| blob | 2ff8d3c45438fb01a661034c56116f7556230838 |
1 /* 32 and 64-bit millicode, original author Hewlett-Packard
2 adapted for gcc by Paul Bame <bame@debian.org>
3 and Alan Modra <alan@linuxcare.com.au>.
5 Copyright 2001, 2002, 2003 Free Software Foundation, Inc.
7 This file is part of GCC and is released under the terms of
8 of the GNU General Public License as published by the Free Software
9 Foundation; either version 2, or (at your option) any later version.
10 See the file COPYING in the top-level GCC source directory for a copy
11 of the license. */
14 #ifdef pa64
15 .level 2.0w
16 #endif
18 /* Hardware General Registers. */
19 r0: .reg %r0
20 r1: .reg %r1
21 r2: .reg %r2
22 r3: .reg %r3
23 r4: .reg %r4
24 r5: .reg %r5
25 r6: .reg %r6
26 r7: .reg %r7
27 r8: .reg %r8
28 r9: .reg %r9
29 r10: .reg %r10
30 r11: .reg %r11
31 r12: .reg %r12
32 r13: .reg %r13
33 r14: .reg %r14
34 r15: .reg %r15
35 r16: .reg %r16
36 r17: .reg %r17
37 r18: .reg %r18
38 r19: .reg %r19
39 r20: .reg %r20
40 r21: .reg %r21
41 r22: .reg %r22
42 r23: .reg %r23
43 r24: .reg %r24
44 r25: .reg %r25
45 r26: .reg %r26
46 r27: .reg %r27
47 r28: .reg %r28
48 r29: .reg %r29
49 r30: .reg %r30
50 r31: .reg %r31
52 /* Hardware Space Registers. */
53 sr0: .reg %sr0
54 sr1: .reg %sr1
55 sr2: .reg %sr2
56 sr3: .reg %sr3
57 sr4: .reg %sr4
58 sr5: .reg %sr5
59 sr6: .reg %sr6
60 sr7: .reg %sr7
62 /* Hardware Floating Point Registers. */
63 fr0: .reg %fr0
64 fr1: .reg %fr1
65 fr2: .reg %fr2
66 fr3: .reg %fr3
67 fr4: .reg %fr4
68 fr5: .reg %fr5
69 fr6: .reg %fr6
70 fr7: .reg %fr7
71 fr8: .reg %fr8
72 fr9: .reg %fr9
73 fr10: .reg %fr10
74 fr11: .reg %fr11
75 fr12: .reg %fr12
76 fr13: .reg %fr13
77 fr14: .reg %fr14
78 fr15: .reg %fr15
80 /* Hardware Control Registers. */
81 cr11: .reg %cr11
82 sar: .reg %cr11 /* Shift Amount Register */
84 /* Software Architecture General Registers. */
85 rp: .reg r2 /* return pointer */
86 #ifdef pa64
87 mrp: .reg r2 /* millicode return pointer */
88 #else
89 mrp: .reg r31 /* millicode return pointer */
90 #endif
91 ret0: .reg r28 /* return value */
92 ret1: .reg r29 /* return value (high part of double) */
93 sp: .reg r30 /* stack pointer */
94 dp: .reg r27 /* data pointer */
95 arg0: .reg r26 /* argument */
96 arg1: .reg r25 /* argument or high part of double argument */
97 arg2: .reg r24 /* argument */
98 arg3: .reg r23 /* argument or high part of double argument */
100 /* Software Architecture Space Registers. */
101 /* sr0 ; return link from BLE */
102 sret: .reg sr1 /* return value */
103 sarg: .reg sr1 /* argument */
104 /* sr4 ; PC SPACE tracker */
105 /* sr5 ; process private data */
107 /* Frame Offsets (millicode convention!) Used when calling other
108 millicode routines. Stack unwinding is dependent upon these
109 definitions. */
110 r31_slot: .equ -20 /* "current RP" slot */
111 sr0_slot: .equ -16 /* "static link" slot */
112 #if defined(pa64)
113 mrp_slot: .equ -16 /* "current RP" slot */
114 psp_slot: .equ -8 /* "previous SP" slot */
115 #else
116 mrp_slot: .equ -20 /* "current RP" slot (replacing "r31_slot") */
117 #endif
120 #define DEFINE(name,value)name: .EQU value
121 #define RDEFINE(name,value)name: .REG value
122 #ifdef milliext
123 #define MILLI_BE(lbl) BE lbl(sr7,r0)
124 #define MILLI_BEN(lbl) BE,n lbl(sr7,r0)
125 #define MILLI_BLE(lbl) BLE lbl(sr7,r0)
126 #define MILLI_BLEN(lbl) BLE,n lbl(sr7,r0)
127 #define MILLIRETN BE,n 0(sr0,mrp)
128 #define MILLIRET BE 0(sr0,mrp)
129 #define MILLI_RETN BE,n 0(sr0,mrp)
130 #define MILLI_RET BE 0(sr0,mrp)
131 #else
132 #define MILLI_BE(lbl) B lbl
133 #define MILLI_BEN(lbl) B,n lbl
134 #define MILLI_BLE(lbl) BL lbl,mrp
135 #define MILLI_BLEN(lbl) BL,n lbl,mrp
136 #define MILLIRETN BV,n 0(mrp)
137 #define MILLIRET BV 0(mrp)
138 #define MILLI_RETN BV,n 0(mrp)
139 #define MILLI_RET BV 0(mrp)
140 #endif
142 #ifdef __STDC__
143 #define CAT(a,b) a##b
144 #else
145 #define CAT(a,b) a/**/b
146 #endif
148 #ifdef ELF
149 #define SUBSPA_MILLI .section .text
150 #define SUBSPA_MILLI_DIV .section .text.div,"ax",@progbits! .align 16
151 #define SUBSPA_MILLI_MUL .section .text.mul,"ax",@progbits! .align 16
152 #define ATTR_MILLI
153 #define SUBSPA_DATA .section .data
154 #define ATTR_DATA
155 #define GLOBAL $global$
156 #define GSYM(sym) !sym:
157 #define LSYM(sym) !CAT(.L,sym:)
158 #define LREF(sym) CAT(.L,sym)
160 #else
162 #ifdef coff
163 /* This used to be .milli but since link32 places different named
164 sections in different segments millicode ends up a long ways away
165 from .text (1meg?). This way they will be a lot closer.
167 The SUBSPA_MILLI_* specify locality sets for certain millicode
168 modules in order to ensure that modules that call one another are
169 placed close together. Without locality sets this is unlikely to
170 happen because of the Dynamite linker library search algorithm. We
171 want these modules close together so that short calls always reach
172 (we don't want to require long calls or use long call stubs). */
174 #define SUBSPA_MILLI .subspa .text
175 #define SUBSPA_MILLI_DIV .subspa .text$dv,align=16
176 #define SUBSPA_MILLI_MUL .subspa .text$mu,align=16
177 #define ATTR_MILLI .attr code,read,execute
178 #define SUBSPA_DATA .subspa .data
179 #define ATTR_DATA .attr init_data,read,write
180 #define GLOBAL _gp
181 #else
182 #define SUBSPA_MILLI .subspa $MILLICODE$,QUAD=0,ALIGN=4,ACCESS=0x2c,SORT=8
183 #define SUBSPA_MILLI_DIV SUBSPA_MILLI
184 #define SUBSPA_MILLI_MUL SUBSPA_MILLI
185 #define ATTR_MILLI
186 #define SUBSPA_DATA .subspa $BSS$,quad=1,align=8,access=0x1f,sort=80,zero
187 #define ATTR_DATA
188 #define GLOBAL $global$
189 #endif
190 #define SPACE_DATA .space $PRIVATE$,spnum=1,sort=16
192 #define GSYM(sym) !sym
193 #define LSYM(sym) !CAT(L$,sym)
194 #define LREF(sym) CAT(L$,sym)
195 #endif
197 #ifdef L_dyncall
198 SUBSPA_MILLI
199 ATTR_DATA
200 GSYM($$dyncall)
201 .export $$dyncall,millicode
202 .proc
203 .callinfo millicode
204 .entry
205 bb,>=,n %r22,30,LREF(1) ; branch if not plabel address
206 depi 0,31,2,%r22 ; clear the two least significant bits
207 ldw 4(%r22),%r19 ; load new LTP value
208 ldw 0(%r22),%r22 ; load address of target
209 LSYM(1)
210 #if defined(LINUX) || defined(NETBSD)
211 bv %r0(%r22) ; branch to the real target
212 #else
213 ldsid (%sr0,%r22),%r1 ; get the "space ident" selected by r22
214 mtsp %r1,%sr0 ; move that space identifier into sr0
215 be 0(%sr0,%r22) ; branch to the real target
216 #endif
217 stw %r2,-24(%r30) ; save return address into frame marker
218 .exit
219 .procend
220 #endif
222 #ifdef L_divI
223 /* ROUTINES: $$divI, $$divoI
225 Single precision divide for signed binary integers.
227 The quotient is truncated towards zero.
228 The sign of the quotient is the XOR of the signs of the dividend and
229 divisor.
230 Divide by zero is trapped.
231 Divide of -2**31 by -1 is trapped for $$divoI but not for $$divI.
233 INPUT REGISTERS:
234 . arg0 == dividend
235 . arg1 == divisor
236 . mrp == return pc
237 . sr0 == return space when called externally
239 OUTPUT REGISTERS:
240 . arg0 = undefined
241 . arg1 = undefined
242 . ret1 = quotient
244 OTHER REGISTERS AFFECTED:
245 . r1 = undefined
247 SIDE EFFECTS:
248 . Causes a trap under the following conditions:
249 . divisor is zero (traps with ADDIT,= 0,25,0)
250 . dividend==-2**31 and divisor==-1 and routine is $$divoI
251 . (traps with ADDO 26,25,0)
252 . Changes memory at the following places:
253 . NONE
255 PERMISSIBLE CONTEXT:
256 . Unwindable.
257 . Suitable for internal or external millicode.
258 . Assumes the special millicode register conventions.
260 DISCUSSION:
261 . Branchs to other millicode routines using BE
262 . $$div_# for # being 2,3,4,5,6,7,8,9,10,12,14,15
263 .
264 . For selected divisors, calls a divide by constant routine written by
265 . Karl Pettis. Eligible divisors are 1..15 excluding 11 and 13.
266 .
267 . The only overflow case is -2**31 divided by -1.
