linux/arch/m68k/math-emu/fp_scan.S
<<
>>
Prefs
   1/*
   2 * fp_scan.S
   3 *
   4 * Copyright Roman Zippel, 1997.  All rights reserved.
   5 *
   6 * Redistribution and use in source and binary forms, with or without
   7 * modification, are permitted provided that the following conditions
   8 * are met:
   9 * 1. Redistributions of source code must retain the above copyright
  10 *    notice, and the entire permission notice in its entirety,
  11 *    including the disclaimer of warranties.
  12 * 2. Redistributions in binary form must reproduce the above copyright
  13 *    notice, this list of conditions and the following disclaimer in the
  14 *    documentation and/or other materials provided with the distribution.
  15 * 3. The name of the author may not be used to endorse or promote
  16 *    products derived from this software without specific prior
  17 *    written permission.
  18 *
  19 * ALTERNATIVELY, this product may be distributed under the terms of
  20 * the GNU General Public License, in which case the provisions of the GPL are
  21 * required INSTEAD OF the above restrictions.  (This clause is
  22 * necessary due to a potential bad interaction between the GPL and
  23 * the restrictions contained in a BSD-style copyright.)
  24 *
  25 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  26 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  27 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  28 * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  29 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  30 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  31 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  33 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
  35 * OF THE POSSIBILITY OF SUCH DAMAGE.
  36 */
  37
  38#include "fp_emu.h"
  39#include "fp_decode.h"
  40
  41        .globl  fp_scan, fp_datasize
  42
  43        .data
  44
  45| %d2 - first two instr words
  46| %d1 - operand size
  47
  48/* operand formats are:
  49
  50        Long = 0,               i.e. fmove.l
  51        Single,                 i.e. fmove.s
  52        Extended,               i.e. fmove.x
  53        Packed-BCD,             i.e. fmove.p
  54        Word,                   i.e. fmove.w
  55        Double,                 i.e. fmove.d
  56*/
  57
  58        .text
  59
  60| On entry:
  61| FPDATA - base of emulated FPU registers
  62
  63fp_scan:
  64| normal fpu instruction? (this excludes fsave/frestore)
  65        fp_get_pc %a0
  66        printf  PDECODE,"%08x: ",1,%a0
  67        getuser.b (%a0),%d0,fp_err_ua1,%a0
  68#if 1
  69        cmp.b   #0xf2,%d0               | cpid = 1
  70#else
  71        cmp.b   #0xfc,%d0               | cpid = 6
  72#endif
  73        jne     fp_nonstd
  74| first two instruction words are kept in %d2
  75        getuser.l (%a0)+,%d2,fp_err_ua1,%a0
  76        fp_put_pc %a0
  77fp_decode_cond:                         | separate conditional instr
  78        fp_decode_cond_instr_type
  79
  80        .long   fp_decode_move, fp_fscc
  81        .long   fp_fbccw, fp_fbccl
  82
  83fp_decode_move:                         | separate move instr
  84        fp_decode_move_instr_type
  85
  86        .long   fp_fgen_fp, fp_ill
  87        .long   fp_fgen_ea, fp_fmove_fp2mem
  88        .long   fp_fmovem_cr, fp_fmovem_cr
  89        .long   fp_fmovem_fp, fp_fmovem_fp
  90
  91| now all arithmetic instr and a few move instr are left
  92fp_fgen_fp:                             | source is a fpu register
  93        clr.b   (FPD_FPSR+2,FPDATA)     | clear the exception byte
  94        fp_decode_sourcespec
  95        printf  PDECODE,"f<op>.x fp%d",1,%d0
  96        fp_get_fp_reg
  97        lea     (FPD_TEMPFP1,FPDATA),%a1 | copy src into a temp location
  98        move.l  (%a0)+,(%a1)+
  99        move.l  (%a0)+,(%a1)+
 100        move.l  (%a0),(%a1)
 101        lea     (-8,%a1),%a0
 102        jra     fp_getdest
 103
 104fp_fgen_ea:                             | source is <ea>
 105        clr.b   (FPD_FPSR+2,FPDATA)     | clear the exception byte
 106        | sort out fmovecr, keep data size in %d1
 107        fp_decode_sourcespec
 108        cmp.w   #7,%d0
 109        jeq     fp_fmovecr
 110        move.w  %d0,%d1                 | store data size twice in %d1
 111        swap    %d1                     | one can be trashed below
 112        move.w  %d0,%d1
 113#ifdef FPU_EMU_DEBUG
 114        lea     0f,%a0
 115        clr.l   %d0
 116        move.b  (%a0,%d1.w),%d0
 117        printf  PDECODE,"f<op>.%c ",1,%d0
 118
 119        .data
 1200:      .byte   'l','s','x','p','w','d','b',0
 121        .previous
 122#endif
 123
 124/*
 125        fp_getsource, fp_getdest
 126
 127        basically, we end up with a pointer to the source operand in
 128        %a1, and a pointer to the destination operand in %a0.  both
 129        are, of course, 96-bit extended floating point numbers.
