1 \input texinfo @c -*- texinfo -*-
3 @setfilename tcc-doc.info
4 @settitle Tiny C Compiler Reference Documentation
13 @center @titlefont{Tiny C Compiler Reference Documentation}
24 @node Top, Introduction, (dir), (dir)
25 @top Tiny C Compiler Reference Documentation
27 This manual documents version @value{VERSION} of the Tiny C Compiler.
30 * Introduction:: Introduction to tcc.
31 * Invoke:: Invocation of tcc (command line, options).
32 * Bounds:: Automatic bounds-checking of C code.
33 * Libtcc:: The libtcc library.
40 TinyCC (aka TCC) is a small but hyper fast C compiler. Unlike other C
41 compilers, it is meant to be self-relying: you do not need an
42 external assembler or linker because TCC does that for you.
44 TCC compiles so @emph{fast} that even for big projects @code{Makefile}s may
47 TCC not only supports ANSI C, but also most of the new ISO C99
48 standard and many GNUC extensions including inline assembly.
50 TCC can also be used to make @emph{C scripts}, i.e. pieces of C source
51 that you run as a Perl or Python script. Compilation is so fast that
52 your script will be as fast as if it was an executable.
54 TCC can also automatically generate memory and bound checks
55 (@pxref{Bounds}) while allowing all C pointers operations. TCC can do
56 these checks even if non patched libraries are used.
58 With @code{libtcc}, you can use TCC as a backend for dynamic code
59 generation (@pxref{Libtcc}).
61 TCC mainly supports the i386 target. There are alpha ports for the ARM
62 (@code{arm-tcc}) and the TMS320C67xx targets (@code{c67-tcc}). More
63 information about the ARM port is available at
64 @url{http://lists.gnu.org/archive/html/tinycc-devel/2003-10/msg00044.html}.
67 @chapter Command line invocation
69 [This manual documents version @value{VERSION} of the Tiny C Compiler]
75 usage: tcc [options] [@var{infile1} @var{infile2}@dots{}] [@option{-run} @var{infile} @var{args}@dots{}]
80 @c man begin DESCRIPTION
81 TCC options are a very much like gcc options. The main difference is that TCC
82 can also execute directly the resulting program and give it runtime
85 Here are some examples to understand the logic:
88 @item @samp{tcc -run a.c}
89 Compile @file{a.c} and execute it directly
91 @item @samp{tcc -run a.c arg1}
92 Compile a.c and execute it directly. arg1 is given as first argument to
93 the @code{main()} of a.c.
95 @item @samp{tcc a.c -run b.c arg1}
96 Compile @file{a.c} and @file{b.c}, link them together and execute them. arg1 is given
97 as first argument to the @code{main()} of the resulting program. Because
98 multiple C files are specified, @option{--} are necessary to clearly separate the
99 program arguments from the TCC options.
101 @item @samp{tcc -o myprog a.c b.c}
102 Compile @file{a.c} and @file{b.c}, link them and generate the executable @file{myprog}.
104 @item @samp{tcc -o myprog a.o b.o}
105 link @file{a.o} and @file{b.o} together and generate the executable @file{myprog}.
107 @item @samp{tcc -c a.c}
108 Compile @file{a.c} and generate object file @file{a.o}.
110 @item @samp{tcc -c asmfile.S}
111 Preprocess with C preprocess and assemble @file{asmfile.S} and generate
112 object file @file{asmfile.o}.
114 @item @samp{tcc -c asmfile.s}
115 Assemble (but not preprocess) @file{asmfile.s} and generate object file
118 @item @samp{tcc -r -o ab.o a.c b.c}
119 Compile @file{a.c} and @file{b.c}, link them together and generate the object file @file{ab.o}.
125 TCC can be invoked from @emph{scripts}, just as shell scripts. You just
126 need to add @code{#!/usr/local/bin/tcc -run} at the start of your C source:
129 #!/usr/local/bin/tcc -run
134 printf("Hello World\n");
140 @section Option summary
147 Display current TCC version.
150 Generate an object file (@option{-o} option must also be given).
153 Put object file, executable, or dll into output file @file{outfile}.
156 Set the path where the tcc internal libraries can be found (default is
157 @file{PREFIX/lib/tcc}).
