| 1 | /* ix87 specific implementation of pow function.
|
|---|
| 2 | Copyright (C) 1996, 1997, 1998, 1999, 2001, 2004, 2005
|
|---|
| 3 | Free Software Foundation, Inc.
|
|---|
| 4 | This file is part of the GNU C Library.
|
|---|
| 5 | Contributed by Ulrich Drepper <[email protected]>, 1996.
|
|---|
| 6 |
|
|---|
| 7 | The GNU C Library is free software; you can redistribute it and/or
|
|---|
| 8 | modify it under the terms of the GNU Lesser General Public
|
|---|
| 9 | License as published by the Free Software Foundation; either
|
|---|
| 10 | version 2.1 of the License, or (at your option) any later version.
|
|---|
| 11 |
|
|---|
| 12 | The GNU C Library is distributed in the hope that it will be useful,
|
|---|
| 13 | but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|---|
| 14 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|---|
| 15 | Lesser General Public License for more details.
|
|---|
| 16 |
|
|---|
| 17 | You should have received a copy of the GNU Lesser General Public
|
|---|
| 18 | License along with the GNU C Library; if not, write to the Free
|
|---|
| 19 | Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
|
|---|
| 20 | 02111-1307 USA. */
|
|---|
| 21 |
|
|---|
| 22 | /*#include <machine/asm.h>*/
|
|---|
| 23 | #include <iprt/cdefs.h>
|
|---|
| 24 |
|
|---|
| 25 | #ifdef __MINGW32__
|
|---|
| 26 | # define ASM_TYPE_DIRECTIVE(name,typearg)
|
|---|
| 27 | # define ASM_SIZE_DIRECTIVE(name)
|
|---|
| 28 | # define cfi_adjust_cfa_offset(a)
|
|---|
| 29 | # define C_LABEL(name) _ ## name:
|
|---|
| 30 | # define C_SYMBOL_NAME(name) _ ## name
|
|---|
| 31 | # define ASM_GLOBAL_DIRECTIVE .global
|
|---|
| 32 | # define ALIGNARG(log2) 1<<log2
|
|---|
| 33 | #elif __APPLE__
|
|---|
| 34 | # define ASM_TYPE_DIRECTIVE(name,typearg)
|
|---|
| 35 | # define ASM_SIZE_DIRECTIVE(name)
|
|---|
| 36 | # define cfi_adjust_cfa_offset(a)
|
|---|
| 37 | # define C_LABEL(name) _ ## name:
|
|---|
| 38 | # define C_SYMBOL_NAME(name) _ ## name
|
|---|
| 39 | # define ASM_GLOBAL_DIRECTIVE .globl
|
|---|
| 40 | # define ALIGNARG(log2) log2
|
|---|
| 41 | #else
|
|---|
| 42 | # define ASM_TYPE_DIRECTIVE(name,typearg) .type name,typearg;
|
|---|
| 43 | # define ASM_SIZE_DIRECTIVE(name) .size name,.-name;
|
|---|
| 44 | # define C_LABEL(name) name:
|
|---|
| 45 | # define C_SYMBOL_NAME(name) name
|
|---|
| 46 | # /* figure this one out. */
|
|---|
| 47 | # define cfi_adjust_cfa_offset(a)
|
|---|
| 48 | # define ASM_GLOBAL_DIRECTIVE .global
|
|---|
| 49 | # define ALIGNARG(log2) 1<<log2