 130*/
 131
 132fp_getsource:
 133        | decode addressing mode for source
 134        fp_decode_addr_mode
 135
 136        .long   fp_data, fp_ill
 137        .long   fp_indirect, fp_postinc
 138        .long   fp_predecr, fp_disp16
 139        .long   fp_extmode0, fp_extmode1
 140
 141        | addressing mode: data register direct
 142fp_data:
 143        fp_mode_data_direct
 144        jsr     fp_get_data_reg
 145        lea     (FPD_TEMPFP1,FPDATA),%a0
 146        jmp     ([0f:w,%pc,%d1.w*4])
 147
 148        .align  4
 1490:
 150        .long   fp_data_long, fp_data_single
 151        .long   fp_ill, fp_ill
 152        .long   fp_data_word, fp_ill
 153        .long   fp_data_byte, fp_ill
 154
 155        | data types that fit in an integer data register
 156fp_data_byte:
 157        extb.l  %d0
 158        jra     fp_data_long
 159
 160fp_data_word:
 161        ext.l   %d0
 162
 163fp_data_long:
 164        jsr     fp_conv_long2ext
 165        jra     fp_getdest
 166
 167fp_data_single:
 168        jsr     fp_conv_single2ext
 169        jra     fp_getdest
 170
 171        | addressing mode: address register indirect
 172fp_indirect:
 173        fp_mode_addr_indirect
 174        jra     fp_fetchsource
 175
 176        | addressing mode: address register indirect with postincrement
 177fp_postinc:
 178        fp_mode_addr_indirect_postinc
 179        jra     fp_fetchsource
 180
 181        | addressing mode: address register indirect with predecrement
 182fp_predecr:
 183        fp_mode_addr_indirect_predec
 184        jra     fp_fetchsource
 185
 186        | addressing mode: address register/programm counter indirect
 187        |                  with 16bit displacement
 188fp_disp16:
 189        fp_mode_addr_indirect_disp16
 190        jra     fp_fetchsource
 191
 192        | all other indirect addressing modes will finally end up here
 193fp_extmode0:
 194        fp_mode_addr_indirect_extmode0
 195        jra     fp_fetchsource
 196
 197| all pc relative addressing modes and immediate/absolute modes end up here
 198| the first ones are sent to fp_extmode0 or fp_disp16
 199| and only the latter are handled here
 200fp_extmode1:
 201        fp_decode_addr_reg
 202        jmp     ([0f:w,%pc,%d0*4])
 203
 204        .align  4
 2050:
 206        .long   fp_abs_short, fp_abs_long
 207        .long   fp_disp16, fp_extmode0
 208        .long   fp_immediate, fp_ill
 209        .long   fp_ill, fp_ill
 210
 211        | addressing mode: absolute short
 212fp_abs_short:
 213        fp_mode_abs_short
 214        jra     fp_fetchsource
 215
 216        | addressing mode: absolute long
 217fp_abs_long:
 218        fp_mode_abs_long
 219        jra     fp_fetchsource
 220
 221        | addressing mode: immediate data
 222fp_immediate:
 223        printf  PDECODE,"#"
 224        fp_get_pc %a0