160 Output compilation statistics.
162 @item -run source [args...]
164 Compile file @var{source} and run it with the command line arguments
165 @var{args}. In order to be able to give more than one argument to a
166 script, several TCC options can be given @emph{after} the
167 @option{-run} option, separated by spaces. Example:
170 tcc "-run -L/usr/X11R6/lib -lX11" ex4.c
173 In a script, it gives the following header:
176 #!/usr/local/bin/tcc -run -L/usr/X11R6/lib -lX11
178 int main(int argc, char **argv)
186 Preprocessor options:
190 Specify an additional include path. Include paths are searched in the
191 order they are specified.
193 System include paths are always searched after. The default system
194 include paths are: @file{/usr/local/include}, @file{/usr/include}
195 and @file{PREFIX/lib/tcc/include}. (@file{PREFIX} is usually
196 @file{/usr} or @file{/usr/local}).
199 Define preprocessor symbol @samp{sym} to
200 val. If val is not present, its value is @samp{1}. Function-like macros can
201 also be defined: @option{-DF(a)=a+1}
204 Undefine preprocessor symbol @samp{sym}.
209 Note: each of the following warning options has a negative form beginning with
213 @item -funsigned-char
214 Let the @code{char} type be unsigned.
217 Let the @code{char} type be signed.
220 Do not generate common symbols for uninitialized data.
228 Disable all warnings.
232 Note: each of the following warning options has a negative form beginning with
236 @item -Wimplicit-function-declaration
237 Warn about implicit function declaration.
240 Warn about unsupported GCC features that are ignored by TCC.
242 @item -Wwrite-strings
243 Make string constants be of type @code{const char *} instead of @code{char
247 Abort compilation if warnings are issued.
250 Activate all warnings, except @option{-Werror}, @option{-Wunusupported} and
251 @option{-Wwrite-strings}.
259 Specify an additional static library path for the @option{-l} option. The
260 default library paths are @file{/usr/local/lib}, @file{/usr/lib} and @file{/lib}.
263 Link your program with dynamic library libxxx.so or static library
264 libxxx.a. The library is searched in the paths specified by the
268 Generate a shared library instead of an executable (@option{-o} option
272 Generate a statically linked executable (default is a shared linked
273 executable) (@option{-o} option must also be given).
276 Export global symbols to the dynamic linker. It is useful when a library
277 opened with @code{dlopen()} needs to access executable symbols.
280 Generate an object file combining all input files (@option{-o} option must
283 @item -Wl,-Ttext,address
284 Set the start of the .text section to @var{address}.
286 @item -Wl,--oformat,fmt
287 Use @var{fmt} as output format. The supported output formats are:
290 ELF output format (default)
292 Binary image (only for executable output)
294 COFF output format (only for executable output for TMS320C67xx target)
303 Generate run time debug information so that you get clear run time
304 error messages: @code{ test.c:68: in function 'test5()': dereferencing
305 invalid pointer} instead of the laconic @code{Segmentation
309 Generate additional support code to check
310 memory allocations and array/pointer bounds. @option{-g} is implied. Note
311 that the generated code is slower and bigger in this case.
314 Display N callers in stack traces. This is useful with @option{-g} or
319 Note: GCC options @option{-Ox}, @option{-fx} and @option{-mx} are
326 @settitle Tiny C Compiler
338 @chapter C language support
342 TCC implements all the ANSI C standard, including structure bit fields
343 and floating point numbers (@code{long double}, @code{double}, and
344 @code{float} fully supported).
346 @section ISOC99 extensions
348 TCC implements many features of the new C standard: ISO C99. Currently
349 missing items are: complex and imaginary numbers and variable length
352 Currently implemented ISOC99 features:
356 @item 64 bit @code{long long} types are fully supported.
358 @item The boolean type @code{_Bool} is supported.