|
|---|
| 50 | #endif
|
|---|
| 51 |
|
|---|
| 52 | #define ENTRY(name) \
|
|---|
| 53 | ASM_GLOBAL_DIRECTIVE C_SYMBOL_NAME(name); \
|
|---|
| 54 | ASM_TYPE_DIRECTIVE (C_SYMBOL_NAME(name),@function) \
|
|---|
| 55 | .align ALIGNARG(4); \
|
|---|
| 56 | C_LABEL(name)
|
|---|
| 57 |
|
|---|
| 58 | #undef END
|
|---|
| 59 | #define END(name) \
|
|---|
| 60 | ASM_SIZE_DIRECTIVE(name)
|
|---|
| 61 |
|
|---|
| 62 | #ifdef __ELF__
|
|---|
| 63 | .section .rodata
|
|---|
| 64 | #else
|
|---|
| 65 | .text
|
|---|
| 66 | #endif
|
|---|
| 67 |
|
|---|
| 68 | .align ALIGNARG(4)
|
|---|
| 69 | ASM_TYPE_DIRECTIVE(infinity,@object)
|
|---|
| 70 | inf_zero:
|
|---|
| 71 | infinity:
|
|---|
| 72 | .byte 0, 0, 0, 0, 0, 0, 0xf0, 0x7f
|
|---|
| 73 | ASM_SIZE_DIRECTIVE(infinity)
|
|---|
| 74 | ASM_TYPE_DIRECTIVE(zero,@object)
|
|---|
| 75 | zero: .double 0.0
|
|---|
| 76 | ASM_SIZE_DIRECTIVE(zero)
|
|---|
| 77 | ASM_TYPE_DIRECTIVE(minf_mzero,@object)
|
|---|
| 78 | minf_mzero:
|
|---|
| 79 | minfinity:
|
|---|
| 80 | .byte 0, 0, 0, 0, 0, 0, 0xf0, 0xff
|
|---|
| 81 | mzero:
|
|---|
| 82 | .byte 0, 0, 0, 0, 0, 0, 0, 0x80
|
|---|
| 83 | ASM_SIZE_DIRECTIVE(minf_mzero)
|
|---|
| 84 | ASM_TYPE_DIRECTIVE(one,@object)
|
|---|
| 85 | one: .double 1.0
|
|---|
| 86 | ASM_SIZE_DIRECTIVE(one)
|
|---|
| 87 | ASM_TYPE_DIRECTIVE(limit,@object)
|
|---|
| 88 | limit: .double 0.29
|
|---|
| 89 | ASM_SIZE_DIRECTIVE(limit)
|
|---|
| 90 | ASM_TYPE_DIRECTIVE(p63,@object)
|
|---|
| 91 | p63: .byte 0, 0, 0, 0, 0, 0, 0xe0, 0x43
|
|---|
| 92 | ASM_SIZE_DIRECTIVE(p63)
|
|---|
| 93 |
|
|---|
| 94 | #ifdef PIC
|
|---|
| 95 | #define MO(op) op##@GOTOFF(%ecx)
|
|---|
| 96 | #define MOX(op,x,f) op##@GOTOFF(%ecx,x,f)
|
|---|
| 97 | #else
|
|---|
| 98 | #define MO(op) op
|
|---|
| 99 | #define MOX(op,x,f) op(,x,f)
|
|---|
| 100 | #endif
|
|---|
| 101 |
|
|---|
| 102 | .text
|
|---|
| 103 | //ENTRY(__ieee754_powl)
|
|---|
| 104 | ENTRY(RT_NOCRT(powl))
|
|---|
| 105 | #ifdef __DARWIN__ /* 16-byte long double with 8 byte alignment requirements */
|
|---|
| 106 | fldt 20(%esp) // y
|
|---|
| 107 | #else
|
|---|
| 108 | fldt 16(%esp) // y
|
|---|
| 109 | #endif
|
|---|
| 110 | fxam
|
|---|
| 111 |
|
|---|
| 112 | #ifdef PIC
|
|---|
| 113 | LOAD_PIC_REG (cx)
|
|---|
| 114 | #endif
|
|---|
| 115 |
|
|---|
| 116 | fnstsw
|
|---|
| 117 | movb %ah, %dl
|
|---|
| 118 | andb $0x45, %ah
|
|---|
| 119 | cmpb $0x40, %ah // is y == 0 ?
|
|---|
| 120 | je .L11
|
|---|
| 121 |
|
|---|
| 122 | cmpb $0x05, %ah // is y == ±inf ?