268 . Both routines return -2**31 but only $$divoI traps. */
270 RDEFINE(temp,r1)
271 RDEFINE(retreg,ret1) /* r29 */
272 RDEFINE(temp1,arg0)
273 SUBSPA_MILLI_DIV
274 ATTR_MILLI
275 .import $$divI_2,millicode
276 .import $$divI_3,millicode
277 .import $$divI_4,millicode
278 .import $$divI_5,millicode
279 .import $$divI_6,millicode
280 .import $$divI_7,millicode
281 .import $$divI_8,millicode
282 .import $$divI_9,millicode
283 .import $$divI_10,millicode
284 .import $$divI_12,millicode
285 .import $$divI_14,millicode
286 .import $$divI_15,millicode
287 .export $$divI,millicode
288 .export $$divoI,millicode
289 .proc
290 .callinfo millicode
291 .entry
292 GSYM($$divoI)
293 comib,=,n -1,arg1,LREF(negative1) /* when divisor == -1 */
294 GSYM($$divI)
295 ldo -1(arg1),temp /* is there at most one bit set ? */
296 and,<> arg1,temp,r0 /* if not, don't use power of 2 divide */
297 addi,> 0,arg1,r0 /* if divisor > 0, use power of 2 divide */
298 b,n LREF(neg_denom)
299 LSYM(pow2)
300 addi,>= 0,arg0,retreg /* if numerator is negative, add the */
301 add arg0,temp,retreg /* (denominaotr -1) to correct for shifts */
302 extru,= arg1,15,16,temp /* test denominator with 0xffff0000 */
303 extrs retreg,15,16,retreg /* retreg = retreg >> 16 */
304 or arg1,temp,arg1 /* arg1 = arg1 | (arg1 >> 16) */
305 ldi 0xcc,temp1 /* setup 0xcc in temp1 */
306 extru,= arg1,23,8,temp /* test denominator with 0xff00 */
307 extrs retreg,23,24,retreg /* retreg = retreg >> 8 */
308 or arg1,temp,arg1 /* arg1 = arg1 | (arg1 >> 8) */
309 ldi 0xaa,temp /* setup 0xaa in temp */
310 extru,= arg1,27,4,r0 /* test denominator with 0xf0 */
311 extrs retreg,27,28,retreg /* retreg = retreg >> 4 */
312 and,= arg1,temp1,r0 /* test denominator with 0xcc */
313 extrs retreg,29,30,retreg /* retreg = retreg >> 2 */
314 and,= arg1,temp,r0 /* test denominator with 0xaa */
315 extrs retreg,30,31,retreg /* retreg = retreg >> 1 */
316 MILLIRETN
317 LSYM(neg_denom)
318 addi,< 0,arg1,r0 /* if arg1 >= 0, it's not power of 2 */
319 b,n LREF(regular_seq)
320 sub r0,arg1,temp /* make denominator positive */
321 comb,=,n arg1,temp,LREF(regular_seq) /* test against 0x80000000 and 0 */
322 ldo -1(temp),retreg /* is there at most one bit set ? */
323 and,= temp,retreg,r0 /* if so, the denominator is power of 2 */
324 b,n LREF(regular_seq)
325 sub r0,arg0,retreg /* negate numerator */
326 comb,=,n arg0,retreg,LREF(regular_seq) /* test against 0x80000000 */
327 copy retreg,arg0 /* set up arg0, arg1 and temp */
328 copy temp,arg1 /* before branching to pow2 */
329 b LREF(pow2)
330 ldo -1(arg1),temp
331 LSYM(regular_seq)
332 comib,>>=,n 15,arg1,LREF(small_divisor)
333 add,>= 0,arg0,retreg /* move dividend, if retreg < 0, */
334 LSYM(normal)
335 subi 0,retreg,retreg /* make it positive */
336 sub 0,arg1,temp /* clear carry, */
337 /* negate the divisor */
338 ds 0,temp,0 /* set V-bit to the comple- */
339 /* ment of the divisor sign */
340 add retreg,retreg,retreg /* shift msb bit into carry */
341 ds r0,arg1,temp /* 1st divide step, if no carry */
342 addc retreg,retreg,retreg /* shift retreg with/into carry */
343 ds temp,arg1,temp /* 2nd divide step */
344 addc retreg,retreg,retreg /* shift retreg with/into carry */
345 ds temp,arg1,temp /* 3rd divide step */
346 addc retreg,retreg,retreg /* shift retreg with/into carry */
347 ds temp,arg1,temp /* 4th divide step */
348 addc retreg,retreg,retreg /* shift retreg with/into carry */
349 ds temp,arg1,temp /* 5th divide step */
350 addc retreg,retreg,retreg /* shift retreg with/into carry */
351 ds temp,arg1,temp /* 6th divide step */
352 addc retreg,retreg,retreg /* shift retreg with/into carry */
353 ds temp,arg1,temp /* 7th divide step */
354 addc retreg,retreg,retreg /* shift retreg with/into carry */
355 ds temp,arg1,temp /* 8th divide step */
356 addc retreg,retreg,retreg /* shift retreg with/into carry */
357 ds temp,arg1,temp /* 9th divide step */
358 addc retreg,retreg,retreg /* shift retreg with/into carry */
359 ds temp,arg1,temp /* 10th divide step */
360 addc retreg,retreg,retreg /* shift retreg with/into carry */
361 ds temp,arg1,temp /* 11th divide step */
362 addc retreg,retreg,retreg /* shift retreg with/into carry */
363 ds temp,arg1,temp /* 12th divide step */
364 addc retreg,retreg,retreg /* shift retreg with/into carry */
365 ds temp,arg1,temp /* 13th divide step */
366 addc retreg,retreg,retreg /* shift retreg with/into carry */
367 ds temp,arg1,temp /* 14th divide step */
368 addc retreg,retreg,retreg /* shift retreg with/into carry */
369 ds temp,arg1,temp /* 15th divide step */
370 addc retreg,retreg,retreg /* shift retreg with/into carry */
371 ds temp,arg1,temp /* 16th divide step */
372 addc retreg,retreg,retreg /* shift retreg with/into carry */
373 ds temp,arg1,temp /* 17th divide step */
374 addc retreg,retreg,retreg /* shift retreg with/into carry */
375 ds temp,arg1,temp /* 18th divide step */
376 addc retreg,retreg,retreg /* shift retreg with/into carry */
377 ds temp,arg1,temp /* 19th divide step */
378 addc retreg,retreg,retreg /* shift retreg with/into carry */
379 ds temp,arg1,temp /* 20th divide step */
380 addc retreg,retreg,retreg /* shift retreg with/into carry */
381 ds temp,arg1,temp /* 21st divide step */
382 addc retreg,retreg,retreg /* shift retreg with/into carry */
383 ds temp,arg1,temp /* 22nd divide step */
384 addc retreg,retreg,retreg /* shift retreg with/into carry */
385 ds temp,arg1,temp /* 23rd divide step */
386 addc retreg,retreg,retreg /* shift retreg with/into carry */
387 ds temp,arg1,temp /* 24th divide step */
388 addc retreg,retreg,retreg /* shift retreg with/into carry */
389 ds temp,arg1,temp /* 25th divide step */
390 addc retreg,retreg,retreg /* shift retreg with/into carry */
391 ds temp,arg1,temp /* 26th divide step */
392 addc retreg,retreg,retreg /* shift retreg with/into carry */
393 ds temp,arg1,temp /* 27th divide step */
394 addc retreg,retreg,retreg /* shift retreg with/into carry */
395 ds temp,arg1,temp /* 28th divide step */
396 addc retreg,retreg,retreg /* shift retreg with/into carry */
397 ds temp,arg1,temp /* 29th divide step */
398 addc retreg,retreg,retreg /* shift retreg with/into carry */
399 ds temp,arg1,temp /* 30th divide step */
400 addc retreg,retreg,retreg /* shift retreg with/into carry */
401 ds temp,arg1,temp /* 31st divide step */
402 addc retreg,retreg,retreg /* shift retreg with/into carry */
403 ds temp,arg1,temp /* 32nd divide step, */
404 addc retreg,retreg,retreg /* shift last retreg bit into retreg */
405 xor,>= arg0,arg1,0 /* get correct sign of quotient */
406 sub 0,retreg,retreg /* based on operand signs */
407 MILLIRETN
408 nop
410 LSYM(small_divisor)
412 #if defined(pa64)
413 /* Clear the upper 32 bits of the arg1 register. We are working with */
414 /* small divisors (and 32 bit integers) We must not be mislead */
415 /* by "1" bits left in the upper 32 bits. */
416 depd %r0,31,32,%r25
417 #endif
418 blr,n arg1,r0
419 nop
420 /* table for divisor == 0,1, ... ,15 */
421 addit,= 0,arg1,r0 /* trap if divisor == 0 */
422 nop
423 MILLIRET /* divisor == 1 */
424 copy arg0,retreg
425 MILLI_BEN($$divI_2) /* divisor == 2 */
426 nop
427 MILLI_BEN($$divI_3) /* divisor == 3 */
428 nop
429 MILLI_BEN($$divI_4) /* divisor == 4 */
430 nop
431 MILLI_BEN($$divI_5) /* divisor == 5 */
432 nop
433 MILLI_BEN($$divI_6) /* divisor == 6 */
434 nop
435 MILLI_BEN($$divI_7) /* divisor == 7 */
436 nop
437 MILLI_BEN($$divI_8) /* divisor == 8 */
438 nop
439 MILLI_BEN($$divI_9) /* divisor == 9 */
440 nop
441 MILLI_BEN($$divI_10) /* divisor == 10 */
442 nop
443 b LREF(normal) /* divisor == 11 */
444 add,>= 0,arg0,retreg
445 MILLI_BEN($$divI_12) /* divisor == 12 */
446 nop
447 b LREF(normal) /* divisor == 13 */
448 add,>= 0,arg0,retreg
449 MILLI_BEN($$divI_14) /* divisor == 14 */
450 nop
451 MILLI_BEN($$divI_15) /* divisor == 15 */
452 nop
454 LSYM(negative1)
455 sub 0,arg0,retreg /* result is negation of dividend */
456 MILLIRET
457 addo arg0,arg1,r0 /* trap iff dividend==0x80000000 && divisor==-1 */
458 .exit
459 .procend
460 .end
461 #endif
463 #ifdef L_divU
464 /* ROUTINE: $$divU
465 .
466 . Single precision divide for unsigned integers.
467 .
468 . Quotient is truncated towards zero.
469 . Traps on divide by zero.