360 @item @code{__func__} is a string variable containing the current
363 @item Variadic macros: @code{__VA_ARGS__} can be used for
364 function-like macros:
366 #define dprintf(level, __VA_ARGS__) printf(__VA_ARGS__)
370 @code{dprintf} can then be used with a variable number of parameters.
372 @item Declarations can appear anywhere in a block (as in C++).
374 @item Array and struct/union elements can be initialized in any order by
377 struct @{ int x, y; @} st[10] = @{ [0].x = 1, [0].y = 2 @};
379 int tab[10] = @{ 1, 2, [5] = 5, [9] = 9@};
382 @item Compound initializers are supported:
384 int *p = (int [])@{ 1, 2, 3 @};
386 to initialize a pointer pointing to an initialized array. The same
387 works for structures and strings.
389 @item Hexadecimal floating point constants are supported:
391 double d = 0x1234p10;
395 is the same as writing
397 double d = 4771840.0;
400 @item @code{inline} keyword is ignored.
402 @item @code{restrict} keyword is ignored.
405 @section GNU C extensions
407 TCC implements some GNU C extensions:
411 @item array designators can be used without '=':
413 int a[10] = @{ [0] 1, [5] 2, 3, 4 @};
416 @item Structure field designators can be a label:
418 struct @{ int x, y; @} st = @{ x: 1, y: 1@};
422 struct @{ int x, y; @} st = @{ .x = 1, .y = 1@};
425 @item @code{\e} is ASCII character 27.
427 @item case ranges : ranges can be used in @code{case}s:
431 printf("range 1 to 9\n");
434 printf("unexpected\n");
439 @item The keyword @code{__attribute__} is handled to specify variable or
440 function attributes. The following attributes are supported:
442 @item @code{aligned(n)}: align data to n bytes (must be a power of two).
444 @item @code{section(name)}: generate function or data in assembly
445 section name (name is a string containing the section name) instead
446 of the default section.
448 @item @code{unused}: specify that the variable or the function is unused.
450 @item @code{cdecl}: use standard C calling convention.
452 @item @code{stdcall}: use Pascal-like calling convention.
456 Here are some examples:
458 int a __attribute__ ((aligned(8), section(".mysection")));
462 align variable @code{a} to 8 bytes and put it in section @code{.mysection}.
465 int my_add(int a, int b) __attribute__ ((section(".mycodesection")))
472 generate function @code{my_add} in section @code{.mycodesection}.
474 @item GNU style variadic macros:
476 #define dprintf(fmt, args@dots{}) printf(fmt, ## args)
479 dprintf("one arg %d\n", 1);
482 @item @code{__FUNCTION__} is interpreted as C99 @code{__func__}
483 (so it has not exactly the same semantics as string literal GNUC
484 where it is a string literal).
486 @item The @code{__alignof__} keyword can be used as @code{sizeof}
487 to get the alignment of a type or an expression.
489 @item The @code{typeof(x)} returns the type of @code{x}.
490 @code{x} is an expression or a type.
492 @item Computed gotos: @code{&&label} returns a pointer of type
493 @code{void *} on the goto label @code{label}. @code{goto *expr} can be
494 used to jump on the pointer resulting from @code{expr}.
496 @item Inline assembly with asm instruction:
497 @cindex inline assembly
498 @cindex assembly, inline
501 static inline void * my_memcpy(void * to, const void * from, size_t n)
504 __asm__ __volatile__(
509 "1:\ttestb $1,%b4\n\t"
513 : "=&c" (d0), "=&D" (d1), "=&S" (d2)
514 :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from)
522 TCC includes its own x86 inline assembler with a @code{gas}-like (GNU
523 assembler) syntax. No intermediate files are generated. GCC 3.x named
524 operands are supported.
526 @item @code{__builtin_types_compatible_p()} and @code{__builtin_constant_p()}
531 @section TinyCC extensions
535 @item @code{__TINYC__} is a predefined macro to @code{1} to
536 indicate that you use TCC.
538 @item @code{#!} at the start of a line is ignored to allow scripting.
540 @item Binary digits can be entered (@code{0b101} instead of
543 @item @code{__BOUNDS_CHECKING_ON} is defined if bound checking is activated.