|
|---|
| 123 | je .L12
|
|---|
| 124 |
|
|---|
| 125 | cmpb $0x01, %ah // is y == NaN ?
|
|---|
| 126 | je .L30
|
|---|
| 127 |
|
|---|
| 128 | fldt 4(%esp) // x : y
|
|---|
| 129 |
|
|---|
| 130 | subl $8,%esp
|
|---|
| 131 | cfi_adjust_cfa_offset (8)
|
|---|
| 132 |
|
|---|
| 133 | fxam
|
|---|
| 134 | fnstsw
|
|---|
| 135 | movb %ah, %dh
|
|---|
| 136 | andb $0x45, %ah
|
|---|
| 137 | cmpb $0x40, %ah
|
|---|
| 138 | je .L20 // x is ±0
|
|---|
| 139 |
|
|---|
| 140 | cmpb $0x05, %ah
|
|---|
| 141 | je .L15 // x is ±inf
|
|---|
| 142 |
|
|---|
| 143 | fxch // y : x
|
|---|
| 144 |
|
|---|
| 145 | /* fistpll raises invalid exception for |y| >= 1L<<63. */
|
|---|
| 146 | fld %st // y : y : x
|
|---|
| 147 | fabs // |y| : y : x
|
|---|
| 148 | fcompl MO(p63) // y : x
|
|---|
| 149 | fnstsw
|
|---|
| 150 | sahf
|
|---|
| 151 | jnc .L2
|
|---|
| 152 |
|
|---|
| 153 | /* First see whether `y' is a natural number. In this case we
|
|---|
| 154 | can use a more precise algorithm. */
|
|---|
| 155 | fld %st // y : y : x
|
|---|
| 156 | fistpll (%esp) // y : x
|
|---|
| 157 | fildll (%esp) // int(y) : y : x
|
|---|
| 158 | fucomp %st(1) // y : x
|
|---|
| 159 | fnstsw
|
|---|
| 160 | sahf
|
|---|
| 161 | jne .L2
|
|---|
| 162 |
|
|---|
| 163 | /* OK, we have an integer value for y. */
|
|---|
| 164 | popl %eax
|
|---|
| 165 | cfi_adjust_cfa_offset (-4)
|
|---|
| 166 | popl %edx
|
|---|
| 167 | cfi_adjust_cfa_offset (-4)
|
|---|
| 168 | orl $0, %edx
|
|---|
| 169 | fstp %st(0) // x
|
|---|
| 170 | jns .L4 // y >= 0, jump
|
|---|
| 171 | fdivrl MO(one) // 1/x (now referred to as x)
|
|---|
| 172 | negl %eax
|
|---|
| 173 | adcl $0, %edx
|
|---|
| 174 | negl %edx
|
|---|
| 175 | .L4: fldl MO(one) // 1 : x
|
|---|
| 176 | fxch
|
|---|
| 177 |
|
|---|
| 178 | .L6: shrdl $1, %edx, %eax
|
|---|
| 179 | jnc .L5
|
|---|
| 180 | fxch
|
|---|
| 181 | fmul %st(1) // x : ST*x
|
|---|
| 182 | fxch
|
|---|
| 183 | .L5: fmul %st(0), %st // x*x : ST*x
|
|---|
| 184 | shrl $1, %edx
|
|---|
| 185 | movl %eax, %ecx
|
|---|
| 186 | orl %edx, %ecx
|
|---|
| 187 | jnz .L6
|
|---|
| 188 | fstp %st(0) // ST*x
|
|---|
| 189 | ret
|
|---|
| 190 |
|
|---|
| 191 | /* y is ±NAN */
|
|---|
| 192 | .L30: fldt 4(%esp) // x : y
|
|---|
| 193 | fldl MO(one) // 1.0 : x : y
|
|---|
| 194 | fucomp %st(1) // x : y
|
|---|
| 195 | fnstsw
|
|---|
| 196 | sahf
|
|---|
| 197 | je .L31
|
|---|
| 198 | fxch // y : x
|
|---|
| 199 | .L31: fstp %st(1)
