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}).
62 @chapter Command line invocation
64 [This manual documents version @value{VERSION} of the Tiny C Compiler]
70 usage: tcc [options] [@var{infile1} @var{infile2}@dots{}] [@option{-run} @var{infile} @var{args}@dots{}]
75 @c man begin DESCRIPTION
76 TCC options are a very much like gcc options. The main difference is that TCC
77 can also execute directly the resulting program and give it runtime
80 Here are some examples to understand the logic:
83 @item @samp{tcc -run a.c}
84 Compile @file{a.c} and execute it directly
86 @item @samp{tcc -run a.c arg1}
87 Compile a.c and execute it directly. arg1 is given as first argument to
88 the @code{main()} of a.c.
90 @item @samp{tcc a.c -run b.c arg1}
91 Compile @file{a.c} and @file{b.c}, link them together and execute them. arg1 is given
92 as first argument to the @code{main()} of the resulting program. Because
93 multiple C files are specified, @option{--} are necessary to clearly separate the
94 program arguments from the TCC options.
96 @item @samp{tcc -o myprog a.c b.c}
97 Compile @file{a.c} and @file{b.c}, link them and generate the executable @file{myprog}.
99 @item @samp{tcc -o myprog a.o b.o}
100 link @file{a.o} and @file{b.o} together and generate the executable @file{myprog}.
102 @item @samp{tcc -c a.c}
103 Compile @file{a.c} and generate object file @file{a.o}.
105 @item @samp{tcc -c asmfile.S}
106 Preprocess with C preprocess and assemble @file{asmfile.S} and generate
107 object file @file{asmfile.o}.
109 @item @samp{tcc -c asmfile.s}
110 Assemble (but not preprocess) @file{asmfile.s} and generate object file
113 @item @samp{tcc -r -o ab.o a.c b.c}
114 Compile @file{a.c} and @file{b.c}, link them together and generate the object file @file{ab.o}.
120 TCC can be invoked from @emph{scripts}, just as shell scripts. You just
121 need to add @code{#!/usr/local/bin/tcc -run} at the start of your C source:
124 #!/usr/local/bin/tcc -run
129 printf("Hello World\n");
135 @section Option summary
142 Display current TCC version.
145 Generate an object file (@option{-o} option must also be given).
148 Put object file, executable, or dll into output file @file{outfile}.
151 Set the path where the tcc internal libraries can be found (default is
152 @file{PREFIX/lib/tcc}).
155 Output compilation statistics.
161 Preprocessor options:
165 Specify an additional include path. Include paths are searched in the
166 order they are specified.
168 System include paths are always searched after. The default system
169 include paths are: @file{/usr/local/include}, @file{/usr/include}
170 and @file{PREFIX/lib/tcc/include}. (@file{PREFIX} is usually
171 @file{/usr} or @file{/usr/local}).
174 Define preprocessor symbol @samp{sym} to
175 val. If val is not present, its value is @samp{1}. Function-like macros can
176 also be defined: @option{-DF(a)=a+1}
179 Undefine preprocessor symbol @samp{sym}.
186 Disable all warnings.
190 Note: each of the following warning options has a negative form beginning with
195 Warn about unsupported GCC features that are ignored by TCC.
197 @item -Wwrite-strings
198 Make string constants being of type @code{const char *} intead of @code{char
202 Abort compilation if warnings are issued.
205 Activate all warnings, except @option{-Werror}, @option{-Wunusupported} and
206 @option{-Wwrite-strings} (currently not useful).
214 Specify an additional static library path for the @option{-l} option. The
215 default library paths are @file{/usr/local/lib}, @file{/usr/lib} and @file{/lib}.
218 Link your program with dynamic library libxxx.so or static library
219 libxxx.a. The library is searched in the paths specified by the
223 Generate a shared library instead of an executable (@option{-o} option
227 Generate a statically linked executable (default is a shared linked
228 executable) (@option{-o} option must also be given).
231 Export global symbols to the dynamic linker. It is useful when a library
232 opened with @code{dlopen()} needs to access executable symbols.