 225        move.w  (fp_datasize,%d1.w*2),%d0
 226        addq.w  #1,%d0
 227        and.w   #-2,%d0
 228#ifdef FPU_EMU_DEBUG
 229        movem.l %d0/%d1,-(%sp)
 230        movel   %a0,%a1
 231        clr.l   %d1
 232        jra     2f
 2331:      getuser.b (%a1)+,%d1,fp_err_ua1,%a1
 234        printf  PDECODE,"%02x",1,%d1
 2352:      dbra    %d0,1b
 236        movem.l (%sp)+,%d0/%d1
 237#endif
 238        lea     (%a0,%d0.w),%a1
 239        fp_put_pc %a1
 240|       jra     fp_fetchsource
 241
 242fp_fetchsource:
 243        move.l  %a0,%a1
 244        swap    %d1
 245        lea     (FPD_TEMPFP1,FPDATA),%a0
 246        jmp     ([0f:w,%pc,%d1.w*4])
 247
 248        .align  4
 2490:      .long   fp_long, fp_single
 250        .long   fp_ext, fp_pack
 251        .long   fp_word, fp_double
 252        .long   fp_byte, fp_ill
 253
 254fp_long:
 255        getuser.l (%a1),%d0,fp_err_ua1,%a1
 256        jsr     fp_conv_long2ext
 257        jra     fp_getdest
 258
 259fp_single:
 260        getuser.l (%a1),%d0,fp_err_ua1,%a1
 261        jsr     fp_conv_single2ext
 262        jra     fp_getdest
 263
 264fp_ext:
 265        getuser.l (%a1)+,%d0,fp_err_ua1,%a1
 266        lsr.l   #8,%d0
 267        lsr.l   #7,%d0
 268        lsr.w   #1,%d0
 269        move.l  %d0,(%a0)+
 270        getuser.l (%a1)+,%d0,fp_err_ua1,%a1
 271        move.l  %d0,(%a0)+
 272        getuser.l (%a1),%d0,fp_err_ua1,%a1
 273        move.l  %d0,(%a0)
 274        subq.l  #8,%a0
 275        jra     fp_getdest
 276
 277fp_pack:
 278        /* not supported yet */
 279        jra     fp_ill
 280
 281fp_word:
 282        getuser.w (%a1),%d0,fp_err_ua1,%a1
 283        ext.l   %d0
 284        jsr     fp_conv_long2ext
 285        jra     fp_getdest
 286
 287fp_double:
 288        jsr     fp_conv_double2ext
 289        jra     fp_getdest
 290
 291fp_byte:
 292        getuser.b (%a1),%d0,fp_err_ua1,%a1
 293        extb.l  %d0
 294        jsr     fp_conv_long2ext
 295|       jra     fp_getdest
 296
 297fp_getdest:
 298        move.l  %a0,%a1
 299        bfextu  %d2{#22,#3},%d0
 300        printf  PDECODE,",fp%d\n",1,%d0
 301        fp_get_fp_reg
 302        movem.l %a0/%a1,-(%sp)
 303        pea     fp_finalrounding
 304        bfextu  %d2{#25,#7},%d0
 305        jmp     ([0f:w,%pc,%d0*4])
 306
 307        .align  4
 3080:
 309        .long   fp_fmove_mem2fp, fp_fint, fp_fsinh, fp_fintrz
 310        .long   fp_fsqrt, fp_ill, fp_flognp1, fp_ill
 311        .long   fp_fetoxm1, fp_ftanh, fp_fatan, fp_ill
 312        .long   fp_fasin, fp_fatanh, fp_fsin, fp_ftan
 313        .long   fp_fetox, fp_ftwotox, fp_ftentox, fp_ill
 314        .long   fp_flogn, fp_flog10, fp_flog2, fp_ill
 315        .long   fp_fabs, fp_fcosh, fp_fneg, fp_ill
 316        .long   fp_facos, fp_fcos, fp_fgetexp, fp_fgetman
 317        .long   fp_fdiv, fp_fmod, fp_fadd, fp_fmul