|
|---|
| 200 | ret
|
|---|
| 201 |
|
|---|
| 202 | cfi_adjust_cfa_offset (8)
|
|---|
| 203 | .align ALIGNARG(4)
|
|---|
| 204 | .L2: /* y is a real number. */
|
|---|
| 205 | fxch // x : y
|
|---|
| 206 | fldl MO(one) // 1.0 : x : y
|
|---|
| 207 | fld %st(1) // x : 1.0 : x : y
|
|---|
| 208 | fsub %st(1) // x-1 : 1.0 : x : y
|
|---|
| 209 | fabs // |x-1| : 1.0 : x : y
|
|---|
| 210 | fcompl MO(limit) // 1.0 : x : y
|
|---|
| 211 | fnstsw
|
|---|
| 212 | fxch // x : 1.0 : y
|
|---|
| 213 | sahf
|
|---|
| 214 | ja .L7
|
|---|
| 215 | fsub %st(1) // x-1 : 1.0 : y
|
|---|
| 216 | fyl2xp1 // log2(x) : y
|
|---|
| 217 | jmp .L8
|
|---|
| 218 |
|
|---|
| 219 | .L7: fyl2x // log2(x) : y
|
|---|
| 220 | .L8: fmul %st(1) // y*log2(x) : y
|
|---|
| 221 | fxam
|
|---|
| 222 | fnstsw
|
|---|
| 223 | andb $0x45, %ah
|
|---|
| 224 | cmpb $0x05, %ah // is y*log2(x) == ±inf ?
|
|---|
| 225 | je .L28
|
|---|
| 226 | fst %st(1) // y*log2(x) : y*log2(x)
|
|---|
| 227 | frndint // int(y*log2(x)) : y*log2(x)
|
|---|
| 228 | fsubr %st, %st(1) // int(y*log2(x)) : fract(y*log2(x))
|
|---|
| 229 | fxch // fract(y*log2(x)) : int(y*log2(x))
|
|---|
| 230 | f2xm1 // 2^fract(y*log2(x))-1 : int(y*log2(x))
|
|---|
| 231 | faddl MO(one) // 2^fract(y*log2(x)) : int(y*log2(x))
|
|---|
| 232 | fscale // 2^fract(y*log2(x))*2^int(y*log2(x)) : int(y*log2(x))
|
|---|
| 233 | addl $8, %esp
|
|---|
| 234 | cfi_adjust_cfa_offset (-8)
|
|---|
| 235 | fstp %st(1) // 2^fract(y*log2(x))*2^int(y*log2(x))
|
|---|
| 236 | ret
|
|---|
| 237 |
|
|---|
| 238 | cfi_adjust_cfa_offset (8)
|
|---|
| 239 | .L28: fstp %st(1) // y*log2(x)
|
|---|
| 240 | fldl MO(one) // 1 : y*log2(x)
|
|---|
| 241 | fscale // 2^(y*log2(x)) : y*log2(x)
|
|---|
| 242 | addl $8, %esp
|
|---|
| 243 | cfi_adjust_cfa_offset (-8)
|
|---|
| 244 | fstp %st(1) // 2^(y*log2(x))
|
|---|
| 245 | ret
|
|---|
| 246 |
|
|---|
| 247 | // pow(x,±0) = 1
|
|---|
| 248 | .align ALIGNARG(4)
|
|---|
| 249 | .L11: fstp %st(0) // pop y
|
|---|
| 250 | fldl MO(one)
|
|---|
| 251 | ret
|
|---|
| 252 |
|
|---|
| 253 | // y == ±inf
|
|---|
| 254 | .align ALIGNARG(4)
|
|---|
| 255 | .L12: fstp %st(0) // pop y
|
|---|
| 256 | fldt 4(%esp) // x
|
|---|
| 257 | fabs
|
|---|
| 258 | fcompl MO(one) // < 1, == 1, or > 1
|
|---|
| 259 | fnstsw
|
|---|
| 260 | andb $0x45, %ah
|
|---|
| 261 | cmpb $0x45, %ah
|
|---|
| 262 | je .L13 // jump if x is NaN
|
|---|
| 263 |
|
|---|
| 264 | cmpb $0x40, %ah
