~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/arch/m68k/ifpsp060/src/ilsp.S

Version: ~ [ linux-6.12-rc7 ] ~ [ linux-6.11.7 ] ~ [ linux-6.10.14 ] ~ [ linux-6.9.12 ] ~ [ linux-6.8.12 ] ~ [ linux-6.7.12 ] ~ [ linux-6.6.60 ] ~ [ linux-6.5.13 ] ~ [ linux-6.4.16 ] ~ [ linux-6.3.13 ] ~ [ linux-6.2.16 ] ~ [ linux-6.1.116 ] ~ [ linux-6.0.19 ] ~ [ linux-5.19.17 ] ~ [ linux-5.18.19 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.171 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.229 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.285 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.323 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.336 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.337 ] ~ [ linux-4.4.302 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.12 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

Diff markup

Differences between /arch/m68k/ifpsp060/src/ilsp.S (Architecture m68k) and /arch/i386/ifpsp060/src/ilsp.S (Architecture i386)


  1 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~    
  2 MOTOROLA MICROPROCESSOR & MEMORY TECHNOLOGY GR    
  3 M68000 Hi-Performance Microprocessor Division     
  4 M68060 Software Package                           
  5 Production Release P1.00 -- October 10, 1994      
  6                                                   
  7 M68060 Software Package Copyright © 1993, 199    
  8                                                   
  9 THE SOFTWARE is provided on an "AS IS" basis a    
 10 To the maximum extent permitted by applicable     
 11 MOTOROLA DISCLAIMS ALL WARRANTIES WHETHER EXPR    
 12 INCLUDING IMPLIED WARRANTIES OF MERCHANTABILIT    
 13 and any warranty against infringement with reg    
 14 (INCLUDING ANY MODIFIED VERSIONS THEREOF) and     
 15                                                   
 16 To the maximum extent permitted by applicable     
 17 IN NO EVENT SHALL MOTOROLA BE LIABLE FOR ANY D    
 18 (INCLUDING WITHOUT LIMITATION, DAMAGES FOR LOS    
 19 BUSINESS INTERRUPTION, LOSS OF BUSINESS INFORM    
 20 ARISING OF THE USE OR INABILITY TO USE THE SOF    
 21 Motorola assumes no responsibility for the mai    
 22                                                   
 23 You are hereby granted a copyright license to     
 24 so long as this entire notice is retained with    
 25 redistributed versions, and that such modified    
 26 No licenses are granted by implication, estopp    
 27 or trademarks of Motorola, Inc.                   
 28 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~    
 29 # litop.s:                                        
 30 #       This file is appended to the top of th    
 31 # and contains the entry points into the packa    
 32 # effect, branches to one of the branch table     
 33 #                                                 
 34                                                   
 35         bra.l   _060LSP__idivs64_                 
 36         short   0x0000                            
 37         bra.l   _060LSP__idivu64_                 
 38         short   0x0000                            
 39                                                   
 40         bra.l   _060LSP__imuls64_                 
 41         short   0x0000                            
 42         bra.l   _060LSP__imulu64_                 
 43         short   0x0000                            
 44                                                   
 45         bra.l   _060LSP__cmp2_Ab_                 
 46         short   0x0000                            
 47         bra.l   _060LSP__cmp2_Aw_                 
 48         short   0x0000                            
 49         bra.l   _060LSP__cmp2_Al_                 
 50         short   0x0000                            
 51         bra.l   _060LSP__cmp2_Db_                 
 52         short   0x0000                            
 53         bra.l   _060LSP__cmp2_Dw_                 
 54         short   0x0000                            
 55         bra.l   _060LSP__cmp2_Dl_                 
 56         short   0x0000                            
 57                                                   
 58 # leave room for future possible aditions.        
 59         align   0x200                             
 60                                                   
 61 ##############################################    
 62 # XDEF ***************************************    
 63 #       _060LSP__idivu64_(): Emulate 64-bit un    
 64 #       _060LSP__idivs64_(): Emulate 64-bit si    
 65 #                                                 
 66 #       This is the library version which is a    
 67 #       and therefore does not work exactly li    
 68 #       64-bit divide instruction.                
 69 #                                                 
 70 # XREF ***************************************    
 71 #       None.                                     
 72 #                                                 
 73 # INPUT **************************************    
 74 #       0x4(sp)  = divisor                        
 75 #       0x8(sp)  = hi(dividend)                   
 76 #       0xc(sp)  = lo(dividend)                   
 77 #       0x10(sp) = pointer to location to plac    
 78 #                                                 
 79 # OUTPUT *************************************    
 80 #       0x10(sp) = points to location of remai    
 81 #                  remainder is in first longw    
 82 #                                                 
 83 # ALGORITHM **********************************    
 84 #       If the operands are signed, make them     
 85 # sign info for later. Separate out special ca    
 86 # or 32-bit divides if possible. Else, use a s    
 87 # to calculate the result.                        