547 @chapter TinyCC Assembler
549 Since version 0.9.16, TinyCC integrates its own assembler. TinyCC
550 assembler supports a gas-like syntax (GNU assembler). You can
551 desactivate assembler support if you want a smaller TinyCC executable
552 (the C compiler does not rely on the assembler).
554 TinyCC Assembler is used to handle files with @file{.S} (C
555 preprocessed assembler) and @file{.s} extensions. It is also used to
556 handle the GNU inline assembler with the @code{asm} keyword.
560 TinyCC Assembler supports most of the gas syntax. The tokens are the
565 @item C and C++ comments are supported.
567 @item Identifiers are the same as C, so you cannot use '.' or '$'.
569 @item Only 32 bit integer numbers are supported.
577 @item Integers in decimal, octal and hexa are supported.
579 @item Unary operators: +, -, ~.
581 @item Binary operators in decreasing priority order:
589 @item A value is either an absolute number or a label plus an offset.
590 All operators accept absolute values except '+' and '-'. '+' or '-' can be
591 used to add an offset to a label. '-' supports two labels only if they
592 are the same or if they are both defined and in the same section.
600 @item All labels are considered as local, except undefined ones.
602 @item Numeric labels can be used as local @code{gas}-like labels.
603 They can be defined several times in the same source. Use 'b'
604 (backward) or 'f' (forward) as suffix to reference them:
608 jmp 1b /* jump to '1' label before */
609 jmp 1f /* jump to '1' label after */
616 @cindex assembler directives
617 @cindex directives, assembler
634 All directives are preceeded by a '.'. The following directives are
638 @item .align n[,value]
639 @item .skip n[,value]
640 @item .space n[,value]
641 @item .byte value1[,value2...]
642 @item .word value1[,value2...]
643 @item .short value1[,value2...]
644 @item .int value1[,value2...]
645 @item .long value1[,value2...]
649 @item .section section
653 @item .fill repeat[,size[,value]]
656 @item .string string[,...]
657 @item .asciz string[,...]
658 @item .ascii string[,...]
661 @section X86 Assembler
664 All X86 opcodes are supported. Only ATT syntax is supported (source
665 then destination operand order). If no size suffix is given, TinyCC
666 tries to guess it from the operand sizes.
668 Currently, MMX opcodes are supported but not SSE ones.
670 @chapter TinyCC Linker
673 @section ELF file generation
676 TCC can directly output relocatable ELF files (object files),
677 executable ELF files and dynamic ELF libraries without relying on an
680 Dynamic ELF libraries can be output but the C compiler does not generate
681 position independent code (PIC). It means that the dynamic library
682 code generated by TCC cannot be factorized among processes yet.
684 TCC linker eliminates unreferenced object code in libraries. A single pass is
685 done on the object and library list, so the order in which object files and
686 libraries are specified is important (same constraint as GNU ld). No grouping
687 options (@option{--start-group} and @option{--end-group}) are supported.
689 @section ELF file loader
691 TCC can load ELF object files, archives (.a files) and dynamic
694 @section GNU Linker Scripts
695 @cindex scripts, linker
696 @cindex linker scripts
697 @cindex GROUP, linker command
698 @cindex FILE, linker command
699 @cindex OUTPUT_FORMAT, linker command
700 @cindex TARGET, linker command
702 Because on many Linux systems some dynamic libraries (such as
703 @file{/usr/lib/libc.so}) are in fact GNU ld link scripts (horrible!),
704 the TCC linker also supports a subset of GNU ld scripts.
706 The @code{GROUP} and @code{FILE} commands are supported. @code{OUTPUT_FORMAT}
707 and @code{TARGET} are ignored.
709 Example from @file{/usr/lib/libc.so}:
712 Use the shared library, but some functions are only in
713 the static library, so try that secondarily. */
714 GROUP ( /lib/libc.so.6 /usr/lib/libc_nonshared.a )
718 @chapter TinyCC Memory and Bound checks
720 @cindex memory checks
722 This feature is activated with the @option{-b} (@pxref{Invoke}).
724 Note that pointer size is @emph{unchanged} and that code generated
725 with bound checks is @emph{fully compatible} with unchecked
726 code. When a pointer comes from unchecked code, it is assumed to be
727 valid. Even very obscure C code with casts should work correctly.