|
|---|
| 265 | je .L14 // jump if |x| == 1
|
|---|
| 266 |
|
|---|
| 267 | shlb $1, %ah
|
|---|
| 268 | xorb %ah, %dl
|
|---|
| 269 | andl $2, %edx
|
|---|
| 270 | fldl MOX(inf_zero, %edx, 4)
|
|---|
| 271 | ret
|
|---|
| 272 |
|
|---|
| 273 | .align ALIGNARG(4)
|
|---|
| 274 | .L14: fldl MO(one)
|
|---|
| 275 | ret
|
|---|
| 276 |
|
|---|
| 277 | .align ALIGNARG(4)
|
|---|
| 278 | .L13: fldt 4(%esp) // load x == NaN
|
|---|
| 279 | ret
|
|---|
| 280 |
|
|---|
| 281 | cfi_adjust_cfa_offset (8)
|
|---|
| 282 | .align ALIGNARG(4)
|
|---|
| 283 | // x is ±inf
|
|---|
| 284 | .L15: fstp %st(0) // y
|
|---|
| 285 | testb $2, %dh
|
|---|
| 286 | jz .L16 // jump if x == +inf
|
|---|
| 287 |
|
|---|
| 288 | // We must find out whether y is an odd integer.
|
|---|
| 289 | fld %st // y : y
|
|---|
| 290 | fistpll (%esp) // y
|
|---|
| 291 | fildll (%esp) // int(y) : y
|
|---|
| 292 | fucompp // <empty>
|
|---|
| 293 | fnstsw
|
|---|
| 294 | sahf
|
|---|
| 295 | jne .L17
|
|---|
| 296 |
|
|---|
| 297 | // OK, the value is an integer, but is it odd?
|
|---|
| 298 | popl %eax
|
|---|
| 299 | cfi_adjust_cfa_offset (-4)
|
|---|
| 300 | popl %edx
|
|---|
| 301 | cfi_adjust_cfa_offset (-4)
|
|---|
| 302 | andb $1, %al
|
|---|
| 303 | jz .L18 // jump if not odd
|
|---|
| 304 | // It's an odd integer.
|
|---|
| 305 | shrl $31, %edx
|
|---|
| 306 | fldl MOX(minf_mzero, %edx, 8)
|
|---|
| 307 | ret
|
|---|
| 308 |
|
|---|
| 309 | cfi_adjust_cfa_offset (8)
|
|---|
| 310 | .align ALIGNARG(4)
|
|---|
| 311 | .L16: fcompl MO(zero)
|
|---|
| 312 | addl $8, %esp
|
|---|
| 313 | cfi_adjust_cfa_offset (-8)
|
|---|
| 314 | fnstsw
|
|---|
| 315 | shrl $5, %eax
|
|---|
| 316 | andl $8, %eax
|
|---|
| 317 | fldl MOX(inf_zero, %eax, 1)
|
|---|
| 318 | ret
|
|---|
| 319 |
|
|---|
| 320 | cfi_adjust_cfa_offset (8)
|
|---|
| 321 | .align ALIGNARG(4)
|
|---|
| 322 | .L17: shll $30, %edx // sign bit for y in right position
|
|---|
| 323 | addl $8, %esp
|
|---|
| 324 | cfi_adjust_cfa_offset (-8)
|
|---|
| 325 | .L18: shrl $31, %edx
|
|---|
| 326 | fldl MOX(inf_zero, %edx, 8)
|
|---|
| 327 | ret
|
|---|
| 328 |
|
|---|
| 329 | cfi_adjust_cfa_offset (8)
|
|---|
| 330 | .align ALIGNARG(4)
|
|---|
| 331 | // x is ±0
|
|---|
| 332 | .L20: fstp %st(0) // y
|
|---|
| 333 | testb $2, %dl
|
|---|
| 334 | jz .L21 // y > 0
|
|---|
| 335 |
|
|---|
| 336 | // x is ±0 and y is < 0. We must find out whether y is an odd integer.