 88 #       Restore sign info if signed instructio    
 89 # codes before performing the final "rts". If     
 90 # zero, then perform a divide-by-zero using a     
 91 # divide instruction. This way, the operating     
 92 # the event occurred even though it may not po    
 93 #                                                 
 94 ##############################################    
 95                                                   
 96 set     POSNEG,         -1                        
 97 set     NDIVISOR,       -2                        
 98 set     NDIVIDEND,      -3                        
 99 set     DDSECOND,       -4                        
100 set     DDNORMAL,       -8                        
101 set     DDQUOTIENT,     -12                       
102 set     DIV64_CC,       -16                       
103                                                   
104 ##########                                        
105 # divs.l #                                        
106 ##########                                        
107         global          _060LSP__idivs64_         
108 _060LSP__idivs64_:                                
109 # PROLOGUE BEGIN #############################    
110         link.w          %a6,&-16                  
111         movm.l          &0x3f00,-(%sp)            
112 #       fmovm.l         &0x0,-(%sp)               
113 # PROLOGUE END ###############################    
114                                                   
115         mov.w           %cc,DIV64_CC(%a6)         
116         st              POSNEG(%a6)               
117         bra.b           ldiv64_cont               
118                                                   
119 ##########                                        
120 # divu.l #                                        
121 ##########                                        
122         global          _060LSP__idivu64_         
123 _060LSP__idivu64_:                                
124 # PROLOGUE BEGIN #############################    
125         link.w          %a6,&-16                  
126         movm.l          &0x3f00,-(%sp)            
127 #       fmovm.l         &0x0,-(%sp)               
128 # PROLOGUE END ###############################    
129                                                   
130         mov.w           %cc,DIV64_CC(%a6)         
131         sf              POSNEG(%a6)               
132                                                   
133 ldiv64_cont:                                      
134         mov.l           0x8(%a6),%d7              
135                                                   
136         beq.w           ldiv64eq0                 
137                                                   
138         mov.l           0xc(%a6), %d5             
139         mov.l           0x10(%a6), %d6            
140                                                   
141 # separate signed and unsigned divide             
142         tst.b           POSNEG(%a6)               
143         beq.b           ldspecialcases            
144                                                   
145 # save the sign of the divisor                    
146 # make divisor unsigned if it's negative          
147         tst.l           %d7                       
148         slt             NDIVISOR(%a6)             
149         bpl.b           ldsgndividend             
150         neg.l           %d7                       
151                                                   
152 # save the sign of the dividend                   
153 # make dividend unsigned if it's negative         
154 ldsgndividend:                                    
155         tst.l           %d5                       
156         slt             NDIVIDEND(%a6)            
157         bpl.b           ldspecialcases            
158                                                   
159         mov.w           &0x0, %cc                 
160         negx.l          %d6                       
161         negx.l          %d5                       
162                                                   
163 # extract some special cases:                     
164 #       - is (dividend == 0) ?                    
165 #       - is (hi(dividend) == 0 && (divisor <=    
166 ldspecialcases:                                   
167         tst.l           %d5                       
168         bne.b           ldnormaldivide            
169                                                   
170         tst.l           %d6                       
171         beq.w           lddone                    
172                                                   
173         cmp.l           %d7,%d6                   
174         bls.b           ld32bitdivide             
175                                                   
176         exg             %d5,%d6                   
177         bra.w           ldivfinish                
178                                                   
179 ld32bitdivide:                                    
180         tdivu.l         %d7, %d5:%d6              
181                                                   
182         bra.b           ldivfinish                
183                                                   
184 ldnormaldivide:                                   
185 # last special case:                              
186 #       - is hi(dividend) >= divisor ? if yes,    
187         cmp.l           %d7,%d5                   
188         bls.b           lddovf                    
189                                                   
190 # perform the divide algorithm:                   
191         bsr.l           ldclassical               
192                                                   
193 # separate into signed and unsigned finishes.     