471 INPUT REGISTERS:
472 . arg0 == dividend
473 . arg1 == divisor
474 . mrp == return pc
475 . sr0 == return space when called externally
477 OUTPUT REGISTERS:
478 . arg0 = undefined
479 . arg1 = undefined
480 . ret1 = quotient
482 OTHER REGISTERS AFFECTED:
483 . r1 = undefined
485 SIDE EFFECTS:
486 . Causes a trap under the following conditions:
487 . divisor is zero
488 . Changes memory at the following places:
489 . NONE
491 PERMISSIBLE CONTEXT:
492 . Unwindable.
493 . Does not create a stack frame.
494 . Suitable for internal or external millicode.
495 . Assumes the special millicode register conventions.
497 DISCUSSION:
498 . Branchs to other millicode routines using BE:
499 . $$divU_# for 3,5,6,7,9,10,12,14,15
500 .
501 . For selected small divisors calls the special divide by constant
502 . routines written by Karl Pettis. These are: 3,5,6,7,9,10,12,14,15. */
504 RDEFINE(temp,r1)
505 RDEFINE(retreg,ret1) /* r29 */
506 RDEFINE(temp1,arg0)
507 SUBSPA_MILLI_DIV
508 ATTR_MILLI
509 .export $$divU,millicode
510 .import $$divU_3,millicode
511 .import $$divU_5,millicode
512 .import $$divU_6,millicode
513 .import $$divU_7,millicode
514 .import $$divU_9,millicode
515 .import $$divU_10,millicode
516 .import $$divU_12,millicode
517 .import $$divU_14,millicode
518 .import $$divU_15,millicode
519 .proc
520 .callinfo millicode
521 .entry
522 GSYM($$divU)
523 /* The subtract is not nullified since it does no harm and can be used
524 by the two cases that branch back to "normal". */
525 ldo -1(arg1),temp /* is there at most one bit set ? */
526 and,= arg1,temp,r0 /* if so, denominator is power of 2 */
527 b LREF(regular_seq)
528 addit,= 0,arg1,0 /* trap for zero dvr */
529 copy arg0,retreg
530 extru,= arg1,15,16,temp /* test denominator with 0xffff0000 */
531 extru retreg,15,16,retreg /* retreg = retreg >> 16 */
532 or arg1,temp,arg1 /* arg1 = arg1 | (arg1 >> 16) */
533 ldi 0xcc,temp1 /* setup 0xcc in temp1 */
534 extru,= arg1,23,8,temp /* test denominator with 0xff00 */
535 extru retreg,23,24,retreg /* retreg = retreg >> 8 */
536 or arg1,temp,arg1 /* arg1 = arg1 | (arg1 >> 8) */
537 ldi 0xaa,temp /* setup 0xaa in temp */
538 extru,= arg1,27,4,r0 /* test denominator with 0xf0 */
539 extru retreg,27,28,retreg /* retreg = retreg >> 4 */
540 and,= arg1,temp1,r0 /* test denominator with 0xcc */
541 extru retreg,29,30,retreg /* retreg = retreg >> 2 */
542 and,= arg1,temp,r0 /* test denominator with 0xaa */
543 extru retreg,30,31,retreg /* retreg = retreg >> 1 */
544 MILLIRETN
545 nop
546 LSYM(regular_seq)
547 comib,>= 15,arg1,LREF(special_divisor)
548 subi 0,arg1,temp /* clear carry, negate the divisor */
549 ds r0,temp,r0 /* set V-bit to 1 */
550 LSYM(normal)
551 add arg0,arg0,retreg /* shift msb bit into carry */
552 ds r0,arg1,temp /* 1st divide step, if no carry */
553 addc retreg,retreg,retreg /* shift retreg with/into carry */
554 ds temp,arg1,temp /* 2nd divide step */
555 addc retreg,retreg,retreg /* shift retreg with/into carry */
556 ds temp,arg1,temp /* 3rd divide step */
557 addc retreg,retreg,retreg /* shift retreg with/into carry */
558 ds temp,arg1,temp /* 4th divide step */
559 addc retreg,retreg,retreg /* shift retreg with/into carry */
560 ds temp,arg1,temp /* 5th divide step */
561 addc retreg,retreg,retreg /* shift retreg with/into carry */
562 ds temp,arg1,temp /* 6th divide step */
563 addc retreg,retreg,retreg /* shift retreg with/into carry */
564 ds temp,arg1,temp /* 7th divide step */
565 addc retreg,retreg,retreg /* shift retreg with/into carry */
566 ds temp,arg1,temp /* 8th divide step */
567 addc retreg,retreg,retreg /* shift retreg with/into carry */
568 ds temp,arg1,temp /* 9th divide step */
569 addc retreg,retreg,retreg /* shift retreg with/into carry */
570 ds temp,arg1,temp /* 10th divide step */
571 addc retreg,retreg,retreg /* shift retreg with/into carry */
572 ds temp,arg1,temp /* 11th divide step */
573 addc retreg,retreg,retreg /* shift retreg with/into carry */
574 ds temp,arg1,temp /* 12th divide step */
575 addc retreg,retreg,retreg /* shift retreg with/into carry */
576 ds temp,arg1,temp /* 13th divide step */
577 addc retreg,retreg,retreg /* shift retreg with/into carry */
578 ds temp,arg1,temp /* 14th divide step */
579 addc retreg,retreg,retreg /* shift retreg with/into carry */
580 ds temp,arg1,temp /* 15th divide step */
581 addc retreg,retreg,retreg /* shift retreg with/into carry */
582 ds temp,arg1,temp /* 16th divide step */
583 addc retreg,retreg,retreg /* shift retreg with/into carry */
584 ds temp,arg1,temp /* 17th divide step */
585 addc retreg,retreg,retreg /* shift retreg with/into carry */
586 ds temp,arg1,temp /* 18th divide step */
587 addc retreg,retreg,retreg /* shift retreg with/into carry */
588 ds temp,arg1,temp /* 19th divide step */
589 addc retreg,retreg,retreg /* shift retreg with/into carry */
590 ds temp,arg1,temp /* 20th divide step */
591 addc retreg,retreg,retreg /* shift retreg with/into carry */
592 ds temp,arg1,temp /* 21st divide step */
593 addc retreg,retreg,retreg /* shift retreg with/into carry */
594 ds temp,arg1,temp /* 22nd divide step */
595 addc retreg,retreg,retreg /* shift retreg with/into carry */
596 ds temp,arg1,temp /* 23rd divide step */
597 addc retreg,retreg,retreg /* shift retreg with/into carry */
598 ds temp,arg1,temp /* 24th divide step */
599 addc retreg,retreg,retreg /* shift retreg with/into carry */
600 ds temp,arg1,temp /* 25th divide step */
601 addc retreg,retreg,retreg /* shift retreg with/into carry */
602 ds temp,arg1,temp /* 26th divide step */
603 addc retreg,retreg,retreg /* shift retreg with/into carry */
604 ds temp,arg1,temp /* 27th divide step */
605 addc retreg,retreg,retreg /* shift retreg with/into carry */
606 ds temp,arg1,temp /* 28th divide step */
607 addc retreg,retreg,retreg /* shift retreg with/into carry */
608 ds temp,arg1,temp /* 29th divide step */
609 addc retreg,retreg,retreg /* shift retreg with/into carry */
610 ds temp,arg1,temp /* 30th divide step */
611 addc retreg,retreg,retreg /* shift retreg with/into carry */
612 ds temp,arg1,temp /* 31st divide step */
613 addc retreg,retreg,retreg /* shift retreg with/into carry */
614 ds temp,arg1,temp /* 32nd divide step, */
615 MILLIRET
616 addc retreg,retreg,retreg /* shift last retreg bit into retreg */
618 /* Handle the cases where divisor is a small constant or has high bit on. */
619 LSYM(special_divisor)
620 /* blr arg1,r0 */
621 /* comib,>,n 0,arg1,LREF(big_divisor) ; nullify previous instruction */
623 /* Pratap 8/13/90. The 815 Stirling chip set has a bug that prevents us from
624 generating such a blr, comib sequence. A problem in nullification. So I
625 rewrote this code. */
627 #if defined(pa64)
628 /* Clear the upper 32 bits of the arg1 register. We are working with
629 small divisors (and 32 bit unsigned integers) We must not be mislead
630 by "1" bits left in the upper 32 bits. */
631 depd %r0,31,32,%r25
632 #endif
633 comib,> 0,arg1,LREF(big_divisor)
634 nop
635 blr arg1,r0
636 nop
638 LSYM(zero_divisor) /* this label is here to provide external visibility */
639 addit,= 0,arg1,0 /* trap for zero dvr */
640 nop
641 MILLIRET /* divisor == 1 */
642 copy arg0,retreg
643 MILLIRET /* divisor == 2 */
644 extru arg0,30,31,retreg
645 MILLI_BEN($$divU_3) /* divisor == 3 */
646 nop
647 MILLIRET /* divisor == 4 */
648 extru arg0,29,30,retreg
649 MILLI_BEN($$divU_5) /* divisor == 5 */
650 nop
651 MILLI_BEN($$divU_6) /* divisor == 6 */
652 nop
653 MILLI_BEN($$divU_7) /* divisor == 7 */
654 nop
655 MILLIRET /* divisor == 8 */
656 extru arg0,28,29,retreg
657 MILLI_BEN($$divU_9) /* divisor == 9 */
658 nop
659 MILLI_BEN($$divU_10) /* divisor == 10 */
660 nop
661 b LREF(normal) /* divisor == 11 */
662 ds r0,temp,r0 /* set V-bit to 1 */
663 MILLI_BEN($$divU_12) /* divisor == 12 */
664 nop
665 b LREF(normal) /* divisor == 13 */
666 ds r0,temp,r0 /* set V-bit to 1 */
667 MILLI_BEN($$divU_14) /* divisor == 14 */
668 nop
669 MILLI_BEN($$divU_15) /* divisor == 15 */
670 nop
672 /* Handle the case where the high bit is on in the divisor.
673 Compute: if( dividend>=divisor) quotient=1; else quotient=0;
674 Note: dividend>==divisor iff dividend-divisor does not borrow
675 and not borrow iff carry. */
676 LSYM(big_divisor)
677 sub arg0,arg1,r0
678 MILLIRET
679 addc r0,r0,retreg
680 .exit
681 .procend
682 .end
683 #endif
685 #ifdef L_remI
686 /* ROUTINE: $$remI
688 DESCRIPTION:
689 . $$remI returns the remainder of the division of two signed 32-bit
690 . integers. The sign of the remainder is the same as the sign of
691 . the dividend.