|
|---|
| 337 | testb $2, %dh
|
|---|
| 338 | jz .L25
|
|---|
| 339 |
|
|---|
| 340 | fld %st // y : y
|
|---|
| 341 | fistpll (%esp) // y
|
|---|
| 342 | fildll (%esp) // int(y) : y
|
|---|
| 343 | fucompp // <empty>
|
|---|
| 344 | fnstsw
|
|---|
| 345 | sahf
|
|---|
| 346 | jne .L26
|
|---|
| 347 |
|
|---|
| 348 | // OK, the value is an integer, but is it odd?
|
|---|
| 349 | popl %eax
|
|---|
| 350 | cfi_adjust_cfa_offset (-4)
|
|---|
| 351 | popl %edx
|
|---|
| 352 | cfi_adjust_cfa_offset (-4)
|
|---|
| 353 | andb $1, %al
|
|---|
| 354 | jz .L27 // jump if not odd
|
|---|
| 355 | // It's an odd integer.
|
|---|
| 356 | // Raise divide-by-zero exception and get minus infinity value.
|
|---|
| 357 | fldl MO(one)
|
|---|
| 358 | fdivl MO(zero)
|
|---|
| 359 | fchs
|
|---|
| 360 | ret
|
|---|
| 361 |
|
|---|
| 362 | cfi_adjust_cfa_offset (8)
|
|---|
| 363 | .L25: fstp %st(0)
|
|---|
| 364 | .L26: addl $8, %esp
|
|---|
| 365 | cfi_adjust_cfa_offset (-8)
|
|---|
| 366 | .L27: // Raise divide-by-zero exception and get infinity value.
|
|---|
| 367 | fldl MO(one)
|
|---|
| 368 | fdivl MO(zero)
|
|---|
| 369 | ret
|
|---|
| 370 |
|
|---|
| 371 | cfi_adjust_cfa_offset (8)
|
|---|
| 372 | .align ALIGNARG(4)
|
|---|
| 373 | // x is ±0 and y is > 0. We must find out whether y is an odd integer.
|
|---|
| 374 | .L21: testb $2, %dh
|
|---|
| 375 | jz .L22
|
|---|
| 376 |
|
|---|
| 377 | fld %st // y : y
|
|---|
| 378 | fistpll (%esp) // y
|
|---|
| 379 | fildll (%esp) // int(y) : y
|
|---|
| 380 | fucompp // <empty>
|
|---|
| 381 | fnstsw
|
|---|
| 382 | sahf
|
|---|
| 383 | jne .L23
|
|---|
| 384 |
|
|---|
| 385 | // OK, the value is an integer, but is it odd?
|
|---|
| 386 | popl %eax
|
|---|
| 387 | cfi_adjust_cfa_offset (-4)
|
|---|
| 388 | popl %edx
|
|---|
| 389 | cfi_adjust_cfa_offset (-4)
|
|---|
| 390 | andb $1, %al
|
|---|
| 391 | jz .L24 // jump if not odd
|
|---|
| 392 | // It's an odd integer.
|
|---|
| 393 | fldl MO(mzero)
|
|---|
| 394 | ret
|
|---|
| 395 |
|
|---|
| 396 | cfi_adjust_cfa_offset (8)
|
|---|
| 397 | .L22: fstp %st(0)
|
|---|
| 398 | .L23: addl $8, %esp // Don't use 2 x pop
|
|---|
| 399 | cfi_adjust_cfa_offset (-8)
|
|---|
| 400 | .L24: fldl MO(zero)
|
|---|
| 401 | ret
|
|---|
| 402 |
|
|---|
| 403 | END(RT_NOCRT(powl))
|
|---|
| 404 | //END(__ieee754_powl)
|
|---|