235 Generate an object file combining all input files (@option{-o} option must
244 Generate run time debug information so that you get clear run time
245 error messages: @code{ test.c:68: in function 'test5()': dereferencing
246 invalid pointer} instead of the laconic @code{Segmentation
250 Generate additional support code to check
251 memory allocations and array/pointer bounds. @option{-g} is implied. Note
252 that the generated code is slower and bigger in this case.
255 Display N callers in stack traces. This is useful with @option{-g} or
260 Note: GCC options @option{-Ox}, @option{-fx} and @option{-mx} are
267 @settitle Tiny C Compiler
279 @chapter C language support
283 TCC implements all the ANSI C standard, including structure bit fields
284 and floating point numbers (@code{long double}, @code{double}, and
285 @code{float} fully supported).
287 @section ISOC99 extensions
289 TCC implements many features of the new C standard: ISO C99. Currently
290 missing items are: complex and imaginary numbers and variable length
293 Currently implemented ISOC99 features:
297 @item 64 bit @code{long long} types are fully supported.
299 @item The boolean type @code{_Bool} is supported.
301 @item @code{__func__} is a string variable containing the current
304 @item Variadic macros: @code{__VA_ARGS__} can be used for
305 function-like macros:
307 #define dprintf(level, __VA_ARGS__) printf(__VA_ARGS__)
311 @code{dprintf} can then be used with a variable number of parameters.
313 @item Declarations can appear anywhere in a block (as in C++).
315 @item Array and struct/union elements can be initialized in any order by
318 struct @{ int x, y; @} st[10] = @{ [0].x = 1, [0].y = 2 @};
320 int tab[10] = @{ 1, 2, [5] = 5, [9] = 9@};
323 @item Compound initializers are supported:
325 int *p = (int [])@{ 1, 2, 3 @};
327 to initialize a pointer pointing to an initialized array. The same
328 works for structures and strings.
330 @item Hexadecimal floating point constants are supported:
332 double d = 0x1234p10;
336 is the same as writing
338 double d = 4771840.0;
341 @item @code{inline} keyword is ignored.
343 @item @code{restrict} keyword is ignored.
346 @section GNU C extensions
348 TCC implements some GNU C extensions:
352 @item array designators can be used without '=':
354 int a[10] = @{ [0] 1, [5] 2, 3, 4 @};
357 @item Structure field designators can be a label:
359 struct @{ int x, y; @} st = @{ x: 1, y: 1@};
363 struct @{ int x, y; @} st = @{ .x = 1, .y = 1@};
366 @item @code{\e} is ASCII character 27.
368 @item case ranges : ranges can be used in @code{case}s:
372 printf("range 1 to 9\n");
375 printf("unexpected\n");
380 @item The keyword @code{__attribute__} is handled to specify variable or
381 function attributes. The following attributes are supported:
383 @item @code{aligned(n)}: align data to n bytes (must be a power of two).
385 @item @code{section(name)}: generate function or data in assembly
386 section name (name is a string containing the section name) instead
387 of the default section.
389 @item @code{unused}: specify that the variable or the function is unused.
391 @item @code{cdecl}: use standard C calling convention.
393 @item @code{stdcall}: use Pascal-like calling convention.
397 Here are some examples:
399 int a __attribute__ ((aligned(8), section(".mysection")));
403 align variable @code{a} to 8 bytes and put it in section @code{.mysection}.
406 int my_add(int a, int b) __attribute__ ((section(".mycodesection")))
413 generate function @code{my_add} in section @code{.mycodesection}.
415 @item GNU style variadic macros:
417 #define dprintf(fmt, args@dots{}) printf(fmt, ## args)
420 dprintf("one arg %d\n", 1);
423 @item @code{__FUNCTION__} is interpreted as C99 @code{__func__}
424 (so it has not exactly the same semantics as string literal GNUC
425 where it is a string literal).
427 @item The @code{__alignof__} keyword can be used as @code{sizeof}
428 to get the alignment of a type or an expression.
430 @item The @code{typeof(x)} returns the type of @code{x}.
431 @code{x} is an expression or a type.
433 @item Computed gotos: @code{&&label} returns a pointer of type
434 @code{void *} on the goto label @code{label}. @code{goto *expr} can be
435 used to jump on the pointer resulting from @code{expr}.