694 INPUT REGISTERS:
695 . arg0 == dividend
696 . arg1 == divisor
697 . mrp == return pc
698 . sr0 == return space when called externally
700 OUTPUT REGISTERS:
701 . arg0 = destroyed
702 . arg1 = destroyed
703 . ret1 = remainder
705 OTHER REGISTERS AFFECTED:
706 . r1 = undefined
708 SIDE EFFECTS:
709 . Causes a trap under the following conditions: DIVIDE BY ZERO
710 . Changes memory at the following places: NONE
712 PERMISSIBLE CONTEXT:
713 . Unwindable
714 . Does not create a stack frame
715 . Is usable for internal or external microcode
717 DISCUSSION:
718 . Calls other millicode routines via mrp: NONE
719 . Calls other millicode routines: NONE */
721 RDEFINE(tmp,r1)
722 RDEFINE(retreg,ret1)
724 SUBSPA_MILLI
725 ATTR_MILLI
726 .proc
727 .callinfo millicode
728 .entry
729 GSYM($$remI)
730 GSYM($$remoI)
731 .export $$remI,MILLICODE
732 .export $$remoI,MILLICODE
733 ldo -1(arg1),tmp /* is there at most one bit set ? */
734 and,<> arg1,tmp,r0 /* if not, don't use power of 2 */
735 addi,> 0,arg1,r0 /* if denominator > 0, use power */
736 /* of 2 */
737 b,n LREF(neg_denom)
738 LSYM(pow2)
739 comb,>,n 0,arg0,LREF(neg_num) /* is numerator < 0 ? */
740 and arg0,tmp,retreg /* get the result */
741 MILLIRETN
742 LSYM(neg_num)
743 subi 0,arg0,arg0 /* negate numerator */
744 and arg0,tmp,retreg /* get the result */
745 subi 0,retreg,retreg /* negate result */
746 MILLIRETN
747 LSYM(neg_denom)
748 addi,< 0,arg1,r0 /* if arg1 >= 0, it's not power */
749 /* of 2 */
750 b,n LREF(regular_seq)
751 sub r0,arg1,tmp /* make denominator positive */
752 comb,=,n arg1,tmp,LREF(regular_seq) /* test against 0x80000000 and 0 */
753 ldo -1(tmp),retreg /* is there at most one bit set ? */
754 and,= tmp,retreg,r0 /* if not, go to regular_seq */
755 b,n LREF(regular_seq)
756 comb,>,n 0,arg0,LREF(neg_num_2) /* if arg0 < 0, negate it */
757 and arg0,retreg,retreg
758 MILLIRETN
759 LSYM(neg_num_2)
760 subi 0,arg0,tmp /* test against 0x80000000 */
761 and tmp,retreg,retreg
762 subi 0,retreg,retreg
763 MILLIRETN
764 LSYM(regular_seq)
765 addit,= 0,arg1,0 /* trap if div by zero */
766 add,>= 0,arg0,retreg /* move dividend, if retreg < 0, */
767 sub 0,retreg,retreg /* make it positive */
768 sub 0,arg1, tmp /* clear carry, */
769 /* negate the divisor */
770 ds 0, tmp,0 /* set V-bit to the comple- */
771 /* ment of the divisor sign */
772 or 0,0, tmp /* clear tmp */
773 add retreg,retreg,retreg /* shift msb bit into carry */
774 ds tmp,arg1, tmp /* 1st divide step, if no carry */
775 /* out, msb of quotient = 0 */
776 addc retreg,retreg,retreg /* shift retreg with/into carry */
777 LSYM(t1)
778 ds tmp,arg1, tmp /* 2nd divide step */
779 addc retreg,retreg,retreg /* shift retreg with/into carry */
780 ds tmp,arg1, tmp /* 3rd divide step */
781 addc retreg,retreg,retreg /* shift retreg with/into carry */
782 ds tmp,arg1, tmp /* 4th divide step */
783 addc retreg,retreg,retreg /* shift retreg with/into carry */
784 ds tmp,arg1, tmp /* 5th divide step */
785 addc retreg,retreg,retreg /* shift retreg with/into carry */
786 ds tmp,arg1, tmp /* 6th divide step */
787 addc retreg,retreg,retreg /* shift retreg with/into carry */
788 ds tmp,arg1, tmp /* 7th divide step */
789 addc retreg,retreg,retreg /* shift retreg with/into carry */
790 ds tmp,arg1, tmp /* 8th divide step */
791 addc retreg,retreg,retreg /* shift retreg with/into carry */
792 ds tmp,arg1, tmp /* 9th divide step */
793 addc retreg,retreg,retreg /* shift retreg with/into carry */
794 ds tmp,arg1, tmp /* 10th divide step */
795 addc retreg,retreg,retreg /* shift retreg with/into carry */
796 ds tmp,arg1, tmp /* 11th divide step */
797 addc retreg,retreg,retreg /* shift retreg with/into carry */
798 ds tmp,arg1, tmp /* 12th divide step */
799 addc retreg,retreg,retreg /* shift retreg with/into carry */
800 ds tmp,arg1, tmp /* 13th divide step */
801 addc retreg,retreg,retreg /* shift retreg with/into carry */
802 ds tmp,arg1, tmp /* 14th divide step */
803 addc retreg,retreg,retreg /* shift retreg with/into carry */
804 ds tmp,arg1, tmp /* 15th divide step */
805 addc retreg,retreg,retreg /* shift retreg with/into carry */
806 ds tmp,arg1, tmp /* 16th divide step */
807 addc retreg,retreg,retreg /* shift retreg with/into carry */
808 ds tmp,arg1, tmp /* 17th divide step */
809 addc retreg,retreg,retreg /* shift retreg with/into carry */
810 ds tmp,arg1, tmp /* 18th divide step */
811 addc retreg,retreg,retreg /* shift retreg with/into carry */
812 ds tmp,arg1, tmp /* 19th divide step */
813 addc retreg,retreg,retreg /* shift retreg with/into carry */
814 ds tmp,arg1, tmp /* 20th divide step */
815 addc retreg,retreg,retreg /* shift retreg with/into carry */
816 ds tmp,arg1, tmp /* 21st divide step */
817 addc retreg,retreg,retreg /* shift retreg with/into carry */
818 ds tmp,arg1, tmp /* 22nd divide step */
819 addc retreg,retreg,retreg /* shift retreg with/into carry */
820 ds tmp,arg1, tmp /* 23rd divide step */
821 addc retreg,retreg,retreg /* shift retreg with/into carry */
822 ds tmp,arg1, tmp /* 24th divide step */
823 addc retreg,retreg,retreg /* shift retreg with/into carry */
824 ds tmp,arg1, tmp /* 25th divide step */
825 addc retreg,retreg,retreg /* shift retreg with/into carry */
826 ds tmp,arg1, tmp /* 26th divide step */
827 addc retreg,retreg,retreg /* shift retreg with/into carry */
828 ds tmp,arg1, tmp /* 27th divide step */
829 addc retreg,retreg,retreg /* shift retreg with/into carry */
830 ds tmp,arg1, tmp /* 28th divide step */
831 addc retreg,retreg,retreg /* shift retreg with/into carry */
832 ds tmp,arg1, tmp /* 29th divide step */
833 addc retreg,retreg,retreg /* shift retreg with/into carry */
834 ds tmp,arg1, tmp /* 30th divide step */
835 addc retreg,retreg,retreg /* shift retreg with/into carry */
836 ds tmp,arg1, tmp /* 31st divide step */
837 addc retreg,retreg,retreg /* shift retreg with/into carry */
838 ds tmp,arg1, tmp /* 32nd divide step, */
839 addc retreg,retreg,retreg /* shift last bit into retreg */
840 movb,>=,n tmp,retreg,LREF(finish) /* branch if pos. tmp */
841 add,< arg1,0,0 /* if arg1 > 0, add arg1 */
842 add,tr tmp,arg1,retreg /* for correcting remainder tmp */
843 sub tmp,arg1,retreg /* else add absolute value arg1 */
844 LSYM(finish)
845 add,>= arg0,0,0 /* set sign of remainder */
846 sub 0,retreg,retreg /* to sign of dividend */
847 MILLIRET
848 nop
849 .exit
850 .procend
851 #ifdef milliext
852 .origin 0x00000200
853 #endif
854 .end
855 #endif
857 #ifdef L_remU
858 /* ROUTINE: $$remU
859 . Single precision divide for remainder with unsigned binary integers.
860 .
861 . The remainder must be dividend-(dividend/divisor)*divisor.
862 . Divide by zero is trapped.
864 INPUT REGISTERS:
865 . arg0 == dividend
866 . arg1 == divisor
867 . mrp == return pc
868 . sr0 == return space when called externally
870 OUTPUT REGISTERS:
871 . arg0 = undefined
872 . arg1 = undefined
873 . ret1 = remainder
875 OTHER REGISTERS AFFECTED:
876 . r1 = undefined
878 SIDE EFFECTS:
879 . Causes a trap under the following conditions: DIVIDE BY ZERO
880 . Changes memory at the following places: NONE
882 PERMISSIBLE CONTEXT:
883 . Unwindable.
884 . Does not create a stack frame.
885 . Suitable for internal or external millicode.
886 . Assumes the special millicode register conventions.
