Home | History | Annotate | Line # | Download | only in softfloat
softfloat.h revision 1.11
      1 /*	$NetBSD: softfloat.h,v 1.11 2025/09/17 11:37:38 nat Exp $	*/
      2 
      3 /* This is a derivative work. */
      4 
      5 /*
      6 ===============================================================================
      7 
      8 This C header file is part of the SoftFloat IEC/IEEE Floating-point
      9 Arithmetic Package, Release 2a.
     10 
     11 Written by John R. Hauser.  This work was made possible in part by the
     12 International Computer Science Institute, located at Suite 600, 1947 Center
     13 Street, Berkeley, California 94704.  Funding was partially provided by the
     14 National Science Foundation under grant MIP-9311980.  The original version
     15 of this code was written as part of a project to build a fixed-point vector
     16 processor in collaboration with the University of California at Berkeley,
     17 overseen by Profs. Nelson Morgan and John Wawrzynek.  More information
     18 is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/
     19 arithmetic/SoftFloat.html'.
     20 
     21 THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE.  Although reasonable effort
     22 has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT
     23 TIMES RESULT IN INCORRECT BEHAVIOR.  USE OF THIS SOFTWARE IS RESTRICTED TO
     24 PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY
     25 AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE.
     26 
     27 Derivative works are acceptable, even for commercial purposes, so long as
     28 (1) they include prominent notice that the work is derivative, and (2) they
     29 include prominent notice akin to these four paragraphs for those parts of
     30 this code that are retained.
     31 
     32 ===============================================================================
     33 */
     34 
     35 /*
     36 -------------------------------------------------------------------------------
     37 The macro `FLOATX80' must be defined to enable the extended double-precision
     38 floating-point format `floatx80'.  If this macro is not defined, the
     39 `floatx80' type will not be defined, and none of the functions that either
     40 input or output the `floatx80' type will be defined.  The same applies to
     41 the `FLOAT128' macro and the quadruple-precision format `float128'.
     42 -------------------------------------------------------------------------------
     43 */
     44 #ifndef __mcoldfire__
     45 #define FLOATX80
     46 #endif
     47 /* #define FLOAT128 */
     48 
     49 #include <machine/ieeefp.h>
     50 
     51 /*
     52 -------------------------------------------------------------------------------
     53 Software IEC/IEEE floating-point types.
     54 -------------------------------------------------------------------------------
     55 */
     56 typedef unsigned int float32;
     57 typedef unsigned long long float64;
     58 #ifdef FLOATX80
     59 #define X80SHIFT	16
     60 
     61 typedef struct {
     62     unsigned long high;
     63     unsigned long long low;
     64 }__packed floatx80;
     65 #endif
     66 #ifdef FLOAT128
     67 typedef struct {
     68     unsigned long long high, low;
     69 } float128;
     70 #endif
     71 
     72 /*
     73 -------------------------------------------------------------------------------
     74 Software IEC/IEEE floating-point underflow tininess-detection mode.
     75 -------------------------------------------------------------------------------
     76 */
     77 #ifndef SOFTFLOAT_FOR_GCC
     78 extern int8 float_detect_tininess;
     79 #endif
     80 enum {
     81     float_tininess_after_rounding  = 0,
     82     float_tininess_before_rounding = 1
     83 };
     84 
     85 /*
     86 -------------------------------------------------------------------------------
     87 Software IEC/IEEE floating-point rounding mode.
     88 -------------------------------------------------------------------------------
     89 */
     90 extern fp_rnd float_rounding_mode;
     91 #define float_round_nearest_even FP_RN
     92 #define float_round_to_zero      FP_RZ
     93 #define float_round_down         FP_RM
     94 #define float_round_up           FP_RP
     95 
     96 /*
     97 -------------------------------------------------------------------------------
     98 Software IEC/IEEE floating-point exception flags.
     99 -------------------------------------------------------------------------------
    100 */
    101 extern fp_except float_exception_flags;
    102 extern fp_except float_exception_mask;
    103 enum {
    104     float_flag_inexact   = FP_X_IMP,
    105     float_flag_underflow = FP_X_UFL,
    106     float_flag_overflow  = FP_X_OFL,
    107     float_flag_divbyzero = FP_X_DZ,
    108     float_flag_invalid   = FP_X_INV
    109 };
    110 
    111 /*
    112 -------------------------------------------------------------------------------
    113 Routine to raise any or all of the software IEC/IEEE floating-point
    114 exception flags.