437 @item Inline assembly with asm instruction:
438 @cindex inline assembly
439 @cindex assembly, inline
442 static inline void * my_memcpy(void * to, const void * from, size_t n)
445 __asm__ __volatile__(
450 "1:\ttestb $1,%b4\n\t"
454 : "=&c" (d0), "=&D" (d1), "=&S" (d2)
455 :"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from)
463 TCC includes its own x86 inline assembler with a @code{gas}-like (GNU
464 assembler) syntax. No intermediate files are generated. GCC 3.x named
465 operands are supported.
467 @item @code{__builtin_types_compatible_p()} and @code{__builtin_constant_p()}
472 @section TinyCC extensions
476 @item @code{__TINYC__} is a predefined macro to @code{1} to
477 indicate that you use TCC.
479 @item @code{#!} at the start of a line is ignored to allow scripting.
481 @item Binary digits can be entered (@code{0b101} instead of
484 @item @code{__BOUNDS_CHECKING_ON} is defined if bound checking is activated.
488 @chapter TinyCC Assembler
490 Since version 0.9.16, TinyCC integrates its own assembler. TinyCC
491 assembler supports a gas-like syntax (GNU assembler). You can
492 desactivate assembler support if you want a smaller TinyCC executable
493 (the C compiler does not rely on the assembler).
495 TinyCC Assembler is used to handle files with @file{.S} (C
496 preprocessed assembler) and @file{.s} extensions. It is also used to
497 handle the GNU inline assembler with the @code{asm} keyword.
501 TinyCC Assembler supports most of the gas syntax. The tokens are the
506 @item C and C++ comments are supported.
508 @item Identifiers are the same as C, so you cannot use '.' or '$'.
510 @item Only 32 bit integer numbers are supported.
518 @item Integers in decimal, octal and hexa are supported.
520 @item Unary operators: +, -, ~.
522 @item Binary operators in decreasing priority order:
530 @item A value is either an absolute number or a label plus an offset.
531 All operators accept absolute values except '+' and '-'. '+' or '-' can be
532 used to add an offset to a label. '-' supports two labels only if they
533 are the same or if they are both defined and in the same section.
541 @item All labels are considered as local, except undefined ones.
543 @item Numeric labels can be used as local @code{gas}-like labels.
544 They can be defined several times in the same source. Use 'b'
545 (backward) or 'f' (forward) as suffix to reference them:
549 jmp 1b /* jump to '1' label before */
550 jmp 1f /* jump to '1' label after */
557 @cindex assembler directives
558 @cindex directives, assembler
574 All directives are preceeded by a '.'. The following directives are
578 @item .align n[,value]
579 @item .skip n[,value]
580 @item .space n[,value]
581 @item .byte value1[,value2...]
582 @item .word value1[,value2...]
583 @item .short value1[,value2...]
584 @item .int value1[,value2...]
585 @item .long value1[,value2...]
588 @item .section section
594 @section X86 Assembler
597 All X86 opcodes are supported. Only ATT syntax is supported (source
598 then destination operand order). If no size suffix is given, TinyCC
599 tries to guess it from the operand sizes.
601 Currently, MMX opcodes are supported but not SSE ones.
603 @chapter TinyCC Linker
606 @section ELF file generation
609 TCC can directly output relocatable ELF files (object files),
610 executable ELF files and dynamic ELF libraries without relying on an
613 Dynamic ELF libraries can be output but the C compiler does not generate
614 position independent code (PIC). It means that the dynamic library
615 code generated by TCC cannot be factorized among processes yet.
617 TCC linker eliminates unreferenced object code in libraries. A single pass is
618 done on the object and library list, so the order in which object files and
619 libraries are specified is important (same constraint as GNU ld). No grouping
620 options (@option{--start-group} and @option{--end-group}) are supported.
622 @section ELF file loader
624 TCC can load ELF object files, archives (.a files) and dynamic
627 @section GNU Linker Scripts
628 @cindex scripts, linker
629 @cindex linker scripts
630 @cindex GROUP, linker command
631 @cindex FILE, linker command
632 @cindex OUTPUT_FORMAT, linker command
633 @cindex TARGET, linker command
635 Because on many Linux systems some dynamic libraries (such as
636 @file{/usr/lib/libc.so}) are in fact GNU ld link scripts (horrible!),
637 the TCC linker also supports a subset of GNU ld scripts.