888 DISCUSSION:
889 . Calls other millicode routines using mrp: NONE
890 . Calls other millicode routines: NONE */
893 RDEFINE(temp,r1)
894 RDEFINE(rmndr,ret1) /* r29 */
895 SUBSPA_MILLI
896 ATTR_MILLI
897 .export $$remU,millicode
898 .proc
899 .callinfo millicode
900 .entry
901 GSYM($$remU)
902 ldo -1(arg1),temp /* is there at most one bit set ? */
903 and,= arg1,temp,r0 /* if not, don't use power of 2 */
904 b LREF(regular_seq)
905 addit,= 0,arg1,r0 /* trap on div by zero */
906 and arg0,temp,rmndr /* get the result for power of 2 */
907 MILLIRETN
908 LSYM(regular_seq)
909 comib,>=,n 0,arg1,LREF(special_case)
910 subi 0,arg1,rmndr /* clear carry, negate the divisor */
911 ds r0,rmndr,r0 /* set V-bit to 1 */
912 add arg0,arg0,temp /* shift msb bit into carry */
913 ds r0,arg1,rmndr /* 1st divide step, if no carry */
914 addc temp,temp,temp /* shift temp with/into carry */
915 ds rmndr,arg1,rmndr /* 2nd divide step */
916 addc temp,temp,temp /* shift temp with/into carry */
917 ds rmndr,arg1,rmndr /* 3rd divide step */
918 addc temp,temp,temp /* shift temp with/into carry */
919 ds rmndr,arg1,rmndr /* 4th divide step */
920 addc temp,temp,temp /* shift temp with/into carry */
921 ds rmndr,arg1,rmndr /* 5th divide step */
922 addc temp,temp,temp /* shift temp with/into carry */
923 ds rmndr,arg1,rmndr /* 6th divide step */
924 addc temp,temp,temp /* shift temp with/into carry */
925 ds rmndr,arg1,rmndr /* 7th divide step */
926 addc temp,temp,temp /* shift temp with/into carry */
927 ds rmndr,arg1,rmndr /* 8th divide step */
928 addc temp,temp,temp /* shift temp with/into carry */
929 ds rmndr,arg1,rmndr /* 9th divide step */
930 addc temp,temp,temp /* shift temp with/into carry */
931 ds rmndr,arg1,rmndr /* 10th divide step */
932 addc temp,temp,temp /* shift temp with/into carry */
933 ds rmndr,arg1,rmndr /* 11th divide step */
934 addc temp,temp,temp /* shift temp with/into carry */
935 ds rmndr,arg1,rmndr /* 12th divide step */
936 addc temp,temp,temp /* shift temp with/into carry */
937 ds rmndr,arg1,rmndr /* 13th divide step */
938 addc temp,temp,temp /* shift temp with/into carry */
939 ds rmndr,arg1,rmndr /* 14th divide step */
940 addc temp,temp,temp /* shift temp with/into carry */
941 ds rmndr,arg1,rmndr /* 15th divide step */
942 addc temp,temp,temp /* shift temp with/into carry */
943 ds rmndr,arg1,rmndr /* 16th divide step */
944 addc temp,temp,temp /* shift temp with/into carry */
945 ds rmndr,arg1,rmndr /* 17th divide step */
946 addc temp,temp,temp /* shift temp with/into carry */
947 ds rmndr,arg1,rmndr /* 18th divide step */
948 addc temp,temp,temp /* shift temp with/into carry */
949 ds rmndr,arg1,rmndr /* 19th divide step */
950 addc temp,temp,temp /* shift temp with/into carry */
951 ds rmndr,arg1,rmndr /* 20th divide step */
952 addc temp,temp,temp /* shift temp with/into carry */
953 ds rmndr,arg1,rmndr /* 21st divide step */
954 addc temp,temp,temp /* shift temp with/into carry */
955 ds rmndr,arg1,rmndr /* 22nd divide step */
956 addc temp,temp,temp /* shift temp with/into carry */
957 ds rmndr,arg1,rmndr /* 23rd divide step */
958 addc temp,temp,temp /* shift temp with/into carry */
959 ds rmndr,arg1,rmndr /* 24th divide step */
960 addc temp,temp,temp /* shift temp with/into carry */
961 ds rmndr,arg1,rmndr /* 25th divide step */
962 addc temp,temp,temp /* shift temp with/into carry */
963 ds rmndr,arg1,rmndr /* 26th divide step */
964 addc temp,temp,temp /* shift temp with/into carry */
965 ds rmndr,arg1,rmndr /* 27th divide step */
966 addc temp,temp,temp /* shift temp with/into carry */
967 ds rmndr,arg1,rmndr /* 28th divide step */
968 addc temp,temp,temp /* shift temp with/into carry */
969 ds rmndr,arg1,rmndr /* 29th divide step */
970 addc temp,temp,temp /* shift temp with/into carry */
971 ds rmndr,arg1,rmndr /* 30th divide step */
972 addc temp,temp,temp /* shift temp with/into carry */
973 ds rmndr,arg1,rmndr /* 31st divide step */
974 addc temp,temp,temp /* shift temp with/into carry */
975 ds rmndr,arg1,rmndr /* 32nd divide step, */
976 comiclr,<= 0,rmndr,r0
977 add rmndr,arg1,rmndr /* correction */
978 MILLIRETN
979 nop
981 /* Putting >= on the last DS and deleting COMICLR does not work! */
982 LSYM(special_case)
983 sub,>>= arg0,arg1,rmndr
984 copy arg0,rmndr
985 MILLIRETN
986 nop
987 .exit
988 .procend
989 .end
990 #endif
992 #ifdef L_div_const
993 /* ROUTINE: $$divI_2
994 . $$divI_3 $$divU_3
995 . $$divI_4
996 . $$divI_5 $$divU_5
997 . $$divI_6 $$divU_6
998 . $$divI_7 $$divU_7
999 . $$divI_8
1000 . $$divI_9 $$divU_9
1001 . $$divI_10 $$divU_10
1002 .
1003 . $$divI_12 $$divU_12
1004 .
1005 . $$divI_14 $$divU_14
1006 . $$divI_15 $$divU_15
1007 . $$divI_16
1008 . $$divI_17 $$divU_17
1009 .
1010 . Divide by selected constants for single precision binary integers.
1012 INPUT REGISTERS:
1013 . arg0 == dividend
1014 . mrp == return pc
1015 . sr0 == return space when called externally
1017 OUTPUT REGISTERS:
1018 . arg0 = undefined
1019 . arg1 = undefined
1020 . ret1 = quotient
1022 OTHER REGISTERS AFFECTED:
1023 . r1 = undefined
1025 SIDE EFFECTS:
1026 . Causes a trap under the following conditions: NONE
1027 . Changes memory at the following places: NONE
1029 PERMISSIBLE CONTEXT:
1030 . Unwindable.
1031 . Does not create a stack frame.
1032 . Suitable for internal or external millicode.
1033 . Assumes the special millicode register conventions.
1035 DISCUSSION:
1036 . Calls other millicode routines using mrp: NONE
1037 . Calls other millicode routines: NONE */
1040 /* TRUNCATED DIVISION BY SMALL INTEGERS
1042 We are interested in q(x) = floor(x/y), where x >= 0 and y > 0
1043 (with y fixed).
1045 Let a = floor(z/y), for some choice of z. Note that z will be
1046 chosen so that division by z is cheap.
1048 Let r be the remainder(z/y). In other words, r = z - ay.
1050 Now, our method is to choose a value for b such that
1052 q'(x) = floor((ax+b)/z)
1054 is equal to q(x) over as large a range of x as possible. If the
1055 two are equal over a sufficiently large range, and if it is easy to
1056 form the product (ax), and it is easy to divide by z, then we can
1057 perform the division much faster than the general division algorithm.
1059 So, we want the following to be true:
1061 . For x in the following range:
1062 .
1063 . ky <= x < (k+1)y
1064 .
1065 . implies that
1066 .
1067 . k <= (ax+b)/z < (k+1)
1069 We want to determine b such that this is true for all k in the
1070 range {0..K} for some maximum K.
1072 Since (ax+b) is an increasing function of x, we can take each
1073 bound separately to determine the "best" value for b.
1075 (ax+b)/z < (k+1) implies
1077 (a((k+1)y-1)+b < (k+1)z implies
1079 b < a + (k+1)(z-ay) implies
1081 b < a + (k+1)r
1083 This needs to be true for all k in the range {0..K}. In
1084 particular, it is true for k = 0 and this leads to a maximum
1085 acceptable value for b.
1087 b < a+r or b <= a+r-1
1089 Taking the other bound, we have
1091 k <= (ax+b)/z implies
1093 k <= (aky+b)/z implies
1095 k(z-ay) <= b implies
1097 kr <= b
1099 Clearly, the largest range for k will be achieved by maximizing b,
1100 when r is not zero. When r is zero, then the simplest choice for b
1101 is 0. When r is not 0, set
1103 . b = a+r-1
1105 Now, by construction, q'(x) = floor((ax+b)/z) = q(x) = floor(x/y)
1106 for all x in the range:
1108 . 0 <= x < (K+1)y
1110 We need to determine what K is. Of our two bounds,
1112 . b < a+(k+1)r is satisfied for all k >= 0, by construction.
1114 The other bound is
1116 . kr <= b
1118 This is always true if r = 0. If r is not 0 (the usual case), then
1119 K = floor((a+r-1)/r), is the maximum value for k.
1121 Therefore, the formula q'(x) = floor((ax+b)/z) yields the correct
1122 answer for q(x) = floor(x/y) when x is in the range
1124 (0,(K+1)y-1) K = floor((a+r-1)/r)
1126 To be most useful, we want (K+1)y-1 = (max x) >= 2**32-1 so that
1127 the formula for q'(x) yields the correct value of q(x) for all x
1128 representable by a single word in HPPA.
1130 We are also constrained in that computing the product (ax), adding
1131 b, and dividing by z must all be done quickly, otherwise we will be
1132 better off going through the general algorithm using the DS
1133 instruction, which uses approximately 70 cycles.
1135 For each y, there is a choice of z which satisfies the constraints
1136 for (K+1)y >= 2**32. We may not, however, be able to satisfy the
1137 timing constraints for arbitrary y. It seems that z being equal to
1138 a power of 2 or a power of 2 minus 1 is as good as we can do, since
1139 it minimizes the time to do division by z. We want the choice of z
1140 to also result in a value for (a) that minimizes the computation of
1141 the product (ax). This is best achieved if (a) has a regular bit
1142 pattern (so the multiplication can be done with shifts and adds).
1143 The value of (a) also needs to be less than 2**32 so the product is
1144 always guaranteed to fit in 2 words.
1146 In actual practice, the following should be done:
1148 1) For negative x, you should take the absolute value and remember
1149 . the fact so that the result can be negated. This obviously does
1150 . not apply in the unsigned case.
1151 2) For even y, you should factor out the power of 2 that divides y
1152 . and divide x by it. You can then proceed by dividing by the
1153 . odd factor of y.
1155 Here is a table of some odd values of y, and corresponding choices
1156 for z which are "good".
1158 y z r a (hex) max x (hex)
1160 3 2**32 1 55555555 100000001
1161 5 2**32 1 33333333 100000003
1162 7 2**24-1 0 249249 (infinite)
1163 9 2**24-1 0 1c71c7 (infinite)
1164 11 2**20-1 0 1745d (infinite)
1165 13 2**24-1 0 13b13b (infinite)
1166 15 2**32 1 11111111 10000000d
1167 17 2**32 1 f0f0f0f 10000000f
1169 If r is 1, then b = a+r-1 = a. This simplifies the computation
1170 of (ax+b), since you can compute (x+1)(a) instead. If r is 0,
1171 then b = 0 is ok to use which simplifies (ax+b).
