. Some are recognized for compatibility only, and have no effect.
A compiler directive is always specified as a comment using curly braces:
{$DIRECTIVE [value]}
The older (* *) notation, or the C++ notation // cannot be used for directives.
The fact that a directive is specified as a comment means that it is not a pascal instruction. That in turn means that if a non-existing directive is specified, the compiler does not give an error, it just gives a warning about an invalid/unknown directive.
There is a distinction between local and global directives:
Some directives can only take a boolean value, a ’+’ to switch them on, or a ’-’ to switch them off. These directives are also known as switches. Many switches have a long form also. If they do, then the name of the long form is given also.
For long switches, the + or - character to switch the option on or off, may be replaced by the ON or OFF keywords.
Thus {$I+} is equivalent to {$IOCHECKS ON} or
{$IOCHECKS +} and
{$C-} is equivalent to {$ASSERTIONS OFF} or
{$ASSERTIONS -}
The long forms of the switches are the same as their Delphi counterparts.
Local directives can occur more than once in a unit or program, If they have a command line counterpart, the command line argument is restored as the default for each compiled file. The local directives influence the compiler’s behaviour from the moment they’re encountered until the moment another switch annihilates their behaviour, or the end of the current unit or program is reached.
The {$ALIGN directive can be used to select the data alignment strategy of the compiler for records. It takes a numerical argument which can be 1, 2, 4, 8, 16 or 32, specifying the alignment boundary in bytes. For these values, it has the same effect as the {$PACKRECORDS} directive (see section ?? ).
Thus, the following
{$A 8}
is equivalent to
{$PACKRECORDS 8}
and specifies to the compiler that all data inside a record should be aligned on 8 byte boundaries.
In MACPAS mode, additionally it can have the following values:
These values are not available in the {$PACKRECORDS} directive.
These directives are the same as the $PACKRECORDS directive (see section ?? ), but they have the alignment specifier embedded in the directive. Thus the following:
{$A8 }
is equivalent to
{$PACKRECORDS 8}
Note that the special cases of $PACKRECORDS cannot be expressed this way.
The {$ASMMODE XXX} directive informs the compiler what kind of assembler it can expect in an asm block. The XXX should be replaced by one of the following:
These switches are local, and retain their value to the end of the unit that is compiled, unless they are replaced by another directive of the same type. The command line switch that corresponds to this switch is -R.
The default assembler reader is the AT&T reader.
By default, the compiler uses shortcut boolean evaluation, i.e., the evaluation of a boolean expression is stopped once the result of the total exression is known with certainty. The {$B } switch can be used to change this behaviour: if its argument is ON, then the compiler will always evaluate all terms in the expression. If it is OFF (the default) then the compiler will only evaluate as many terms as are necessary to determine the result of the complete expression.
So, in the following example, the function Bofu, which has a boolean result, will never get called.
If False and Bofu then ...
A consequence of this is that any additional actions that are done by Bofu are not executed. If compiled with {$B ON}, then BoFu will be called anyway.
The {$ASSERTIONS} switch determines if assert statements are compiled into the binary or not. If the switch is on, the statement
Assert(BooleanExpression,AssertMessage);
Will be compiled in the binary. If the BooleanExpression evaluates to False, the RTL will check if the AssertErrorProc is set. If it is set, it will be called with as parameters the AssertMessage message, the name of the file, the LineNumber and the address. If it is not set, a runtime error 227 is generated.
The AssertErrorProc is defined as
Type
TAssertErrorProc=procedure(Const msg,fname : String;
lineno,erroraddr : Longint);
Var
AssertErrorProc = TAssertErrorProc;
This can be used mainly for debugging purposes. The system unit sets the AssertErrorProc to a handler that displays a message on stderr and simply exits with a run-time error 227. The sysutils unit catches the run-time error 227 and raises an EAssertionFailed exception.
The $BITPACKING directive tells the compiler whether it should use bitpacking or not when it encounters the packed keyword for a structured type. The possible values are ON and OFF. If ON, then the compiler will bitpack structures when it encounters the Packed keyword.
In the following example, the TMyRecord record will be bitpacked:
{$BITPACKING ON}
Type
TMyRecord = packed record
B1,B2,B3,B4 : Boolean;
end;
Note that:
The {$CALLING } directive tells the compiler which calling convention should be used if none is specified:
{$CALLING REGISTER}
By default it is REGISTER. The following calling conventions exist:
For a more detailed explanation of calling conventions, see section ?? . As a special case, DEFAULT can be used, to restore the default calling convention.
The {$CHECKPOINTER} directive turns heap pointer checking on (value ON) or off (value OFF). If heap pointer checking is on and the code is compiled with the -gh (heaptrace) option on, then a check is inserted when dereferencing a pointer. The check will verify that the pointer contains a valid value, i.e. points to a location that is reachable by the program: the stack or a location in the heap. If not, a run-time error 216 or 204 is raised.
If the code is compiled without -gh switch, then this directive has no effect. Note that this considerably slows down the code.
This switch sets the code alignment. It takes an argument which is the alignment in bytes.
{$CODEALIGN 8}
There are some more arguments which can be specified, to tune the behaviour even more. The general form is
{$CODEALIGN PARAM=VALUE}
Where PARAM is the parameter to tune, and VAR value is a numerical value specifying an alignment. PARAM can be one of the following strings:
By default the size of a data structure determines the alignment:
With the above switches the minimum required alignment and a maximum used alignment can be specified. The maximum allowed alignment is only meaningful if it is smaller than the natural size. i.e. setting the maximum alignment (e.g. VARMAX) to 4, the alignment is forced to be at most 4 bytes: The Int64 will then also be aligned at 4 bytes. The SmallInt will still be aligned at 2 bytes.
These values can also be specified on the command line as
-OaPARAM=VALUE
This boolean switch determines whether C like assignment operators are allowed. By default, these assignments are not allowed. After the following statement:
{$COPERATORS ON}
The following operators are allowed:
Var I : Integer; begin I:=1; I+=3; // Add 3 to I and assign the result to I; I-=2; // Substract 2 from I and assign the result to I; I*=2; // Multiply I with 2 and assign the result to I; I/=2; // Divide I with 2 and assign the result to I; end;
The directive
{$DEFINE name}
defines the symbol name. This symbol remains defined until the end of the current module (i.e. unit or program), or until a $UNDEF name directive is encountered.
If name is already defined, this has no effect. Name is case insensitive.
The symbols that are defined in a unit, are not saved in the unit file, so they are also not exported from a unit.
Under Mac Pascal mode, the $DEFINEC directive is equivalent to the $DEFINE directive and is provided for Mac Pascal compatibility.
{$DEFINE} can also be used to define macros or compile-time constants:
{$DEFINE cdecl:=stdcall}
will redefine the standard modifier cdecl as stdcall.
More information about macros and compile-time constants, can be found in section ??
The {$DEFINE} directive has a command-line equivalent, -d
-dNAME
will define the symbol Name. Using
-dcdecl:=stdcall
one can redefine the standard modifier cdecl as stdcall from the command-line.
The {$ELSE} switches between compiling and ignoring the source text delimited by the preceding {$IFxxx} and following {$ENDIF}. Any text after the ELSE keyword but before the brace is ignored:
{$ELSE some ignored text}
is the same as
{$ELSE}
This is useful for indication what switch is meant.
In MACPAS mode, this directive can be used as an alternative to the $ELSE directive. It is supported for compatibility with existing pascal compilers.
This directive can be used as a shortcut for a new {$IF } directive inside an {$ELSE } clause:
{$IF XXX}
// XXX Code here
{$ELSEIF YYY}
// YYY code here
{$ELSE}
// And default code here
{$ENDIF}
is equivalent to
{$IF XXX}
// XXX Code here
{$ELSE }
{$IF YYY}
// YYY code here
{$ELSE}
// And default code here
{$ENDIF}
{$ENDIF}
The directive is followed by an expression like the ones recognized by the {$IF } directive.
The {$ELIFC } variant is allowed only in MACPAS mode.
In MACPAS mode, this directive can be used as an alternative to the $ENDIF directive. It is supported for compatibility with existing pascal compilers.
The {$ENDIF} directive ends the conditional compilation initiated by the last {$IFxxx} directive. Any text after the ENDIF keyword but before the closing brace is ignored:
{$ENDIF some ignored text}
is the same as
{$ENDIF}
This is useful for indication what switch is meant to be ended.
The following code
{$ERROR This code is erroneous !}
will display an error message when the compiler encounters it, and increase the error count of the compiler. The compiler will continue to compile, but no code will be emitted.
The $ERRORC variant is supplied for Mac Pascal compatibility.
This directive is parsed for Delphi compatibility but otherwise ignored. In Delphi, it marks the end of a collapsible region in the IDE.
This directive is parsed for Delphi compatibility but otherwise ignored. A warning will be displayed when this directive is encountered.
This directive is parsed for Delphi compatibility but otherwise ignored.
This directive is recognized for compatibility with Turbo Pascal. Under the 32-bit and 64-bit programming models, the concept of near and far calls have no meaning, hence the directive is ignored. A warning is printed to the screen, as a reminder.
As an example, the following piece of code:
{$F+}
Procedure TestProc;
begin
Writeln ('Hello From TestProc');
end;
begin
testProc
end.
Generates the following compiler output:
malpertuus: >pp -vw testf Compiler: ppc386 Units are searched in: /home/michael;/usr/bin/;/usr/lib/ppc/0.9.1/linuxunits Target OS: Linux Compiling testf.pp testf.pp(1) Warning: illegal compiler switch 7739 kB free Calling assembler... Assembled... Calling linker... 12 lines compiled, 1.00000000000000E+0000
One can see that the verbosity level was set to display warnings.
When declaring a function as Far (this has the same effect as setting it between {$F+} …{$F-} directives), the compiler also generates a warning:
testf.pp(3) Warning: FAR ignored
The same story is true for procedures declared as Near. The warning displayed in that case is:
testf.pp(3) Warning: NEAR ignored
The following code
{$FATAL This code is erroneous !}
will display an error message when the compiler encounters it, and the compiler will immediatly stop the compilation process.
This is mainly useful in conjunction wih {$IFDEF} or {$IFOPT} statements.
This directive selects the type of coprocessor used to do floating point calculations. The directive must be followed by the type of floating point unit. The allowed values depend on the target CPU:
This directive corresponds to the -Cf command line option.
If {$GOTO ON} is specified, the compiler will support Goto statements and Label declarations. By default, $GOTO OFF is assumed. This directive corresponds to the -Sg command line option.
As an example, the following code can be compiled:
{$GOTO ON}
label Theend;
begin
If ParamCount=0 then
GoTo TheEnd;
Writeln ('You specified command line options');
TheEnd:
end.
Remark: When compiling assembler code using the inline assembler readers, any labels used in the assembler code must be declared, and the {$GOTO ON} directive should be used.
If {$LONGSTRINGS ON} is specified, the keyword String (no length specifier) will be treated as AnsiString, and the compiler will treat the corresponding variable as an ansistring, and will generate corresponding code. This switch corresponds to the -Sh command line option.
By default, the use of ansistrings is off, corresponding to {$H-}. The system unit is compiled without ansistrings, all its functions accept shortstring arguments. The same is true for all RTL units, except the sysutils unit, which is compiled with ansistrings.
However, the {$MODE} statement influences the default value of {$H}: a {$MODE DELPHI} directive implies a {$H+} statement, all other modes switch it off. As a result, you should always put {$H+} after a mode directive. This behaviour has changed, in older Free Pascal versions this was not so.
If the generation of hints is turned on, through the -vh command line option or the {$HINTS ON} directive, then
{$Hint This code should be optimized }
will display a hint message when the compiler encounters it.
By default, no hints are generated.
{$HINTS ON} switches the generation of hints on. {$HINTS OFF} switches the generation of hints off. Contrary to the command line option -vh this is a local switch, this is useful for checking parts of the code.
This directive is parsed for Delphi compatibility but otherwise ignored.
The directive {$IF expr} will continue the compilation if the boolean expression expr evaluates to True. If the compilation evaluates to False, then the source is skipped to the first {$ELSE} or {$ENDIF} directive.
The compiler must be able to evaluate the expression at parse time. This means that variables or constants that are defined in the source cannot be used. Macros and symbols may be used, however.
More information on this can be found in the section about conditionals, ?? on page ??.
In MACPAS mode, this directive can be used as an alternative to the $IF directive. It is supported for compatibility with existing pascal compilers.
If the symbol Name is not defined then the {$IFDEF name} will skip the compilation of the text that follows it to the first {$ELSE} or {$ENDIF} directive. If Name is defined, then compilation continues as if the directive wasn’t there.
If the symbol Name is defined then the {$IFNDEF name} will skip the compilation of the text that follows it to the first {$ELSE} or {$ENDIF} directive. If it is not defined, then compilation continues as if the directive wasn’t there.
The {$IFOPT switch} will compile the text that follows it if the switch switch is currently in the specified state. If it isn’t in the specified state, then compilation continues after the corresponding {$ELSE} or {$ENDIF} directive.
As an example:
{$IFOPT M+}
Writeln ('Compiled with type information');
{$ENDIF}
Will compile the Writeln statement only if generation of type information is on.
Remark: The {$IFOPT} directive accepts only short options, i.e. {$IFOPT TYPEINFO} will not be accepted.
The compiler generates an implicit try...finally frame around each procedure that needs initialization or finalization of variables, and finalizes the variables in the finally block. This slows down these procedures (up to 5-10% sometimes). With this directive, the generation of such frames can be disabled. One should be careful with this directive, because it can lead to memory leaks if an exception occurs inside the routine. Therefore, it is set to ON by default.
If the generation of info is turned on, through the -vi command line option, then
{$INFO This was coded on a rainy day by Bugs Bunny}
will display an info message when the compiler encounters it.
This is useful in conjunction with the {$IFDEF} directive, to show information about which part of the code is being compiled.
The {$INLINE ON} directive tells the compiler that the Inline procedure modifier should be allowed. Procedures that are declared inline are copied to the places where they are called. This has the effect that there is no actual procedure call, the code of the procedure is just copied to where the procedure is needed, this results in faster execution speed if the function or procedure is used a lot.
By default, Inline procedures are not allowed. This directive must be specified to use inlined code. The directive is equivalent to the command line switch -Si. For more information on inline routines, consult the Reference guide .
The {$INTERFACES} directive tells the compiler what it should take as the parent interface of an interface declaration which does not explicitly specify a parent interface. By default the Windows COM IUnknown interface is used. Other implementations of interfaces (CORBA or Java) do not necessarily have this interface, and for such cases, this directive can be used. It accepts the following three values:
The {$I-} or {$IOCHECKS OFF} directive tells the compiler not to generate input/output checking code in the program. By default, the compiler generates I/O checking code. This behaviour can be controlled globally with the -Ci switch.
When compiling using the -Ci compiler switch, the Free Pascal compiler inserts input/output checking code after every input/output call in the code. If an error occurred during input or output, then a run-time error will be generated. This switch can also be used to avoid this behaviour.
If no I/O checking code is generated, to check if something went wrong, the IOResult function can be used to see if everything went without problems.
Conversely, {$I+} will turn error-checking back on, until another directive is encountered which turns it off again.
The most common use for this switch is to check if the opening of a file went without problems, as in the following piece of code:
assign (f,'file.txt');
{$I-}
rewrite (f);
{$I+}
if IOResult<>0 then
begin
Writeln ('Error opening file: "file.txt"');
exit
end;
See the IOResult function explanation in Reference guide for a detailed description of the possible errors that can occur when using input/output checking.
This boolean directive switches IEEE (floating point) error checking for constants on or off. It is the local equivalent of the global -C3 command-line switch.
The following turns on IEEE (floating point) error checking for constants:
{$IEEEERRORS ON}
The {$I filename} or {$INCLUDE filename} directive tells the compiler to read further statements from the file filename. The statements read there will be inserted as if they occurred in the current file.
If the file with the given filename exists, it will be included. If no extension is given, the compiler will append the .pp extension to the file and try with that filename. No other extensions are tried.
The filename can be placed between single quotes, they will not be regarded as part of the file’s name. Indeed, if the filename contains a space, then it must be surrounded by single quotes:
{$I 'my file name'}
will try to include the file my file name or my file name.pp.
{$I my file name}
will try to include the file my or my.pp.
If the filename is an asterisk (*) then the compiler will use the unit or program name as filename and try to include that. The following code
unit testi;
interface
{$I *}
implementation
end.
will include the file testi or testi.pp if they exist.
Type A = Integer;
Care should be taken with this mechanism, because the unit name should already match the unit filename, meaning most likely the unit will include itself recursively.
Include files can be nested, but not infinitely deep. The number of files is restricted to the number of file descriptors available to the Free Pascal compiler.
Contrary to Turbo Pascal, include files can cross blocks. I.e. a block can start in one file (with a Begin keyword) and can end in another (with a End keyword). The smallest entity in an include file must be a token, i.e. an identifier, keyword or operator.
The compiler will look for the file to include in the following places:
Directories can be added to the include file search path with the -Fi command line option.
In this form:
{$INCLUDE %XXX%}
the {$INCLUDE} directive inserts a string constant in the source code.
Here XXX can be one of the following:
If XXX is none of the above, then it is assumed to be the name of an environment variable. Its value will be fetched from the environment, if it exists, otherwise an empty string is inserted. As a result, this will generate a macro with the value of the XXX specifier, as if it were a string (or, in the case of LINENUM, an integer).
For example, the following program
Program InfoDemo;
Const User = {$I %USER%};
begin
Write ('This program was compiled at ',{$I %TIME%});
Writeln (' on ',{$I %DATE%});
Writeln ('By ',User);
Writeln ('Compiler version: ',{$I %FPCVERSION%});
Writeln ('Target CPU: ',{$I %FPCTARGET%});
end.
Creates the following output:
This program was compiled at 17:40:18 on 1998/09/09 By michael Compiler version: 0.99.7 Target CPU: i386
This boolean switch tells the compiler whether or not assignments to typed constants are allowed. The default is to allow assignments to typed constants.
The following statement will switch off assignments to typed constants:
{$WRITEABLECONST OFF}
After this switch, the following statement will no longer compile:
Const MyString : String = 'Some nice string'; begin MyString:='Some Other string'; end.
But an initialized variable will still compile:
Var MyString : String = 'Some nice string'; begin MyString:='Some Other string'; end.
The {$L filename} or {$LINK filename} directive tells the compiler that the file filename should be linked to the program. This cannot be used for libraries, see section ?? for that.
The compiler will look for this file in the following locations:
Directories can be added to the object file search path with the -Fo command line option.
On linux systems and on operating systems with case-sensitive filesystems (such as unix systems), the name is case sensitive, and must be typed exactly as it appears on your system.
Remark: Take care that the object file you’re linking is in a format the linker understands. Which format this is, depends on the platform you’re on. Typing ld or ld --help on the command line gives a list of formats ld knows about.
Other files and options can be passed to the linker using the -k command line option. More than one of these options can be used, and they will be passed to the linker, in the order that they were specified on the command line, just before the names of the object files that must be linked.
This directive is recognized for Darwin pascal compilers, but is otherwise ignored.
The {$LINKFRAMEWORK name} will link to a framework named name. This switch is available only on the Darwin platform.
The {$LINKLIB name} will link to a library name. This has the effect of passing -lname to the linker.
As an example, consider the following unit:
unit getlen;
interface
{$LINKLIB c}
function strlen (P : pchar) : longint;cdecl;
implementation
function strlen (P : pchar) : longint;cdecl;external;
end.
If one would issue the command
ppc386 foo.pp
where foo.pp has the above unit in its uses clause, then the compiler would link the program to the c library, by passing the linker the -lc option.
The same can be obtained by removing the linklib directive in the above unit, and specify -k-lc on the command line:
ppc386 -k-lc foo.pp
Note that the linker will look for the library in the linker library search path: one should never specify a complete path to the library. The linker library search path can be set with the -Fl command line option.
For classes that are compiled in the {$M+} or {$TYPEINFO ON} state, the compiler will generate Run-Time Type Information (RTTI). All descendent class of a class that was compiled in the {$M+} state will get RTTI information too. Any class that is used as a field or property in a published section will also get RTTI information.
By default, no Run-Time Type Information is generated for published sections, making them equivalent to public sections. Only when a class (or one of its parent classes) was compiled in the {$M+} state, the compiler will generate RTTI for the methods and properties in the published section.
The TPersistent object that is present in the classes unit (part of the RTL) is generated in the {$M+} state. The generation of RTTI allows programmers to stream objects, and to access published properties of objects, without knowing the actual class of the object.
The run-time type information is accessible through the TypInfo unit, which is part of the Free Pascal Run-Time Library.
Remark: The streaming system implemented by Free Pascal requires that all streamable components be descendent from TPersistent. It is possible to create classes with published sections that do not descend from TPersistent, but those classes will not be streamed correctly by the streaming system of the Classes unit.
In the {$MACRO ON} state, the compiler allows the use of C-style (although not as elaborate) macros. Macros provide a means for simple text substitution. This directive is equivalent to the command line option -Sm. By default, macros are not allowed.
More information on using macros can be found in section ?? .
The {$MAXFPUREGISTERS XXX} directive tells the compiler how much floating point variables can be kept in the floating point processor registers on an Intel X86 processor. This switch is ignored unless the -Or (use register variables) optimization is used.
This is quite tricky because the Intel FPU stack is limited to 8 entries. The compiler uses a heuristic algorithm to determine how much variables should be put onto the stack: in leaf procedures it is limited to 3 and in non leaf procedures to 1. But in case of a deep call tree or, even worse, a recursive procedure, this can still lead to a FPU stack overflow, so the user can tell the compiler how much (floating point) variables should be kept in registers.