194 ldivfinish:                                       
195         tst.b           POSNEG(%a6)               
196         beq.b           lddone                    
197                                                   
198 # it was a divs.l, so ccode setting is a littl    
199         tst.b           NDIVIDEND(%a6)            
200         beq.b           ldcc                      
201         neg.l           %d5                       
202 ldcc:                                             
203         mov.b           NDIVISOR(%a6), %d0        
204         eor.b           %d0, NDIVIDEND(%a6)       
205         beq.b           ldqpos                    
206                                                   
207 # 0x80000000 is the largest number representab    
208 # number. the negative of 0x80000000 is 0x8000    
209         cmpi.l          %d6, &0x80000000          
210         bhi.b           lddovf                    
211                                                   
212         neg.l           %d6                       
213                                                   
214         bra.b           lddone                    
215                                                   
216 ldqpos:                                           
217         btst            &0x1f, %d6                
218         bne.b           lddovf                    
219                                                   
220 lddone:                                           
221 # if the register numbers are the same, only t    
222 # so, if we always save the quotient second, w    
223         andi.w          &0x10,DIV64_CC(%a6)       
224         mov.w           DIV64_CC(%a6),%cc         
225         tst.l           %d6                       
226                                                   
227 # here, the result is in d1 and d0. the curren    
228 # the values at the location pointed to by a0.    
229 # use movm here to not disturb the condition c    
230 ldexit:                                           
231         movm.l          &0x0060,([0x14,%a6])      
232                                                   
233 # EPILOGUE BEGIN #############################    
234 #       fmovm.l         (%sp)+,&0x0               
235         movm.l          (%sp)+,&0x00fc            
236         unlk            %a6                       
237 # EPILOGUE END ###############################    
238                                                   
239         rts                                       
240                                                   
241 # the result should be the unchanged dividend     
242 lddovf:                                           
243         mov.l           0xc(%a6), %d5             
244         mov.l           0x10(%a6), %d6            
245                                                   
246         andi.w          &0x1c,DIV64_CC(%a6)       
247         ori.w           &0x02,DIV64_CC(%a6)       
248         mov.w           DIV64_CC(%a6),%cc         
249                                                   
250         bra.b           ldexit                    
251                                                   
252 ldiv64eq0:                                        
253         mov.l           0xc(%a6),([0x14,%a6])     
254         mov.l           0x10(%a6),([0x14,%a6],    
255                                                   
256         mov.w           DIV64_CC(%a6),%cc         
257                                                   
258 # EPILOGUE BEGIN #############################    
259 #       fmovm.l         (%sp)+,&0x0               
260         movm.l          (%sp)+,&0x00fc            
261         unlk            %a6                       
262 # EPILOGUE END ###############################    
263                                                   
264         divu.w          &0x0,%d0                  
265         rts                                       
266                                                   
267 ##############################################    
268 ##############################################    
269 # This routine uses the 'classical' Algorithm     
270 # Art of Computer Programming, vol II, Seminum    
271 # For this implementation b=2**16, and the tar    
272 # where U,V are words of the quadword dividend    
273 # and U1, V1 are the most significant words.      
274 #                                                 
275 # The most sig. longword of the 64 bit dividen    
276 # in %d6. The divisor must be in the variable     
277 # signed/unsigned flag ddusign must be set (0=    
278 # The quotient is returned in %d6, remainder i    
279 # v (overflow) bit is set in the saved %ccr. I    
280 # is unchanged.                                   
281 ##############################################    
282 ldclassical:                                      
283 # if the divisor msw is 0, use simpler algorit    
284 # one at ddknuth:                                 
285                                                   
286         cmpi.l          %d7, &0xffff              
287         bhi.b           lddknuth                  
288                                                   
289 # Since the divisor is only a word (and larger    
290 # a simpler algorithm may be used :               
291 # In the general case, four quotient words wou    
292 # dividing the divisor word into each dividend    
293 # the first two quotient words must be zero, o    
294 # Since we already checked this case above, we    
295 # longword of the dividend as (0) remainder (s    
296 # the last two divisions to get a quotient lon    
297                                                   
298         clr.l           %d1                       
299         swap            %d5                       
300         swap            %d6                       
301         mov.w           %d6, %d5                  
302                                                   
303         divu.w          %d7, %d5                  
304                                                   
305         mov.w           %d5, %d1                  
306         swap            %d6                       
307         mov.w           %d6, %d5                  
308                                                   
309         divu.w          %d7, %d5                  
310                                                   
311         swap            %d1                       
312         mov.w           %d5, %d1                  
313         clr.w           %d5                       
314         swap            %d5                       
315         mov.l           %d1, %d6                  
316                                                   
317         rts                                       
318                                                   
319 lddknuth:                                         
320 # In this algorithm, the divisor is treated as    
321 # which is divided into a 3 digit (word) divid    
322 # digit (word). After subtraction, the dividen    
323 # process repeated. Before beginning, the divi    
324 # 'normalized' so that the process of estimati    
325 # will yield verifiably correct results..         