1173 The bit patterns for 55555555, 33333333, and 11111111 are obviously
1174 very regular. The bit patterns for the other values of a above are:
1176 y (hex) (binary)
1178 7 249249 001001001001001001001001 << regular >>
1179 9 1c71c7 000111000111000111000111 << regular >>
1180 11 1745d 000000010111010001011101 << irregular >>
1181 13 13b13b 000100111011000100111011 << irregular >>
1183 The bit patterns for (a) corresponding to (y) of 11 and 13 may be
1184 too irregular to warrant using this method.
1186 When z is a power of 2 minus 1, then the division by z is slightly
1187 more complicated, involving an iterative solution.
1189 The code presented here solves division by 1 through 17, except for
1190 11 and 13. There are algorithms for both signed and unsigned
1191 quantities given.
1193 TIMINGS (cycles)
1195 divisor positive negative unsigned
1197 . 1 2 2 2
1198 . 2 4 4 2
1199 . 3 19 21 19
1200 . 4 4 4 2
1201 . 5 18 22 19
1202 . 6 19 22 19
1203 . 8 4 4 2
1204 . 10 18 19 17
1205 . 12 18 20 18
1206 . 15 16 18 16
1207 . 16 4 4 2
1208 . 17 16 18 16
1210 Now, the algorithm for 7, 9, and 14 is an iterative one. That is,
1211 a loop body is executed until the tentative quotient is 0. The
1212 number of times the loop body is executed varies depending on the
1213 dividend, but is never more than two times. If the dividend is
1214 less than the divisor, then the loop body is not executed at all.
1215 Each iteration adds 4 cycles to the timings.
1217 divisor positive negative unsigned
1219 . 7 19+4n 20+4n 20+4n n = number of iterations
1220 . 9 21+4n 22+4n 21+4n
1221 . 14 21+4n 22+4n 20+4n
1223 To give an idea of how the number of iterations varies, here is a
1224 table of dividend versus number of iterations when dividing by 7.
1226 smallest largest required
1227 dividend dividend iterations
1229 . 0 6 0
1230 . 7 0x6ffffff 1
1231 0x1000006 0xffffffff 2
1233 There is some overlap in the range of numbers requiring 1 and 2
1234 iterations. */
1236 RDEFINE(t2,r1)
1237 RDEFINE(x2,arg0) /* r26 */
1238 RDEFINE(t1,arg1) /* r25 */
1239 RDEFINE(x1,ret1) /* r29 */
1241 SUBSPA_MILLI_DIV
1242 ATTR_MILLI
1244 .proc
1245 .callinfo millicode
1246 .entry
1247 /* NONE of these routines require a stack frame
1248 ALL of these routines are unwindable from millicode */
1250 GSYM($$divide_by_constant)
1251 .export $$divide_by_constant,millicode
1252 /* Provides a "nice" label for the code covered by the unwind descriptor
1253 for things like gprof. */
1255 /* DIVISION BY 2 (shift by 1) */
1256 GSYM($$divI_2)
1257 .export $$divI_2,millicode
1258 comclr,>= arg0,0,0
1259 addi 1,arg0,arg0
1260 MILLIRET
1261 extrs arg0,30,31,ret1
1264 /* DIVISION BY 4 (shift by 2) */
1265 GSYM($$divI_4)
1266 .export $$divI_4,millicode
1267 comclr,>= arg0,0,0
1268 addi 3,arg0,arg0
1269 MILLIRET
1270 extrs arg0,29,30,ret1
1273 /* DIVISION BY 8 (shift by 3) */
1274 GSYM($$divI_8)
1275 .export $$divI_8,millicode
1276 comclr,>= arg0,0,0
1277 addi 7,arg0,arg0
1278 MILLIRET
1279 extrs arg0,28,29,ret1
1281 /* DIVISION BY 16 (shift by 4) */
1282 GSYM($$divI_16)
1283 .export $$divI_16,millicode
1284 comclr,>= arg0,0,0
1285 addi 15,arg0,arg0
1286 MILLIRET
1287 extrs arg0,27,28,ret1
1289 /****************************************************************************
1290 *
1291 * DIVISION BY DIVISORS OF FFFFFFFF, and powers of 2 times these
1292 *
1293 * includes 3,5,15,17 and also 6,10,12
1294 *
1295 ****************************************************************************/
1297 /* DIVISION BY 3 (use z = 2**32; a = 55555555) */
1299 GSYM($$divI_3)
1300 .export $$divI_3,millicode
1301 comb,<,N x2,0,LREF(neg3)
1303 addi 1,x2,x2 /* this cannot overflow */
1304 extru x2,1,2,x1 /* multiply by 5 to get started */
1305 sh2add x2,x2,x2
1306 b LREF(pos)
1307 addc x1,0,x1
1309 LSYM(neg3)
1310 subi 1,x2,x2 /* this cannot overflow */
1311 extru x2,1,2,x1 /* multiply by 5 to get started */
1312 sh2add x2,x2,x2
1313 b LREF(neg)
1314 addc x1,0,x1
1316 GSYM($$divU_3)
1317 .export $$divU_3,millicode
1318 addi 1,x2,x2 /* this CAN overflow */
1319 addc 0,0,x1
1320 shd x1,x2,30,t1 /* multiply by 5 to get started */
1321 sh2add x2,x2,x2
1322 b LREF(pos)
1323 addc x1,t1,x1
1325 /* DIVISION BY 5 (use z = 2**32; a = 33333333) */
1327 GSYM($$divI_5)
1328 .export $$divI_5,millicode
1329 comb,<,N x2,0,LREF(neg5)
1331 addi 3,x2,t1 /* this cannot overflow */
1332 sh1add x2,t1,x2 /* multiply by 3 to get started */
1333 b LREF(pos)
1334 addc 0,0,x1
1336 LSYM(neg5)
1337 sub 0,x2,x2 /* negate x2 */
1338 addi 1,x2,x2 /* this cannot overflow */
1339 shd 0,x2,31,x1 /* get top bit (can be 1) */
1340 sh1add x2,x2,x2 /* multiply by 3 to get started */
1341 b LREF(neg)
1342 addc x1,0,x1
1344 GSYM($$divU_5)
1345 .export $$divU_5,millicode
1346 addi 1,x2,x2 /* this CAN overflow */
1347 addc 0,0,x1
1348 shd x1,x2,31,t1 /* multiply by 3 to get started */
1349 sh1add x2,x2,x2
1350 b LREF(pos)
1351 addc t1,x1,x1
1353 /* DIVISION BY 6 (shift to divide by 2 then divide by 3) */
1354 GSYM($$divI_6)
1355 .export $$divI_6,millicode
1356 comb,<,N x2,0,LREF(neg6)
1357 extru x2,30,31,x2 /* divide by 2 */
1358 addi 5,x2,t1 /* compute 5*(x2+1) = 5*x2+5 */
1359 sh2add x2,t1,x2 /* multiply by 5 to get started */
1360 b LREF(pos)
1361 addc 0,0,x1
1363 LSYM(neg6)
1364 subi 2,x2,x2 /* negate, divide by 2, and add 1 */
1365 /* negation and adding 1 are done */
1366 /* at the same time by the SUBI */
1367 extru x2,30,31,x2
1368 shd 0,x2,30,x1
1369 sh2add x2,x2,x2 /* multiply by 5 to get started */
1370 b LREF(neg)
1371 addc x1,0,x1
1373 GSYM($$divU_6)
1374 .export $$divU_6,millicode
1375 extru x2,30,31,x2 /* divide by 2 */
1376 addi 1,x2,x2 /* cannot carry */
1377 shd 0,x2,30,x1 /* multiply by 5 to get started */
1378 sh2add x2,x2,x2
1379 b LREF(pos)
1380 addc x1,0,x1
1382 /* DIVISION BY 10 (shift to divide by 2 then divide by 5) */
1383 GSYM($$divU_10)
1384 .export $$divU_10,millicode
1385 extru x2,30,31,x2 /* divide by 2 */
1386 addi 3,x2,t1 /* compute 3*(x2+1) = (3*x2)+3 */
1387 sh1add x2,t1,x2 /* multiply by 3 to get started */
1388 addc 0,0,x1
1389 LSYM(pos)
1390 shd x1,x2,28,t1 /* multiply by 0x11 */
1391 shd x2,0,28,t2
1392 add x2,t2,x2
1393 addc x1,t1,x1
1394 LSYM(pos_for_17)
1395 shd x1,x2,24,t1 /* multiply by 0x101 */
1396 shd x2,0,24,t2
1397 add x2,t2,x2
1398 addc x1,t1,x1
1400 shd x1,x2,16,t1 /* multiply by 0x10001 */
1401 shd x2,0,16,t2
1402 add x2,t2,x2
1403 MILLIRET
1404 addc x1,t1,x1
1406 GSYM($$divI_10)
1407 .export $$divI_10,millicode
1408 comb,< x2,0,LREF(neg10)
1409 copy 0,x1
1410 extru x2,30,31,x2 /* divide by 2 */
1411 addib,TR 1,x2,LREF(pos) /* add 1 (cannot overflow) */
1412 sh1add x2,x2,x2 /* multiply by 3 to get started */
1414 LSYM(neg10)
1415 subi 2,x2,x2 /* negate, divide by 2, and add 1 */
1416 /* negation and adding 1 are done */
1417 /* at the same time by the SUBI */
1418 extru x2,30,31,x2
1419 sh1add x2,x2,x2 /* multiply by 3 to get started */
1420 LSYM(neg)
1421 shd x1,x2,28,t1 /* multiply by 0x11 */
1422 shd x2,0,28,t2
1423 add x2,t2,x2
1424 addc x1,t1,x1
1425 LSYM(neg_for_17)
1426 shd x1,x2,24,t1 /* multiply by 0x101 */
1427 shd x2,0,24,t2
1428 add x2,t2,x2
1429 addc x1,t1,x1
1431 shd x1,x2,16,t1 /* multiply by 0x10001 */
1432 shd x2,0,16,t2
1433 add x2,t2,x2
1434 addc x1,t1,x1
1435 MILLIRET
1436 sub 0,x1,x1
1438 /* DIVISION BY 12 (shift to divide by 4 then divide by 3) */
1439 GSYM($$divI_12)
1440 .export $$divI_12,millicode
1441 comb,< x2,0,LREF(neg12)
1442 copy 0,x1
1443 extru x2,29,30,x2 /* divide by 4 */
1444 addib,tr 1,x2,LREF(pos) /* compute 5*(x2+1) = 5*x2+5 */
1445 sh2add x2,x2,x2 /* multiply by 5 to get started */
1447 LSYM(neg12)
1448 subi 4,x2,x2 /* negate, divide by 4, and add 1 */
1449 /* negation and adding 1 are done */
1450 /* at the same time by the SUBI */
1451 extru x2,29,30,x2
1452 b LREF(neg)
1453 sh2add x2,x2,x2 /* multiply by 5 to get started */
1455 GSYM($$divU_12)
1456 .export $$divU_12,millicode
1457 extru x2,29,30,x2 /* divide by 4 */
1458 addi 5,x2,t1 /* cannot carry */
1459 sh2add x2,t1,x2 /* multiply by 5 to get started */
1460 b LREF(pos)
1461 addc 0,0,x1
1463 /* DIVISION BY 15 (use z = 2**32; a = 11111111) */
1464 GSYM($$divI_15)
1465 .export $$divI_15,millicode
1466 comb,< x2,0,LREF(neg15)
1467 copy 0,x1
1468 addib,tr 1,x2,LREF(pos)+4
1469 shd x1,x2,28,t1
1471 LSYM(neg15)
1472 b LREF(neg)
1473 subi 1,x2,x2
1475 GSYM($$divU_15)
1476 .export $$divU_15,millicode
1477 addi 1,x2,x2 /* this CAN overflow */
1478 b LREF(pos)
1479 addc 0,0,x1
1481 /* DIVISION BY 17 (use z = 2**32; a = f0f0f0f) */
1482 GSYM($$divI_17)
1483 .export $$divI_17,millicode
1484 comb,<,n x2,0,LREF(neg17)
1485 addi 1,x2,x2 /* this cannot overflow */
1486 shd 0,x2,28,t1 /* multiply by 0xf to get started */
1487 shd x2,0,28,t2
1488 sub t2,x2,x2
1489 b LREF(pos_for_17)
1490 subb t1,0,x1
1492 LSYM(neg17)
1493 subi 1,x2,x2 /* this cannot overflow */
1494 shd 0,x2,28,t1 /* multiply by 0xf to get started */
1495 shd x2,0,28,t2
1496 sub t2,x2,x2
1497 b LREF(neg_for_17)
1498 subb t1,0,x1
1500 GSYM($$divU_17)
1501 .export $$divU_17,millicode
1502 addi 1,x2,x2 /* this CAN overflow */
1503 addc 0,0,x1
1504 shd x1,x2,28,t1 /* multiply by 0xf to get started */
1505 LSYM(u17)
1506 shd x2,0,28,t2
1507 sub t2,x2,x2
1508 b LREF(pos_for_17)
1509 subb t1,x1,x1
1512 /* DIVISION BY DIVISORS OF FFFFFF, and powers of 2 times these
1513 includes 7,9 and also 14
1516 z = 2**24-1
1517 r = z mod x = 0
1519 so choose b = 0
1521 Also, in order to divide by z = 2**24-1, we approximate by dividing
1522 by (z+1) = 2**24 (which is easy), and then correcting.