The directive accepts the following arguments:
Remark: This directive is valid until the end of the current procedure.
If the generation of info is turned on, through the -vi command line option, then
{$MESSAGE This was coded on a rainy day by Bugs Bunny }
will display an info message when the compiler encounters it. The effect is the same as the {$INFO} directive.
This directive is provided for Delphi compatibility: it has the same effect as the $PACKENUM directive (see section ?? ).
This switch is the equivalent of the -CF command line switch. Normally, the compiler will set the precision of a floating point constant to the minimally required precision to represent it exactly. This switch can be used to ensure that the compiler never lowers the precision belowed the specified value. Supported values are 32, 64 and DEFAULT. 80 is not supported for implementation reasons.
Note that this has nothing to do with the actual precision used by calculations: there the type of the variable will determine what precision is used. This switch determines only with what precision a constant declaration is stored:
{$MINFPCONSTPREC 64}
Const
MyFloat = 0.5;
The type of the above constant will be double even though it can be represented exactly using single.
Note that a value of 80 (Extended precision) is not supported.
Free Pascal supports optimization for the MMX Intel processor (see also chapter ??).
This optimizes certain code parts for the MMX Intel processor, thus greatly improving speed. The speed is noticed mostly when moving large amounts of data. Things that change are
Remark: MMX support is NOT emulated on non-MMX systems, i.e. if the processor doesn’t have the MMX extensions, the MMX optimizations cannot be used.
When MMX support is on, it is not allowed to do floating point arithmetic. It is allowed to move floating point data, but no arithmetic can be done. If floating point math must be done anyway, first MMX support must be switched off and the FPU must be cleared using the emms function of the cpu unit.
The following example will make this more clear:
Program MMXDemo;
uses mmx;
var
d1 : double;
a : array[0..10000] of double;
i : longint;
begin
d1:=1.0;
{$mmx+}
{ floating point data is used, but we do _no_ arithmetic }
for i:=0 to 10000 do
a[i]:=d2; { this is done with 64 bit moves }
{$mmx-}
emms; { clear fpu }
{ now we can do floating point arithmetic }
...
end.
See the chapter on MMX (??) for more information on this topic.
This directive is parsed for Delphi compatibility but is otherwise ignored.
If the generation of notes is turned on, through the -vn command line option or the {$NOTES ON} directive, then
{$NOTE Ask Santa Claus to look at this code}
will display a note message when the compiler encounters it.
{$NOTES ON} switches the generation of notes on. {$NOTES OFF} switches the generation of notes off. Contrary to the command line option -vn this is a local switch, this is useful for checking parts of the code.
By default, {$NOTES} is off.
This boolean switch determines whether code to test the SELF pointer is inserted in methods. By default it is OFF. For example:
{$OBJECTCHECKS ON}
If the SELF pointer is NIL a run-time error 210 (range check) will be generated.
This switch is also activated by the -CR command line option.
This switch enables optimization. It can have the following possible values:
The following strings are supported:
Example:
{$OPTIMIZATION ON}
is equivalent to
{$OPTIMIZATION LEVEL2}
Multiple optimizations can be specified with:
{$OPTIMIZATION REGVAR,SIZE,LEVEL2}
This switch is also activated by the -Ooxxx command line switch. Note the small ’o’: it is -Oo followed by the switch name.
This directive tells the compiler the minimum number of bytes it should use when storing enumerated types. It is of the following form:
{$PACKENUM xxx}
{$MINENUMSIZE xxx}
Where the form with $MINENUMSIZE is for Delphi compatibility. xxx can be one of 1,2 or 4, or NORMAL or DEFAULT.
The default enumeration size depends on the compiler mode:
As an alternative form one can use {$Z1}, {$Z2} {$Z4}. The {$Z} form takes a boolean argument, where ON is equivalent to {$Z4} and OFF is equivalent to {$Z1}.
So the following code
{$PACKENUM 1}
Type
Days = (monday, tuesday, wednesday, thursday, friday,
saturday, sunday);
will use 1 byte to store a variable of type Days, whereas it nomally would use 4 bytes. The above code is equivalent to
{$Z1}
Type
Days = (monday, tuesday, wednesday, thursday, friday,
saturday, sunday);
or equivalent to
{$Z OFF}
Type
Days = (monday, tuesday, wednesday, thursday, friday,
saturday, sunday);
This directive controls the byte alignment of the elements in a record, object or class type definition.
It is of the following form:
{$PACKRECORDS n}
Where n is one of 1, 2, 4, 8, 16, C, NORMAL or DEFAULT. This means that the elements of a record which have size greater than n will be aligned on n byte boundaries. Elements with size less than or equal to n will be aligned to a natural boundary, i.e. to a power of two that is equal to or larger than the element’s size. The special value C is used to specify alignment as by the GNU CC compiler. It should be used only when making import units for C routines.
The default alignment (which can be selected with DEFAULT) is natural alignment, contrary to Turbo Pascal, where it is 1.
More information on this and an example program can be found in the reference guide, in the section about record types.
The following shorthands can be used for this directive:
{$A1 }
{$A2 }
{$A4 }
{$A8 }
The $PACKSET directive takes a numeric argument of 1, 2, 4 or 8. This number determines the number of bytes used to store a set: The compiler rounds the number of bytes needed to store the set down/up to the closest multiple of the PACKSET setting, with the exception that 3-byte sets are always rounded up to 4-byte sets.
Other allowed values are FIXED, DEFAULT, or NORMAL. With these values, the compiler stores sets with less than 32 elements in 4 bytes, and sets with less than 256 elements in 32 bytes.
The $POP directive restores the values of all local compiler directives with the last values that were stored on the settings stack. The settings are then deleted from the stack.
The settings can be stored on the stack with the $PUSH directive (see section ?? ).
Note that global settings are not restored by this directive.
The $PUSH directive saves the values of all local compiler directives that were stored on the settings stack. Up to 20 sets of settings can be stored on the stack.
The settings can be restored from the stack using the $POP directive (see section ?? ).
Note that global [settings (search paths etc.) are not saved by this directive.
The settings stack is preserved accross units, i.e. when the compiler starts compiling a new unit, the stack is not emptied.
The {$Q+} or {$OV+} (MACPAS mode only) or {$OVERFLOWCHECKS ON} directive turns on integer overflow checking. This means that the compiler inserts code to check for overflow when doing computations with integers. When an overflow occurs, the run-time library will generate a run-time error 215: It prints a message Overflow at xxx, and exits the program with exit code 215.
Remark: Overflow checking behaviour is not the same as in Turbo Pascal since all arithmetic operations are done via 32-bit or 64-bit values. Furthermore, the Inc() and Dec standard system procedures are checked for overflow in Free Pascal , while in Turbo Pascal they are not.
Using the {$Q-} switch (or the {$OV-} switch in MACPAS mode) switches off the overflow checking code generation.
The generation of overflow checking code can also be controlled using the -Co command line compiler option (see the Users guide ).
In Delphi, overflow checking is only switchable on a procedure level. In Free Pascal, the {$Q } directive can be used on an expression-level.
By default, the compiler doesn’t generate code to check the ranges of array indices, enumeration types, subrange types, etc. Specifying the {$R+} switch tells the computer to generate code to check these indices. If, at run-time, an index or enumeration type is specified that is out of the declared range of the compiler, then a run-time error is generated, and the program exits with exit code 201. This can happen when doing a typecast (implicit or explicit) on an enumeration type or subrange type.
The {$RANGECHECKS OFF} switch tells the compiler not to generate range checking code. This may result in faulty program behaviour, but no run-time errors will be generated.
Remark: The standard functions val and Read will also check ranges when the call is compiled in {$R+} mode.
In Delphi, range checking is only switchable on a procedure level. In Free Pascal, the {$R } directive can be used on an expression-level.
The $REGION directive is recognised for Delphi compatibility only. In Delphi it serves to mark the beginning of a collapsible region in the IDE.
This directive includes a resource in the binary. The argument to this directive is the resource file to include in the binary:
{$R icons.res}
Will include the file icons.res as a resource in the binary. Up to version 2.2.N, resources are supported only for Windows (native resources are used) and for platforms using ELF binaries (linux, BSD). As of version 2.3.1, resources have been implemented for all supported platforms.
The asterix can be used as a placeholder for the current unit/program filename:
unit myunit;
{$R *.res}
will include myunit.res.
This works only on the intel compiler, and MMX support must be on ({$MMX +}) for this to have any effect. See the section on saturation support (section ?? ) for more information on the effect of this directive.
This boolean directives controls how the compiler generates code to store FPU values. If set to ON, the compiler inserts a FWAIT opcode after the store of a floating point value, so any errors are reported at once. This slows down code, but makes sure that the errors are reported on the location where the instruction was executed.
The boolean $SCOPEDENUMS directive controls how enumeration type values are inserted in the symbol tables. In its default state (OFF) the enumeration values are inserted directly in the symbol table of the current scope. If the directive is set to ON then the values are inserted inside a scope with the name of the enumeration type.
Practically this means that the following is the default behaviour:
{$SCOPEDENUMS OFF}
Type
TMyEnum = (one,two,three);
Var
A : TMyEnum;
begin
A:=one;
end.
The value one can be referenced directly. The following will give an error:
begin A:=TMyEnum.one; end.
However, if the SCOPEDENUMS directive is set to ON, then the assignment must be made as follows:
{$SCOPEDENUMS ON}
Type
TMyEnum = (one,two,three);
Var
A : TMyEnum;
begin
A:=TMyEnum.one;
end.
i.e. the value must be prefixed with the type name.
In MACPAS mode, this directive can be used to define compiler symbols. It is an alternative to the $DEFINE directive for macros. It is supported for compatibility with existing Pascal compilers. It will define a symbol with a certain value (called a compiler variable expression).
The expression syntax is similar to expressions used in macros, but the expression must be evaluated at compile-time by the compiler. This means that only some basic arithmetic and logical operators can be used, and some extra possibilities such as the TRUE,FALSE and UNDEFINED operators:
{$SETC TARGET_CPU_PPC := NOT UNDEFINED CPUPOWERPC}
{$SETC TARGET_CPU_68K := NOT UNDEFINED CPUM68K}
{$SETC TARGET_CPU_X86 := NOT UNDEFINED CPUI386}
{$SETC TARGET_CPU_MIPS := FALSE}
{$SETC TARGET_OS_UNIX := (NOT UNDEFINED UNIX) AND (UNDEFINED DARWIN)}
The := assignment symbol may be replaced with the = symbol.
Note that this command works only in MACPAS mode, but independent of the -Sm command line option or {$MACRO } directive.
If you specify the {$STATIC ON} directive, then Static methods are allowed for objects. Static objects methods do not require a Self variable. They are equivalent to Class methods for classes. By default, Static methods are not allowed. Class methods are always allowed. Note that also static fields can be defined.
This directive is equivalent to the -St command line option.
The following code
{$STOP This code is erroneous !}
will display an error message when the compiler encounters it. The compiler will immediatly stop the compilation process.
It has the same effect as the {$FATAL} directive.
This directive is recognized for Delphi compatibility, but is currently ignored. In Delphi, it controls the generation of code that checks the sanity of string variables and arguments.
In the {$T+} or {$TYPEDADDRESS ON} state, the @ operator, when applied to a variable, returns a result of type ^T, if the type of the variable is T. In the {$T-} state, the result is always an untyped pointer, which is assignment compatible with all other pointer types.
For example, the following code will not compile:
{$T+}
Var
I : Integer;
P : PChar;
begin
P:=@I;
end.
The compiler will give a type mismatch error:
testt.pp(8,6) Error: Incompatible types: got "^SmallInt" expected "PChar"
By default however, the address operator returns an untyped pointer.
The directive
{$UNDEF name}
un-defines the symbol name if it was previously defined. Name is case insensitive.
In Mac Pascal mode, $UNDEFC is equivalent to $UNDEF, and is provided for Mac Pascal compatibility.
The {$VARSTRINGCHECKS } determines how strict the compiler is when checking string type compatibility for strings passed by reference. When in the + or ON state, the compiler checks that strings passed as parameters are of the string type as the declared parameters of the procedure.
By default, the compiler assumes that all short strings are type compatible. That is, the following code will compile:
Procedure MyProcedure(var Arg: String[10]);
begin
Writeln('Arg ',Arg);
end;
Var
S : String[12];
begin
S:='123456789012';
Myprocedure(S);
end.
The types of Arg and S are strictly speaking not compatible: The Arg parameter is a string of length 10, and the variable S is a string of length 12: The value will be silently truncated to a string of length 10.
In the {$V+} state, this code will trigger a compiler error:
testv.pp(14,16) Error: string types doesn't match, because of $V+ mode
Note that this is only for strings passed by reference, not for strings passed by value.
The {$W} switch directive controls the generation of stackframes. In the on state, the compiler will generate a stackframe for every procedure or function.
In the off state, the compiler will omit the generation of a stackframe if the following conditions are satisfied:
If these conditions are satisfied, the stack frame will be omitted.
If the compiler encounters a
{$WAIT}
directive, it will resume compiling only after the user has pressed the enter key. If the generation of info messages is turned on, then the compiler will display the following message:
Press <return> to continue
before waiting for a keypress.
Remark: This may interfere with automatic compilation processes. It should be used only for compiler debugging purposes.
This directive allows to selectively turn on or off the emission of warnings. It takes the following form
{$WARN IDENTIFIER ON}
{$WARN IDENTIFIER OFF}
{$WARN IDENTIFIER +}
{$WARN IDENTIFIER -}
{$WARN IDENTIFIER ERROR}
ON or + turns on emission of the warning. The OFF or - values suppress the warning. ERROR promotes the warning to an error, and the compiler will treat it as such.
The IDENTIFIER is the name of a warning message. The following names are recognized:
Besides the above text identifiers, the identifier can also be a message number. The numbers of the messages are displayed when the -vq command-line option is used.
If the generation of warnings is turned on, through the -vw command line option or the {$WARNINGS ON} directive, then
{$WARNING This is dubious code}
will display a warning message when the compiler encounters it.
{$WARNINGS ON} switches the generation of warnings on. {$WARNINGS OFF} switches the generation of warnings off. Contrary to the command line option -vw this is a local switch, this is useful for checking parts of your code.
By default, no warnings are emitted.
This switch is an equivalent of the {$PACKENUM } switch (see section ?? ).
Global directives affect the whole of the compilation process. That is why they also have a command line counterpart. The command line counterpart is given for each of the directives. They must be specified before the unit or program clause in a source file, or they will have no effect.
Used on the PALM os only, it can be set to specify the application name, which can be viewed on the Palm only. This directive only makes sense in a program source file, not in a unit.
{$APPID MyApplication}
Used on the PALM os only, it can be set to specify the application name which can be viewed on the Palm only. This directive only makes sense in a program source file, not in a unit.
{$APPNAME My Application, compiled using Free Pascal.}
This directive is currently only supported on the following targets: Win32, Mac, OS2 and AmigaOS. On other targets, the directive is ignored.
The {$APPTYPE XXX} accepts one argument which specifies what kind of application is compiled. It can have the following values:
Note that on such applications cannot take command line options, nor return a result code. They will run in a special terminal window, implemented as a SIOW application, see the MPW documentation for details.
On os/2 , these applications can run both full-screen and in a terminal window.
linux applications are always console applications. The application itself can decide to close the standard files, though.
On os/2 and , the GUI application type creates a GUI application, as on Windows. On os/2 , this is a real Presentation Manager application.
Care should be taken when compiling GUI applications; the Input and Output files are not available in a GUI application, and attempting to read from or write to them will result in a run-time error.
It is possible to determine the application type of a Windows or Amiga application at runtime. The IsConsole constant, declared in the Win32 and Amiga system units as
Const IsConsole : Boolean;
contains True if the application is a console application, False if the application is a GUI application.
This directive allows specifying the default calling convention used by the compiler, when no calling convention is specified for a procedure or function declaration. It can be one of the following values:
This directive is equivalent to the -Cc command line option.
This switch sets the codepage of the rest of the source file. The codepage is only taken into account when interpreting literal strings, the actual code must be in US-ASCII. The argument to this switch is the name of the code page to be used.
{$CODEPAGE UTF8}
The ’UTF-8’ codepage can be specified as ’UTF-8’ or ’UTF8’. The list of supported codepages is the list of codepages supported by the charset unit of the RTL.
This is intended for the NETWARE version of the compiler: it specifies the copyright information that can be viewed on a module for a Netware OS.
For example:
{$COPYRIGHT GNU copyleft. compiled using Free Pascal}
When this switch is on, the compiler inserts GNU debugging information in the executable. The effect of this switch is the same as the command line switch -g.
By default, insertion of debugging information is off.
This switch is recognised for compatibility only, but is ignored completely by the compiler. At a later stage, this switch may be activated.
This directive controls the emulation of the coprocessor. There is no command line counterpart for this directive.
When this switch is enabled, all floating point instructions which are not supported by standard coprocessor emulators will give out a warning.
The compiler itself doesn’t do the emulation of the coprocessor.
To use coprocessor emulation under dos (go32v2) you must use the emu387 unit, which contains correct initialization code for the emulator.
Under linux and most unix ’es, the kernel takes care of the coprocessor support, so this switch is not necessary on those platforms.
When the switch is on, no floating point opcodes are emitted by the code generator. Instead, internal run-time library routines are called to do the necessary calculations. In this case all real types are mapped to the single IEEE floating point type.
Remark: By default, emulation is on for non-unix targets. For unix targets, floating point emulation (if required) is handled by the operating system, and by default it is off.
This directive can appear in a program or library source file: it sets the extension for the generated binary/library. The initial dot (sometimes associated with the extension) may not be present.
Example:
{$EXTENSTION 'cgi'}
will set the extension to ’cgi’.
This option serves to specify the framework search path on Darwin, where the compiler looks for framework files. Used as
{$FRAMEWORKPATH XXX}
it will add XXX to the framework path. The value XXX can contain one or more paths, separated by semi-colons or colons.
This option is recognised for Turbo Pascal compatibility, but is ignored, since the compiler works only on 32-bit and 64-bit processors.
This option can be used to set the base location for a DLL on windows-based systems. The directive needs a memory location as an option (use $ to specify a hexadecimal value). It is the equivalent of the -WB command-line option.
The following sets the base location to 0x00400000
{$IMAGEBASE $00400000}
This option serves to specify the include path, where the compiler looks for include files. Used as
{$INCLUDEPATH XXX}
it will add XXX to the include path. The value XXX can contain one or more paths, separated by semi-colons or colons.
For example:
{$INCLUDEPATH ../inc;../i386}
{$I strings.inc}
will add the directories ../inc and ../i386 to the include path of the compiler. The compiler will look for the file strings.inc in both these directories, and will include the first found file. This directive is equivalent to the -Fi command line switch.
Caution is in order when using this directive: If you distribute files, the places of the files may not be the same as on your machine; moreover, the directory structure may be different. In general it would be fair to say that you should avoid using absolute paths. Instead, one should use relative paths only, as in the example above.
This switch (not to be confused with the local {$L file} file linking directive) is recognised for Turbo Pascal compatibility, but is ignored. Generation of symbol information is controlled by the $D switch.
The LIBPREFIX directive has a similar function as the {$EXTENSION } and {$LIBSUFFIX } compiler directives: It sets the prefix of the library. This is by default ’lib’ on unices, and empty on windows. The name is simply prepended to the filename.
Example:
library tl;
{$LIBPREFIX 'library'}
begin
end.
will result in a filename librarytl.so on linux, or librarytl.dll on windows.
This option serves to specify the library path, where the linker looks for static or dynamic libraries. {$LIBRARYPATH XXX} will add XXX to the library path. XXX can contain one or more paths, separated by semi-colons or colons.
For example:
{$LIBRARYPATH /usr/X11/lib;/usr/local/lib}
{$LINKLIB X11}
will add the directories /usr/X11/lib and /usr/local/lib to the linker library path. The linker will look for the library libX11.so in both these directories, and use the first found file. This directive is equivalent to the -Fl command line switch.
Caution is in order when using this directive: If you distribute files, the places of the libraries may not be the same as on your machine; moreover, the directory structure may be different. In general it would be fair to say that you should avoid using this directive. If you are not sure, it is better practice to use makefiles and makefile variables.
The LIBSUFFIX directive has a similar function as the {$EXTENSION } and {$LIBPREFIX } compiler directives. directive: it sets the suffix of the library. This is usually used for version numbers: it is simply added to the outputfilename, before the extension.
Example:
library tl;
{$LIBSUFFIX '-1.2.3'}
begin
end.
will result in a filename libtl-1.2.3.so on linux, or libtl-1.2.3.dll on windows.
The {$MINSTACKSIZE} sets the maximum stack size for an executable on Windows-based systems. It needs an argument, the size (in bytes) of the stack. The maximum value is $7FFFFFFF, the minimum value is 2048 or the value of $MINSTACKSIZE if it was specified.
The following example sets the maximum stack size to$FFFFFF bytes:
{$MINSTACKSIZE $FFFFFF}
This switch can be used to set the heap and stacksize. Its format is as follows:
{$M StackSize,HeapSize}
where StackSize and HeapSize should be two integer values, greater than 1024. The first number sets the size of the stack, and the second the size of the heap. The stack size setting is ignored on Unix platforms unless stack checking is enabled: in that case the stack checking code will use the size set here as maximum stack size.