326                                                   
327         clr.l           DDNORMAL(%a6)             
328         clr.b           DDSECOND(%a6)             
329         clr.l           %d1                       
330 lddnchk:                                          
331         btst            &31, %d7                  
332         bne.b           lddnormalized             
333         addq.l          &0x1, DDNORMAL(%a6)       
334         lsl.l           &0x1, %d7                 
335         lsl.l           &0x1, %d6                 
336         roxl.l          &0x1, %d5                 
337         bra.w           lddnchk                   
338 lddnormalized:                                    
339                                                   
340 # Now calculate an estimate of the quotient wo    
341 # The comments use subscripts for the first qu    
342         mov.l           %d7, %d3                  
343         mov.l           %d5, %d2                  
344         swap            %d2                       
345         swap            %d3                       
346         cmp.w           %d2, %d3                  
347         bne.b           lddqcalc1                 
348         mov.w           &0xffff, %d1              
349         bra.b           lddadj0                   
350 lddqcalc1:                                        
351         mov.l           %d5, %d1                  
352                                                   
353         divu.w          %d3, %d1                  
354                                                   
355         andi.l          &0x0000ffff, %d1          
356 lddadj0:                                          
357                                                   
358 # now test the trial quotient and adjust. This    
359 # normalization assures (according to Knuth) t    
360 # quotient will be at worst 1 too large.          
361         mov.l           %d6, -(%sp)               
362         clr.w           %d6                       
363         swap            %d6                       
364 lddadj1: mov.l          %d7, %d3                  
365         mov.l           %d1, %d2                  
366         mulu.w          %d7, %d2                  
367         swap            %d3                       
368         mulu.w          %d1, %d3                  
369         mov.l           %d5, %d4                  
370         sub.l           %d3, %d4                  
371                                                   
372         swap            %d4                       
373                                                   
374         mov.w           %d4,%d0                   
375         mov.w           %d6,%d4                   
376                                                   
377         tst.w           %d0                       
378         bne.w           lddadjd1                  
379                                                   
380 #       add.l           %d6, %d4                  
381                                                   
382         cmp.l           %d2, %d4                  
383         bls.b           lddadjd1                  
384         subq.l          &0x1, %d1                 
385         bra.b           lddadj1                   
386 lddadjd1:                                         
387 # now test the word by multiplying it by the d    
388 # the 3 digit (word) result with the current d    
389         mov.l           %d5, -(%sp)               
390         mov.l           %d1, %d6                  
391         swap            %d6                       
392         mov.l           %d7, %d5                  
393         bsr.l           ldmm2                     
394         mov.l           %d5, %d2                  
395         mov.l           %d6, %d3                  
396         mov.l           (%sp)+, %d5               
397         mov.l           (%sp)+, %d6               
398         sub.l           %d3, %d6                  
399         subx.l          %d2, %d5                  
400         bcc             ldd2nd                    
401         subq.l          &0x1, %d1                 
402 # need to add back divisor longword to current    
403 # - according to Knuth, this is done only 2 ou    
404 # divisor, dividend selection.                    
405         clr.l           %d2                       
406         mov.l           %d7, %d3                  
407         swap            %d3                       
408         clr.w           %d3                       
409         add.l           %d3, %d6                  
410         addx.l          %d2, %d5                  
411         mov.l           %d7, %d3                  
412         clr.w           %d3                       
413         swap            %d3                       
414         add.l           %d3, %d5                  
415 ldd2nd:                                           
416         tst.b           DDSECOND(%a6)   # both    
417         bne.b           lddremain                 
418 # first quotient digit now correct. store digi    
419 # (subtracted) dividend                           
420         mov.w           %d1, DDQUOTIENT(%a6)      
421         clr.l           %d1                       
422         swap            %d5                       
423         swap            %d6                       
424         mov.w           %d6, %d5                  
425         clr.w           %d6                       
426         st              DDSECOND(%a6)             
427         bra.w           lddnormalized             
428 lddremain:                                        
429 # add 2nd word to quotient, get the remainder.    
430         mov.w           %d1, DDQUOTIENT+2(%a6)    
431 # shift down one word/digit to renormalize rem    
432         mov.w           %d5, %d6                  
433         swap            %d6                       
434         swap            %d5                       
435         mov.l           DDNORMAL(%a6), %d7        
436         beq.b           lddrn                     
437         subq.l          &0x1, %d7                 
438 lddnlp:                                           
439         lsr.l           &0x1, %d5                 
440         roxr.l          &0x1, %d6                 
441         dbf             %d7, lddnlp               
442 lddrn:                                            
443         mov.l           %d6, %d5                  
444         mov.l           DDQUOTIENT(%a6), %d6      
445                                                   
446         rts                                       
447 ldmm2:                                            
448 # factors for the 32X32->64 multiplication are    
449 # returns 64 bit result in %d5 (hi) %d6(lo).      
