1/* This file contains the definitions and documentation for the 2 machine modes used in the GNU compiler. 3 Copyright (C) 1987, 1992, 1994, 1997, 1998, 2000, 2003, 2004, 2005 4 Free Software Foundation, Inc. 5 6This file is part of GCC. 7 8GCC is free software; you can redistribute it and/or modify it under 9the terms of the GNU General Public License as published by the Free 10Software Foundation; either version 2, or (at your option) any later 11version. 12 13GCC is distributed in the hope that it will be useful, but WITHOUT ANY 14WARRANTY; without even the implied warranty of MERCHANTABILITY or 15FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 16for more details. 17 18You should have received a copy of the GNU General Public License 19along with GCC; see the file COPYING. If not, write to the Free 20Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 2102110-1301, USA. */ 22 23 24/* This file defines all the MACHINE MODES used by GCC. 25 26 A machine mode specifies a size and format of data 27 at the machine level. 28 29 Each RTL expression has a machine mode. 30 31 At the syntax tree level, each ..._TYPE and each ..._DECL node 32 has a machine mode which describes data of that type or the 33 data of the variable declared. */ 34 35/* This file is included by the genmodes program. Its text is the 36 body of a function. Do not rely on this, it will change in the 37 future. 38 39 The following statements can be used in this file -- all have 40 the form of a C macro call. In their arguments: 41 42 A CLASS argument must be one of the constants defined in 43 mode-classes.def, less the leading MODE_ prefix; some statements 44 that take CLASS arguments have restrictions on which classes are 45 acceptable. For instance, INT. 46 47 A MODE argument must be the printable name of a machine mode, 48 without quotation marks or trailing "mode". For instance, SI. 49 50 A PRECISION, BYTESIZE, or COUNT argument must be a positive integer 51 constant. 52 53 A FORMAT argument must be one of the real_mode_format structures 54 declared in real.h, or else a literal 0. Do not put a leading & 55 on the argument. 56 57 An EXPR argument must be a syntactically valid C expression. 58 If an EXPR contains commas, you may need to write an extra pair of 59 parentheses around it, so it appears to be a single argument to the 60 statement. 61 62 This file defines only those modes which are of use on almost all 63 machines. Other modes can be defined in the target-specific 64 mode definition file, config/ARCH/ARCH-modes.def. 65 66 Order matters in this file in so far as statements which refer to 67 other modes must appear after the modes they refer to. However, 68 statements which do not refer to other modes may appear in any 69 order. 70 71 RANDOM_MODE (MODE); 72 declares MODE to be of class RANDOM. 73 74 CC_MODE (MODE); 75 declares MODE to be of class CC. 76 77 INT_MODE (MODE, BYTESIZE); 78 declares MODE to be of class INT and BYTESIZE bytes wide. 79 All of the bits of its representation are significant. 80 81 FRACTIONAL_INT_MODE (MODE, PRECISION, BYTESIZE); 82 declares MODE to be of class INT, BYTESIZE bytes wide in 83 storage, but with only PRECISION significant bits. 84 85 FLOAT_MODE (MODE, BYTESIZE, FORMAT); 86 declares MODE to be of class FLOAT and BYTESIZE bytes wide, 87 using floating point format FORMAT. 88 All of the bits of its representation are significant. 89 90 DECIMAL FLOAT_MODE (MODE, BYTESIZE); 91 declares MODE to be of class DECIMAL_FLOAT and BYTESIZE bytes 92 wide. All of the bits of its representation are significant. 93 94 FRACTIONAL_FLOAT_MODE (MODE, PRECISION, BYTESIZE, FORMAT); 95 declares MODE to be of class FLOAT, BYTESIZE bytes wide in 96 storage, but with only PRECISION significant bits, using 97 floating point format FORMAT. 