On those systems, in addition to the stack size set here, the operating system or the run environment may have set other (possibly more strict) limits on stack size using the OS’es ulimit system calls.
The two numbers can be set on the command line using the -Ch and -Cs switches.
The {$MINSTACKSIZE} sets the minimum stack size for an executable on Windows-based systems. It needs an argument, the size (in bytes) of the stack. This must be a number larger than 1024.
The following example sets the minimum stack size to 2048 bytes:
{$MINSTACKSIZE 2048}
The {$MODE} sets the compatibility mode of the compiler. This is equivalent to setting one of the command line options -So, -Sd, -Sp or -S2. it has the following arguments:
For an exact description of each of these modes, see appendix ??, on page ??.
As of FPC 2.3.1, the {$MODESWITCH} directive selects some of the features that a {$MODE } directive selects: it can be used to use features that would otherwise not be available in the current mode. For instance, one wishes to program in TP mode, but would like to use the ’Out’ parameter, an option available only in Delphi mode. The {$MODESWITCH } directive allows to activate or deactivate some individual mode features, while not changing the current compiler mode.
This switch is a global switch, and can be used wherever the {$MODE} switch can be used.
The syntax is as follows:
{$MODESWITCH XXX}
{$MODESWITCH XXX+}
{$MODESWITCH XXX-}
The first two will switch on feature XXX, the last one will switch it off.
The feature XXX can be one of the following:
Hence, the following:
{$MODE TP}
{$MODESWITCH OUT}
Will switch on the support for the out parameter type in TP mode. It is equivalent to
{$MODE TP}
{$MODESWITCH OUT+}
This switch is recognised for Turbo Pascal compatibility, but is otherwise ignored, since the compiler always uses the coprocessor for floating point mathematics.
In earlier versions of FPC, this switch was recognised for Turbo Pascal compatibility, but was otherwise ignored: The concept of overlay code is not needed in 32-bit or 64-bit programs.
In newer versions of FPC (certainly as of 2.0.0), this switch became a Delphi compatible switch: it has the same meaning as the {$OPTMIZATIONS ON/OFF} switch, switching on or off level 2 optimizations.
See section ?? on page ?? for more explanations and more detailed optimization settings.
This option serves to specify the object path, where the compiler looks for object files. {$OBJECTPATH XXX} will add XXX to the object path. XXX can contain one or more paths, separated by semi-colons or colons.
For example:
{$OBJECTPATH ../inc;../i386}
{$L strings.o}
will add the directories ../inc and ../i386 to the object path of the compiler. The compiler will look for the file strings.o in both these directories, and will link the first found file in the program. This directive is equivalent to the -Fo command line switch.
Caution is in order when using this directive: If you distribute files, the places of the files may not be the same as on your machine; moreover, the directory structure may be different. In general it would be fair to say that you should avoid using absolute paths, instead use relative paths, as in the example above. Only use this directive if you are certain of the places where the files reside. If you are not sure, it is better practice to use makefiles and makefile variables.
If this switch is on, all function or procedure parameters of type string are considered to be open string parameters; this parameter only has effect for short strings, not for ansistrings.
When using openstrings, the declared type of the string can be different from the type of string that is actually passed, even for strings that are passed by reference. The declared size of the string passed can be examined with the High(P) call.
By default, the use of openstrings is off.
The {$PASCALMAINNAME NNN} directive sets the assembler symbol name of the program or library entry point to NNN. This directive is the equivalent of the -XM command line switch.
Under normal circumstances, it should not be necessary to use this switch.
The {$PIC } directive takes a boolean argument and tells the compiler whether it should generate PIC (Position Indepedent Code) or not. This directive is the equivalent of the -Cg command line switch.
This directive is only useful on Unix platforms: Units should be compiled using PIC code if they are supposed to be in a library. For programs, using PIC code is not needed, but it doesn’t hurt either (although PIC code is slower).
The following
{$PIC ON}
unit MyUnit;
tells the compiler to compile myunit using PIC code.
This boolean directive enables or disables the use of pointer arithmatics in expressions nivolving pointers. When enabled, it allows to take the difference of 2 pointers, or to add an integer value to pointers? By default, POINTERMATH is on.
The following
{$POINTERMATH OFF}
unit MyUnit;
tells the compiler to give an error whenever pointer math appears in an expression.
This directive turns the generation of profiling code on (or off). It is equivalent to the -gp command line option. Default is OFF. This directive only makes sense in a program source file, not in a unit.
The {$S+} directive tells the compiler to generate stack checking code. This generates code to check if a stack overflow occurred, i.e. to see whether the stack has grown beyond its maximally allowed size. If the stack grows beyond the maximum size, then a run-time error is generated, and the program will exit with exit code 202.
Specifying {$S-} will turn generation of stack-checking code off.
The command line compiler switch -Ct has the same effect as the {$S+} directive.
By default, no stack checking is performed.
Remark: Stack checking can only be used to provide help during debugging, to try and track routines that use an excessive amount of local memory. It is not intended and cannot be used to actually safely handle such errors. It does not matter whether the error handling is through exception handling or otherwise.
When a stack error occurs, this is a fatal error and the application cannot be kept running correctly, neither in a production environment, nor under debugging.
This directive can be used for the Novell netware targets to specify the screen name. The argument is the screen name to be used.
{$SCREENNAME My Nice Screen}
Will set the screenname of the current application to ’My Nice Screen’.
The $SETPEFLAGS executable sets the PE flags on Windows. These flags are written to the binary PE header file. It expects a numerical constant as an argument, this constant will be written to the header block.
Under normal circumstances, the compiler itself will determine the values of the PE flags to write to the binary file, but this directive can be used to override the compiler’s behaviour.
A unit that is compiled in the {$SMARTLINK ON} state will be compiled in such a way that it can be used for smartlinking. This means that the unit is chopped in logical pieces: each procedure is put in its own object file, and all object files are put together in a big archive. When using such a unit, only the pieces of code that you really need or call will be linked in your program, thus reducing the size of your executable substantially.
Beware: using smartlinked units slows down the compilation process, because a separate object file must be created for each procedure. If you have units with many functions and procedures, this can be a time consuming process, even more so if you use an external assembler (the assembler is called to assemble each procedure or function code block separately).
The smartlinking directive should be specified before the unit declaration part:
{$SMARTLINK ON}
Unit MyUnit;
Interface
...
This directive is equivalent to the -CX command line switch.
This directive sets the system calling convention to use on a MorphOS or Amiga system. The directive needs an argument, which should be one of the following:
The directive will generate a warning if used on another system.
This directive can be set to specify the thread name when compiling for Netware.
This option serves to specify the unit path, where the compiler looks for unit files. {$UNITPATH XXX} will add XXX to the unit path. XXX can contain one or more paths, separated by semi-colons or colons.
For example:
{$UNITPATH ../units;../i386/units}
Uses strings;
will add the directories ../units and ../i386/units to the unit path of the compiler. The compiler will look for the file strings.ppu in both these directories, and will link the first found file in the program. This directive is equivalent to the -Fu command line switch.
Caution is in order when using this directive: If you distribute files, the places of the files may not be the same as on your machine; moreover, the directory structure may be different. In general it would be fair to say that you should avoid using absolute paths, instead use relative paths, as in the example above. Only use this directive if you are certain of the places where the files reside. If you are not sure, it is better practice to use makefiles and makefile variables.
Note that this switch does not propagate to other units, i.e. it’s scope is limited to the current unit.
This boolean directive is meant to import COM interfaces. Sometimes COM interfaces have property setters which accept arguments that are not by value, but by reference. These setters are normally forbidden. This flag enables the use of property setters with var, const, out arguments. By default it is OFF. The effect is on interface declarations, but also on class definitions.
The following example only compiles in the ON state:
{$VARPROPSETTER ON}
Type
TMyInterface = Interface
Procedure SetP(Var AValue : Integer);
Function GetP : Integer;
Property MyP : Integer Read GetP Write SetP;
end;
In the OFF state, the following error will be generated:
testvp.pp(7,48) Error: Illegal symbol for property access
On Windows , this can be used to specify a version number for a library. This version number will be used when the library is installed, and can be viewed in the Windows Explorer by opening the property sheet of the DLL and looking on the tab ’Version’. The version number consists of minimally one, maximum 3 numbers:
{$VERSION 1}
Or:
{$VERSION 1.1}
And even:
{$VERSION 1.1.1}
This cannot yet be used for executables on Windows, but may be activated in the future.
This switch is parsed for Delphi compatibility but is otherwise ignored. The compiler will write a warning when it is encountered.
Extended syntax allows you to drop the result of a function. This means that you can use a function call as if it were a procedure. By default this feature is on. You can switch it off using the {$X-} or {$EXTENDEDSYNTAX OFF}directive.
The following, for instance, will not compile:
function Func (var Arg : sometype) : longint;
begin
... { declaration of Func }
end;
...
{$X-}
Func (A);
The reason this construct is supported is that you may wish to call a function for certain side-effects it has, but you don’t need the function result. In this case you don’t need to assign the function result, saving you an extra variable.
The command line compiler switch -Sa1 has the same effect as the {$X+} directive.
By default, extended syntax is assumed.
This switch controls the generation of browser information. It is recognized for compatibility with Turbo Pascal and Delphi only, as Browser information generation is not yet fully supported.
The Free Pascal compiler supports conditionals as in normal Turbo Pascal , Delphior Mac OS Pascal. It does, however, more than that. It allows you to make macros which can be used in your code, and it allows you to define messages or errors which will be displayed when compiling. It also has support for compile-time variables and compile-time expressions, as commonly found in compilers.
The various conditional compilation directives ($IF, $IFDEF, $IFOPT are used in combination with $DEFINE to allow the programmer to choose at compile time which portions of the code should be compiled. This can be used for instance
These options are then chosen when the program is compiled, including or excluding parts of the code as needed. This is opposed to using normal variables and running through selected portions of code at run time, in which case extra code is included in the executable.
The rules for using conditional symbols are the same as under Turbo Pascal or Delphi. Defining a symbol goes as follows:
{$define Symbol}
From this point on in your code, the compiler knows the symbol Symbol. Symbols are, like the Pascal language, case insensitive.
You can also define a symbol on the command line. the -dSymbol option defines the symbol Symbol. You can specify as many symbols on the command line as you want.
Undefining an existing symbol is done in a similar way:
{$undef Symbol}
If the symbol didn’t exist yet, this doesn’t do anything. If the symbol
existed previously, the symbol will be erased, and will not be recognized
any more in the code following the {$undef ...} statement.
You can also undefine symbols from the command line with the -u command line switch.
To compile code conditionally, depending on whether a symbol is defined or
not, you can enclose the code in a {$ifdef Symbol} …{$endif}
pair. For instance the following code will never be compiled:
{$undef MySymbol}
{$ifdef Mysymbol}
DoSomething;
...
{$endif}
Similarly, you can enclose your code in a {$ifndef Symbol} …{$endif}
pair. Then the code between the pair will only be compiled when the used
symbol doesn’t exist. For example, in the following code, the call to the
DoSomething will always be compiled:
{$undef MySymbol}
{$ifndef Mysymbol}
DoSomething;
...
{$endif}
You can combine the two alternatives in one structure, namely as follows
{$ifdef Mysymbol}
DoSomething;
{$else}
DoSomethingElse
{$endif}
In this example, if MySymbol exists, then the call to DoSomething will be compiled. If it doesn’t exist, the call to DoSomethingElse is compiled.
The Free Pascal compiler defines some symbols before starting to compile your program or unit. You can use these symbols to differentiate between different versions of the compiler, and between different compilers. To get all the possible defines when starting compilation, see appendix ??
Remark: Symbols, even when they’re defined in the interface part of a unit, are not available outside that unit.
Macros are very much like symbols or compile-time variables in their syntax, the difference is that macros have a value whereas a symbol simply is defined or is not defined. Furthermore, following the definition of a macro, any occurrence of the macro in the pascal source will be replaced with the value of the macro (much like the macro support in the C preprocessor). If macro support is required, the -Sm command line switch must be used to switch it on, or the directive must be inserted:
{$MACRO ON}
otherwise macros will be regarded as a symbol.
Defining a macro in a program is done in the same way as defining a symbol; in a {$define} preprocessor statement1:
{$define ident:=expr}
If the compiler encounters ident in the rest of the source file, it will be replaced immediately by expr. This replacement works recursive, meaning that when the compiler expanded one macro, it will look at the resulting expression again to see if another replacement can be made. This means that care should be taken when using macros, because an infinite loop can occur in this manner.
Here are two examples which illustrate the use of macros:
{$define sum:=a:=a+b;}
...
sum { will be expanded to 'a:=a+b;'
remark the absence of the semicolon}
...
{$define b:=100}
sum { Will be expanded recursively to a:=a+100; }
...
The previous example could go wrong:
{$define sum:=a:=a+b;}
...
sum { will be expanded to 'a:=a+b;'
remark the absence of the semicolon}
...
{$define b=sum} { DON'T do this !!!}
sum { Will be infinitely recursively expanded... }
...
On my system, the last example results in a heap error, causing the compiler to exit with a run-time error 203.
Remark: Macros defined in the interface part of a unit are not available outside that unit! They can just be used as a notational convenience, or in conditional compiles.
By default the compiler predefines three macros, containing the version number, the release number and the patch number. They are listed in table (??) .
Table 2.1: Predefined macros
Symbol Contains FPC_FULLVERSION An integer version number of the compiler. FPC_VERSION The version number of the compiler. FPC_RELEASE The release number of the compiler. FPC_PATCH The patch number of the compiler.
The FPC_FULLVERSION macro contains a version number which always uses 2 digits for the RELEASE and PATCH version numbers. This means that version 2.3.1 will result in FPC_FULLVERSION=20301. This number makes it easier to determine minimum versions.
Remark: Don’t forget that macro support isn’t on by default. It must be turned on with the -Sm command line switch or using the {$MACRO ON} directive.
In MacPas mode, compile time variables can be defined. They are distinct from symbols in that they have a value, and they are distinct from macros, in that they cannot be used to replace portions of the source text with their value. Their behaviour are compatible with compile time variables found in popular pascal compilers for Macintosh.
A compile time variable is defined like this:
{$SETC ident:= expression}
The expression is a so-called compile time expression, which is evaluated once, at the point where the {$SETC } directve is encountered in the source. The resulting value is then assigned to the compile time variable.
A second {$SETC } directive for the same variable overwrites the previous value.
Contrary to macros and symbols, compile time variables defined in the Interface part of a unit are exported. This means their value will be available in units which uses the unit in which the variable is defined. This requires that both units are compiled in macpas mode.
The big difference between macros and compile time variables is that the former is a pure text substitution mechanism (much like in C), where the latter resemble normal programming language variables, but they are available to the compiler only.
In mode MacPas, compile time variables are always enabled.
Except for the regular Turbo Pascal constructs for conditional compilation, the Free Pascal compiler also supports a stronger conditional compile mechanism: The {$IF} construct, which can be used to evaluate compile-time expressions.
The prototype of this construct is as follows:
{$if expr}
CompileTheseLines;
{$else}
BetterCompileTheseLines;
{$endif}
The content of an expression is restricted to what can be evaluated at compile-time:
The symbols are replaced with their value. For macros recursive substitution might occur.
The following boolean operators are available:
=, <>, >, <, >=, <=, AND, NOT, OR, IN
The IN operator tests for presence of a compile-time variable in a set.
The following functions are also available:
{$IF DEFINED(MySym)}
is equivalent to
{$IF DEFINED MySym}
In expressions, the following rules are used for evaluation:
If the complete expression evaluates to ’0’, then it is considered False and rejected. Otherwise it is considered True and accepted. This may have unexpected consequences:
{$if 0}
will evaluate to False and be rejected, while
{$if 00}
will evaluate to True.
The basic usage of compile time expressions is as follows:
{$if expr}
CompileTheseLines;
{$endif}
If expr evaluates to TRUE, then CompileTheseLines will be included in the source.
Like in regular pascal, it is possible to use {$ELSE }:
{$if expr}
CompileTheseLines;
{$else}
BetterCompileTheseLines;
{$endif}
If expr evaluates to True, CompileTheseLines will be compiled. Otherwise, BetterCompileTheseLines will be compiled.
Additionally, it is possible to use {$ELSEIF}
{$IF expr}
// ...
{$ELSEIF expr}
// ...
{$ELSEIF expr}
// ...
{$ELSE}
// ...
{$ENDIF}
In addition to the above constructs, which are also supported by Delphi, the above is completely equivalent to the following construct in MacPas mode:
{$IFC expr}
//...
{$ELIFC expr}
...
{$ELIFC expr}
...
{$ELSEC}
...
{$ENDC}
that is, IFC corresponds to IF, ELIFC corresponds to ELSEIF, ELSEC is equivalent with ELSE and ENDC is the equivalent of ENDIF. Additionally, IFEND is equivalent to ENDIF:
{$IF EXPR}
CompileThis;
{$ENDIF}
In MacPas mode it is possible to mix these constructs.
The following example shows some of the possibilities:
{$ifdef fpc}
var
y : longint;
{$else fpc}
var
z : longint;
{$endif fpc}
var
x : longint;
begin
{$IF (FPC_VERSION > 2) or
((FPC_VERSION = 2)
and ((FPC_RELEASE > 0) or
((FPC_RELEASE = 0) and (FPC_PATCH >= 1))))}
{$DEFINE FPC_VER_201_PLUS}
{$ENDIF}
{$ifdef FPC_VER_201_PLUS}
{$info At least this is version 2.0.1}
{$else}
{$fatal Problem with version check}
{$endif}
{$define x:=1234}
{$if x=1234}
{$info x=1234}
{$else}
{$fatal x should be 1234}
{$endif}
{$if 12asdf and 12asdf}
{$info $if 12asdf and 12asdf is ok}
{$else}
{$fatal $if 12asdf and 12asdf rejected}
{$endif}
{$if 0 or 1}
{$info $if 0 or 1 is ok}
{$else}
{$fatal $if 0 or 1 rejected}
{$endif}
{$if 0}
{$fatal $if 0 accepted}
{$else}
{$info $if 0 is ok}
{$endif}
{$if 12=12}
{$info $if 12=12 is ok}
{$else}
{$fatal $if 12=12 rejected}
{$endif}
{$if 12<>312}
{$info $if 12<>312 is ok}
{$else}
{$fatal $if 12<>312 rejected}
{$endif}
{$if 12<=312}
{$info $if 12<=312 is ok}
{$else}
{$fatal $if 12<=312 rejected}
{$endif}
{$if 12<312}
{$info $if 12<312 is ok}
{$else}
{$fatal $if 12<312 rejected}
{$endif}
{$if a12=a12}
{$info $if a12=a12 is ok}
{$else}
{$fatal $if a12=a12 rejected}
{$endif}
{$if a12<=z312}
{$info $if a12<=z312 is ok}
{$else}
{$fatal $if a12<=z312 rejected}
{$endif}
{$if a12<z312}
{$info $if a12<z312 is ok}
{$else}
{$fatal $if a12<z312 rejected}
{$endif}
{$if not(0)}
{$info $if not(0) is OK}
{$else}
{$fatal $if not(0) rejected}
{$endif}
{$IF NOT UNDEFINED FPC}
// Detect FPC stuff when compiling on MAC.
{$SETC TARGET_RT_MAC_68881:= FALSE}
{$SETC TARGET_OS_MAC := (NOT UNDEFINED MACOS)
OR (NOT UNDEFINED DARWIN)}
{$SETC TARGET_OS_WIN32 := NOT UNDEFINED WIN32}
{$SETC TARGET_OS_UNIX := (NOT UNDEFINED UNIX)
AND (UNDEFINED DARWIN)}
{$SETC TYPE_EXTENDED := TRUE}
{$SETC TYPE_LONGLONG := FALSE}
{$SETC TYPE_BOOL := FALSE}
{$ENDIF}
{$info *************************************************}
{$info * Now have to follow at least 2 error messages: *}
{$info *************************************************}
{$if not(0}
{$endif}
{$if not(<}
{$endif}
end.
As you can see from the example, this construct isn’t useful when used with normal symbols, only if you use macros, which are explained in section ?? . They can be very useful. When trying this example, you must switch on macro support, with the -Sm command line switch.
The following example works only in MacPas mode:
{$SETC TARGET_OS_MAC := (NOT UNDEFINED MACOS) OR (NOT UNDEFINED DARWIN)}
{$SETC DEBUG := TRUE}
{$SETC VERSION := 4}
{$SETC NEWMODULEUNDERDEVELOPMENT := (VERSION >= 4) OR DEBUG}
{$IFC NEWMODULEUNDERDEVELOPMENT}
{$IFC TARGET_OS_MAC}
... new mac code
{$ELSEC}
... new other code
{$ENDC}
{$ELSEC}
... old code
{$ENDC}
Free Pascal lets you define normal, warning and error messages in your code. Messages can be used to display useful information, such as copyright notices, a list of symbols that your code reacts on etc.
Warnings can be used if you think some part of your code is still buggy, or if you think that a certain combination of symbols isn’t useful.
Error messages can be useful if you need a certain symbol to be defined, to warn that a certain variable isn’t defined, or when the compiler version isn’t suitable for your code.
The compiler treats these messages as if they were generated by the compiler. This means that if you haven’t turned on warning messages, the warning will not be displayed. Errors are always displayed, and the compiler stops if 50 errors have occurred. After a fatal error, the compiler stops at once.