1524 (ax) = (z+1)q' + r
1525 . = zq' + (q'+r)
1527 So to compute (ax)/z, compute q' = (ax)/(z+1) and r = (ax) mod (z+1)
1528 Then the true remainder of (ax)/z is (q'+r). Repeat the process
1529 with this new remainder, adding the tentative quotients together,
1530 until a tentative quotient is 0 (and then we are done). There is
1531 one last correction to be done. It is possible that (q'+r) = z.
1532 If so, then (q'+r)/(z+1) = 0 and it looks like we are done. But,
1533 in fact, we need to add 1 more to the quotient. Now, it turns
1534 out that this happens if and only if the original value x is
1535 an exact multiple of y. So, to avoid a three instruction test at
1536 the end, instead use 1 instruction to add 1 to x at the beginning. */
1538 /* DIVISION BY 7 (use z = 2**24-1; a = 249249) */
1539 GSYM($$divI_7)
1540 .export $$divI_7,millicode
1541 comb,<,n x2,0,LREF(neg7)
1542 LSYM(7)
1543 addi 1,x2,x2 /* cannot overflow */
1544 shd 0,x2,29,x1
1545 sh3add x2,x2,x2
1546 addc x1,0,x1
1547 LSYM(pos7)
1548 shd x1,x2,26,t1
1549 shd x2,0,26,t2
1550 add x2,t2,x2
1551 addc x1,t1,x1
1553 shd x1,x2,20,t1
1554 shd x2,0,20,t2
1555 add x2,t2,x2
1556 addc x1,t1,t1
1558 /* computed <t1,x2>. Now divide it by (2**24 - 1) */
1560 copy 0,x1
1561 shd,= t1,x2,24,t1 /* tentative quotient */
1562 LSYM(1)
1563 addb,tr t1,x1,LREF(2) /* add to previous quotient */
1564 extru x2,31,24,x2 /* new remainder (unadjusted) */
1566 MILLIRETN
1568 LSYM(2)
1569 addb,tr t1,x2,LREF(1) /* adjust remainder */
1570 extru,= x2,7,8,t1 /* new quotient */
1572 LSYM(neg7)
1573 subi 1,x2,x2 /* negate x2 and add 1 */
1574 LSYM(8)
1575 shd 0,x2,29,x1
1576 sh3add x2,x2,x2
1577 addc x1,0,x1
1579 LSYM(neg7_shift)
1580 shd x1,x2,26,t1
1581 shd x2,0,26,t2
1582 add x2,t2,x2
1583 addc x1,t1,x1
1585 shd x1,x2,20,t1
1586 shd x2,0,20,t2
1587 add x2,t2,x2
1588 addc x1,t1,t1
1590 /* computed <t1,x2>. Now divide it by (2**24 - 1) */
1592 copy 0,x1
1593 shd,= t1,x2,24,t1 /* tentative quotient */
1594 LSYM(3)
1595 addb,tr t1,x1,LREF(4) /* add to previous quotient */
1596 extru x2,31,24,x2 /* new remainder (unadjusted) */
1598 MILLIRET
1599 sub 0,x1,x1 /* negate result */
1601 LSYM(4)
1602 addb,tr t1,x2,LREF(3) /* adjust remainder */
1603 extru,= x2,7,8,t1 /* new quotient */
1605 GSYM($$divU_7)
1606 .export $$divU_7,millicode
1607 addi 1,x2,x2 /* can carry */
1608 addc 0,0,x1
1609 shd x1,x2,29,t1
1610 sh3add x2,x2,x2
1611 b LREF(pos7)
1612 addc t1,x1,x1
1614 /* DIVISION BY 9 (use z = 2**24-1; a = 1c71c7) */
1615 GSYM($$divI_9)
1616 .export $$divI_9,millicode
1617 comb,<,n x2,0,LREF(neg9)
1618 addi 1,x2,x2 /* cannot overflow */
1619 shd 0,x2,29,t1
1620 shd x2,0,29,t2
1621 sub t2,x2,x2
1622 b LREF(pos7)
1623 subb t1,0,x1
1625 LSYM(neg9)
1626 subi 1,x2,x2 /* negate and add 1 */
1627 shd 0,x2,29,t1
1628 shd x2,0,29,t2
1629 sub t2,x2,x2
1630 b LREF(neg7_shift)
1631 subb t1,0,x1
1633 GSYM($$divU_9)
1634 .export $$divU_9,millicode
1635 addi 1,x2,x2 /* can carry */
1636 addc 0,0,x1
1637 shd x1,x2,29,t1
1638 shd x2,0,29,t2
1639 sub t2,x2,x2
1640 b LREF(pos7)
1641 subb t1,x1,x1
1643 /* DIVISION BY 14 (shift to divide by 2 then divide by 7) */
1644 GSYM($$divI_14)
1645 .export $$divI_14,millicode
1646 comb,<,n x2,0,LREF(neg14)
1647 GSYM($$divU_14)
1648 .export $$divU_14,millicode
1649 b LREF(7) /* go to 7 case */
1650 extru x2,30,31,x2 /* divide by 2 */
1652 LSYM(neg14)
1653 subi 2,x2,x2 /* negate (and add 2) */
1654 b LREF(8)
1655 extru x2,30,31,x2 /* divide by 2 */
1656 .exit
1657 .procend
1658 .end
1659 #endif
1661 #ifdef L_mulI
1662 /* VERSION "@(#)$$mulI $ Revision: 12.4 $ $ Date: 94/03/17 17:18:51 $" */
1663 /******************************************************************************
1664 This routine is used on PA2.0 processors when gcc -mno-fpregs is used
1666 ROUTINE: $$mulI
1669 DESCRIPTION:
1671 $$mulI multiplies two single word integers, giving a single
1672 word result.