    115 -------------------------------------------------------------------------------
    116 */
    117 void float_raise( fp_except );
    118 
    119 /*
    120 -------------------------------------------------------------------------------
    121 Software IEC/IEEE integer-to-floating-point conversion routines.
    122 -------------------------------------------------------------------------------
    123 */
    124 float32 int32_to_float32( int32 );
    125 float32 uint32_to_float32( uint32 );
    126 float64 int32_to_float64( int32 );
    127 float64 uint32_to_float64( uint32 );
    128 #ifdef FLOATX80
    129 floatx80 int32_to_floatx80( int32 );
    130 floatx80 uint32_to_floatx80( uint32 );
    131 #endif
    132 #ifdef FLOAT128
    133 float128 int32_to_float128( int32 );
    134 float128 uint32_to_float128( uint32 );
    135 #endif
    136 float32 int64_to_float32( long long );
    137 float64 int64_to_float64( long long );
    138 #ifdef FLOATX80
    139 floatx80 int64_to_floatx80( long long );
    140 #endif
    141 #ifdef FLOAT128
    142 float128 int64_to_float128( long long );
    143 #endif
    144 
    145 /*
    146 -------------------------------------------------------------------------------
    147 Software IEC/IEEE single-precision conversion routines.
    148 -------------------------------------------------------------------------------
    149 */
    150 int float32_to_int32( float32 );
    151 int float32_to_int32_round_to_zero( float32 );
    152 unsigned int float32_to_uint32_round_to_zero( float32 );
    153 long long float32_to_int64( float32 );
    154 long long float32_to_int64_round_to_zero( float32 );
    155 float64 float32_to_float64( float32 );
    156 #ifdef FLOATX80
    157 floatx80 float32_to_floatx80( float32 );
    158 #endif
    159 #ifdef FLOAT128
    160 float128 float32_to_float128( float32 );
    161 #endif
    162 
    163 /*
    164 -------------------------------------------------------------------------------
    165 Software IEC/IEEE single-precision operations.
    166 -------------------------------------------------------------------------------
    167 */
    168 float32 float32_round_to_int( float32 );
    169 float32 float32_add( float32, float32 );
    170 float32 float32_sub( float32, float32 );
    171 float32 float32_mul( float32, float32 );
    172 float32 float32_div( float32, float32 );
    173 float32 float32_rem( float32, float32 );
    174 float32 float32_sqrt( float32 );
    175 flag float32_eq( float32, float32 );
    176 flag float32_le( float32, float32 );
    177 flag float32_lt( float32, float32 );
    178 flag float32_eq_signaling( float32, float32 );
    179 flag float32_le_quiet( float32, float32 );
    180 flag float32_lt_quiet( float32, float32 );
    181 #if !defined(SOFTFLOAT_FOR_GCC) || defined(SOFTFLOATM68K_FOR_GCC)
    182 flag float32_is_signaling_nan( float32 );
    183 #endif
    184 
    185 /*
    186 -------------------------------------------------------------------------------
    187 Software IEC/IEEE double-precision conversion routines.
    188 -------------------------------------------------------------------------------
    189 */
    190 int float64_to_int32( float64 );
    191 int float64_to_int32_round_to_zero( float64 );
    192 unsigned int float64_to_uint32_round_to_zero( float64 );
    193 long long float64_to_int64( float64 );
    194 long long float64_to_int64_round_to_zero( float64 );
    195 float32 float64_to_float32( float64 );
    196 #ifdef FLOATX80
    197 floatx80 float64_to_floatx80( float64 );
    198 #endif
    199 #ifdef FLOAT128
    200 float128 float64_to_float128( float64 );
    201 #endif
    202 
    203 /*
    204 -------------------------------------------------------------------------------