450 # destroys %d2,%d3,%d4.                           
451                                                   
452 # multiply hi,lo words of each factor to get 4    
453         mov.l           %d6, %d2                  
454         mov.l           %d6, %d3                  
455         mov.l           %d5, %d4                  
456         swap            %d3                       
457         swap            %d4                       
458         mulu.w          %d5, %d6                  
459         mulu.w          %d3, %d5                  
460         mulu.w          %d4, %d2                  
461         mulu.w          %d4, %d3                  
462 # now use swap and addx to consolidate to two     
463         clr.l           %d4                       
464         swap            %d6                       
465         add.w           %d5, %d6                  
466         addx.w          %d4, %d3                  
467         add.w           %d2, %d6                  
468         addx.w          %d4, %d3                  
469         swap            %d6                       
470         clr.w           %d5                       
471         clr.w           %d2                       
472         swap            %d5                       
473         swap            %d2                       
474         add.l           %d2, %d5                  
475         add.l           %d3, %d5        # %d5     
476         rts                                       
477                                                   
478 ##############################################    
479 # XDEF ***************************************    
480 #       _060LSP__imulu64_(): Emulate 64-bit un    
481 #       _060LSP__imuls64_(): Emulate 64-bit si    
482 #                                                 
483 #       This is the library version which is a    
484 #       and therefore does not work exactly li    
485 #       64-bit multiply instruction.              
486 #                                                 
487 # XREF ***************************************    
488 #       None                                      
489 #                                                 
490 # INPUT **************************************    
491 #       0x4(sp) = multiplier                      
492 #       0x8(sp) = multiplicand                    
493 #       0xc(sp) = pointer to location to place    
494 #                                                 
495 # OUTPUT *************************************    
496 #       0xc(sp) = points to location of 64-bit    
497 #                                                 
498 # ALGORITHM **********************************    
499 #       Perform the multiply in pieces using 1    
500 # multiplies and "add" instructions.              
501 #       Set the condition codes as appropriate    
502 # "rts".                                          
503 #                                                 
504 ##############################################    
505                                                   
506 set MUL64_CC, -4                                  
507                                                   
508         global          _060LSP__imulu64_         
509 _060LSP__imulu64_:                                
510                                                   
511 # PROLOGUE BEGIN #############################    
512         link.w          %a6,&-4                   
513         movm.l          &0x3800,-(%sp)            
514 #       fmovm.l         &0x0,-(%sp)               
515 # PROLOGUE END ###############################    
516                                                   
517         mov.w           %cc,MUL64_CC(%a6)         
518                                                   
519         mov.l           0x8(%a6),%d0              
520         beq.w           mulu64_zero               
521                                                   
522         mov.l           0xc(%a6),%d1              
523         beq.w           mulu64_zero               
524                                                   
525 ##############################################    
526 #       63                         32             
527 #       ----------------------------              
528 #       | hi(mplier) * hi(mplicand)|              
529 #       ----------------------------              
530 #                    -------------------------    
531 #                    | hi(mplier) * lo(mplican    
532 #                    -------------------------    
533 #                    -------------------------    
534 #                    | lo(mplier) * hi(mplican    
535 #                    -------------------------    
536 #         |                        -----------    
537 #       --|--                      | lo(mplier    
538 #         |                        -----------    
539 #       ======================================    
540 #       --------------------------------------    
541 #       |       hi(result)         |        lo    
542 #       --------------------------------------    
543 ##############################################    
544 mulu64_alg:                                       
545 # load temp registers with operands               
546         mov.l           %d0,%d2                   
547         mov.l           %d0,%d3                   
548         mov.l           %d1,%d4                   
549         swap            %d3                       
550         swap            %d4                       
551                                                   
552 # complete necessary multiplies:                  
553         mulu.w          %d1,%d0                   
554         mulu.w          %d3,%d1                   
555         mulu.w          %d4,%d2                   
556         mulu.w          %d4,%d3                   
557                                                   
558 # add lo portions of [2],[3] to hi portion of     
559 # add carries produced from these adds to [4].    