For messages, the syntax is as follows:
{$Message Message text}
or
{$Info Message text}
For notes:
{$Note Message text}
For warnings:
{$Warning Warning Message text}
For hints:
{$Hint Warning Message text}
For errors:
{$Error Error Message text}
Lastly, for fatal errors:
{$Fatal Error Message text}
or
{$Stop Error Message text}
The difference between $Error and $FatalError or $Stop messages is that when the compiler encounters an error, it still continues to compile. With a fatal error, the compiler stops.
Remark: You cannot use the ’}’ character in your message, since this will be treated as the closing brace of the message.
As an example, the following piece of code will generate an error when neither of the symbols RequiredVar1 or RequiredVar2 are defined:
{$IFNDEF RequiredVar1}
{$IFNDEF RequiredVar2}
{$Error One of Requiredvar1 or Requiredvar2 must be defined}
{$ENDIF}
{$ENDIF}
But the compiler will continue to compile. It will not, however, generate a unit file or a program (since an error occurred).
Free Pascal supports inserting assembler statements in between Pascal code. The mechanism for this is the same as under Turbo Pascal and Delphi. There are, however some substantial differences, as will be explained in the following sections.
There are essentially 2 ways to embed assembly code in the pascal source. The first one is the simplest, by using an asm block:
Var I : Integer; begin I:=3; asm movl I,%eax end; end;
Everything between the asm and end block is inserted as assembler in the generated code. Depending on the assembler reader mode, the compiler performs substitution of certain names with their addresses.
The second way is implementing a complete procedure or function in assembler. This is done by adding a assembler modifier to the function or procedure header:
function geteipasebx : pointer;assembler; asm movl (%esp),%ebx ret end;
It’s still possible to declare variables in an assembler procedure:
procedure Move(const source;var dest;count:SizeInt);assembler; var saveesi,saveedi : longint; asm movl %edi,saveedi end;
The compiler will reserve space on the stack for these variables, it inserts some commands for this.
Note that the assembler name of an assembler function will still be ’mangled’ by the compiler, i.e. the label for this function will not be the name of the function as declared. To change this, an Alias modifier can be used:
function geteipasebx : pointer;assembler;[alias:'FPC_GETEIPINEBX']; asm movl (%esp),%ebx ret end;
To make the function available in assembler code outside the current unit, the Public modifier can be added:
function geteipasebx : pointer;assembler;[public,alias:'FPC_GETEIPINEBX']; asm movl (%esp),%ebx ret end;
Free Pascal supports Intel syntax for the Intel family of Ix86 processors in its asm blocks.
The Intel syntax in your asm block is converted to AT&T syntax by the compiler, after which it is inserted in the compiled source. The supported assembler constructs are a subset of the normal assembly syntax. In what follows we specify what constructs are not supported in Free Pascal , but which exist in Turbo Pascal :
mov al, byte ptr MyWord -- allowed, mov al, byte(MyWord) -- allowed, mov al, shortint(MyWord) -- not allowed.
const s= 10; const t = 32767;in Turbo Pascal:
mov al, byte(s) -- useless typecast. mov al, byte(t) -- syntax error!In this parser, either of those cases will give out a syntax error.
mov al,byte ptr ['c'] -- not allowed. mov al,byte ptr [100h] -- not allowed.(This is due to the limitation of the GNU Assembler).
mov al,[ds:bx] -- not alloweduse instead:
mov al,ds:[bx]
const myscale = 1; ... mov al,byte ptr [esi+ebx*myscale] -- not allowed.use:
mov al, byte ptr [esi+ebx*1]
@: -- not alloweduse instead:
@1: -- allowed
lds si,@mylabel -- not allowed
mov al,[byte ptr myvar] -- not allowed.use:
mov al,byte ptr [myvar] -- allowed.
The Intel inline assembler supports the following macros:
In earlier versions, Free Pascal used only the gnu as assembler to generate its object files for the Intel x86 processors. Only after some time, an internal assembler was created, which wrote directly to an object file.
Since the gnu assembler uses AT&T assembly syntax, the code you write should use the same syntax. The differences between AT&T and Intel syntax as used in Turbo Pascal are summarized in the following:
Section:[Base + Index*Scale + Offs]is written in AT&T syntax as:
Section:Offs(Base,Index,Scale)Where Base and Index are optional 32-bit base and index registers, and Scale is used to multiply Index. It can take the values 1,2,4 and 8. The Section is used to specify an optional section register for the memory operand.
More information about the AT&T syntax can be found in the as manual, although the following differences with normal AT&T assembly must be taken into account:
const myid = 10; ... movl $myid,%eax -- allowed movl myid(%esi),%eax -- not allowed.
The AT&T inline assembler supports the following macros:
The inline assembler reader for the Motorola 680x0 family of processors uses the Motorola Assembler syntax (q.v). A few differences do exist:
@MyLabel:
The inline assembler supports the following macros:
When the compiler uses variables, it sometimes stores them, or the result of some calculations, in the processor registers. If you insert assembler code in your program that modifies the processor registers, then this may interfere with the compiler’s idea about the registers. To avoid this problem, Free Pascal allows you to tell the compiler which registers have changed in an asm block. The compiler will then save and reload these registers if it was using them. Telling the compiler which registers have changed is done by specifying a set of register names behind an assembly block, as follows:
asm ... end ['R1', ... ,'Rn'];
Here R1 to Rn are the names of the registers you modify in your assembly code.
As an example:
asm movl BP,%eax movl 4(%eax),%eax movl %eax,__RESULT end ['EAX'];
This example tells the compiler that the EAX register was modified.
For assembler routines, i.e., routines that are written completely in assembler, the ABI of the processor & platform must be respected, i.e. the routine itself must know what registers to save and what not, but it can tell the compiler using the same method what registers were changed or not. The compiler will save specified registers to the stack on entry and restore them on routine exit.
The only thing the compiler normally does, is create a minimal stack frame if needed (e.g. when variables are declared). All the rest is up to the programmer.
As noted in the previous chapter, older Free Pascal compilers relied on the gnu assembler to make object files. The compiler only generated an assembly language file which was then passed on to the assembler. In the following two sections, we discuss what is generated when you compile a unit or a program.
When you compile a unit, the Free Pascal compiler generates 2 files:
The assembly language file contains the actual source code for the statements in your unit, and the necessary memory allocations for any variables you use in your unit. This file is converted by the assembler to an object file (with extension .o) which can then be linked to other units and your program, to form an executable.
By default, the assembly file is removed after it has been compiled. Only in the case of the -s command line option, the assembly file will be left on disk, so the assembler can be called later. You can disable the erasing of the assembler file with the -a switch.
The unit file contains all the information the compiler needs to use the unit:
The detailed contents and structure of this file are described in the first appendix. You can examine a unit description file using the ppudump program, which shows the contents of the file.
If you want to distribute a unit without source code, you must provide both the unit description file and the object file.
You can also provide a C header file to go with the object file. In that case, your unit can be used by someone who wishes to write his programs in C. However, you must make this header file yourself since the Free Pascal compiler doesn’t make one for you.
When you compile a program, the compiler produces again 2 files:
The link response file is, by default, removed from the disk. Only when you specify the -s command line option or when linking fails, then the file is left on the disk. It is named link.res.
The assembly language file is converted to an object file by the assembler, and then linked together with the rest of the units and a program header, to form your final program.
The program header file is a small assembly program which provides the entry point for the program. This is where the execution of your program starts, so it depends on the operating system, because operating systems pass parameters to executables in wildly different ways.
By default, its name is prt0.o, and the source file resides in prt0.as or some variant of this name: Which file is actually used depends on the system, and on linux systems, whether the C library is used or not.
It usually resides where the system unit source for your system resides. Its main function is to save the environment and command line arguments and set up the stack. Then it calls the main program.
Free Pascal supports the new MMX (Multi-Media extensions) instructions of Intel processors. The idea of MMX is to process multiple data with one instruction, for example the processor can add simultaneously 4 words. To implement this efficiently, the Pascal language needs to be extended. So Free Pascal allows to add for example two array[0..3] of word, if MMX support is switched on. The operation is done by the MMX unit and allows people without assembler knowledge to take advantage of the MMX extensions.
Here is an example:
uses
MMX; { include some predefined data types }
const
{ tmmxword = array[0..3] of word;, declared by unit MMX }
w1 : tmmxword = (111,123,432,4356);
w2 : tmmxword = (4213,63456,756,4);
var
w3 : tmmxword;
l : longint;
begin
if is_mmx_cpu then { is_mmx_cpu is exported from unit mmx }
begin
{$mmx+} { turn mmx on }
w3:=w1+w2;
{$mmx-}
end
else
begin
for i:=0 to 3 do
w3[i]:=w1[i]+w2[i];
end;
end.
One important point of MMX is the support of saturated operations. If a operation would cause an overflow, the value stays at the highest or lowest possible value for the data type: If you use byte values you get normally 250+12=6. This is very annoying when doing color manipulations or changing audio samples, when you have to do a word add and check if the value is greater than 255. The solution is saturation: 250+12 gives 255. Saturated operations are supported by the MMX unit. If you want to use them, you have simple turn the switch saturation on: $saturation+
Here is an example:
Program SaturationDemo;
{
example for saturation, scales data (for example audio)
with 1.5 with rounding to negative infinity
}
uses mmx;
var
audio1 : tmmxword;
i: smallint;
const
helpdata1 : tmmxword = ($c000,$c000,$c000,$c000);
helpdata2 : tmmxword = ($8000,$8000,$8000,$8000);
begin
{ audio1 contains four 16 bit audio samples }
{$mmx+}
{ convert it to $8000 is defined as zero, multiply data with 0.75 }
audio1:=(audio1+helpdata2)*(helpdata1);
{$saturation+}
{ avoid overflows (all values>$ffff becomes $ffff) }
audio1:=(audio1+helpdata2)-helpdata2;
{$saturation-}
{ now mupltily with 2 and change to integer }
for i:=0 to 3 do
audio1[i] := audio1[i] shl 1;
audio1:=audio1-helpdata2;
{$mmx-}
end.
In the beginning of 1997 the MMX instructions were introduced in the Pentium processors, so multitasking systems wouldn’t save the newly introduced MMX registers. To work around that problem, Intel mapped the MMX registers to the FPU register.
The consequence is that you can’t mix MMX and floating point operations. After using MMX operations and before using floating point operations, you have to call the routine EMMS of the MMX unit. This routine restores the FPU registers.
Careful: The compiler doesn’t warn if you mix floating point and MMX operations, so be careful.
The MMX instructions are optimized for multimedia operations (what else?). So it isn’t possible to perform all possible operations: some operations give a type mismatch, see section ?? for the supported MMX operations.
An important restriction is that MMX operations aren’t range or overflow checked, even when you turn range and overflow checking on. This is due to the nature of MMX operations.
The MMX unit must always be used when doing MMX operations because the exit code of this unit clears the MMX unit. If it wouldn’t do that, other program will crash. A consequence of this is that you can’t use MMX operations in the exit code of your units or programs, since they would interfere with the exit code of the MMX unit. The compiler can’t check this, so you are responsible for this!
The following operations are supported in the compiler when MMX extensions are enabled:
Here are some helpful hints to get optimal performance:
This chapter gives detailed information on the generated code by Free Pascal . It can be useful to write external object files which will be linked to Free Pascal created code blocks.
The compiler has different register conventions, depending on the target processor used; some of the registers have specific uses during the code generation. The following section describes the generic names of the registers on a platform per platform basis. It also indicates what registers are used as scratch registers, and which can be freely used in assembler blocks.
The accumulator register is at least a 32-bit integer hardware register, and is used to return results of function calls which return integral values.
The accumulator 64-bit register is used in 32-bit environments and is defined as the group of registers which will be used when returning 64-bit integral results in function calls. This is a register pair.
This register is used for returning floating point values from functions.
The self register contains a pointer to the actual object or class. This register gives access to the data of the object or class, and the VMT pointer of that object or class.
The frame pointer register is used to access parameters in subroutines, as well as to access local variables. References to the pushed prameters and local variables are constructed using the frame pointer. 1.
The stack pointer is used to give the address of the stack area, where the local variables and parameters to subroutines are stored.
Scratch registers are those which can be used in assembler blocks, or in external object files without requiring any saving before usage.
This indicates what registers are used for what purposes on each of the processors supported by Free Pascal . It also indicates which registers can be used as scratch registers.
Table 6.1: Intel 80x86 Register table
Generic register name CPU Register name accumulator EAX accumulator (64-bit) high / low EDX:EAX float result FP(0) self ESI frame pointer EBP stack pointer ESP scratch regs. N/A
Table 6.2: Motorola 680x0 Register table
Generic register name CPU Register name accumulator D02 accumulator (64-bit) high / low D0:D1 float result FP03 self A5 frame pointer A6 stack pointer A7 scratch regs. D0, D1, A0, A1, FP0, FP1
Contrary to most C compilers and assemblers, all labels generated to pascal variables and routines have mangled names4. This is done so that the compiler can do stronger type checking when parsing the Pascal code. It also permits function and procedure overloading.
The rules for mangled names for variables and typed constants are as follows:
Examples:
unit testvars; interface const publictypedconst : integer = 0; var publicvar : integer; implementation const privatetypedconst : integer = 1; var privatevar : integer; end.
Will result in the following assembler code for the gnu assembler :
.file "testvars.pas" .text .data # [6] publictypedconst : integer = 0; .globl TC__TESTVARS$$_PUBLICTYPEDCONST TC__TESTVARS$$_PUBLICTYPEDCONST: .short 0 # [12] privatetypedconst : integer = 1; TC__TESTVARS$$_PRIVATETYPEDCONST: .short 1 .bss # [8] publicvar : integer; .comm U_TESTVARS_PUBLICVAR,2 # [14] privatevar : integer; .lcomm _PRIVATEVAR,2
The rules for mangled names for routines are as follows:
The following constructs
unit testman; interface type myobject = object constructor init; procedure mymethod; end; implementation constructor myobject.init; begin end; procedure myobject.mymethod; begin end; function myfunc: pointer; begin end; procedure myprocedure(var x: integer; y: longint; z : pchar); begin end; end.
will result in the following assembler file for the Intel 80x86 target:
.file "testman.pas" .text .balign 16 .globl _TESTMAN$$_$$_MYOBJECT_$$_INIT _TESTMAN$$_$$_MYOBJECT_$$_INIT: pushl %ebp movl %esp,%ebp subl $4,%esp movl $0,%edi call FPC_HELP_CONSTRUCTOR jz .L5 jmp .L7 .L5: movl 12(%ebp),%esi movl $0,%edi call FPC_HELP_FAIL .L7: movl %esi,%eax testl %esi,%esi leave ret $8 .balign 16 .globl _TESTMAN$$_$$_MYOBJECT_$$_MYMETHOD _TESTMAN$$_$$_MYOBJECT_$$_MYMETHOD: pushl %ebp movl %esp,%ebp leave ret $4 .balign 16 _TESTMAN$$_MYFUNC: pushl %ebp movl %esp,%ebp subl $4,%esp movl -4(%ebp),%eax leave ret .balign 16 _TESTMAN$$_MYPROCEDURE$INTEGER$LONGINT$PCHAR: pushl %ebp movl %esp,%ebp leave ret $12
To make the symbols externally accessible, it is possible to give nicknames to mangled names, or to change the mangled name directly. Two modifiers can be used:
The prototype for an aliased function or procedure is as follows:
Procedure AliasedProc; alias : 'AliasName';
The procedure AliasedProc will also be known as AliasName. Take care, the name you specify is case sensitive (as C is).
Furthermore, the exports section of a library is also used to declare the names that will be exported by the shared library. The names in the exports section are case-sensitive (while the actual declaration of the routine is not). For more information on the creating shared libraries, chapter ?? .
By default, the calling mechanism the compiler uses is register, that is, the compiler will try to pass as much parameters as posible by storing them in a free register. Not all registers are used, because some registers have a special meaning, but this depends on the CPU.
Function results are returned in the accumulator (first register), if they fit in the register. Method calls (from either objects or clases) have an additional invisible parameter which is self.
When the procedure or function exits, it clears the stack.
Other calling methods are available for linking with external object files and libraries, these are summarized in table (??) . The first column lists the modifier you specify for a procedure declaration. The second one lists the order the paramaters are pushed on the stack. The third column specifies who is responsible for cleaning the stack: the caller or the called function. The alignment column indicates the alignment of the parameters sent to the stack area.
Subroutines will modify a number of registers (the volatile registers). The list of registers that are modified is highly dependent on the processor, calling convention and ABI of the target platform.
Table 6.3: Calling mechanisms in Free Pascal
Modifier Pushing order Stack cleaned by alignment <none> Left-to-right Callee default register Left-to-right Callee default cdecl Right-to-left Caller GCC alignment interrupt Right-to-left Callee default pascal Left-to-right Callee default safecall Right-to-left Callee default stdcall Right-to-left Callee GCC alignment oldfpccall Right-to-left Callee default
Note that the oldfpccall calling convention equals the default calling convention on processors other than 32-bit Intel 386 or higher.
More about this can be found in chapter ?? on linking. Information on GCC registers saved, GCC stack alignment and general stack alignment on an operating system basis is beyond the scope of this manual.
As of version 2.0 (actually, in 1.9.x somewhere) , the register modifier is the default calling convention, prior to that, it was the oldfpccall convention.
The default calling convention, i.e., the calling convention used when none is specified explicitly, can be set using the {$calling } directive, section ?? . The default calling convention for the current platform can be specified with
{$CALLING DEFAULT}
Remark: The popstack modifier is no longer supported as of version 2.0, but has been renamed to oldfpccall. The saveregisters modifier can no longer be used.
When a routine is declared within the scope of a procedure or function, it is said to be nested. In this case, an additional invisible parameter is passed to the nested routine. This additional parameter is the frame pointer address of the parent routine. This permits the nested routine to access the local variables and parameters of the calling routine.
The resulting stack frame after the entry code of a simple nested procedure has been executed is shown in table (??) .
Table 6.4: Stack frame when calling a nested procedure (32-bit processors)
Offset from frame pointer What is stored +x parameters +8 Frame pointer of parent routine +4 Return address +0 Saved frame pointer
Constructor and destructors have special invisible parameters which are passed to them. These invisible parameters are used internally to instantiate the objects and classes.
The actual invisible declaration of an object constructor is as follows:
constructor init(_vmt : pointer; _self : pointer ...);
Where _vmt is a pointer to the virtual method table for this object. This value is nil if a constructor is called within the object instance (such as calling an ihnerited constructor).
_self is either nil if the instance must be allocated dynamically (object is declared as pointer), or the address of the object instance if the object is declared as a normal object (stored in the data area) or if the object instance has already been allocated.
The allocated instance (if allocated via new) (self) is returned in the accumulator.
The declaration of a destructor is as follows:
destructor done(_vmt : pointer; _self : pointer ...);
Where _vmt is a pointer to the virtual method table for this object. This value is nil if a destructor is called within the object instance (such as calling an ihnerited constructor), or if the object instance is a variable and not a pointer.
_self is the address of the object instance.
The actual invisible declaration of a class constructoir is as follows:
constructor init(_vmt: pointer; flag : longint; ..);
_vmt is either nil if called from the instance or if calling an inherited constructor, otherwise it points to the address of the virtual method table.
Where flag is zero if the constructor is called within the object instance or with an instance qualifier otherwise this flag is set to one.
The allocated instance (self) is returned in the accumulator.
The declaration of a destructor is as follows:
destructor done(_self : pointer; flag : longint ...);
_self is the address of the object instance.
flag is zero if the destructor is called within the object instance or with an instance qualifier otherwise this flag is set to one.
Each Pascal procedure and function begins and ends with standard epilogue and prologue code.
Standard entry code for procedures and functions is as follows on the 80x86 architecture:
pushl %ebp movl %esp,%ebp
The generated exit sequence for procedure and functions looks as follows:
leave ret $xx
Where xx is the total size of the pushed parameters.
To have more information on function return values take a look at section ?? .
Standard entry code for procedures and functions is as follows on the 680x0 architecture:
move.l a6,-(sp) move.l sp,a6
The generated exit sequence for procedure and functions looks as follows (in the default processor mode):
unlk a6 rtd #xx
Where xx is the total size of the pushed parameters.
To have more information on function return values take a look at section ?? .
When a function or procedure is called, then the following is done by the compiler:
The resulting stack frame upon entering looks as in table (??) .
Table 6.5: Stack frame when calling a procedure (32-bit model)
Offset What is stored Optional? +x extra parameters Yes +12 function result Yes +8 self Yes +4 Return address No +0 Frame pointer of parent procedure Yes
Each parameter passed to a routine is guaranteed to decrement the stack pointer by a certain minimum amount. This behavior varies from one operating system to another. For example, passing a byte as a value parameter to a routine could either decrement the stack pointer by 1, 2, 4 or even 8 bytes depending on the target operating system and processor.
For example, on FreeBSD , all parameters passed to a routine guarantee a minimal stack decrease of four bytes per parameter, even if the parameter actually takes less then 4 bytes to store on the stack (such as passing a byte value parameter to the stack).
Certain processors have limitations on the size of the parameters and local variables in routines. This is shown in table (??) .
Table 6.6: Maximum limits for processors
Processor Parameters Local variables Intel 80x86 (all) 64K No limit Motorola 68020 (default) 32K No limit Motorola 68000 32K 32K
Furthermore, the m68k compiler, in 68000 mode, limits the size of data elements to 32K (arrays, records, objects, etc.). This restriction does not exist in 68020 mode.
When you only use Pascal code, and Pascal units, then you will not see much of the part that the linker plays in creating your executable. The linker is only called when you compile a program. When compiling units, the linker isn’t invoked.