    205 Software IEC/IEEE double-precision operations.
    206 -------------------------------------------------------------------------------
    207 */
    208 float64 float64_round_to_int( float64 );
    209 float64 float64_add( float64, float64 );
    210 float64 float64_sub( float64, float64 );
    211 float64 float64_mul( float64, float64 );
    212 float64 float64_div( float64, float64 );
    213 float64 float64_rem( float64, float64 );
    214 float64 float64_sqrt( float64 );
    215 flag float64_eq( float64, float64 );
    216 flag float64_le( float64, float64 );
    217 flag float64_lt( float64, float64 );
    218 flag float64_eq_signaling( float64, float64 );
    219 flag float64_le_quiet( float64, float64 );
    220 flag float64_lt_quiet( float64, float64 );
    221 #if !defined(SOFTFLOAT_FOR_GCC) || defined(SOFTFLOATM68K_FOR_GCC)
    222 flag float64_is_signaling_nan( float64 );
    223 #endif
    224 
    225 #ifdef FLOATX80
    226 
    227 /*
    228 -------------------------------------------------------------------------------
    229 Software IEC/IEEE extended double-precision conversion routines.
    230 -------------------------------------------------------------------------------
    231 */
    232 int floatx80_to_int32( floatx80 );
    233 int floatx80_to_int32_round_to_zero( floatx80 );
    234 long long floatx80_to_int64( floatx80 );
    235 long long floatx80_to_int64_round_to_zero( floatx80 );
    236 float32 floatx80_to_float32( floatx80 );
    237 float64 floatx80_to_float64( floatx80 );
    238 #ifdef FLOAT128
    239 float128 floatx80_to_float128( floatx80 );
    240 #endif
    241 
    242 /*
    243 -------------------------------------------------------------------------------
    244 Software IEC/IEEE extended double-precision rounding precision.  Valid
    245 values are 32, 64, and 80.
    246 -------------------------------------------------------------------------------
    247 */
    248 extern int floatx80_rounding_precision;
    249 
    250 /*
    251 -------------------------------------------------------------------------------
    252 Software IEC/IEEE extended double-precision operations.
    253 -------------------------------------------------------------------------------
    254 */
    255 floatx80 floatx80_round_to_int( floatx80 );
    256 floatx80 floatx80_add( floatx80, floatx80 );
    257 floatx80 floatx80_sub( floatx80, floatx80 );
    258 floatx80 floatx80_mul( floatx80, floatx80 );
    259 floatx80 floatx80_div( floatx80, floatx80 );
    260 floatx80 floatx80_rem( floatx80, floatx80 );
    261 floatx80 floatx80_sqrt( floatx80 );
    262 flag floatx80_eq( floatx80, floatx80 );
    263 flag floatx80_le( floatx80, floatx80 );
    264 flag floatx80_lt( floatx80, floatx80 );
    265 flag floatx80_eq_signaling( floatx80, floatx80 );
    266 flag floatx80_le_quiet( floatx80, floatx80 );
    267 flag floatx80_lt_quiet( floatx80, floatx80 );
    268 flag floatx80_is_signaling_nan( floatx80 );
    269 flag floatx80_is_nan( floatx80 );
    270 
    271 #endif
    272 
    273 #ifdef FLOAT128
    274 
    275 /*
    276 -------------------------------------------------------------------------------
    277 Software IEC/IEEE quadruple-precision conversion routines.
    278 -------------------------------------------------------------------------------
    279 */
    280 int float128_to_int32( float128 );
    281 int float128_to_int32_round_to_zero( float128 );
    282 long long float128_to_int64( float128 );
    283 long long float128_to_int64_round_to_zero( float128 );
    284 float32 float128_to_float32( float128 );
    285 float64 float128_to_float64( float128 );
    286 #ifdef FLOATX80
    287 floatx80 float128_to_floatx80( float128 );
    288 #endif
    289 
    290 /*
    291 -------------------------------------------------------------------------------
    292 Software IEC/IEEE quadruple-precision operations.
    293 -------------------------------------------------------------------------------
    294 */
    295 float128 float128_round_to_int( float128 );
    296 float128 float128_add( float128, float128 );
    297 float128 float128_sub( float128, float128 );
    298 float128 float128_mul( float128, float128 );
    299 float128 float128_div( float128, float128 );
    300 float128 float128_rem( float128, float128 );
    301 float128 float128_sqrt( float128 );
    302 flag float128_eq( float128, float128 );
    303 flag float128_le( float128, float128 );
    304 flag float128_lt( float128, float128 );
    305 flag float128_eq_signaling( float128, float128 );
    306 flag float128_le_quiet( float128, float128 );
    307 flag float128_lt_quiet( float128, float128 );
    308 flag float128_is_signaling_nan( float128 );
    309 
    310 #endif
    311