639 The @code{GROUP} and @code{FILE} commands are supported. @code{OUTPUT_FORMAT}
640 and @code{TARGET} are ignored.
642 Example from @file{/usr/lib/libc.so}:
645 Use the shared library, but some functions are only in
646 the static library, so try that secondarily. */
647 GROUP ( /lib/libc.so.6 /usr/lib/libc_nonshared.a )
651 @chapter TinyCC Memory and Bound checks
653 @cindex memory checks
655 This feature is activated with the @option{-b} (@pxref{Invoke}).
657 Note that pointer size is @emph{unchanged} and that code generated
658 with bound checks is @emph{fully compatible} with unchecked
659 code. When a pointer comes from unchecked code, it is assumed to be
660 valid. Even very obscure C code with casts should work correctly.
662 For more information about the ideas behind this method, see
663 @url{http://www.doc.ic.ac.uk/~phjk/BoundsChecking.html}.
665 Here are some examples of caught errors:
669 @item Invalid range with standard string function:
677 @item Out of bounds-error in global or local arrays:
687 @item Out of bounds-error in malloc'ed data:
691 tab = malloc(20 * sizeof(int));
699 @item Access of freed memory:
703 tab = malloc(20 * sizeof(int));
715 tab = malloc(20 * sizeof(int));
724 @chapter The @code{libtcc} library
726 The @code{libtcc} library enables you to use TCC as a backend for
727 dynamic code generation.
729 Read the @file{libtcc.h} to have an overview of the API. Read
730 @file{libtcc_test.c} to have a very simple example.
732 The idea consists in giving a C string containing the program you want
733 to compile directly to @code{libtcc}. Then you can access to any global
734 symbol (function or variable) defined.
736 @chapter Developer's guide
738 This chapter gives some hints to understand how TCC works. You can skip
739 it if you do not intend to modify the TCC code.
741 @section File reading
743 The @code{BufferedFile} structure contains the context needed to read a
744 file, including the current line number. @code{tcc_open()} opens a new
745 file and @code{tcc_close()} closes it. @code{inp()} returns the next
750 @code{next()} reads the next token in the current
751 file. @code{next_nomacro()} reads the next token without macro
754 @code{tok} contains the current token (see @code{TOK_xxx})
755 constants. Identifiers and keywords are also keywords. @code{tokc}
756 contains additional infos about the token (for example a constant value
757 if number or string token).
761 The parser is hardcoded (yacc is not necessary). It does only one pass,
766 @item For initialized arrays with unknown size, a first pass
767 is done to count the number of elements.
769 @item For architectures where arguments are evaluated in
770 reverse order, a first pass is done to reverse the argument order.
776 The types are stored in a single 'int' variable. It was choosen in the
777 first stages of development when tcc was much simpler. Now, it may not
778 be the best solution.
781 #define VT_INT 0 /* integer type */
782 #define VT_BYTE 1 /* signed byte type */
783 #define VT_SHORT 2 /* short type */
784 #define VT_VOID 3 /* void type */
785 #define VT_PTR 4 /* pointer */
786 #define VT_ENUM 5 /* enum definition */
787 #define VT_FUNC 6 /* function type */
788 #define VT_STRUCT 7 /* struct/union definition */
789 #define VT_FLOAT 8 /* IEEE float */
790 #define VT_DOUBLE 9 /* IEEE double */
791 #define VT_LDOUBLE 10 /* IEEE long double */
792 #define VT_BOOL 11 /* ISOC99 boolean type */
793 #define VT_LLONG 12 /* 64 bit integer */
794 #define VT_LONG 13 /* long integer (NEVER USED as type, only
796 #define VT_BTYPE 0x000f /* mask for basic type */
797 #define VT_UNSIGNED 0x0010 /* unsigned type */
798 #define VT_ARRAY 0x0020 /* array type (also has VT_PTR) */
799 #define VT_BITFIELD 0x0040 /* bitfield modifier */
801 #define VT_STRUCT_SHIFT 16 /* structure/enum name shift (16 bits left) */
804 When a reference to another type is needed (for pointers, functions and
805 structures), the @code{32 - VT_STRUCT_SHIFT} high order bits are used to
806 store an identifier reference.