560 # lo([1]) is the final lo 16 bits of the resul    
561         clr.l           %d4                       
562         swap            %d0                       
563         add.w           %d1,%d0                   
564         addx.l          %d4,%d3                   
565         add.w           %d2,%d0                   
566         addx.l          %d4,%d3                   
567         swap            %d0                       
568                                                   
569 # lo portions of [2],[3] have been added in to    
570 # now, clear lo, put hi in lo reg, and add to     
571         clr.w           %d1                       
572         clr.w           %d2                       
573         swap            %d1                       
574         swap            %d2                       
575         add.l           %d2,%d1                   
576         add.l           %d3,%d1                   
577                                                   
578 # now, grab the condition codes. only one that    
579 # 'N' CAN be set if the operation is unsigned     
580         mov.w           MUL64_CC(%a6),%d4         
581         andi.b          &0x10,%d4                 
582         tst.l           %d1                       
583         bpl.b           mulu64_ddone              
584         ori.b           &0x8,%d4                  
585 mulu64_ddone:                                     
586         mov.w           %d4,%cc                   
587                                                   
588 # here, the result is in d1 and d0. the curren    
589 # the values at the location pointed to by a0.    
590 # use movm here to not disturb the condition c    
591 mulu64_end:                                       
592         exg             %d1,%d0                   
593         movm.l          &0x0003,([0x10,%a6])      
594                                                   
595 # EPILOGUE BEGIN #############################    
596 #       fmovm.l         (%sp)+,&0x0               
597         movm.l          (%sp)+,&0x001c            
598         unlk            %a6                       
599 # EPILOGUE END ###############################    
600                                                   
601         rts                                       
602                                                   
603 # one or both of the operands is zero so the r    
604 # save the zero result to the register file an    
605 mulu64_zero:                                      
606         clr.l           %d0                       
607         clr.l           %d1                       
608                                                   
609         mov.w           MUL64_CC(%a6),%d4         
610         andi.b          &0x10,%d4                 
611         ori.b           &0x4,%d4                  
612         mov.w           %d4,%cc                   
613                                                   
614         bra.b           mulu64_end                
615                                                   
616 ##########                                        
617 # muls.l #                                        
618 ##########                                        
619         global          _060LSP__imuls64_         
620 _060LSP__imuls64_:                                
621                                                   
622 # PROLOGUE BEGIN #############################    
623         link.w          %a6,&-4                   
624         movm.l          &0x3c00,-(%sp)            
625 #       fmovm.l         &0x0,-(%sp)               
626 # PROLOGUE END ###############################    
627                                                   
628         mov.w           %cc,MUL64_CC(%a6)         
629                                                   
630         mov.l           0x8(%a6),%d0              
631         beq.b           mulu64_zero               
632                                                   
633         mov.l           0xc(%a6),%d1              
634         beq.b           mulu64_zero               
635                                                   
636         clr.b           %d5                       
637         tst.l           %d0                       
638         bge.b           muls64_chk_md_sgn         
639         neg.l           %d0                       
640                                                   
641         ori.b           &0x1,%d5                  
642                                                   
643 # the result sign is the exclusive or of the o    
644 muls64_chk_md_sgn:                                
645         tst.l           %d1                       
646         bge.b           muls64_alg                
647         neg.l           %d1                       
648                                                   
649         eori.b          &0x1,%d5                  
650                                                   
651 ##############################################    
652 #       63                         32             
653 #       ----------------------------              
654 #       | hi(mplier) * hi(mplicand)|              
655 #       ----------------------------              
656 #                    -------------------------    
657 #                    | hi(mplier) * lo(mplican    
658 #                    -------------------------    
659 #                    -------------------------    
660 #                    | lo(mplier) * hi(mplican    
661 #                    -------------------------    
662 #         |                        -----------    
663 #       --|--                      | lo(mplier    
664 #         |                        -----------    
665 #       ======================================    
666 #       --------------------------------------    
667 #       |       hi(result)         |        lo    
668 #       --------------------------------------    
669 ##############################################    
670 muls64_alg:                                       
671 # load temp registers with operands               
672         mov.l           %d0,%d2                   
673         mov.l           %d0,%d3                   
674         mov.l           %d1,%d4                   
675         swap            %d3                       
676         swap            %d4                       
677                                                   
678 # complete necessary multiplies:                  
679         mulu.w          %d1,%d0                   
680         mulu.w          %d3,%d1                   
681         mulu.w          %d4,%d2                   
682         mulu.w          %d4,%d3                   
683                                                   
684 # add lo portions of [2],[3] to hi portion of     
685 # add carries produced from these adds to [4].    
686 # lo([1]) is the final lo 16 bits of the resul    
687         clr.l           %d4                       
688         swap            %d0                       
689         add.w           %d1,%d0                   
690         addx.l          %d4,%d3                   
691         add.w           %d2,%d0                   
692         addx.l          %d4,%d3                   
693         swap            %d0                       
694                                                   
695 # lo portions of [2],[3] have been added in to    
696 # now, clear lo, put hi in lo reg, and add to     
697         clr.w           %d1                       
698         clr.w           %d2                       
699         swap            %d1                       
700         swap            %d2                       
701         add.l           %d2,%d1                   
702         add.l           %d3,%d1                   
703                                                   
704         tst.b           %d5                       
705         beq.b           muls64_done               
706                                                   
707 # result should be a signed negative number.      