However, there are times that linking to C libraries, or to external object files created by other compilers, may be necessary. The Free Pascal compiler can generate calls to a C function, and can generate functions that can be called from C (exported functions).
In general, there are 3 things you must do to use a function that resides in an external library or object file:
The same holds for variables. To access a variable that resides in an external object file, you must declare it, and tell the compiler where to find it. The following sections attempt to explain how to do this.
The first step in using external code blocks is declaring the function you want to use. Free Pascal supports Delphi syntax, i.e. you must use the external directive. The external directive replaces, in effect, the code block of the function.
The external directive doesn’t specify a calling convention; it just tells the compiler that the code for a procedure or function resides in an external code block. A calling convention modifier should be declared if the external code blocks does not have the same calling conventions as Free Pascal . For more information on the calling conventions section ?? .
There exist four variants of the external directive:
Procedure ProcName (Args : TPRocArgs); external;The external directive tells the compiler that the function resides in an external block of code. You can use this together with the {$L} or {$LinkLib} directives to link to a function or procedure in a library or external object file. Object files are looked for in the object search path (set by -Fo) and libraries are searched for in the linker path (set by -Fl).
Procedure ProcName (Args : TPRocArgs); external 'Name';This tells the compiler that the procedure resides in a library with name ’Name’. This method is equivalent to the following:
Procedure ProcName (Args : TPRocArgs);external;
{$LinkLib 'Name'}
Procedure ProcName (Args : TPRocArgs); external 'Name'
name 'OtherProcName';
This has the same meaning as the previous declaration, only the compiler
will use the name ’OtherProcName’ when linking to the library. This
can be used to give different names to procedures and functions in an
external library. The name of the routine is case-sensitive and should
match exactly the name of the routine in the object file.This method is equivalent to the following code:
Procedure OtherProcName (Args : TProcArgs); external;
{$LinkLib 'Name'}
Procedure ProcName (Args : TPRocArgs);
begin
OtherProcName (Args);
end;
Procedure ProcName (Args : TPRocArgs); external 'Name'
Index SomeIndex;
This tells the compiler that the procedure ProcName resides in a
dynamic link library, with index SomeIndex.Remark: Note that this is only available under Windows and os/2 .
Some libaries or code blocks have variables which they export. You can access these variables much in the same way as external functions. To access an external variable, you declare it as follows:
Var MyVar : MyType; external name 'varname';
The effect of this declaration is twofold:
The variable will be accessible with its declared name, i.e. MyVar in this case.
A second possibility is the declaration:
Var varname : MyType; cvar; external;
The effect of this declaration is twofold as in the previous case:
The first possibility allows you to change the name of the external variable for internal use.
As an example, let’s look at the following C file (in extvar.c):
/* Declare a variable, allocate storage */ int extvar = 12;
And the following program (in extdemo.pp):
Program ExtDemo;
{$L extvar.o}
Var { Case sensitive declaration !! }
extvar : longint; cvar;external;
I : longint; external name 'extvar';
begin
{ Extvar can be used case insensitive !! }
Writeln ('Variable ''extvar'' has value: ',ExtVar);
Writeln ('Variable ''I'' has value: ',i);
end.
Compiling the C file, and the pascal program:
gcc -c -o extvar.o extvar.c ppc386 -Sv extdemo
Will produce a program extdemo which will print
Variable 'extvar' has value: 12 Variable 'I' has value: 12
on your screen.
To make sure that all parameters are correctly passed to the external routines, you should declare it with the correct calling convention modifier. When linking with code blocks compiled with standard C compilers (such as GCC), the cdecl modifier should be used so as to indicate that the external routine uses C type calling conventions. For more information on the supported calling conventions, see section ?? .
As might be expected, external variable declarations do not require any calling convention modifier.
Having declared the external function or variable that resides in an object file, you can use it as if it was defined in your own program or unit. To produce an executable, you must still link the object file in. This can be done with the {$L file.o} directive.
This will cause the linker to link in the object file file.o. On most systems, this filename is case sensitive. The object file is first searched in the current directory, and then the directories specified by the -Fo command line.
You cannot specify libraries in this way, it is for object files only.
Here we present an example. Consider that you have some assembly routine which uses the C calling convention that calculates the nth Fibonacci number:
.text
.align 4
.globl Fibonacci
.type Fibonacci,@function
Fibonacci:
pushl %ebp
movl %esp,%ebp
movl 8(%ebp),%edx
xorl %ecx,%ecx
xorl %eax,%eax
movl $1,%ebx
incl %edx
loop:
decl %edx
je endloop
movl %ecx,%eax
addl %ebx,%eax
movl %ebx,%ecx
movl %eax,%ebx
jmp loop
endloop:
movl %ebp,%esp
popl %ebp
ret
Then you can call this function with the following Pascal Program:
Program FibonacciDemo;
var i : longint;
Function Fibonacci (L : longint):longint;cdecl;external;
{$L fib.o}
begin
For I:=1 to 40 do
writeln ('Fib(',i,') : ',Fibonacci (i));
end.
With just two commands, this can be made into a program:
as -o fib.o fib.s ppc386 fibo.pp
This example supposes that you have your assembler routine in fib.s, and your Pascal program in fibo.pp.
To link your program to a library, the procedure depends on how you declared the external procedure.
In case you used the following syntax to declare your procedure:
Procedure ProcName (Args : TPRocArgs); external 'Name';
You don’t need to take additional steps to link your file in, the compiler will do all that is needed for you. On Windows it will link to name.dll, on linux and most unix ’es your program will be linked to library libname, which can be a static or dynamic library.
In case you used
Procedure ProcName (Args : TPRocArgs); external;
You still need to explicity link to the library. This can be done in 2 ways:
{$LinkLib 'gpm'}
This will link to the gpm library. On unix systems (such as linux ),
you must not specify the extension or ’lib’ prefix of the library. The compiler takes
care of that. On other systems (such as Windows ), you need to specify the full
name.
ppc386 -k'-lgpm' myprog.ppIs equivalent to the above method, and tells the linker to link to the gpm library.
As an example, consider the following program:
program printlength;
{$linklib c} { Case sensitive }
{ Declaration for the standard C function strlen }
Function strlen (P : pchar) : longint; cdecl;external;
begin
Writeln (strlen('Programming is easy !'));
end.
This program can be compiled with:
ppc386 prlen.pp
Supposing, of course, that the program source resides in prlen.pp.
To use functions in C that have a variable number of arguments, you must compile your unit or program in objfpc mode or Delphi mode, and use the Array of const argument, as in the following example:
program testaocc;
{$mode objfpc}
Const
P : Pchar
= 'example';
F : Pchar
= 'This %s uses printf to print numbers (%d) and strings.'#10;
procedure printf(fm: pchar;args: array of const);cdecl;external 'c';
begin
printf(F,[P,123]);
end.
The output of this program looks like this:
This example uses printf to print numbers (123) and strings.
As an alternative, the program can be constructed as follows:
program testaocc;
Const
P : Pchar
= 'example';
F : Pchar
= 'This %s uses printf to print numbers (%d) and strings.'#10;
procedure printf(fm: pchar);cdecl;varargs;external 'c';
begin
printf(F,P,123);
end.
The varargs modifier signals the compiler that the function allows a variable number of arguments (the ellipsis notation in C).
Free Pascal supports making shared or static libraries in a straightforward and easy manner. If you want to make static libraries for other Free Pascal programmers, you just need to provide a command line switch. To make shared libraries, refer to the chapter ?? . If you want C programmers to be able to use your code as well, you will need to adapt your code a little. This process is described first.
When exporting functions from a library, there are 2 things you must take in account:
The calling conventions are controlled by the modifiers cdecl, popstack, pascal, safecall, stdcall and register. See section ?? for more information on the different kinds of calling scheme.
The naming conventions can be controlled by 2 modifiers in the case of static libraries:
For more information on how these different modifiers change the name mangling of the routine section ?? .
Remark: If in your unit, you use functions that are in other units, or system functions, then the C program will need to link in the object files from these units too.
Similarly as when you export functions, you can export variables. When exporting variables, one should only consider the names of the variables. To declare a variable that should be used by a C program, one declares it with the cvar modifier:
Var MyVar : MyTpe; cvar;
This will tell the compiler that the assembler name of the variable (the one which is used by C programs) should be exactly as specified in the declaration, i.e., case sensitive.
It is not allowed to declare multiple variables as cvar in one statement, i.e. the following code will produce an error:
var Z1,Z2 : longint;cvar;
To create shared libraries one should use the library keyword in the main compilation file (the project file). For more information on creating shared libraries, chapter ?? .
The .o object files that the compiler writes when it compiles a unit, are regular object files as they are produced by a C compiler. They can be combined using the ar and ranlib tools into a static library. However, for various reasons, this is a bad idea.
To remedy these (and other) problems requires intimate knowledge of the inner workings of the compiler and RTL, and it is therefor a bad idea to attempt the use of static libraries with Free Pascal.
When you compile a unit, the compiler will by default always look for unit files.
To be able to differentiate between units that have been compiled as static or dynamic libraries, there are 2 switches:
Definition of one symbol will automatically undefine the other.
These two switches can be used in conjunction with the configuration file fpc.cfg. The existence of one of these symbols can be used to decide which unit search path to set. For example, on linux :
# Set unit paths #IFDEF FPC_LINK_STATIC -Up/usr/lib/fpc/linuxunits/staticunits #ENDIF #IFDEF FPC_LINK_DYNAMIC -Up/usr/lib/fpc/linuxunits/sharedunits #ENDIF
With such a configuration file, the compiler will look for its units in different directories, depending on whether -XD or -XS is used.
You can compile your units using smart linking. When you use smartlinking, the compiler creates a series of code blocks that are as small as possible, i.e. a code block will contain only the code for one procedure or function.
When you compile a program that uses a smart-linked unit, the compiler will only link in the code that you actually need, and will leave out all other code. This will result in a smaller binary, which is loaded in memory faster, thus speeding up execution.
To enable smartlinking, one can give the smartlink option on the command line: -Cx, or one can put the {$SMARTLINK ON} directive in the unit file:
Unit Testunit
{SMARTLINK ON}
Interface
...
Smartlinking will slow down the compilation process, especially for large units.
When a unit foo.pp is smartlinked, the name of the codefile is changed to libfoo.a.
Technically speaking, the compiler makes small assembler files for each procedure and function in the unit, as well as for all global defined variables (whether they’re in the interface section or not). It then assembles all these small files, and uses ar to collect the resulting object files in one archive.
Smartlinking and the creation of shared (or dynamic) libraries are mutually exclusive, that is, if you turn on smartlinking, then the creation of shared libraries is turned of. The creation of static libraries is still possible. The reason for this is that it has little sense in making a smartlinked dynamical library. The whole shared library is loaded into memory anyway by the dynamic linker (or the operating system), so there would be no gain in size by making it smartlinked.
The Free Pascal compiler issues 32-bit or 64-bit code. This has several consequences:
This section gives information on the storage space occupied by the different possible types in Free Pascal . Information on internal alignment will also be given.
The storage size of the default integer types are given in Reference guide . In the case of user defined-types, the storage space occupied depends on the bounds of the type:
A char, or a subrange of the char type, is stored as a byte. A WideChar is stored as a word, i.e. 2 bytes.
The Boolean type is stored as a byte and can take a value of true or false.
A ByteBool is stored as a byte, a WordBool type is stored as a word, and a longbool is stored as a longint.
By default all enumerations are stored as a longword (4 bytes), which is equivalent to specifying the {$Z4}, {$PACKENUM 4} or {$PACKENUM DEFAULT} switches.
This default behavior can be changed by compiler switches, and by the compiler mode.
In the tp compiler mode, or while the {$Z1} or {$PACKENUM 1} switches are in effect, the storage space used is shown in table (??) .
Table 8.1: Enumeration storage for tp mode
# Of Elements in Enum. Storage space used 0..255 byte (1 byte) 256..65535 word (2 bytes) > 65535 longword (4 bytes)
When the {$Z2} or {$PACKENUM 2} switches are in effect, the value is stored in 2 bytes (a word), if the enumeration has less or equal than 65535 elements. If there are more elements, the enumeration value is stored as a 4 byte value (a longword).
Floating point type sizes and mapping vary from one processor to another. Except for the Intel 80x86 architecture, the extended type maps to the IEEE double type if a hardware floating point coprocessor is present.
Floating point types have a storage binary format divided into three distinct fields : the mantissa, the exponent and the sign bit which stores the sign of the floating point value.
The single type occupies 4 bytes of storage space, and its memory structure is the same as the IEEE-754 single type. This type is the only type which is guaranteed to be available on all platforms (either emulated via software or directly via hardware).
The memory format of the single format looks like what is shown in Figure ?? .
The double type occupies 8 bytes of storage space, and its memory structure is the same as the IEEE-754 double type.
The memory format of the double format looks like like what is shown in Figure ?? .
On processors which do not support co-processor operations (and which have the {$E+} switch), the double type does not exist.
For Intel 80x86 processors, the extended type has takes up 10 bytes of memory space. For more information on the extended type consult the Intel Programmer’s reference.
For all other processors which support floating point operations, the extended type is a nickname for the type which supports the most precision, this is usually the double type. On processors which do not support co-processor operations (and which have the {$E+} switch), the extended type usually maps to the single type.
For Intel 80x86 processors, the comp type contains a 63-bit integral value, and a sign bit (in the MSB position). The comp type uses 8 bytes of storage space.
On other processors, the comp type is not supported.
Contrary to Turbo Pascal, where the real type had a special internal format, under Free Pascal the real type simply maps to one of the other real types. It maps to the double type on processors which support floating point operations, while it maps to the single type on processors which do not support floating point operations in hardware. See table (??) for more information on this.
Table 8.2: Processor mapping of real type
Processor Real type mapping Intel 80x86 double Motorola 680x0 (with {$E-} switch) double Motorola 680x0 (with {$E+} switch) single
A pointer type is stored as a longword (unsigned 32-bit value) on 32-bit processors, and is stored as a 64-bit unsigned value1 on 64-bit processors.
The ansistring is a dynamically allocated string which has no length limitation. When the string is no longer being referenced (its reference count reaches zero), its memory is automatically freed.
If the ansistring is a constant, then its reference count will be equal to -1, indicating that it should never be freed. The structure in memory for an ansistring is shown in table (??) .
Table 8.3: AnsiString memory structure (32-bit model)
Offset Contains -8 Longint with reference count. -4 Longint with actual string size. 0 Actual array of char, null-terminated.
A shortstring occupies as many bytes as its maximum length plus one. The first byte contains the current dynamic length of the string. The following bytes contain the actual characters (of type char) of the string. The maximum size of a short string is the length byte followed by 255 characters.
A widestring is allocated on the heap, much like an ansistring. Unlike the ansistring, a widestring takes 2 bytes per character, and is terminated with a double null.
A set is stored as an array of bits, where each bit indicates if the element is in the set or excluded from the set. The maximum number of elements in a set is 256.
If a set has less than 32 elements, it is coded as an unsigned 32-bit value. Otherwise it is coded as an array of 8 unsigned 32-bit values (longwords), and hence has a size of 256 bytes.
The longword number of a specific element E is given by :
LongwordNumber = (E div 32);
and the bit number within that 32-bit value is given by:
BitNumber = (E mod 32);
A static array is stored as a contiguous sequence of variables of the components of the array. The components with the lowest indexes are stored first in memory. No alignment is done between each element of the array. A multi-dimensional array is stored with the rightmost dimension increasing first.
A dynamic array is stored as a pointer to a block of memory on the heap. The memory on the heap is a contiguous sequence of variables of the components of the array, just as for a static array. The reference count and memory size are stored in memory just before the actual start of the array, at a negative offset relative to the address the pointer refers to. It should not be used.
Each field of a record is stored in a continguous sequence of variables, where the first field is stored at the lowest address in memory. In case of variant fields in a record, each variant starts at the same address in memory. Fields of record are usually aligned, unless the packed directive is specified when declaring the record type.
For more information on field alignment, consult section ?? .
Objects are stored in memory just as ordinary records with an extra field: a pointer to the Virtual Method Table (VMT). This field is stored first, and all fields in the object are stored in the order they are declared (with possible alignment of field addresses, unless the object was declared as being packed).
The VMT is initialized by the call to the object’s Constructor method. If the new operator was used to call the constructor, the data fields of the object will be stored in heap memory, otherwise they will directly be stored in the data section of the final executable.
If an object doesn’t have virtual methods, no pointer to a VMT is inserted.
The memory allocated looks as in table (??) .
Table 8.4: Object memory layout (32/64-bit model)
Offset What 32-bit 64-bit +0 +0 Pointer to VMT (optional). +4 +8 Data. All fields in the order they’ve been declared. …
The Virtual Method Table (VMT) for each object type consists of 2 check fields (containing the size of the data), a pointer to the object’s ancestor’s VMT (Nil if there is no ancestor), and then the pointers to all virtual methods. The VMT layout is illustrated in table (??) . The VMT is constructed by the compiler.
Table 8.5: Object Virtual Method Table memory layout (32/64-bit model)
Offset What 32-bit 64-bit +0 +0 Size of object type data +4 +8 Minus the size of object type data. Enables determining of valid VMT pointers. +8 +16 Pointer to ancestor VMT, Nil if no ancestor available. +12 +24 Pointers to the virtual methods. …
Just like objects, classes are stored in memory just as ordinary records with an extra field: a pointer to the Virtual Method Table (VMT). This field is stored first, and all fields in the class are stored in the order they are declared.
Contrary to objects, all data fields of a class are always stored in heap memory.
The memory allocated looks as in table (??) .
Table 8.6: Class memory layout (32/64-bit model)
Offset What 32-bit 64-bit +0 +0 Pointer to VMT. +4 +8 Data. All fields in the order they’ve been declared. …
The Virtual Method Table (VMT) of each class consists of several fields, which are used for runtime type information. The VMT layout is illustrated in table (??) . The VMT is constructed by the compiler.
Table 8.7: Class Virtual Method Table memory layout (32/64-bit model)
Offset What 32-bit 64-bit +0 +0 Size of object type data +4 +8 Minus the size of object type data. Enables determining of valid VMT pointers. +8 +16 Pointer to ancestor VMT, Nil if no ancestor available. +12 +24 Pointer to the class name (stored as a shortstring). +16 +32 Pointer to the dynamic method table (using message with integers). +20 +40 Pointer to the method definition table. +24 +48 Pointer to the field definition table. +28 +56 Pointer to type information table. +32 +64 Pointer to instance initialization table. +36 +72 Pointer to the auto table. +40 +80 Pointer to the interface table. +44 +88 Pointer to the dynamic method table (using message with strings). +48 +96 Pointer to the Destroy destructor. +52 +104 Pointer to the NewInstance method. +56 +112 Pointer to the FreeInstance method. +60 +120 Pointer to the SafeCallException method. +64 +128 Pointer to the DefaultHandler method. +68 +136 Pointer to the AfterConstruction method. +72 +144 Pointer to the BeforeDestruction method. +76 +152 Pointer to the DefaultHandlerStr method. +80 +160 Pointer to the Dispatch method. +84 +168 Pointer to the DispatchStr method. +88 +176 Pointer to the Equals method. +92 +184 Pointer to the GetHashCode method. +96 +192 Pointer to the ToString method. +100 +200 Pointers to other virtual methods. …
See the file rtl/inc/objpash.inc for the most current VMT information.
File types are represented as records. Typed files and untyped files are represented as a fixed record:
Const
PrivDataLength=3*SizeOf(SizeInt) + 5*SizeOf(pointer);
Type
filerec = packed record
handle : THandle;
mode : longint;
recsize : Sizeint;
_private : array[1..PrivDataLength] of byte;
userdata : array[1..32] of byte;
name : array[0..filerecnamelength] of char;
End;
Text files are described using the following record:
TextBuf = array[0..255] of char;
textrec = packed record
handle : THandle;
mode : longint;
bufsize : SizeInt;
_private : SizeInt;
bufpos : SizeInt;
bufend : SizeInt;
bufptr : ^textbuf;
openfunc : pointer;
inoutfunc : pointer;
flushfunc : pointer;
closefunc : pointer;
userdata : array[1..32] of byte;
name : array[0..255] of char;
LineEnd : TLineEndStr;
buffer : textbuf;
End;
A procedural type is stored as a generic pointer, which stores the address of the routine.
A procedural type to a normal procedure or function is stored as a generic pointer, which stores the address of the entry point of the routine.
In the case of a method procedural type, the storage consists of two pointers, the first being a pointer to the entry point of the method, and the second one being a pointer to self (the object instance).
All static data (variables and typed constants) which are greater than a byte are usually aligned on a multiple of two boundary. This alignment applies only to the start address of the variables, and not the alignment of fields within structures or objects for example. For more information on structured alignment, section ?? . The alignment is similar across the different target processors.
Table 8.8: Data alignment
Data size (bytes) Alignment (small size) Alignment (fast) 1 1 1 2-3 2 2 4-7 2 4 8+ 2 4
The alignment columns indicates the address alignment of the variable, i.e the start address of the variable will be aligned on that boundary. The small size alignment is valid when the code generated should be optimized for size (-Og compiler option) and not speed, otherwise the fast alignment is used to align the data (this is the default).
By default all elements in a structure are aligned to a 2 byte boundary, unless the $PACKRECORDS directive or packed modifier is used to align the data in another way. For example a record or object having a 1 byte element, will have its size rounded up to 2, so the size of the structure will actually be 2 bytes.
The heap is used to store all dynamic variables, and to store class instances. The interface to the heap is the same as in Turbo Pascal and Delphi although the effects are maybe not the same. The heap is thread-safe, so allocating memory from various threads is not a problem.