808 The @code{VT_UNSIGNED} flag can be set for chars, shorts, ints and long
811 Arrays are considered as pointers @code{VT_PTR} with the flag
814 The @code{VT_BITFIELD} flag can be set for chars, shorts, ints and long
815 longs. If it is set, then the bitfield position is stored from bits
816 VT_STRUCT_SHIFT to VT_STRUCT_SHIFT + 5 and the bit field size is stored
817 from bits VT_STRUCT_SHIFT + 6 to VT_STRUCT_SHIFT + 11.
819 @code{VT_LONG} is never used except during parsing.
821 During parsing, the storage of an object is also stored in the type
825 #define VT_EXTERN 0x00000080 /* extern definition */
826 #define VT_STATIC 0x00000100 /* static variable */
827 #define VT_TYPEDEF 0x00000200 /* typedef definition */
832 All symbols are stored in hashed symbol stacks. Each symbol stack
833 contains @code{Sym} structures.
835 @code{Sym.v} contains the symbol name (remember
836 an idenfier is also a token, so a string is never necessary to store
837 it). @code{Sym.t} gives the type of the symbol. @code{Sym.r} is usually
838 the register in which the corresponding variable is stored. @code{Sym.c} is
839 usually a constant associated to the symbol.
841 Four main symbol stacks are defined:
846 for the macros (@code{#define}s).
849 for the global variables, functions and types.
852 for the local variables, functions and types.
854 @item global_label_stack
855 for the local labels (for @code{goto}).
858 for GCC block local labels (see the @code{__label__} keyword).
862 @code{sym_push()} is used to add a new symbol in the local symbol
863 stack. If no local symbol stack is active, it is added in the global
866 @code{sym_pop(st,b)} pops symbols from the symbol stack @var{st} until
867 the symbol @var{b} is on the top of stack. If @var{b} is NULL, the stack
870 @code{sym_find(v)} return the symbol associated to the identifier
871 @var{v}. The local stack is searched first from top to bottom, then the
876 The generated code and datas are written in sections. The structure
877 @code{Section} contains all the necessary information for a given
878 section. @code{new_section()} creates a new section. ELF file semantics
879 is assumed for each section.
881 The following sections are predefined:
886 is the section containing the generated code. @var{ind} contains the
887 current position in the code section.
890 contains initialized data
893 contains uninitialized data
896 @itemx lbounds_section
897 are used when bound checking is activated
900 @itemx stabstr_section
901 are used when debugging is actived to store debug information
904 @itemx strtab_section
905 contain the exported symbols (currently only used for debugging).
909 @section Code generation
910 @cindex code generation
912 @subsection Introduction
914 The TCC code generator directly generates linked binary code in one
915 pass. It is rather unusual these days (see gcc for example which
916 generates text assembly), but it can be very fast and surprisingly
919 The TCC code generator is register based. Optimization is only done at
920 the expression level. No intermediate representation of expression is
921 kept except the current values stored in the @emph{value stack}.
923 On x86, three temporary registers are used. When more registers are
924 needed, one register is spilled into a new temporary variable on the stack.
926 @subsection The value stack
927 @cindex value stack, introduction
929 When an expression is parsed, its value is pushed on the value stack
930 (@var{vstack}). The top of the value stack is @var{vtop}. Each value
931 stack entry is the structure @code{SValue}.
933 @code{SValue.t} is the type. @code{SValue.r} indicates how the value is
934 currently stored in the generated code. It is usually a CPU register
935 index (@code{REG_xxx} constants), but additional values and flags are
939 #define VT_CONST 0x00f0
940 #define VT_LLOCAL 0x00f1
941 #define VT_LOCAL 0x00f2
942 #define VT_CMP 0x00f3
943 #define VT_JMP 0x00f4
944 #define VT_JMPI 0x00f5
945 #define VT_LVAL 0x0100
946 #define VT_SYM 0x0200
947 #define VT_MUSTCAST 0x0400
948 #define VT_MUSTBOUND 0x0800
949 #define VT_BOUNDED 0x8000
950 #define VT_LVAL_BYTE 0x1000
951 #define VT_LVAL_SHORT 0x2000
952 #define VT_LVAL_UNSIGNED 0x4000
953 #define VT_LVAL_TYPE (VT_LVAL_BYTE | VT_LVAL_SHORT | VT_LVAL_UNSIGNED)
959 indicates that the value is a constant. It is stored in the union
960 @code{SValue.c}, depending on its type.