708 # compute 2's complement of the unsigned numbe    
709 #   -negate all bits and add 1                    
710 muls64_neg:                                       
711         not.l           %d0                       
712         not.l           %d1                       
713         addq.l          &1,%d0                    
714         addx.l          %d4,%d1                   
715                                                   
716 muls64_done:                                      
717         mov.w           MUL64_CC(%a6),%d4         
718         andi.b          &0x10,%d4                 
719         tst.l           %d1                       
720         bpl.b           muls64_ddone              
721         ori.b           &0x8,%d4                  
722 muls64_ddone:                                     
723         mov.w           %d4,%cc                   
724                                                   
725 # here, the result is in d1 and d0. the curren    
726 # the values at the location pointed to by a0.    
727 # use movm here to not disturb the condition c    
728 muls64_end:                                       
729         exg             %d1,%d0                   
730         movm.l          &0x0003,([0x10,%a6])      
731                                                   
732 # EPILOGUE BEGIN #############################    
733 #       fmovm.l         (%sp)+,&0x0               
734         movm.l          (%sp)+,&0x003c            
735         unlk            %a6                       
736 # EPILOGUE END ###############################    
737                                                   
738         rts                                       
739                                                   
740 # one or both of the operands is zero so the r    
741 # save the zero result to the register file an    
742 muls64_zero:                                      
743         clr.l           %d0                       
744         clr.l           %d1                       
745                                                   
746         mov.w           MUL64_CC(%a6),%d4         
747         andi.b          &0x10,%d4                 
748         ori.b           &0x4,%d4                  
749         mov.w           %d4,%cc                   
750                                                   
751         bra.b           muls64_end                
752                                                   
753 ##############################################    
754 # XDEF ***************************************    
755 #       _060LSP__cmp2_Ab_(): Emulate "cmp2.b A    
756 #       _060LSP__cmp2_Aw_(): Emulate "cmp2.w A    
757 #       _060LSP__cmp2_Al_(): Emulate "cmp2.l A    
758 #       _060LSP__cmp2_Db_(): Emulate "cmp2.b D    
759 #       _060LSP__cmp2_Dw_(): Emulate "cmp2.w D    
760 #       _060LSP__cmp2_Dl_(): Emulate "cmp2.l D    
761 #                                                 
762 #       This is the library version which is a    
763 #       and therefore does not work exactly li    
764 #       instruction.                              
765 #                                                 
766 # XREF ***************************************    
767 #       None                                      
768 #                                                 
769 # INPUT **************************************    
770 #       0x4(sp) = Rn                              
771 #       0x8(sp) = pointer to boundary pair        
772 #                                                 
773 # OUTPUT *************************************    
774 #       cc = condition codes are set correctly    
775 #                                                 
776 # ALGORITHM **********************************    
777 #       In the interest of simplicity, all ope    
778 # longword size whether the operation is byte,    
779 # bounds are sign extended accordingly. If Rn     
780 # also sign extended. If Rn is an address regi    
781 # extended since the full register is always u    
782 #       The condition codes are set correctly     
783 #                                                 
784 ##############################################    
785                                                   
786 set     CMP2_CC,        -4                        
787                                                   
788         global          _060LSP__cmp2_Ab_         
789 _060LSP__cmp2_Ab_:                                
790                                                   
791 # PROLOGUE BEGIN #############################    
792         link.w          %a6,&-4                   
793         movm.l          &0x3800,-(%sp)            
794 #       fmovm.l         &0x0,-(%sp)               
795 # PROLOGUE END ###############################    
796                                                   
797         mov.w           %cc,CMP2_CC(%a6)          
798         mov.l           0x8(%a6), %d2             
799                                                   
800         mov.b           ([0xc,%a6],0x0),%d0       
801         mov.b           ([0xc,%a6],0x1),%d1       
802                                                   
803         extb.l          %d0                       
804         extb.l          %d1                       
805         bra.w           l_cmp2_cmp                
806                                                   
807         global          _060LSP__cmp2_Aw_         
808 _060LSP__cmp2_Aw_:                                
809                                                   
810 # PROLOGUE BEGIN #############################    
811         link.w          %a6,&-4                   
812         movm.l          &0x3800,-(%sp)            
813 #       fmovm.l         &0x0,-(%sp)               
814 # PROLOGUE END ###############################    
815                                                   
816         mov.w           %cc,CMP2_CC(%a6)          
817         mov.l           0x8(%a6), %d2             
818                                                   
819         mov.w           ([0xc,%a6],0x0),%d0       
820         mov.w           ([0xc,%a6],0x2),%d1       
821                                                   