The heap is a memory structure which is organized as a stack. The heap bottom is stored in the variable HeapOrg. Initially the heap pointer (HeapPtr) points to the bottom of the heap. When a variable is allocated on the heap, HeapPtr is incremented by the size of the allocated memory block. This has the effect of stacking dynamic variables on top of each other.
Each time a block is allocated, its size is normalized to have a granularity of 16 (or 32 on 64 bit systems) bytes.
When Dispose or FreeMem is called to dispose of a memory block which is not on the top of the heap, the heap becomes fragmented. The deallocation routines also add the freed blocks to the freelist which is actually a linked list of free blocks. Furthermore, if the deallocated block was less then 8K in size, the free list cache is also updated.
The free list cache is actually a cache of free heap blocks which have specific lengths (the adjusted block size divided by 16 gives the index into the free list cache table). It is faster to access then searching through the entire freelist.
The format of an entry in the freelist is as follows:
PFreeRecord = ^TFreeRecord; TFreeRecord = record Size : longint; Next : PFreeRecord; Prev : PFreeRecord; end;
The Next field points to the next free block, while the Prev field points to the previous free block.
The algorithm for allocating memory is as follows:
The heap allocates memory from the operating system on an as-needed basis.
OS memory is requested in blocks: It first tries to increase memory in a 64Kb chunk if the size to allocate is less than 64Kb, or 256Kb or 1024K otherwise. If this fails, it tries to increase the heap by the amount you requested from the heap.
If the attempt to reserve OS memory fails, the value returned depends on the value of the ReturnNilIfGrowHeapFails global variable. This is summarized in table (??) .
Table 8.9: ReturnNilIfGrowHeapFails value
ReturnNilIfGrowHeapFails Default memory value manager action FALSE (The default) Runtime error 203 generated TRUE GetMem, ReallocMem and New returns nil
ReturnNilIfGrowHeapFails can be set to change the behavior of the default memory manager error handler.
Free Pascal provides a unit that allows you to trace allocation and deallocation of heap memory: heaptrc.
If you specify the -gh switch on the command line, or if you include heaptrc as the first unit in your uses clause, the memory manager will trace what is allocated and deallocated, and on exit of your program, a summary will be sent to standard output.
More information on using the heaptrc mechanism can be found in the Users guide and Unit reference .
Free Pascal allows you to write and use your own memory manager. The standard functions GetMem, FreeMem, ReallocMem etc. use a special record in the system unit to do the actual memory management. The system unit initializes this record with the system unit’s own memory manager, but you can read and set this record using the GetMemoryManager and SetMemoryManager calls:
procedure GetMemoryManager(var MemMgr: TMemoryManager); procedure SetMemoryManager(const MemMgr: TMemoryManager);
the TMemoryManager record is defined as follows:
TMemoryManager = record
NeedLock : Boolean;
Getmem : Function(Size:PtrInt):Pointer;
Freemem : Function(var p:pointer):PtrInt;
FreememSize : Function(var p:pointer;Size:PtrInt):PtrInt;
AllocMem : Function(Size:PtrInt):Pointer;
ReAllocMem : Function(var p:pointer;Size:PtrInt):Pointer;
MemSize : function(p:pointer):PtrInt;
InitThread : procedure;
DoneThread : procedure;
RelocateHeap : procedure;
GetHeapStatus : function :THeapStatus;
GetFPCHeapStatus : function :TFPCHeapStatus;
end;
As you can see, the elements of this record are mostly procedural variables. The system unit does nothing but call these various variables when you allocate or deallocate memory.
Each of these fields corresponds to the corresponding call in the system unit. We’ll describe each one of them:
To implement your own memory manager, it is sufficient to construct such a record and to issue a call to SetMemoryManager.
To avoid conflicts with the system memory manager, setting the memory manager should happen as soon as possible in the initialization of your program, i.e. before any call to getmem is processed.
This means in practice that the unit implementing the memory manager should be the first in the uses clause of your program or library, since it will then be initialized before all other units - except the system unit itself, of course.
This also means that it is not possible to use the heaptrc unit in combination with a custom memory manager, since the heaptrc unit uses the system memory manager to do all its allocation. Putting the heaptrc unit after the unit implementing the memory manager would overwrite the memory manager record installed by the custom memory manager, and vice versa.
The following unit shows a straightforward implementation of a custom memory manager using the memory manager of the C library. It is distributed as a package with Free Pascal .
unit cmem;
interface
Const
LibName = 'libc';
Function Malloc (Size : ptrint) : Pointer;
cdecl; external LibName name 'malloc';
Procedure Free (P : pointer);
cdecl; external LibName name 'free';
function ReAlloc (P : Pointer; Size : ptrint) : pointer;
cdecl; external LibName name 'realloc';
Function CAlloc (unitSize,UnitCount : ptrint) : pointer;
cdecl; external LibName name 'calloc';
implementation
type
pptrint = ^ptrint;
Function CGetMem (Size : ptrint) : Pointer;
begin
CGetMem:=Malloc(Size+sizeof(ptrint));
if (CGetMem <> nil) then
begin
pptrint(CGetMem)^ := size;
inc(CGetMem,sizeof(ptrint));
end;
end;
Function CFreeMem (P : pointer) : ptrint;
begin
if (p <> nil) then
dec(p,sizeof(ptrint));
Free(P);
CFreeMem:=0;
end;
Function CFreeMemSize(p:pointer;Size:ptrint):ptrint;
begin
if size<=0 then
begin
if size<0 then
runerror(204);
exit;
end;
if (p <> nil) then
begin
if (size <> pptrint(p-sizeof(ptrint))^) then
runerror(204);
end;
CFreeMemSize:=CFreeMem(P);
end;
Function CAllocMem(Size : ptrint) : Pointer;
begin
CAllocMem:=calloc(Size+sizeof(ptrint),1);
if (CAllocMem <> nil) then
begin
pptrint(CAllocMem)^ := size;
inc(CAllocMem,sizeof(ptrint));
end;
end;
Function CReAllocMem (var p:pointer;Size:ptrint):Pointer;
begin
if size=0 then
begin
if p<>nil then
begin
dec(p,sizeof(ptrint));
free(p);
p:=nil;
end;
end
else
begin
inc(size,sizeof(ptrint));
if p=nil then
p:=malloc(Size)
else
begin
dec(p,sizeof(ptrint));
p:=realloc(p,size);
end;
if (p <> nil) then
begin
pptrint(p)^ := size-sizeof(ptrint);
inc(p,sizeof(ptrint));
end;
end;
CReAllocMem:=p;
end;
Function CMemSize (p:pointer): ptrint;
begin
CMemSize:=pptrint(p-sizeof(ptrint))^;
end;
function CGetHeapStatus:THeapStatus;
var res: THeapStatus;
begin
fillchar(res,sizeof(res),0);
CGetHeapStatus:=res;
end;
function CGetFPCHeapStatus:TFPCHeapStatus;
begin
fillchar(CGetFPCHeapStatus,sizeof(CGetFPCHeapStatus),0);
end;
Const
CMemoryManager : TMemoryManager =
(
NeedLock : false;
GetMem : @CGetmem;
FreeMem : @CFreeMem;
FreememSize : @CFreememSize;
AllocMem : @CAllocMem;
ReallocMem : @CReAllocMem;
MemSize : @CMemSize;
InitThread : Nil;
DoneThread : Nil;
RelocateHeap : Nil;
GetHeapStatus : @CGetHeapStatus;
GetFPCHeapStatus: @CGetFPCHeapStatus;
);
Var
OldMemoryManager : TMemoryManager;
Initialization
GetMemoryManager (OldMemoryManager);
SetMemoryManager (CmemoryManager);
Finalization
SetMemoryManager (OldMemoryManager);
end.
Because Free Pascal for dos is a 32 bit compiler, and uses a dos extender, accessing DOS memory isn’t trivial. What follows is an attempt to an explanation of how to access and use dos or real mode memory3.
In Proteced Mode, memory is accessed through Selectors and Offsets. You can think of Selectors as the protected mode equivalents of segments.
In Free Pascal , a pointer is an offset into the DS selector, which points to the Data of your program.
To access the (real mode) dos memory, somehow you need a selector that points to the dos memory. The go32 unit provides you with such a selector: The DosMemSelector variable, as it is conveniently called.
You can also allocate memory in dos ’s memory space, using the global_dos_alloc function of the go32 unit. This function will allocate memory in a place where dos sees it.
As an example, here is a function that returns memory in real mode dos and returns a selector:offset pair for it.
procedure dosalloc(var selector : word;
var segment : word;
size : longint);
var result : longint;
begin
result := global_dos_alloc(size);
selector := word(result);
segment := word(result shr 16);
end;
(You need to free this memory using the global_dos_free function.)
You can access any place in memory using a selector. You can get a selector using the function:
function allocate_ldt_descriptors(count : word) : word;
and then let this selector point to the physical memory you want using the function
function set_segment_base_address(d : word;s : longint) : boolean;
Its length can be set using the function:
function set_segment_limit(d : word;s : longint) : boolean;
You can manipulate the memory pointed to by the selector using the functions of the GO32 unit. For instance with the seg_fillchar function. After using the selector, you must free it again using the function:
function free_ldt_descriptor(d : word) : boolean;
More information on all this can be found in the Unit reference , the chapter on the go32 unit.
The fact that 16-bit code is no longer used, means that some of the older Turbo Pascal constructs and functions are obsolete. The following is a list of functions which shouldn’t be used anymore:
You shouldn’t use these functions, since they are very non-portable, they’re specific to dos and the 80x86 processor. The Free Pascal compiler is designed to be portable to other platforms, so you should keep your code as portable as possible, and not system specific. That is, unless you’re writing some driver units, of course.
The old Turbo Pascal functions MemAvail and MaxAvail functions are no longer available in Free Pascal as of version 2.0. The reason for this incompatibility is below:
On modern operating systems, 4 the idea of "Available Free Memory" is not valid for an application. The reasons are:
Therefore, programs using MemAvail and MaxAvail functions should be rewritten so they no longer use these functions, because it does not make sense anymore on modern OS’es. There are 3 possibilities:
Resource strings primarily exist to make internationalization of applications easier, by introducing a language construct that provides a uniform way of handling constant strings.
Most applications communicate with the user through some messages on the graphical screen or console. Storing these messages in special constants allows storing them in a uniform way in separate files, which can be used for translation. A programmers interface exists to manipulate the actual values of the constant strings at runtime, and a utility tool comes with the Free Pascal compiler to convert the resource string files to whatever format is wanted by the programmer. Both these things are discussed in the following sections.
When a unit is compiled that contains a resourcestring section, the compiler does 2 things:
This approach has 2 advantages: first of all, the value of the string is always present in the program. If the programmer doesn’t care to translate the strings, the default values are always present in the binary. This also avoids having to provide a file containing the strings. Secondly, having all strings together in a compiler generated file ensures that all strings are together (you can have multiple resourcestring sections in 1 unit or program) and having this file in a fixed format, allows the programmer to choose his way of internationalization.
For each unit that is compiled and that contains a resourcestring section, the compiler generates a file that has the name of the unit, and an extension .rst. The format of this file is as follows:
If the unit contains no resourcestring section, no file is generated.
For example, the following unit:
unit rsdemo;
{$mode delphi}
{$H+}
interface
resourcestring
First = 'First';
Second = 'A Second very long string that should cover more than 1 line';
implementation
end.
Will result in the following resource string file:
# hash value = 5048740 rsdemo.first='First' # hash value = 171989989 rsdemo.second='A Second very long string that should cover more than 1 li'+ 'ne'
The hash value is calculated with the function Hash. It is present in the objpas unit. The value is the same value that the GNU gettext mechanism uses. It is in no way unique, and can only be used to speed up searches.
The rstconv utility that comes with the Free Pascal compiler allows manipulation of these resource string files. At the moment, it can only be used to make a .po file that can be fed to the GNU msgfmt program. If someone wishes to have another format (Win32 resource files spring to mind), one can enhance the rstconv program so it can generate other types of files as well. GNU gettext was chosen because it is available on all platforms, and is already widely used in the Unix and free software community. Since the Free Pascal team doesn’t want to restrict the use of resource strings, the .rst format was chosen to provide a neutral method, not restricted to any tool.
If you use resource strings in your units, and you want people to be able to translate the strings, you must provide the resource string file. Currently, there is no way to extract them from the unit file, though this is in principle possible. It is not required to do this, the program can be compiled without it, but then the translation of the strings isn’t possible.
Having compiled a program with resourcestrings is not enough to internationalize your program. At run-time, the program must initialize the string tables with the correct values for the language that the user selected. By default no such initialization is performed. All strings are initialized with their declared values.
The objpas unit provides the mechanism to correctly initialize the string tables. There is no need to include this unit in a uses clause, since it is automatically loaded when a program or unit is compiled in Delphi or objfpc mode. Since one of these mode is required to use resource strings, the unit is always loaded when needed anyway.
The resource strings are stored in tables, one per unit, and one for the program, if it contains a resourcestring section as well. Each resourcestring is stored with its name, hash value, default value, and the current value, all as AnsiStrings.
The objpas unit offers methods to retrieve the number of resourcestring tables, the number of strings per table, and the above information for each string. It also offers a method to set the current value of the strings.
Here are the declarations of all the functions:
Function ResourceStringTableCount : Longint;
Function ResourceStringCount(TableIndex: longint): longint;
Function GetResourceStringName(TableIndex,
StringIndex: Longint): Ansistring;
Function GetResourceStringHash(TableIndex,
StringIndex: Longint): Longint;
Function GetResourceStringDefaultValue(TableIndex,
StringIndex: Longint): AnsiString;
Function GetResourceStringCurrentValue(TableIndex,
StringIndex: Longint): AnsiString;
Function SetResourceStringValue(TableIndex,
StringIndex : longint;
Value: Ansistring): Boolean;
Procedure SetResourceStrings (SetFunction: TResourceIterator);
Two other function exist, for convenience only:
Function Hash(S: AnsiString): longint; Procedure ResetResourceTables;
Here is a short explanation of what each function does. A more detailed explanation of the functions can be found in the Reference guide .
Two other functions exist, for convenience only:
Given some Translate function, the following code would initialize all resource strings:
Var I,J : Longint;
S : AnsiString;
begin
For I:=0 to ResourceStringTableCount-1 do
For J:=0 to ResourceStringCount(i)-1 do
begin
S:=Translate(GetResourceStringDefaultValue(I,J));
SetResourceStringValue(I,J,S);
end;
end;
Other methods are of course possible, and the Translate function can be implemented in a variety of ways.
The unit gettext provides a way to internationalize an application with the GNU gettext utilities. This unit is supplied with the Free Component Library (FCL). it can be used as follows:
for a given application, the following steps must be followed:
TranslateResourcestrings('intl/restest.%s.mo');
the %s specifier will be replaced by the contents of the LANG
environment variable. This call should happen at program startup.
An example program exists in the FCL-base sources, in the fcl-base/tests directory.
In principle it is possible to translate all resource strings at any time in a running program. However, this change is not communicated to other strings; its change is noticed only when a constant string is being used.
Consider the following example:
Const
help = 'With a little help of a programmer.';
Var
A : AnsiString;
begin
{ lots of code }
A:=Help;
{ Again some code}
TranslateStrings;
{ More code }
After the call to TranslateStrings, the value of A will remain unchanged. This means that the assignment A:=Help must be executed again in order for the change to become visible. This is important, especially for GUI programs which have e.g. a menu. In order for the change in resource strings to become visible, the new values must be reloaded by program code into the menus …
Free Pascal supports thread programming: There is a language construct available for thread-local storage (ThreadVar), and cross-platform low-level thread routines are available for those operating systems that support threads.
All routines for threading are available in the system unit, under the form of a thread manager. A thread manager must implement some basic routines which the RTL needs to be able to support threading. For Windows, a default threading manager is integrated in the system unit. For other platforms, a thread manager must be included explicitly by the programmer. On systems where posix threads are available, the cthreads unit implements a thread manager which uses the C POSIX thread library. No native pascal thread library exists for such systems.
Although it is not forbidden to do so, it is not recommended to use system-specific threading routines: The language support for multithreaded programs will not be enabled, meaning that threadvars will not work, the heap manager will be confused which may lead to severe program errors.
If no threading support is present in the binary, the use of thread routines or the creation of a thread will result in an exception or a run-time error 232.
For linux (and other Unixes), the C thread manager can be enabled by inserting the cthreads unit in the program’s unit clause. Without this, threading programs will give an error when started. It is imperative that the unit be inserted as early in the uses clause as possible.
At a later time, a system thread manager may be implemented which implements threads without Libc support.
The following sections show how to program threads, and how to protect access to data common to all threads using (cross-platform) critical sections. Finally, the thread manager is explained in more detail.
To start a new thread, the BeginThread function should be used. It has one mandatory argument: the function which will be executed in the new thread. The result of the function is the exit result of the thread. The thread function can be passed a pointer, which can be used to acces initialization data: The programmer must make sure that the data is accessible from the thread and does not go out of scope before the thread has accessed it.
Type
TThreadFunc = function(parameter : pointer) : ptrint;
function BeginThread(sa : Pointer;
stacksize : SizeUInt;
ThreadFunction : tthreadfunc;
p : pointer;
creationFlags : dword;
var ThreadId : TThreadID) : TThreadID;
This rather complicated full form of the function also comes in more simplified forms:
function BeginThread(ThreadFunction : tthreadfunc) : TThreadID;
function BeginThread(ThreadFunction : tthreadfunc;
p : pointer) : TThreadID;
function BeginThread(ThreadFunction : tthreadfunc;
p : pointer;
var ThreadId : TThreadID) : TThreadID;
function BeginThread(ThreadFunction : tthreadfunc;
p : pointer;
var ThreadId : TThreadID;
const stacksize: SizeUInt) : TThreadID;
The parameters have the following meaning:
The newly started thread will run until the ThreadFunction exits, or until it explicitly calls the EndThread function:
procedure EndThread(ExitCode : DWord); procedure EndThread;
The exitcode can be examined by the code which started the thread.
The following is a small example of how to program a thread:
{$mode objfpc}
uses
sysutils {$ifdef unix},cthreads{$endif} ;
const
threadcount = 100;
stringlen = 10000;
var
finished : longint;
threadvar
thri : ptrint;
function f(p : pointer) : ptrint;
var
s : ansistring;
begin
Writeln('thread ',longint(p),' started');
thri:=0;
while (thri<stringlen) do
begin
s:=s+'1';
inc(thri);
end;
Writeln('thread ',longint(p),' finished');
InterLockedIncrement(finished);
f:=0;
end;
var
i : longint;
begin
finished:=0;
for i:=1 to threadcount do
BeginThread(@f,pointer(i));
while finished<threadcount do ;
Writeln(finished);
end.
The InterLockedIncrement is a thread-safe version of the standard Inc function.
To provide system-independent support for thread programming, some utility functions are implemented to manipulate threads. To use these functions the thread ID must have been retrieved when the thread was started, because most functions require the ID to identify the thread on which they should act:
function SuspendThread(threadHandle: TThreadID): dword;
function ResumeThread(threadHandle: TThreadID): dword;
function KillThread(threadHandle: TThreadID): dword;
function WaitForThreadTerminate(threadHandle: TThreadID;
TimeoutMs : longint): dword;
function ThreadSetPriority(threadHandle: TThreadID;
Prio: longint): boolean;
function ThreadGetPriority(threadHandle: TThreadID): Integer;
function GetCurrentThreadId: dword;
procedure ThreadSwitch;
The meaning of these functions should be clear:
When programming threads, it is sometimes necessary to avoid concurrent access to certain resources, or to avoid having a certain routine executed by two threads. This can be done using a Critical Section. The FPC heap manager uses critical sections when multithreading is enabled.
The TRTLCriticalSection type is an Opaque type; it depends on the OS on which the code is executed. It should be initialized before it is first used, and should be disposed of when it is no longer necessary.
To protect a piece of code, a call to EnterCriticalSection should be made: When this call returns, it is guaranteed that the current thread is the only thread executing the subsequent code. The call may have suspended the current thread for an indefinite time to ensure this.
When the protected code is finished, LeaveCriticalSection must be called: this will enable other threads to start executing the protected code. To minimize waiting time for the threads, it is important to keep the protected block as small as possible.
The definition of these calls is as follows:
procedure InitCriticalSection(var cs: TRTLCriticalSection); procedure DoneCriticalSection(var cs: TRTLCriticalSection); procedure EnterCriticalSection(var cs: TRTLCriticalSection); procedure LeaveCriticalSection(var cs: TRTLCriticalSection);
The meaning of these calls is again almost obvious:
Note that the LeaveCriticalsection call must be executed. Failing to do so will prevent all other threads from executing the code in the critical section. It is therefore good practice to enclose the critical section in a Try..finally block. Typically, the code will look as follows:
Var
MyCS : TRTLCriticalSection;
Procedure CriticalProc;
begin
EnterCriticalSection(MyCS);
Try
// Protected Code
Finally
LeaveCriticalSection(MyCS);
end;
end;
Procedure ThreadProcedure;
begin
// Code executed in threads...
CriticalProc;
// More Code executed in threads...
end;
begin
InitCriticalSection(MyCS);
// Code to start threads.
DoneCriticalSection(MyCS);
end.
Just like the heap is implemented using a heap manager, and widestring management is left to a widestring manager, the threads have been implemented using a thread manager. This means that there is a record which has fields of procedural type for all possible functions used in the thread routines. The thread routines use these fields to do the actual work.