963 indicates a local variable pointer at offset @code{SValue.c.i} in the
967 indicates that the value is actually stored in the CPU flags (i.e. the
968 value is the consequence of a test). The value is either 0 or 1. The
969 actual CPU flags used is indicated in @code{SValue.c.i}.
971 If any code is generated which destroys the CPU flags, this value MUST be
972 put in a normal register.
976 indicates that the value is the consequence of a conditional jump. For VT_JMP,
977 it is 1 if the jump is taken, 0 otherwise. For VT_JMPI it is inverted.
979 These values are used to compile the @code{||} and @code{&&} logical
982 If any code is generated, this value MUST be put in a normal
983 register. Otherwise, the generated code won't be executed if the jump is
987 is a flag indicating that the value is actually an lvalue (left value of
988 an assignment). It means that the value stored is actually a pointer to
991 Understanding the use @code{VT_LVAL} is very important if you want to
992 understand how TCC works.
996 @itemx VT_LVAL_UNSIGNED
997 if the lvalue has an integer type, then these flags give its real
998 type. The type alone is not enough in case of cast optimisations.
1001 is a saved lvalue on the stack. @code{VT_LLOCAL} should be eliminated
1002 ASAP because its semantics are rather complicated.
1005 indicates that a cast to the value type must be performed if the value
1006 is used (lazy casting).
1009 indicates that the symbol @code{SValue.sym} must be added to the constant.
1013 are only used for optional bound checking.
1017 @subsection Manipulating the value stack
1020 @code{vsetc()} and @code{vset()} pushes a new value on the value
1021 stack. If the previous @var{vtop} was stored in a very unsafe place(for
1022 example in the CPU flags), then some code is generated to put the
1023 previous @var{vtop} in a safe storage.
1025 @code{vpop()} pops @var{vtop}. In some cases, it also generates cleanup
1026 code (for example if stacked floating point registers are used as on
1029 The @code{gv(rc)} function generates code to evaluate @var{vtop} (the
1030 top value of the stack) into registers. @var{rc} selects in which
1031 register class the value should be put. @code{gv()} is the @emph{most
1032 important function} of the code generator.
1034 @code{gv2()} is the same as @code{gv()} but for the top two stack
1037 @subsection CPU dependent code generation
1038 @cindex CPU dependent
1039 See the @file{i386-gen.c} file to have an example.
1044 must generate the code needed to load a stack value into a register.
1047 must generate the code needed to store a register into a stack value
1051 @itemx gfunc_param()
1053 should generate a function call
1055 @item gfunc_prolog()
1056 @itemx gfunc_epilog()
1057 should generate a function prolog/epilog.
1060 must generate the binary integer operation @var{op} on the two top
1061 entries of the stack which are guaranted to contain integer types.
1063 The result value should be put on the stack.
1066 same as @code{gen_opi()} for floating point operations. The two top
1067 entries of the stack are guaranted to contain floating point values of
1070 @item gen_cvt_itof()
1071 integer to floating point conversion.
1073 @item gen_cvt_ftoi()
1074 floating point to integer conversion.
1076 @item gen_cvt_ftof()
1077 floating point to floating point of different size conversion.
1079 @item gen_bounded_ptr_add()
1080 @item gen_bounded_ptr_deref()
1081 are only used for bounds checking.
1085 @section Optimizations done
1086 @cindex optimizations
1087 @cindex constant propagation
1088 @cindex strength reduction
1089 @cindex comparison operators
1090 @cindex caching processor flags
1091 @cindex flags, caching
1092 @cindex jump optimization
1093 Constant propagation is done for all operations. Multiplications and
1094 divisions are optimized to shifts when appropriate. Comparison
1095 operators are optimized by maintaining a special cache for the
1096 processor flags. &&, || and ! are optimized by maintaining a special
1097 'jump target' value. No other jump optimization is currently performed
1098 because it would require to store the code in a more abstract fashion.
1100 @unnumbered Concept Index
1107 @c texinfo-column-for-description: 32