822         ext.l           %d0                       
823         ext.l           %d1                       
824         bra.w           l_cmp2_cmp                
825                                                   
826         global          _060LSP__cmp2_Al_         
827 _060LSP__cmp2_Al_:                                
828                                                   
829 # PROLOGUE BEGIN #############################    
830         link.w          %a6,&-4                   
831         movm.l          &0x3800,-(%sp)            
832 #       fmovm.l         &0x0,-(%sp)               
833 # PROLOGUE END ###############################    
834                                                   
835         mov.w           %cc,CMP2_CC(%a6)          
836         mov.l           0x8(%a6), %d2             
837                                                   
838         mov.l           ([0xc,%a6],0x0),%d0       
839         mov.l           ([0xc,%a6],0x4),%d1       
840         bra.w           l_cmp2_cmp                
841                                                   
842         global          _060LSP__cmp2_Db_         
843 _060LSP__cmp2_Db_:                                
844                                                   
845 # PROLOGUE BEGIN #############################    
846         link.w          %a6,&-4                   
847         movm.l          &0x3800,-(%sp)            
848 #       fmovm.l         &0x0,-(%sp)               
849 # PROLOGUE END ###############################    
850                                                   
851         mov.w           %cc,CMP2_CC(%a6)          
852         mov.l           0x8(%a6), %d2             
853                                                   
854         mov.b           ([0xc,%a6],0x0),%d0       
855         mov.b           ([0xc,%a6],0x1),%d1       
856                                                   
857         extb.l          %d0                       
858         extb.l          %d1                       
859                                                   
860 # operation is a data register compare.           
861 # sign extend byte to long so we can do simple    
862         extb.l          %d2                       
863         bra.w           l_cmp2_cmp                
864                                                   
865         global          _060LSP__cmp2_Dw_         
866 _060LSP__cmp2_Dw_:                                
867                                                   
868 # PROLOGUE BEGIN #############################    
869         link.w          %a6,&-4                   
870         movm.l          &0x3800,-(%sp)            
871 #       fmovm.l         &0x0,-(%sp)               
872 # PROLOGUE END ###############################    
873                                                   
874         mov.w           %cc,CMP2_CC(%a6)          
875         mov.l           0x8(%a6), %d2             
876                                                   
877         mov.w           ([0xc,%a6],0x0),%d0       
878         mov.w           ([0xc,%a6],0x2),%d1       
879                                                   
880         ext.l           %d0                       
881         ext.l           %d1                       
882                                                   
883 # operation is a data register compare.           
884 # sign extend word to long so we can do simple    
885         ext.l           %d2                       
886         bra.w           l_cmp2_cmp                
887                                                   
888         global          _060LSP__cmp2_Dl_         
889 _060LSP__cmp2_Dl_:                                
890                                                   
891 # PROLOGUE BEGIN #############################    
892         link.w          %a6,&-4                   
893         movm.l          &0x3800,-(%sp)            
894 #       fmovm.l         &0x0,-(%sp)               
895 # PROLOGUE END ###############################    
896                                                   
897         mov.w           %cc,CMP2_CC(%a6)          
898         mov.l           0x8(%a6), %d2             
899                                                   
900         mov.l           ([0xc,%a6],0x0),%d0       
901         mov.l           ([0xc,%a6],0x4),%d1       
902                                                   
903 #                                                 
904 # To set the ccodes correctly:                    
905 #       (1) save 'Z' bit from (Rn - lo)           
906 #       (2) save 'Z' and 'N' bits from ((hi -     
907 #       (3) keep 'X', 'N', and 'V' from before    
908 #       (4) combine ccodes                        
909 #                                                 
910 l_cmp2_cmp:                                       
911         sub.l           %d0, %d2                  
912         mov.w           %cc, %d3                  
913         andi.b          &0x4, %d3                 
914         sub.l           %d0, %d1                  
915         cmp.l           %d1,%d2                   
916                                                   
917         mov.w           %cc, %d4                  
918         or.b            %d4, %d3                  
919         andi.b          &0x5, %d3                 
920                                                   
921         mov.w           CMP2_CC(%a6), %d4         
922         andi.b          &0x1a, %d4                
923         or.b            %d3, %d4                  
924         mov.w           %d4,%cc                   
925                                                   
926 # EPILOGUE BEGIN #############################    
927 #       fmovm.l         (%sp)+,&0x0               
928         movm.l          (%sp)+,&0x001c            
929         unlk            %a6                       
930 # EPILOGUE END ###############################    
931                                                   
932         rts                                       
                                                      

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

sflogo.php