 318        .long   fpa_fsgldiv, fp_frem, fp_fscale, fpa_fsglmul
 319        .long   fp_fsub, fp_ill, fp_ill, fp_ill
 320        .long   fp_ill, fp_ill, fp_ill, fp_ill
 321        .long   fp_fsincos0, fp_fsincos1, fp_fsincos2, fp_fsincos3
 322        .long   fp_fsincos4, fp_fsincos5, fp_fsincos6, fp_fsincos7
 323        .long   fp_fcmp, fp_ill, fp_ftst, fp_ill
 324        .long   fp_ill, fp_ill, fp_ill, fp_ill
 325        .long   fp_fsmove, fp_fssqrt, fp_ill, fp_ill
 326        .long   fp_fdmove, fp_fdsqrt, fp_ill, fp_ill
 327        .long   fp_ill, fp_ill, fp_ill, fp_ill
 328        .long   fp_ill, fp_ill, fp_ill, fp_ill
 329        .long   fp_ill, fp_ill, fp_ill, fp_ill
 330        .long   fp_ill, fp_ill, fp_ill, fp_ill
 331        .long   fp_fsabs, fp_ill, fp_fsneg, fp_ill
 332        .long   fp_fdabs, fp_ill, fp_fdneg, fp_ill
 333        .long   fp_fsdiv, fp_ill, fp_fsadd, fp_fsmul
 334        .long   fp_fddiv, fp_ill, fp_fdadd, fp_fdmul
 335        .long   fp_fssub, fp_ill, fp_ill, fp_ill
 336        .long   fp_fdsub, fp_ill, fp_ill, fp_ill
 337        .long   fp_ill, fp_ill, fp_ill, fp_ill
 338        .long   fp_ill, fp_ill, fp_ill, fp_ill
 339        .long   fp_ill, fp_ill, fp_ill, fp_ill
 340        .long   fp_ill, fp_ill, fp_ill, fp_ill
 341
 342        | Instructions follow
 343
 344        | Move an (emulated) ROM constant
 345fp_fmovecr:
 346        bfextu  %d2{#27,#5},%d0
 347        printf  PINSTR,"fp_fmovecr #%d",1,%d0
 348        move.l  %d0,%d1
 349        add.l   %d0,%d0
 350        add.l   %d1,%d0
 351        lea     (fp_constants,%d0*4),%a0
 352        move.l  #0x801cc0ff,%d0
 353        addq.l  #1,%d1
 354        lsl.l   %d1,%d0
 355        jcc     1f
 356        fp_set_sr FPSR_EXC_INEX2                        | INEX2 exception
 3571:      moveq   #-128,%d0                               | continue with fmove
 358        and.l   %d0,%d2
 359        jra     fp_getdest
 360
 361        .data
 362        .align  4
 363fp_constants:
 364        .long   0x00004000,0xc90fdaa2,0x2168c235        | pi
 365        .extend 0,0,0,0,0,0,0,0,0,0
 366        .long   0x00003ffd,0x9a209a84,0xfbcff798        | log10(2)
 367        .long   0x00004000,0xadf85458,0xa2bb4a9a        | e
 368        .long   0x00003fff,0xb8aa3b29,0x5c17f0bc        | log2(e)
 369        .long   0x00003ffd,0xde5bd8a9,0x37287195        | log10(e)
 370        .long   0x00000000,0x00000000,0x00000000        | 0.0
 371        .long   0x00003ffe,0xb17217f7,0xd1cf79ac        | 1n(2)
 372        .long   0x00004000,0x935d8ddd,0xaaa8ac17        | 1n(10)
 373        | read this as "1.0 * 2^0" - note the high bit in the mantissa
 374        .long   0x00003fff,0x80000000,0x00000000        | 10^0
 375        .long   0x00004002,0xa0000000,0x00000000        | 10^1
 376        .long   0x00004005,0xc8000000,0x00000000        | 10^2