98 99 RESET_FLOAT_FORMAT (MODE, FORMAT); 100 changes the format of MODE, which must be class FLOAT, 101 to FORMAT. Use in an ARCH-modes.def to reset the format 102 of one of the float modes defined in this file. 103 104 PARTIAL_INT_MODE (MODE); 105 declares a mode of class PARTIAL_INT with the same size as 106 MODE (which must be an INT mode). The name of the new mode 107 is made by prefixing a P to the name MODE. This statement 108 may grow a PRECISION argument in the future. 109 110 VECTOR_MODE (CLASS, MODE, COUNT); 111 Declare a vector mode whose component mode is MODE (of class 112 CLASS) with COUNT components. CLASS must be INT or FLOAT. 113 The name of the vector mode takes the form VnX where n is 114 COUNT in decimal and X is MODE. 115 116 VECTOR_MODES (CLASS, WIDTH); 117 For all modes presently declared in class CLASS, construct 118 corresponding vector modes having width WIDTH. Modes whose 119 byte sizes do not evenly divide WIDTH are ignored, as are 120 modes that would produce vector modes with only one component, 121 and modes smaller than one byte (if CLASS is INT) or smaller 122 than two bytes (if CLASS is FLOAT). CLASS must be INT or 123 FLOAT. The names follow the same rule as VECTOR_MODE uses. 124 125 COMPLEX_MODES (CLASS); 126 For all modes presently declared in class CLASS, construct 127 corresponding complex modes. Modes smaller than one byte 128 are ignored. For FLOAT modes, the names are derived by 129 replacing the 'F' in the mode name with a 'C'. (It is an 130 error if there is no 'F'. For INT modes, the names are 131 derived by prefixing a C to the name. 132 133 ADJUST_BYTESIZE (MODE, EXPR); 134 ADJUST_ALIGNMENT (MODE, EXPR); 135 ADJUST_FLOAT_FORMAT (MODE, EXPR); 136 Arrange for the byte size, alignment, or floating point format 137 of MODE to be adjustable at run time. EXPR will be executed 138 once after processing all command line options, and should 139 evaluate to the desired byte size, alignment, or format. 140 141 Unlike a FORMAT argument, if you are adjusting a float format 142 you must put an & in front of the name of each format structure. 143 144 Note: If a mode is ever made which is more than 255 bytes wide, 145 machmode.h and genmodes.c will have to be changed to allocate 146 more space for the mode_size and mode_alignment arrays. */ 147 148/* VOIDmode is used when no mode needs to be specified, 149 as for example on CONST_INT RTL expressions. */ 150RANDOM_MODE (VOID); 151 152/* BLKmode is used for structures, arrays, etc. 153 that fit no more specific mode. */ 154RANDOM_MODE (BLK); 155 156/* Single bit mode used for booleans. */ 157FRACTIONAL_INT_MODE (BI, 1, 1); 158 159/* Basic integer modes. We go up to TI in generic code (128 bits). 160 The name OI is reserved for a 256-bit type (needed by some back ends). 161 FIXME TI shouldn't be generically available either. */ 162INT_MODE (QI, 1); 163INT_MODE (HI, 2); 164INT_MODE (SI, 4); 165INT_MODE (DI, 8); 166INT_MODE (TI, 16); 167 168/* No partial integer modes are defined by default. */ 169 170/* Basic floating point modes. SF and DF are the only modes provided 171 by default. The names QF, HF, XF, and TF are reserved for targets 172 that need 1-word, 2-word, 80-bit, or 128-bit float types respectively. 173 174 These are the IEEE mappings. They can be overridden with 175 RESET_FLOAT_FORMAT or at runtime (in OVERRIDE_OPTIONS). */ 176 177FLOAT_MODE (SF, 4, ieee_single_format); 178FLOAT_MODE (DF, 8, ieee_double_format); 179 180/* Basic CC modes. 181 FIXME define this only for targets that need it. */ 182CC_MODE (CC); 183 184/* Allow the target to specify additional modes of various kinds. */ 185#if HAVE_EXTRA_MODES 186# include EXTRA_MODES_FILE 187#endif 188 189/* Complex modes. */ 190COMPLEX_MODES (INT); 191COMPLEX_MODES (FLOAT); 192 193/* Decimal floating point modes. */ 194DECIMAL_FLOAT_MODE (SD, 4, decimal_single_format); 195DECIMAL_FLOAT_MODE (DD, 8, decimal_double_format); 196DECIMAL_FLOAT_MODE (TD, 16, decimal_quad_format); 197 198/* The symbol Pmode stands for one of the above machine modes (usually SImode). 199 The tm.h file specifies which one. It is not a distinct mode. */ 200 201/* 202Local variables: 203mode:c 204version-control: t 205End: 206*/ 207