729 For more information about the ideas behind this method, see
730 @url{http://www.doc.ic.ac.uk/~phjk/BoundsChecking.html}.
732 Here are some examples of caught errors:
736 @item Invalid range with standard string function:
744 @item Out of bounds-error in global or local arrays:
754 @item Out of bounds-error in malloc'ed data:
758 tab = malloc(20 * sizeof(int));
766 @item Access of freed memory:
770 tab = malloc(20 * sizeof(int));
782 tab = malloc(20 * sizeof(int));
791 @chapter The @code{libtcc} library
793 The @code{libtcc} library enables you to use TCC as a backend for
794 dynamic code generation.
796 Read the @file{libtcc.h} to have an overview of the API. Read
797 @file{libtcc_test.c} to have a very simple example.
799 The idea consists in giving a C string containing the program you want
800 to compile directly to @code{libtcc}. Then you can access to any global
801 symbol (function or variable) defined.
803 @chapter Developer's guide
805 This chapter gives some hints to understand how TCC works. You can skip
806 it if you do not intend to modify the TCC code.
808 @section File reading
810 The @code{BufferedFile} structure contains the context needed to read a
811 file, including the current line number. @code{tcc_open()} opens a new
812 file and @code{tcc_close()} closes it. @code{inp()} returns the next
817 @code{next()} reads the next token in the current
818 file. @code{next_nomacro()} reads the next token without macro
821 @code{tok} contains the current token (see @code{TOK_xxx})
822 constants. Identifiers and keywords are also keywords. @code{tokc}
823 contains additional infos about the token (for example a constant value
824 if number or string token).
828 The parser is hardcoded (yacc is not necessary). It does only one pass,
833 @item For initialized arrays with unknown size, a first pass
834 is done to count the number of elements.
836 @item For architectures where arguments are evaluated in
837 reverse order, a first pass is done to reverse the argument order.
843 The types are stored in a single 'int' variable. It was choosen in the
844 first stages of development when tcc was much simpler. Now, it may not
845 be the best solution.
848 #define VT_INT 0 /* integer type */
849 #define VT_BYTE 1 /* signed byte type */
850 #define VT_SHORT 2 /* short type */
851 #define VT_VOID 3 /* void type */
852 #define VT_PTR 4 /* pointer */
853 #define VT_ENUM 5 /* enum definition */
854 #define VT_FUNC 6 /* function type */
855 #define VT_STRUCT 7 /* struct/union definition */
856 #define VT_FLOAT 8 /* IEEE float */
857 #define VT_DOUBLE 9 /* IEEE double */
858 #define VT_LDOUBLE 10 /* IEEE long double */
859 #define VT_BOOL 11 /* ISOC99 boolean type */
860 #define VT_LLONG 12 /* 64 bit integer */
861 #define VT_LONG 13 /* long integer (NEVER USED as type, only
863 #define VT_BTYPE 0x000f /* mask for basic type */
864 #define VT_UNSIGNED 0x0010 /* unsigned type */
865 #define VT_ARRAY 0x0020 /* array type (also has VT_PTR) */
866 #define VT_BITFIELD 0x0040 /* bitfield modifier */
868 #define VT_STRUCT_SHIFT 16 /* structure/enum name shift (16 bits left) */
871 When a reference to another type is needed (for pointers, functions and
872 structures), the @code{32 - VT_STRUCT_SHIFT} high order bits are used to
873 store an identifier reference.
875 The @code{VT_UNSIGNED} flag can be set for chars, shorts, ints and long
878 Arrays are considered as pointers @code{VT_PTR} with the flag
881 The @code{VT_BITFIELD} flag can be set for chars, shorts, ints and long
882 longs. If it is set, then the bitfield position is stored from bits
883 VT_STRUCT_SHIFT to VT_STRUCT_SHIFT + 5 and the bit field size is stored
884 from bits VT_STRUCT_SHIFT + 6 to VT_STRUCT_SHIFT + 11.