The thread routines install a system thread manager specific for each system. On Windows, the normal Windows routines are used to implement the functions in the thread manager. On Linux and other unices, the system thread manager does nothing: it will generate an error when thread routines are used. The rationale is that the routines for thread management are located in the C library. Implementing the system thread manager would make the RTL dependent on the C library, which is not desirable. To avoid dependency on the C library, the Thread Manager is implemented in a separate unit (cthreads). The initialization code of this unit sets the thread manager to a thread manager record which uses the C (pthreads) routines.
The thread manager record can be retrieved and set just as the record for the heap manager. The record looks (currently) as follows:
TThreadManager = Record InitManager : Function : Boolean; DoneManager : Function : Boolean; BeginThread : TBeginThreadHandler; EndThread : TEndThreadHandler; SuspendThread : TThreadHandler; ResumeThread : TThreadHandler; KillThread : TThreadHandler; ThreadSwitch : TThreadSwitchHandler; WaitForThreadTerminate : TWaitForThreadTerminateHandler; ThreadSetPriority : TThreadSetPriorityHandler; ThreadGetPriority : TThreadGetPriorityHandler; GetCurrentThreadId : TGetCurrentThreadIdHandler; InitCriticalSection : TCriticalSectionHandler; DoneCriticalSection : TCriticalSectionHandler; EnterCriticalSection : TCriticalSectionHandler; LeaveCriticalSection : TCriticalSectionHandler; InitThreadVar : TInitThreadVarHandler; RelocateThreadVar : TRelocateThreadVarHandler; AllocateThreadVars : TAllocateThreadVarsHandler; ReleaseThreadVars : TReleaseThreadVarsHandler; end;
The meaning of most of these functions should be obvious from the descriptions in previous sections.
The InitManager and DoneManager are called when the threadmanager is set (InitManager), or when it is unset (DoneManager). They can be used to initialize the thread manager or to clean up when it is done. If either of them returns False, the operation fails.
There are some special entries in the record, linked to thread variable management:
TInitThreadVarHandler = Procedure(var offset: dword;
size: dword);
The offset parameter indicates the offset in the thread variable
block: All thread variables are located in a single block, one after the
other. The size parameter indicates the size of the thread variable. This
function will be called once for all thread variables in the program.
TRelocateThreadVarHandler = Function(offset : dword) : pointer;It should return the new location for the thread-local variable.
The following sections describe the general optimizations done by the compiler, they are not processor specific. Some of these require some compiler switch override while others are done automatically (those which require a switch will be noted as such).
In Free Pascal , if the operand(s) of an operator are constants, they will be evaluated at compile time.
Example
x:=1+2+3+6+5;
will generate the same code as
x:=17;
Furthermore, if an array index is a constant, the offset will be evaluated at compile time. This means that accessing MyData[5] is as efficient as accessing a normal variable.
Finally, calling Chr, Hi, Lo, Ord, Pred, or Succ functions with constant parameters generates no run-time library calls, instead, the values are evaluated at compile time.
Using the same constant string, floating point value or constant set two or more times generates only one copy of that constant.
Evaluation of boolean expression stops as soon as the result is known, which makes code execute faster then if all boolean operands were evaluated.
Using the in operator is always more efficient then using the
equivalent <>, =, <=, >=, < and >
operators. This is because range comparisons can be done more easily with
the in operator than with normal comparison operators.
Sets which contain less than 33 elements can be directly encoded using a 32-bit value, therefore no run-time library calls to evaluate operands on these sets are required; they are directly encoded by the code generator.
Assignments of constants to variables are range checked at compile time, which removes the need of the generation of runtime range checking code.
When the second operand of a mod on an unsigned value is a constant power of 2, an and instruction is used instead of an integer division. This generates more efficient code.
When one of the operands in a multiplication is a power of two, they are encoded using arithmetic shift instructions, which generates more efficient code.
Similarly, if the divisor in a div operation is a power of two, it is encoded using arithmetic shift instructions.
The same is true when accessing array indexes which are powers of two, the address is calculated using arithmetic shifts instead of the multiply instruction.
By default all variables larger then a byte are guaranteed to be aligned at least on a word boundary.
Alignment on the stack and in the data section is processor dependent.
This feature removes all unreferenced code in the final executable file, making the executable file much smaller.
Smart linking is switched on with the -Cx command line switch, or using the {$SMARTLINK ON} global directive.
The following runtime library routines are coded directly into the final executable: Lo, Hi, High, Sizeof, TypeOf, Length, Pred, Succ, Inc, Dec and Assigned.
Under specific conditions, the stack frame (entry and exit code for the routine, see section section ?? ) will be omitted, and the variable will directly be accessed via the stack pointer.
Conditions for omission of the stack frame:
When using the -Or switch, local variables or parameters which are used very often will be moved to registers for faster access.
This lists the low-level optimizations performed, on a processor per processor basis.
Here follows a listing of the optimizing techniques used in the compiler:
Although you can enable uncertain optimizations in most cases, for people who do not understand the following technical explanation, it might be the safest to leave them off.
Remark: If uncertain optimizations are enabled, the CSE algortihm assumes that
The practical upshot of this is that you cannot use the uncertain optimizations if you both write and read local or global variables directly and through pointers (this includes Var parameters, as those are pointers too).
The following example will produce bad code when you switch on uncertain optimizations:
Var temp: Longint;
Procedure Foo(Var Bar: Longint);
Begin
If (Bar = temp)
Then
Begin
Inc(Bar);
If (Bar <> temp) then Writeln('bug!')
End
End;
Begin
Foo(Temp);
End.
The reason it produces bad code is because you access the global variable Temp both through its name Temp and through a pointer, in this case using the Bar variable parameter, which is nothing but a pointer to Temp in the above code.
On the other hand, you can use the uncertain optimizations if you access global/local variables or parameters through pointers, and only access them through this pointer1.
For example:
Type TMyRec = Record
a, b: Longint;
End;
PMyRec = ^TMyRec;
TMyRecArray = Array [1..100000] of TMyRec;
PMyRecArray = ^TMyRecArray;
Var MyRecArrayPtr: PMyRecArray;
MyRecPtr: PMyRec;
Counter: Longint;
Begin
New(MyRecArrayPtr);
For Counter := 1 to 100000 Do
Begin
MyRecPtr := @MyRecArrayPtr^[Counter];
MyRecPtr^.a := Counter;
MyRecPtr^.b := Counter div 2;
End;
End.
Will produce correct code, because the global variable MyRecArrayPtr is not accessed directly, but only through a pointer (MyRecPtr in this case).
In conclusion, one could say that you can use uncertain optimizations only when you know what you’re doing.
Using the -O2 (the default) switch does several optimizations in the code produced, the most notable being:
This is where the various optimizing switches and their actions are described, grouped per switch.
subl $4,%esp") instead of slower, smaller instructions
("enter $4"). This is the default setting.movzbl (mem), %eax|to a combination of simpler instructions
xorl %eax, %eax movb (mem), %alfor the Pentium.
Here, some general tips for getting better code are presented. They mainly concern coding style.
Here are some tips given to get the smallest code possible.
Traditionally, compilers optimise a program procedure by procedure, or at best compilation unit per compilation unit. Whole program optimisation (WPO) means that the compiler considers all compilation units that make up a program or library and optimises them using the combined knowledge of how they are used together in this particular case.
The way WPO generally works is as follows:
This is the scheme followed by Free Pascal .
The implementation of this scheme is highly compiler dependent. Another implementation could be that the compiler generates some kind of intermediary code (e.g., byte code) and the linker performs all wpo along with the translation to the target machine code
A few general principles have been followed when designing the FPC implementation of WPO:
The first step in WPO is to compile the program (or library) and all of its units as it would be done normally, but specifying in addition the 2 following options on the command-line:
-FW/path/to/feedbackfile.wpo -OW<selected_wpo_options>
The first option tells the compiler where the WPO feedback file should be written, the second option tells the compiler to switch on WPO optimalizations.
The compiler will then, right after the program or library has been linked, collect all necessary information to perform the requested WPO options during a subsequent compilation, and will store this information in the indicated file.
To actually apply the WPO options, the program (or library) and all or some of the units that it uses, must be recompiled using the option
-Fw/path/to/feedbackfile.wpo -Ow<selected_wpo_options>
(Note the small caps in the w). This will tell the compiler to use the feedback file generated in the previous step. The compiler will then read the information collected about the program during the previous compiler run, and use it during the current compilation of units and/or program/library.
Units not recompiled during the second pass will obviously not be optimised, but they will still work correctly when used together with the optimised units and program/library.
Remark: Note that the options must always be specified on the command-line: there is no source directive to turn on WPO, as it makes only sense to use WPO when compiling a complete program.
The -OW and -Ow command-line options require a comma-separated list of whole-program-optimization options. These are strings, each string denotes an option. The following is a list of available options:
There are 2 limitations to this option:
This option has 2 limitations:
Again, there are some limitations:
This information is mainly interesting if external data must be added to the WPO feedback file, e.g. from a profiling tool. For regular use of the WPO feature, the following information is not needed and can be ignored.
The file consists of comments and a number of sections. Comments are lines that start with a #. Each section starts with "% " followed by the name of the section (e.g.,% contextinsensitive_devirtualization).
After that, until either the end of the file or until the next line starting with with "% ", first a human readable description follows of the format of this section (in comments), and then the contents of the section itself.
There are no rules for how the contents of a section should look, except that lines starting with # are reserved for comments and lines starting with % are reserved for section markers.
Free Pascal supports the creation of shared libraries on several operating systems. The following table (table (??) ) indicates which operating systems support the creation of shared libraries.
Table 12.1: Shared library support
Operating systems Library extension Library prefix linux .so lib windows .dll <none> BeOS .so lib FreeBSD .so lib NetBSD .so lib
The library prefix column indicates how the names of the libraries are resolved and created. For example, under linux , the library name will alwaus have the lib prefix when it is created. So if you create a library called mylib, under linux , this will result in the libmylib.so. Furthermore, when importing routines from shared libraries, it is not necessary to give the library prefix or the filename extension.
In the following sections we discuss how to create a library, and how to use these libraries in programs.
Creation of libraries is supported in any mode of the Free Pascal compiler, but it may be that the arguments or return values differ if the library is compiled in 2 different modes. E.g. if your function expects an Integer argument, then the library will expect different integer sizes if you compile it in Delphi mode or in TP mode.
A library can be created just as a program, only it uses the library keyword, and it has an exports section. The following listing demonstrates a simple library:
\input{\exampledir/#1.pp}The function SubStr does not have to be declared in the library file itself. It can also be declared in the interface section of a unit that is used by the library.
Compilation of this source will result in the creation of a library called libsubs.so on unix systems, or subs.dll on Windows or os/2 . The compiler will take care of any additional linking that is required to create a shared library.
The library exports one function: SubStr. The case is important. The case as it appears in the exports clause is used to export the function.
If you want your library to be called from programs compiled with other compilers, it is important to specify the correct calling convention for the exported functions. Since the generated programs by other compilers do not know about the Free Pascal calling conventions, your functions would be called incorrectly, resulting in a corrupted stack.
On Windows , most libraries use the stdcall convention, so it may be better to use that one if your library is to be used on Windows systems. On most unix systems, the C calling convention is used, therefore the cdecl modifier should be used in that case.
In order to use a function that resides in a library, it is sufficient to declare the function as it exists in the library as an external function, with correct arguments and return type. The calling convention used by the function should be declared correctly as well. The compiler will then link the library as specified in the external statement to your program1.
For example, to use the library as defined above from a pascal program, you can use the following pascal program:
\input{\exampledir/#1.pp}As is shown in the example, you must declare the function as external. Here also, it is necessary to specify the correct calling convention (it should always match the convention as used by the function in the library), and to use the correct casing for your declaration. Also notice, that the library importing did not specify the filename extension, nor was the lib prefix added.
This program can be compiled without any additional command-switches, and should run just like that, provided the library is placed where the system can find it. For example, on linux , this is /usr/lib or any directory listed in the /etc/ld.so.conf file. On Windows , this can be the program directory, the Windows system directory, or any directoy mentioned in the PATH.
Using the library in this way links the library to your program at compile time. This means that
Or it may simply be that you don’t know the name of the function to be called, you just know the arguments it expects.
It is therefore also possible to load the library at run-time, store the function address in a procedural variable, and use this procedural variable to access the function in the library.
The following example demonstrates this technique:
\input{\exampledir/#1.pp}As in the case of compile-time linking, the crucial thing in this listing is the declaration of the TSubStrFunc type. It should match the declaration of the function you’re trying to use. Failure to specify a correct definition will result in a faulty stack or, worse still, may cause your program to crash with an access violation.
Remark: The examples in this section assume a linux system; similar commands as the ones below exist for other operating systems, though.
You can also call a Free Pascal generated library from a C program:
\input{\exampledir/#1.c}To compile this example, the following command can be used:
gcc -o ctest ctest.c -lsubs
provided the code is in ctest.c.
The library can also be loaded dynamically from C, as shown in the following example:
\input{\exampledir/#1.c}This can be compiled using the following command:
gcc -o ctest2 ctest2.c -ldl
The -ldl tells gcc that the program needs the libdl.so library to load dynamical libraries.
By default, Free Pascal (actually, the linker used by Free Pascal ) creates libraries that are not relocatable. This means that they must be loaded at a fixed address in memory: this address is called the ImageBase address. If two Free Pascal generated libraries are loaded by a program, there will be a conflict, because the first librarie already occupies the memory location where the second library should be loaded.
There are 2 switches in Free Pascal which control the generation of shared libraries under Windows :
The first option is preferred, as a program may load many libraries present on the system, and they could already be using the ImageBase address. The second option is faster, as no relocation needs to be done if the ImageBase address is not yet in use.
Under Windows and linux (or any platform using ELF binaries) 1, you can include resources in your executable or library using the {$R filename} directive. These resources can then be accessed through the standard Windows API calls: these calls have been made available in the other platforms as well.
When the compiler encounters a resource directive, it just creates an entry in the unit .ppu file; it doesn’t link the resource. Only when it creates a library or executable, it looks for all the resource files for which it encountered a directive, and tries to link them in.
The default extension for resource files is .res. When the filename has as the first character an asterix (*), the compiler will replace the asterix with the name of the current unit, library or program.
Remark: This means that the asterix may only be used after a unit, library or program clause.
The Free Pascal compiler itself doesn’t create any resource files; it just compiles them into the executable. To create resource files, you can use some GUI tools as the Borland resource workshop; but it is also possible to use a Windows resource compiler like gnu windres. windres comes with the gnu binutils, but the Free Pascal distribution also contains a version which you can use.
The usage of windres is straightforward; it reads an input file describing the resources to create and outputs a resource file.
A typical invocation of windres would be
windres -i mystrings.rc -o mystrings.res
this will read the mystrings.rc file and output a mystrings.res resource file.
A complete overview of the windres tools is outside the scope of this document, but here are some things you can use it for:
Some of these will be described below.
String tables can be used to store and retrieve large collections of strings in your application.
A string table looks as follows:
STRINGTABLE { 1, "hello World !"
2, "hello world again !"
3, "last hello world !" }
You can compile this (we assume the file is called tests.rc) as follows:
windres -i tests.rc -o tests.res
And this is the way to retrieve the strings from your program:
program tests;
{$mode objfpc}
Uses Windows;
{$R *.res}
Function LoadResourceString (Index : longint): Shortstring;
begin
SetLength(Result,LoadString(FindResource(0,Nil,RT_STRING),
Index,
@Result[1],
SizeOf(Result))
)
end;
Var
I: longint;
begin
For i:=1 to 3 do
Writeln(LoadResourceString(I));
end.
The call to FindResource searches for the stringtable in the compiled-in resources. The LoadString function then reads the string with index i out of the table, and puts it in a buffer, which can then be used. Both calls are in the windows unit.
The win32 API allows the storing of version information in your binaries. This information can be made visible with the Windows Explorer, by right-clicking on the executable or library, and selecting the ’Properties’ menu. In the tab ’Version’ the version information will be displayed.
Here is how to insert version information in your binary:
1 VERSIONINFO
FILEVERSION 4, 0, 3, 17
PRODUCTVERSION 3, 0, 0, 0
FILEFLAGSMASK 0
FILEOS 0x40000
FILETYPE 1
{
BLOCK "StringFileInfo"
{
BLOCK "040904E4"
{
VALUE "CompanyName", "Free Pascal"
VALUE "FileDescription", "Free Pascal version information extractor"
VALUE "FileVersion", "1.0"
VALUE "InternalName", "Showver"
VALUE "LegalCopyright", "GNU Public License"
VALUE "OriginalFilename", "showver.pp"
VALUE "ProductName", "Free Pascal"
VALUE "ProductVersion", "1.0"
}
}
}
As you can see, you can insert various kinds of information in the version info block. The keyword VERSIONINFO marks the beginning of the version information resource block. The keywords FILEVERSION, PRODUCTVERSION give the actual file version, while the block StringFileInfo gives other information that is displayed in the explorer.
The Free Component Library comes with a unit (fileinfo) that allows to extract and view version information in a straightforward and easy manner; the demo program that comes with it (showver) shows version information for an arbitrary executable or DLL.
When Windows shows an executable in the Explorer, it looks for an icon in the executable to show in front of the filename, the application icon.
Inserting an application icon is very easy and can be done as follows
AppIcon ICON "filename.ico"
This will read the file filename.ico and insert it in the resource file.
Sometimes you want to use symbolic names in your resource file, and use the same names in your program to access the resources. To accomplish this, there exists a preprocessor for windres that understands pascal syntax: fprcp. This preprocessor is shipped with the Free Pascal distribution.
The idea is that the preprocessor reads a pascal unit that has some symbolic constants defined in it, and replaces symbolic names in the resource file by the values of the constants in the unit:
As an example: consider the following unit:
unit myunit; interface Const First = 1; Second = 2: Third = 3; Implementation end.
And the following resource file:
#include "myunit.pp"
STRINGTABLE { First, "hello World !"
Second, "hello world again !"
Third, "last hello world !" }
If you invoke windres with the preprocessor option:
windres --preprocessor fprcp -i myunit.rc -o myunit.res
then the preprocessor will replace the symbolic names ’first’, ’second’ and ’third’ with their actual values.
In your program, you can then refer to the strings by their symbolic names (the constants) instead of using a numeric index.
As described in chapter ?? , unit description files (hereafter called PPU files for short), are used to determine if the unit code must be recompiled or not. In other words, the PPU files act as mini-makefiles, which is used to check dependencies of the different code modules, as well as verify if the modules are up to date or not. Furthermore, it contains all public symbols defined for a module.
The general format of the ppu file format is shown in Figure ?? .
To read or write the ppufile, the ppu unit ppu.pas can be used, which has an object called tppufile which holds all routines that deal with ppufile handling. While describing the layout of a ppufile, the methods which can be used for it are presented as well.
A unit file consists of basically five or six parts:
We will first create an object ppufile which will be used below. We are opening unit test.ppu as an example.
var
ppufile : pppufile;
begin
{ Initialize object }
ppufile:=new(pppufile,init('test.ppu');
{ open the unit and read the header, returns false when it fails }
if not ppufile.openfile then
error('error opening unit test.ppu');
{ here we can read the unit }
{ close unit }
ppufile.closefile;
{ release object }
dispose(ppufile,done);
end;
Note: When a function fails (for example not enough bytes left in an entry) it sets the ppufile.error variable.
The header consists of a record (tppuheader) containing several pieces of information for recompilation. This is shown in table (??) . The header is always stored in little-endian format.
Table A.1: PPU Header
offset size (bytes) description 00h 3 Magic : ’PPU’ in ASCII 03h 3 PPU File format version (e.g : ’021’ in ASCII) 06h 2 Compiler version used to compile this module (major,minor) 08h 2 Code module target processor 0Ah 2 Code module target operating system 0Ch 4 Flags for PPU file 10h 4 Size of PPU file (without header) 14h 4 CRC-32 of the entire PPU file 18h 4 CRC-32 of partial data of PPU file (public data mostly) 1Ch 8 Reserved
The header is already read by the ppufile.openfile command. You can access all fields using ppufile.header which holds the current header record.
Table A.2: PPU CPU Field values
value description 0 unknown 1 Intel 80x86 or compatible 2 Motorola 680x0 or compatible 3 Alpha AXP or compatible 4 PowerPC or compatible
Some of the possible flags in the header are described in table (??) . Not all the flags are described, for more information, read the source code of ppu.pas.
Table A.3: PPU Header Flag values
Symbolic bit flag name Description uf_init Module has an initialization (either Delphi or TP style) section. uf_finalize Module has a finalization section. uf_big_endian All the data stored in the chunks is in big-endian format. uf_has_browser Unit contains symbol browser information. uf_smart_linked The code module has been smartlinked. uf_static_linked The code is statically linked. uf_has_resources Unit has resource section.
Apart from the header section, all the data in the PPU file is separated into data blocks, which permit easily adding additional data blocks, without compromising backward compatibility. This is similar to both Electronic Arts IFF chunk format and Microsoft’s RIFF chunk format.
Each ’chunk’ (tppuentry) has the following format, and can be nested:
Table A.4: chunk data format
offset size (bytes) description 00h 1 Block type (nested (2) or main (1)) 01h 1 Block identifier 02h 4 Size of this data block 06h+ <variable> Data for this block
Each main section chunk must end with an end chunk. Nested chunks are used for record, class or object fields.