 377        .long   0x0000400c,0x9c400000,0x00000000        | 10^4
 378        .long   0x00004019,0xbebc2000,0x00000000        | 10^8
 379        .long   0x00004034,0x8e1bc9bf,0x04000000        | 10^16
 380        .long   0x00004069,0x9dc5ada8,0x2b70b59e        | 10^32
 381        .long   0x000040d3,0xc2781f49,0xffcfa6d5        | 10^64
 382        .long   0x000041a8,0x93ba47c9,0x80e98ce0        | 10^128
 383        .long   0x00004351,0xaa7eebfb,0x9df9de8e        | 10^256
 384        .long   0x000046a3,0xe319a0ae,0xa60e91c7        | 10^512
 385        .long   0x00004d48,0xc9767586,0x81750c17        | 10^1024
 386        .long   0x00005a92,0x9e8b3b5d,0xc53d5de5        | 10^2048
 387        .long   0x00007525,0xc4605202,0x8a20979b        | 10^4096
 388        .previous
 389
 390fp_fmove_mem2fp:
 391        printf  PINSTR,"fmove %p,%p\n",2,%a0,%a1
 392        move.l  (%a1)+,(%a0)+
 393        move.l  (%a1)+,(%a0)+
 394        move.l  (%a1),(%a0)
 395        subq.l  #8,%a0
 396        rts
 397
 398fpa_fsglmul:
 399        move.l  #fp_finalrounding_single_fast,(%sp)
 400        jra     fp_fsglmul
 401
 402fpa_fsgldiv:
 403        move.l  #fp_finalrounding_single_fast,(%sp)
 404        jra     fp_fsgldiv
 405
 406.macro  fp_dosingleprec instr
 407        printf  PINSTR,"single "
 408        move.l  #fp_finalrounding_single,(%sp)
 409        jra     \instr
 410.endm
 411
 412.macro  fp_dodoubleprec instr
 413        printf  PINSTR,"double "
 414        move.l  #fp_finalrounding_double,(%sp)
 415        jra     \instr
 416.endm
 417
 418fp_fsmove:
 419        fp_dosingleprec fp_fmove_mem2fp
 420
 421fp_fssqrt:
 422        fp_dosingleprec fp_fsqrt
 423
 424fp_fdmove:
 425        fp_dodoubleprec fp_fmove_mem2fp
 426
 427fp_fdsqrt:
 428        fp_dodoubleprec fp_fsqrt
 429
 430fp_fsabs:
 431        fp_dosingleprec fp_fabs
 432
 433fp_fsneg:
 434        fp_dosingleprec fp_fneg
 435
 436fp_fdabs:
 437        fp_dodoubleprec fp_fabs
 438
 439fp_fdneg:
 440        fp_dodoubleprec fp_fneg
 441
 442fp_fsdiv:
 443        fp_dosingleprec fp_fdiv
 444
 445fp_fsadd:
 446        fp_dosingleprec fp_fadd
 447
 448fp_fsmul:
 449        fp_dosingleprec fp_fmul
 450
 451fp_fddiv:
 452        fp_dodoubleprec fp_fdiv
 453
 454fp_fdadd:
 455        fp_dodoubleprec fp_fadd
 456
 457fp_fdmul:
 458        fp_dodoubleprec fp_fmul
 459
 460fp_fssub:
 461        fp_dosingleprec fp_fsub
 462
 463fp_fdsub:
 464        fp_dodoubleprec fp_fsub
 465
 466fp_nonstd:
 467        fp_get_pc %a0
 468        getuser.l (%a0),%d0,fp_err_ua1,%a0
 469        printf  ,"nonstd ((%08x)=%08x)\n",2,%a0,%d0
 470        moveq   #-1,%d0
 471        rts
 472
 473        .data
 474        .align  4
 475
 476        | data sizes corresponding to the operand formats
 477fp_datasize:
 478        .word   4, 4, 12, 12, 2, 8, 1, 0
 479