886 @code{VT_LONG} is never used except during parsing.
888 During parsing, the storage of an object is also stored in the type
892 #define VT_EXTERN 0x00000080 /* extern definition */
893 #define VT_STATIC 0x00000100 /* static variable */
894 #define VT_TYPEDEF 0x00000200 /* typedef definition */
899 All symbols are stored in hashed symbol stacks. Each symbol stack
900 contains @code{Sym} structures.
902 @code{Sym.v} contains the symbol name (remember
903 an idenfier is also a token, so a string is never necessary to store
904 it). @code{Sym.t} gives the type of the symbol. @code{Sym.r} is usually
905 the register in which the corresponding variable is stored. @code{Sym.c} is
906 usually a constant associated to the symbol.
908 Four main symbol stacks are defined:
913 for the macros (@code{#define}s).
916 for the global variables, functions and types.
919 for the local variables, functions and types.
921 @item global_label_stack
922 for the local labels (for @code{goto}).
925 for GCC block local labels (see the @code{__label__} keyword).
929 @code{sym_push()} is used to add a new symbol in the local symbol
930 stack. If no local symbol stack is active, it is added in the global
933 @code{sym_pop(st,b)} pops symbols from the symbol stack @var{st} until
934 the symbol @var{b} is on the top of stack. If @var{b} is NULL, the stack
937 @code{sym_find(v)} return the symbol associated to the identifier
938 @var{v}. The local stack is searched first from top to bottom, then the
943 The generated code and datas are written in sections. The structure
944 @code{Section} contains all the necessary information for a given
945 section. @code{new_section()} creates a new section. ELF file semantics
946 is assumed for each section.
948 The following sections are predefined:
953 is the section containing the generated code. @var{ind} contains the
954 current position in the code section.
957 contains initialized data
960 contains uninitialized data
963 @itemx lbounds_section
964 are used when bound checking is activated
967 @itemx stabstr_section
968 are used when debugging is actived to store debug information
971 @itemx strtab_section
972 contain the exported symbols (currently only used for debugging).
976 @section Code generation
977 @cindex code generation
979 @subsection Introduction
981 The TCC code generator directly generates linked binary code in one
982 pass. It is rather unusual these days (see gcc for example which
983 generates text assembly), but it can be very fast and surprisingly
986 The TCC code generator is register based. Optimization is only done at
987 the expression level. No intermediate representation of expression is
988 kept except the current values stored in the @emph{value stack}.
990 On x86, three temporary registers are used. When more registers are
991 needed, one register is spilled into a new temporary variable on the stack.
993 @subsection The value stack
994 @cindex value stack, introduction
996 When an expression is parsed, its value is pushed on the value stack
997 (@var{vstack}). The top of the value stack is @var{vtop}. Each value
998 stack entry is the structure @code{SValue}.
1000 @code{SValue.t} is the type. @code{SValue.r} indicates how the value is
1001 currently stored in the generated code. It is usually a CPU register
1002 index (@code{REG_xxx} constants), but additional values and flags are
1006 #define VT_CONST 0x00f0
1007 #define VT_LLOCAL 0x00f1
1008 #define VT_LOCAL 0x00f2
1009 #define VT_CMP 0x00f3
1010 #define VT_JMP 0x00f4
1011 #define VT_JMPI 0x00f5
1012 #define VT_LVAL 0x0100
1013 #define VT_SYM 0x0200
1014 #define VT_MUSTCAST 0x0400
1015 #define VT_MUSTBOUND 0x0800
1016 #define VT_BOUNDED 0x8000
1017 #define VT_LVAL_BYTE 0x1000
1018 #define VT_LVAL_SHORT 0x2000
1019 #define VT_LVAL_UNSIGNED 0x4000
1020 #define VT_LVAL_TYPE (VT_LVAL_BYTE | VT_LVAL_SHORT | VT_LVAL_UNSIGNED)
1026 indicates that the value is a constant. It is stored in the union
1027 @code{SValue.c}, depending on its type.