To read an entry you can simply call ppufile.readentry:byte, it returns the tppuentry.nr field, which holds the type of the entry. A common way how this works is (example is for the symbols):
repeat
b:=ppufile.readentry;
case b of
ib<etc> : begin
end;
ibendsyms : break;
end;
until false;
The possible entry types are found in ppu.pas, but a short description of the most common ones are shown in table (??) .
Table A.5: Possible PPU Entry types
Symbolic name Location Description ibmodulename General Name of this unit. ibsourcefiles General Name of source files. ibusedmacros General Name and state of macros used. ibloadunit General Modules used by this units. inlinkunitofiles General Object files associated with this unit. iblinkunitstaticlibs General Static libraries associated with this unit. iblinkunitsharedlibs General Shared libraries associated with this unit. ibendinterface General End of General information section. ibstartdefs Interface Start of definitions. ibenddefs Interface End of definitions. ibstartsyms Interface Start of symbol data. ibendsyms Interface End of symbol data. ibendimplementation Implementation End of implementation data. ibendbrowser Browser End of browser section. ibend General End of Unit file.
Then you can parse each entry type yourself. ppufile.readentry will take care of skipping unread bytes in the entry and reads the next entry correctly! A special function is skipuntilentry(untilb:byte):boolean; which will read the ppufile until it finds entry untilb in the main entries.
Parsing an entry can be done with ppufile.getxxx functions. The available functions are:
procedure ppufile.getdata(var b;len:longint); function getbyte:byte; function getword:word; function getlongint:longint; function getreal:ppureal; function getstring:string;
To check if you’re at the end of an entry you can use the following function:
function EndOfEntry:boolean;
notes:
A complete list of entries and what their fields contain can be found in ppudump.pp.
Creating a new ppufile works almost the same as reading one. First you need to init the object and call create:
ppufile:=new(pppufile,init('output.ppu'));
ppufile.createfile;
After that you can simply write all needed entries. You’ll have to take care that you write at least the basic entries for the sections:
ibendinterface ibenddefs ibendsyms ibendbrowser (only when you've set uf_has_browser!) ibendimplementation ibend
Writing an entry is a little different than reading it. You need to first put everything in the entry with ppufile.putxxx:
procedure putdata(var b;len:longint); procedure putbyte(b:byte); procedure putword(w:word); procedure putlongint(l:longint); procedure putreal(d:ppureal); procedure putstring(s:string);
After putting all the things in the entry you need to call ppufile.writeentry(ibnr:byte) where ibnr is the entry number you’re writing.
At the end of the file you need to call ppufile.writeheader to write the new header to the file. This takes automatically care of the new size of the ppufile. When that is also done you can call ppufile.closefile and dispose the object.
Extra functions/variables available for writing are:
ppufile.NewHeader; ppufile.NewEntry;
This will give you a clean header or entry. Normally this is called automatically in ppufile.writeentry, so there should be no need to call these methods. You can call
ppufile.flush;
to flush the current buffers to the disk, and you can set
ppufile.do_crc:boolean;
to False if you don’t want the crc to be updated when writing to disk. This is necessary if you write for example the browser data.
All compiler source files are in several directories, normally the non-processor specific parts are in source/compiler. Subdirectories are present for each of the supported processors and target operating systems.
For more informations about the structure of the compiler have a look at the Compiler Manual which contains also some informations about compiler internals.
The compiler directory also contains a subdirectory utils, which contains mainly the utilities for creation and maintainance of the message files.
The RTL source tree is divided in many subdirectories, but is very structured and easy to understand. It mainly consists of three parts:
There are certain compiler limits inherent to the compiler:
For processor specific compiler limitations refer to the Processor Limitations section in this guide (??).
Here we list the exact effect of the different compiler modes. They can be set with the $Mode switch, or by command line switches.
This mode is selected by the $MODE FPC switch. On the command line, this means that you use none of the other compatibility mode switches. It is the default mode of the compiler (-Mfpc). This means essentially:
This mode is selected by the $MODE TP switch. It tries to emulate, as closely as possible, the behavior of Turbo Pascal 7. On the command line, this mode is selected by the -Mtp switch.
This mode is selected by the $MODE DELPHI switch. It tries to emulate, as closely as possible, the behavior of Delphi 4 or higher. On the command line, this mode is selected by the -Mdelpih switch.
This mode is selected by the $MODE OBJFPC switch. On the command line, this mode is selected by the -Mobjfpc switch.
This mode is selected by the $MODE MACPAS switch. On the command line, this mode is selected by the -Mmacpas switch. It mainly switches on some extra features:
(Note: Macros are called ’Compiler Variables’ in dialects.)
Currently, the following pascal extensions are not yet supported in MACPAS mode:
Free Pascal comes with a special makefile tool, fpcmake, which can be used to construct a Makefile for use with gnu make. All sources from the Free Pascal team are compiled with this system.
fpcmake uses a file Makefile.fpc and constructs a file Makefile from it, based on the settings in Makefile.fpc.
The following sections explain what settings can be set in Makefile.fpc, what variables are set by fpcmake, what variables it expects to be set, and what targets it defines. After that, some settings in the resulting Makefile are explained.
fpcmake generates a makefile, suitable for GNU make, which can be used to
fpcmake knows how the Free Pascal compiler operates, which command line options it uses, how it searches for files and so on; It uses this knowledge to construct sensible command lines.
Specifically, it constructs the following targets in the final makefile:
Each of these targets can be highly configured, or even totally overridden by the configuration file Makefile.fpc.
fpcmake reads a Makefile.fpc and converts it to a Makefile suitable for reading by gnu make to compile your projects. It is similar in functionality to GNU configure or Imake for making X projects.
fpcmake accepts filenames of makefile description files as its command line arguments. For each of these files it will create a Makefile in the same directory where the file is located, overwriting any existing file with that name.
If no options are given, it just attempts to read the file Makefile.fpc in the current directory and tries to construct a Makefile from it if the -m option is given. Any previously existing Makefile will be erased.
if the -p option is given, instead of a Makefile, a Package.fpc is generated. A Package.fpc file describes the package and its dependencies on other packages.
Additionally, the following command line options are recognized:
This section describes the rules that can be present in the file that is processed by fpcmake.
The file Makefile.fpc is a plain ASCII file that contains a number of pre-defined sections as in a Windows .ini-file, or a Samba configuration file.
They look more or less as follows:
[package] name=mysql version=1.0.5 [target] units=mysql_com mysql_version mysql examples=testdb [require] libc=y [install] fpcpackage=y [default] fpcdir=../..
The following sections are recognized (in alphabetical order):
Specifies rules for cleaning the directory of units and programs. The following entries are recognized:
In this section values for various compiler options can be specified, such as the location of several directories and search paths.
The following general keywords are recognised:
The following keys can be used to control the location of the various directories used by the compiler:
The default section contains some default settings. The following keywords are recognized:
The Dist section controls the generation of a distribution package. A distribution package is a set of archive files (zip files or tar files on unix systems) that can be used to distribute the package.
The following keys can be placed in this section:
Contains instructions for installation of the compiled units and programs. The following keywords are recognized:
Units will be installed in the subdirectory units/$(OS_TARGET) of the dirbase entry.
If a package (i.e. a collection of units that work together) is being compiled, then this section is used to keep package information. The following information can be stored:
Anything that is in this section will be inserted as-is in the makefile before the makefile target rules that are generated by fpcmake. This means that any variables that are normally defined by fpcmake rules should not be used in this section.
This section is used to indicate dependency on external packages (i.e units) or tools. The following keywords can be used:
tools=upxwill lead to the definition of a variable with the name UPX which will contain the full path to the upx executable.
In this section dependency rules for the units and any other needed targets can be inserted. It will be included at the end of the generated makefile. Targets or ’default rules’ that are defined by fpcmake can be inserted here; if they are not present, then fpcmake will generate a rule that will call the generic fpc_ version. For a list of standard targets that will be defined by fpcmake, see section ?? .
For example, it is possible to define a target all:. If it is not defined, then fpcmake will generate one which simply calls fpc_all:
all: fpc_all
The fpc_all rule will make all targets as defined in the Target section.
This is the most important section of the makefile.fpc file. Here the files are defined which should be compiled when the ’all’ target is executed.
The following keywords can be used there:
At least the following programs are needed by the generated Makefile to function correctly:
These are standard programs on linux systems, with the possible exception of make. For dos , Windows NT or OS/2 / eComStation, they are distributed as part of Free Pascal releases.
The following programs are optionally needed if you use some special targets. Which ones you need are controlled by the settings in the tools section.
All of these can also be found on the Free Pascal FTP site for dos and Windows NT . ppdep and ppumove are distributed with the Free Pascal compiler.
The makefile generated by fpcmake contains a lot of variables. Some of them are set in the makefile itself, others can be set and are taken into account when set.
These variables can be split in two groups:
Each group will be discussed separately.
The first set of variables controls the directories that are recognised in the makefile. They should not be set in the Makefile.fpc file, but can be specified on the command line.
The following variables can be set on the make command line, they will be recognised and integrated in the compiler command line options.:
The makefile generated by fpcmake contains a lot of makefile variables. fpcmake will write all of the keys in the makefile.fpc as makefile variables in the form SECTION_KEYNAME. This means that the following section:
[package] name=mysql version=1.0.5
will result in the following variable definitions:
override PACKAGE_NAME=mysql override PACKAGE_VERSION=1.0.5
Most targets and rules are constructed using these variables. They will be listed below, together with other variables that are defined by fpcmake.
The following sets of variables are defined:
Each of these sets is discussed in the subsequent:
The following compiler directories are defined by the makefile:
The following directories are used for installs:
The second set of variables controls the targets that are constructed by the makefile. They are created by fpcmake, so you can use them in your rules, but you shouldn’t assign values to them yourself.
The following variables control the compiler command line:
The following variables are program names, used in makefile targets.
The following variables denote extensions of files. These variables include the . (dot) of the extension. They are appended to object names.
The following variables are defined to make targets and rules easier:
The makefile.fpc defines a series of targets, which can be called by your own targets. They have names that resemble default names (such as ’all’, ’clean’), only they have fpc_ prepended.
The makefile makes the following pattern rules:
The following build targets are defined:
The following cleaning targets are defined:
The following archiving targets are defined:
The zip is made uzing the ZIPEXE program. Under linux , a .tar.gz file is created.
There is only one target which produces information about the used variables, rules and targets: fpc_info.
The following information about the makefile is presented:
The Free Pascal team releases at intervals a completely prepared package, with compiler and units all ready to use, the so-called releases. After a release, work on the compiler continues, bugs are fixed and features are added. The Free Pascal team doesn’t make a new release whenever they change something in the compiler, instead the sources are available for anyone to use and compile. There is an automated process that creates compiled versions of RTL and compiler are also made daily, and put on the web (if the build succeeds). Zip files with the sources are also created daily.
There are, nevertheless, circumstances when the compiler must be recompiled manually. When changes are made to compiler code, or when the compiler is downloaded through Subversion.
There are essentially 2 ways of recompiling the compiler: by hand, or using the makefiles. Each of these methods will be discussed.
To compile the compiler easily, it is best to keep the following directory structure (a base directory of /pp/src is supposed, but that may be different):
/pp/src/Makefile
/makefile.fpc
/rtl/linux
/inc
/i386
/...
/compiler
When the makefiles should be used, the above directory tree must be used.
The compiler and rtl source are zipped in such a way that when both are unzipped in the same directory (/pp/src in the above) the above directory tree results.
There are 2 ways to start compiling the compiler and RTL. Both ways must be used, depending on the situation. Usually, the RTL must be compiled first, before compiling the compiler, after which the compiler is compiled using the current compiler. In some special cases the compiler must be compiled first, with a previously compiled RTL.
How to decide which should be compiled first? In general, the answer is that the RTL should be compiled first. There are 2 exceptions to this rule:
How to know if one of these things has occurred? There is no way to know, except by mailing the Free Pascal team. When the compiler cannot be recompiled when first compiling the RTL, then try the other way.
When compiling with make it is necessary to have the above directory structure. Compiling the compiler is achieved with the target cycle.
Under normal circumstances, recompiling the compiler is limited to the following instructions (assuming you start in directory /pp/src):
cd compiler make cycle
This will work only if the makefile is installed correctly and if the needed tools are present in the PATH. Which tools must be installed can be found in appendix ??.
The above instructions will do the following:
The last two steps are repeated 3 times, until three passes have been made or until the generated compiler binary is equal to the binary it was compiled with. This process ensures that the compiler binary is correct.
Compiling for another target: When compiling the compiler for another target, it is necessary to specify the OS_TARGET makefile variable. It can be set to the following values: win32, go32v2, os2 and linux. As an example, cross-compilation for the go32v2 target from the win32 target is chosen:
cd compiler make cycle OS_TARGET=go32v2
This will compile the go32v2 RTL, and compile a go32v2 compiler.
When compiling a new compiler and the compiler should be compiled using an existing compiled RTL, the all target must be used, and another RTL directory than the default (which is the ../rtl/$(OS_TARGET) directory) must be indicated. For instance, assuming that the compiled RTL units are in /pp/rtl/units/i386-linux, typing
cd compiler make clean make all UNITDIR=/pp/rtl/units/i386-linux
should use the RTL from the /pp/rtl/units/i386-linux directory.
This will then compile the compiler using the RTL units in /pp/rtl/units/i386-linux. After this has been done, the ’make cycle’ can be used, starting with this compiler:
make cycle PP=./ppc386
This will do the make cycle from above, but will start with the compiler that was generated by the make all instruction.
In all cases, many options can be passed to make to influence the compile process. In general, the makefiles add any needed compiler options to the command line, so that the RTL and compiler can be compiled. Additional options (e.g. optimization options) can be specified by passing them in OPT.
Compiling by hand is difficult and tedious, but can be done. The compilation of RTL and compiler will be treated separately.
To recompile the RTL, so a new compiler can be built, at least the following units must be built, in the order specified:
To compile these units on a i386, the following statements will do:
ppc386 -Tlinux -b- -Fi../inc -Fi../i386 -FE. -di386 -Us -Sg system.pp ppc386 -Tlinux -b- -Fi../inc -Fi../i386 -FE. -di386 ../inc/strings.pp ppc386 -Tlinux -b- -Fi../inc -Fi../i386 -FE. -di386 dos.pp ppc386 -Tlinux -b- -Fi../inc -Fi../i386 -FE. -di386 ../inc/objects.pp
These are the minimum command line options, needed to compile the RTL.
For another processor, the i386 should be changed into the appropriate processor. For another target OS, the target OS setting (-T) must be set accordingly.
Depending on the target OS there are other units that can be compiled, but which are not strictly needed to recompile the compiler. The following units are available for all plaforms:
Compiling the compiler can be done with one statement. It’s always best to remove all units from the compiler directory first, so something like
rm *.ppu *.o
on linux , and on dos
del *.ppu del *.o
After this, the compiler can be compiled with the following command line:
ppc386 -Tlinux -Fusystems -Fii386/ -Fui386 \ -Fu../rtl/units/i386-linux -di386 -dGDB pp.pas
All the options must be given on a single line, the backslash is needed if you want to specify the options on multiple lines.
So, the minimum options are:
So the absolute minimal command line is
ppc386 -Fusystems -Fii386 -Fui386 -di386 -dGDB -Sg pp.pas
Some other command line options can be used, but the above are the minimum. A list of recognised options can be found in table (??) .
Table F.1: Possible defines when compiling FPC
Define does what GDB Support of the GNU Debugger (required switch). I386 Generate a compiler for the Intel i386+ processor family. M68K Generate a compiler for the M680x0 processor family. X86_64 Generate a compiler for the AMD64 processor family. POWERPC Generate a compiler for the PowerPC processor family. POWERPC64 Generate a compiler for the 64-bit PowerPC processor family. ARM Generate a compiler for the Intel ARM processor family. SPARC Generate a compiler for the SPARC processor family. EXTDEBUG Some extra debug code is executed. MEMDEBUG Some memory usage information is displayed. SUPPORT_MMX only i386: enables the compiler switch MMX which allows the compiler to generate MMX instructions. EXTERN_MSG Don’t compile the msgfiles in the compiler, always use external messagefiles. NOOPT Do not include the optimizer in the compiler. CMEM Use the C memory manager.
This list may be subject to change, the source file pp.pas always contains an up-to-date list.
This appendix describes the possible defines when compiling programs using Free Pascal . A brief explanation of the define, and when it is used is also given.
Table G.1: Possible defines when compiling using FPC
Define description FPC_LINK_DYNAMIC Defined when the output will be linked dynamically. This is defined when using the -XD compiler switch. FPC_LINK_STATIC Defined when the output will be linked statically. This is the default mode. FPC_LINK_SMART Defined when the output will be smartlinked. This is defined when using the -XX compiler switch. FPC_PROFILE Defined when profiling code is added to program. This is defined when using the -pg compiler switch. FPC_CROSSCOMPILING Defined when the target OS/CPU is different from the source OS/CPU. FPC Always defined for Free Pascal . VER2 Always defined for Free Pascal version 2.x.x. VER2_0 Always defined for Free Pascal version 2.0.x. VER2_2 Always defined for Free Pascal version 2.2.x. FPC_VERSION Contains the major version number from FPC. FPC_RELEASE Contains the minor version number from FPC. FPC_PATCH Contains the third part of the version number from FPC. FPC_FULLVERSION Contains the entire version number from FPC as a single number which can be used for comparing. For FPC 2.2.4 it will contain 20204. ENDIAN_LITTLE Defined when the Free Pascal target is a little-endian processor (80x86, Alpha, ARM). ENDIAN_BIG Defined when the Free Pascal target is a big-endian processor (680x0, PowerPC, SPARC, MIPS). FPC_DELPHI Free Pascal is in Delphi mode, either using compiler switch -MDelphi or using the $MODE DELPHI directive. FPC_OBJFPC Free Pascal is in OBJFPC mode, either using compiler switch -Mobjfpc or using the $MODE OBJFPC directive. FPC_TP Free Pascal is in Turbo Pascal mode, either using compiler switch -Mtp or using the $MODE TP directive. FPC_GPC Free Pascal is in GNU Pascal mode, either using compiler switch -SP or using the $MODE GPC directive.
Remark: The ENDIAN_LITTLE and ENDIAN_BIG defines were added starting from Free Pascal version 1.0.5.
Table G.2: Possible CPU defines when compiling using FPC
Define When defined? CPU86 Free Pascal target is an Intel 80x86 or compatible. CPU87 Free Pascal target is an Intel 80x86 or compatible. CPU386 Free Pascal target is an Intel 80386 or later. CPUI386 Free Pascal target is an Intel 80386 or later. CPU68K Free Pascal target is a Motorola 680x0 or compatible. CPUM68K Free Pascal target is a Motorola 680x0 or compatible. CPUM68020 Free Pascal target is a Motorola 68020 or later. CPU68 Free Pascal target is a Motorola 680x0 or compatible. CPUSPARC32 Free Pascal target is a SPARC v7 or compatible. CPUSPARC Free Pascal target is a SPARC v7 or compatible. CPUALPHA Free Pascal target is an Alpha AXP or compatible. CPUPOWERPC Free Pascal target is a 32-bit or 64-bit PowerPC or compatible. CPUPOWERPC32 Free Pascal target is a 32-bit PowerPC or compatible. CPUPOWERPC64 Free Pascal target is a 64-bit PowerPC or compatible. CPUX86_64 Free Pascal target is a AMD64 or Intel 64-bit processor. CPUAMD64 Free Pascal target is a AMD64 or Intel 64-bit processor. CPUIA64 Free Pascal target is a Intel itanium 64-bit processor. CPUARM Free Pascal target is an ARM 32-bit processor. CPUAVR Free Pascal target is an AVR 16-bit processor. CPU16 Free Pascal target is a 16-bit CPU. CPU32 Free Pascal target is a 32-bit CPU. CPU64 Free Pascal target is a 64-bit CPU.
Table G.3: Possible FPU defines when compiling using FPC
Define When defined? FPUSOFT Software emulation of FPU (all types). FPUSSE64 SSE64 FPU on Intel I386 and higher, AMD64. FPUSSE SSE instructions on Intel I386 and higher. FPUSSE2 SSE 2 instructions on Intel I386 and higher. FPUSSE3 SSE 3 instructions on Intel I386 and higher, AMD64. FPULIBGCC GCC library FPU emulation on ARM and M68K. FPU68881 68881 on M68K. FPUFPA FPA on ARM. FPUFPA10 FPA 10 on ARM. FPUFPA11 FPA 11 on ARM. FPUVFP VFP on ARM. FPUX87 X87 FPU on Intel I386 and higher. FPUITANIUM On Intel Itanium. FPUSTANDARD On PowerPC (32/64 bit). FPUHARD On Sparc.
Table G.4: Possible defines when compiling using target OS
Target operating system Defines linux LINUX, UNIX freebsd FREEBSD, BSD, UNIX netbsd NETBSD, BSD, UNIX sunos SUNOS, SOLARIS, UNIX go32v2 GO32V2, DPMI os2 OS2 emx OS2, EMX Windows (all) WINDOWS Windows 32-bit WIN32, MSWINDOWS Windows 64-bit WIN64, MSWINDOWS Windows (winCE) WINCE, UNDER_CE, UNICODE Classic Amiga AMIGA Atari TOS ATARI Classic Macintosh MACOS PalmOS PALMOS BeOS BEOS, UNIX QNX RTP QNX, UNIX Mac OS X BSD, DARWIN, UNIX
Remark: The unix define was added starting from Free Pascal version 1.0.5. The BSD operating systems no longer define LINUX starting with version 1.0.7.
This document was translated from LATEX by HEVEA.