1030 indicates a local variable pointer at offset @code{SValue.c.i} in the
1034 indicates that the value is actually stored in the CPU flags (i.e. the
1035 value is the consequence of a test). The value is either 0 or 1. The
1036 actual CPU flags used is indicated in @code{SValue.c.i}.
1038 If any code is generated which destroys the CPU flags, this value MUST be
1039 put in a normal register.
1043 indicates that the value is the consequence of a conditional jump. For VT_JMP,
1044 it is 1 if the jump is taken, 0 otherwise. For VT_JMPI it is inverted.
1046 These values are used to compile the @code{||} and @code{&&} logical
1049 If any code is generated, this value MUST be put in a normal
1050 register. Otherwise, the generated code won't be executed if the jump is
1054 is a flag indicating that the value is actually an lvalue (left value of
1055 an assignment). It means that the value stored is actually a pointer to
1058 Understanding the use @code{VT_LVAL} is very important if you want to
1059 understand how TCC works.
1062 @itemx VT_LVAL_SHORT
1063 @itemx VT_LVAL_UNSIGNED
1064 if the lvalue has an integer type, then these flags give its real
1065 type. The type alone is not enough in case of cast optimisations.
1068 is a saved lvalue on the stack. @code{VT_LLOCAL} should be eliminated
1069 ASAP because its semantics are rather complicated.
1072 indicates that a cast to the value type must be performed if the value
1073 is used (lazy casting).
1076 indicates that the symbol @code{SValue.sym} must be added to the constant.
1080 are only used for optional bound checking.
1084 @subsection Manipulating the value stack
1087 @code{vsetc()} and @code{vset()} pushes a new value on the value
1088 stack. If the previous @var{vtop} was stored in a very unsafe place(for
1089 example in the CPU flags), then some code is generated to put the
1090 previous @var{vtop} in a safe storage.
1092 @code{vpop()} pops @var{vtop}. In some cases, it also generates cleanup
1093 code (for example if stacked floating point registers are used as on
1096 The @code{gv(rc)} function generates code to evaluate @var{vtop} (the
1097 top value of the stack) into registers. @var{rc} selects in which
1098 register class the value should be put. @code{gv()} is the @emph{most
1099 important function} of the code generator.
1101 @code{gv2()} is the same as @code{gv()} but for the top two stack
1104 @subsection CPU dependent code generation
1105 @cindex CPU dependent
1106 See the @file{i386-gen.c} file to have an example.
1111 must generate the code needed to load a stack value into a register.
1114 must generate the code needed to store a register into a stack value
1118 @itemx gfunc_param()
1120 should generate a function call
1122 @item gfunc_prolog()
1123 @itemx gfunc_epilog()
1124 should generate a function prolog/epilog.
1127 must generate the binary integer operation @var{op} on the two top
1128 entries of the stack which are guaranted to contain integer types.
1130 The result value should be put on the stack.
1133 same as @code{gen_opi()} for floating point operations. The two top
1134 entries of the stack are guaranted to contain floating point values of
1137 @item gen_cvt_itof()
1138 integer to floating point conversion.
1140 @item gen_cvt_ftoi()
1141 floating point to integer conversion.
1143 @item gen_cvt_ftof()
1144 floating point to floating point of different size conversion.
1146 @item gen_bounded_ptr_add()
1147 @item gen_bounded_ptr_deref()
1148 are only used for bounds checking.
1152 @section Optimizations done
1153 @cindex optimizations
1154 @cindex constant propagation
1155 @cindex strength reduction
1156 @cindex comparison operators
1157 @cindex caching processor flags
1158 @cindex flags, caching
1159 @cindex jump optimization
1160 Constant propagation is done for all operations. Multiplications and
1161 divisions are optimized to shifts when appropriate. Comparison
1162 operators are optimized by maintaining a special cache for the
1163 processor flags. &&, || and ! are optimized by maintaining a special
1164 'jump target' value. No other jump optimization is currently performed
1165 because it would require to store the code in a more abstract fashion.
1167 @unnumbered Concept Index
1174 @c texinfo-column-for-description: 32