1675 INPUT REGISTERS:
1677 arg0 = Operand 1
1678 arg1 = Operand 2
1679 r31 == return pc
1680 sr0 == return space when called externally
1683 OUTPUT REGISTERS:
1685 arg0 = undefined
1686 arg1 = undefined
1687 ret1 = result
1689 OTHER REGISTERS AFFECTED:
1691 r1 = undefined
1693 SIDE EFFECTS:
1695 Causes a trap under the following conditions: NONE
1696 Changes memory at the following places: NONE
1698 PERMISSIBLE CONTEXT:
1700 Unwindable
1701 Does not create a stack frame
1702 Is usable for internal or external microcode
1704 DISCUSSION:
1706 Calls other millicode routines via mrp: NONE
1707 Calls other millicode routines: NONE
1709 ***************************************************************************/
1712 #define a0 %arg0
1713 #define a1 %arg1
1714 #define t0 %r1
1715 #define r %ret1
1717 #define a0__128a0 zdep a0,24,25,a0
1718 #define a0__256a0 zdep a0,23,24,a0
1719 #define a1_ne_0_b_l0 comb,<> a1,0,LREF(l0)
1720 #define a1_ne_0_b_l1 comb,<> a1,0,LREF(l1)
1721 #define a1_ne_0_b_l2 comb,<> a1,0,LREF(l2)
1722 #define b_n_ret_t0 b,n LREF(ret_t0)
1723 #define b_e_shift b LREF(e_shift)
1724 #define b_e_t0ma0 b LREF(e_t0ma0)
1725 #define b_e_t0 b LREF(e_t0)
1726 #define b_e_t0a0 b LREF(e_t0a0)
1727 #define b_e_t02a0 b LREF(e_t02a0)
1728 #define b_e_t04a0 b LREF(e_t04a0)
1729 #define b_e_2t0 b LREF(e_2t0)
1730 #define b_e_2t0a0 b LREF(e_2t0a0)
1731 #define b_e_2t04a0 b LREF(e2t04a0)
1732 #define b_e_3t0 b LREF(e_3t0)
1733 #define b_e_4t0 b LREF(e_4t0)
1734 #define b_e_4t0a0 b LREF(e_4t0a0)
1735 #define b_e_4t08a0 b LREF(e4t08a0)
1736 #define b_e_5t0 b LREF(e_5t0)
1737 #define b_e_8t0 b LREF(e_8t0)
1738 #define b_e_8t0a0 b LREF(e_8t0a0)
1739 #define r__r_a0 add r,a0,r
1740 #define r__r_2a0 sh1add a0,r,r
1741 #define r__r_4a0 sh2add a0,r,r
1742 #define r__r_8a0 sh3add a0,r,r
1743 #define r__r_t0 add r,t0,r
1744 #define r__r_2t0 sh1add t0,r,r
1745 #define r__r_4t0 sh2add t0,r,r
1746 #define r__r_8t0 sh3add t0,r,r
1747 #define t0__3a0 sh1add a0,a0,t0
1748 #define t0__4a0 sh2add a0,0,t0
1749 #define t0__5a0 sh2add a0,a0,t0
1750 #define t0__8a0 sh3add a0,0,t0
1751 #define t0__9a0 sh3add a0,a0,t0
1752 #define t0__16a0 zdep a0,27,28,t0
1753 #define t0__32a0 zdep a0,26,27,t0
1754 #define t0__64a0 zdep a0,25,26,t0
1755 #define t0__128a0 zdep a0,24,25,t0
1756 #define t0__t0ma0 sub t0,a0,t0
1757 #define t0__t0_a0 add t0,a0,t0
1758 #define t0__t0_2a0 sh1add a0,t0,t0
1759 #define t0__t0_4a0 sh2add a0,t0,t0
1760 #define t0__t0_8a0 sh3add a0,t0,t0
1761 #define t0__2t0_a0 sh1add t0,a0,t0
1762 #define t0__3t0 sh1add t0,t0,t0
1763 #define t0__4t0 sh2add t0,0,t0
1764 #define t0__4t0_a0 sh2add t0,a0,t0
1765 #define t0__5t0 sh2add t0,t0,t0
1766 #define t0__8t0 sh3add t0,0,t0
1767 #define t0__8t0_a0 sh3add t0,a0,t0
1768 #define t0__9t0 sh3add t0,t0,t0
1769 #define t0__16t0 zdep t0,27,28,t0
1770 #define t0__32t0 zdep t0,26,27,t0
1771 #define t0__256a0 zdep a0,23,24,t0
1774 SUBSPA_MILLI
1775 ATTR_MILLI
1776 .align 16
1777 .proc
1778 .callinfo millicode
1779 .export $$mulI,millicode
1780 GSYM($$mulI)
1781 combt,<<= a1,a0,LREF(l4) /* swap args if unsigned a1>a0 */
1782 copy 0,r /* zero out the result */
1783 xor a0,a1,a0 /* swap a0 & a1 using the */
1784 xor a0,a1,a1 /* old xor trick */
1785 xor a0,a1,a0
1786 LSYM(l4)
1787 combt,<= 0,a0,LREF(l3) /* if a0>=0 then proceed like unsigned */
1788 zdep a1,30,8,t0 /* t0 = (a1&0xff)<<1 ********* */
1789 sub,> 0,a1,t0 /* otherwise negate both and */
1790 combt,<=,n a0,t0,LREF(l2) /* swap back if |a0|<|a1| */
1791 sub 0,a0,a1
1792 movb,tr,n t0,a0,LREF(l2) /* 10th inst. */
1794 LSYM(l0) r__r_t0 /* add in this partial product */
1795 LSYM(l1) a0__256a0 /* a0 <<= 8 ****************** */
1796 LSYM(l2) zdep a1,30,8,t0 /* t0 = (a1&0xff)<<1 ********* */
1797 LSYM(l3) blr t0,0 /* case on these 8 bits ****** */
1798 extru a1,23,24,a1 /* a1 >>= 8 ****************** */
1800 /*16 insts before this. */
1801 /* a0 <<= 8 ************************** */
1802 LSYM(x0) a1_ne_0_b_l2 ! a0__256a0 ! MILLIRETN ! nop
1803 LSYM(x1) a1_ne_0_b_l1 ! r__r_a0 ! MILLIRETN ! nop
1804 LSYM(x2) a1_ne_0_b_l1 ! r__r_2a0 ! MILLIRETN ! nop
1805 LSYM(x3) a1_ne_0_b_l0 ! t0__3a0 ! MILLIRET ! r__r_t0
1806 LSYM(x4) a1_ne_0_b_l1 ! r__r_4a0 ! MILLIRETN ! nop
1807 LSYM(x5) a1_ne_0_b_l0 ! t0__5a0 ! MILLIRET ! r__r_t0
1808 LSYM(x6) t0__3a0 ! a1_ne_0_b_l1 ! r__r_2t0 ! MILLIRETN
1809 LSYM(x7) t0__3a0 ! a1_ne_0_b_l0 ! r__r_4a0 ! b_n_ret_t0
1810 LSYM(x8) a1_ne_0_b_l1 ! r__r_8a0 ! MILLIRETN ! nop
1811 LSYM(x9) a1_ne_0_b_l0 ! t0__9a0 ! MILLIRET ! r__r_t0
1812 LSYM(x10) t0__5a0 ! a1_ne_0_b_l1 ! r__r_2t0 ! MILLIRETN
1813 LSYM(x11) t0__3a0 ! a1_ne_0_b_l0 ! r__r_8a0 ! b_n_ret_t0
1814 LSYM(x12) t0__3a0 ! a1_ne_0_b_l1 ! r__r_4t0 ! MILLIRETN
1815 LSYM(x13) t0__5a0 ! a1_ne_0_b_l0 ! r__r_8a0 ! b_n_ret_t0
1816 LSYM(x14) t0__3a0 ! t0__2t0_a0 ! b_e_shift ! r__r_2t0
1817 LSYM(x15) t0__5a0 ! a1_ne_0_b_l0 ! t0__3t0 ! b_n_ret_t0
1818 LSYM(x16) t0__16a0 ! a1_ne_0_b_l1 ! r__r_t0 ! MILLIRETN
1819 LSYM(x17) t0__9a0 ! a1_ne_0_b_l0 ! t0__t0_8a0 ! b_n_ret_t0
1820 LSYM(x18) t0__9a0 ! a1_ne_0_b_l1 ! r__r_2t0 ! MILLIRETN
1821 LSYM(x19) t0__9a0 ! a1_ne_0_b_l0 ! t0__2t0_a0 ! b_n_ret_t0
1822 LSYM(x20) t0__5a0 ! a1_ne_0_b_l1 ! r__r_4t0 ! MILLIRETN
1823 LSYM(x21) t0__5a0 ! a1_ne_0_b_l0 ! t0__4t0_a0 ! b_n_ret_t0
1824 LSYM(x22) t0__5a0 ! t0__2t0_a0 ! b_e_shift ! r__r_2t0
1825 LSYM(x23) t0__5a0 ! t0__2t0_a0 ! b_e_t0 ! t0__2t0_a0
1826 LSYM(x24) t0__3a0 ! a1_ne_0_b_l1 ! r__r_8t0 ! MILLIRETN
1827 LSYM(x25) t0__5a0 ! a1_ne_0_b_l0 ! t0__5t0 ! b_n_ret_t0
1828 LSYM(x26) t0__3a0 ! t0__4t0_a0 ! b_e_shift ! r__r_2t0
1829 LSYM(x27) t0__3a0 ! a1_ne_0_b_l0 ! t0__9t0 ! b_n_ret_t0
1830 LSYM(x28) t0__3a0 ! t0__2t0_a0 ! b_e_shift ! r__r_4t0
1831 LSYM(x29) t0__3a0 ! t0__2t0_a0 ! b_e_t0 ! t0__4t0_a0
1832 LSYM(x30) t0__5a0 ! t0__3t0 ! b_e_shift ! r__r_2t0
1833 LSYM(x31) t0__32a0 ! a1_ne_0_b_l0 ! t0__t0ma0 ! b_n_ret_t0
1834 LSYM(x32) t0__32a0 ! a1_ne_0_b_l1 ! r__r_t0 ! MILLIRETN
1835 LSYM(x33) t0__8a0 ! a1_ne_0_b_l0 ! t0__4t0_a0 ! b_n_ret_t0
1836 LSYM(x34) t0__16a0 ! t0__t0_a0 ! b_e_shift ! r__r_2t0
1837 LSYM(x35) t0__9a0 ! t0__3t0 ! b_e_t0 ! t0__t0_8a0
1838 LSYM(x36) t0__9a0 ! a1_ne_0_b_l1 ! r__r_4t0 ! MILLIRETN
1839 LSYM(x37) t0__9a0 ! a1_ne_0_b_l0 ! t0__4t0_a0 ! b_n_ret_t0
1840 LSYM(x38) t0__9a0 ! t0__2t0_a0 ! b_e_shift ! r__r_2t0
1841 LSYM(x39) t0__9a0 ! t0__2t0_a0 ! b_e_t0 ! t0__2t0_a0
1842 LSYM(x40) t0__5a0 ! a1_ne_0_b_l1 ! r__r_8t0 ! MILLIRETN
1843 LSYM(x41) t0__5a0 ! a1_ne_0_b_l0 ! t0__8t0_a0 ! b_n_ret_t0
1844 LSYM(x42) t0__5a0 ! t0__4t0_a0 ! b_e_shift ! r__r_2t0
1845 LSYM(x43) t0__5a0 ! t0__4t0_a0 ! b_e_t0 ! t0__2t0_a0
1846 LSYM(x44) t0__5a0 ! t0__2t0_a0 ! b_e_shift ! r__r_4t0
1847 LSYM(x45) t0__9a0 ! a1_ne_0_b_l0 ! t0__5t0 ! b_n_ret_t0
1848 LSYM(x46) t0__9a0 ! t0__5t0 ! b_e_t0 ! t0__t0_a0
1849 LSYM(x47) t0__9a0 ! t0__5t0 ! b_e_t0 ! t0__t0_2a0
1850 LSYM(x48) t0__3a0 ! a1_ne_0_b_l0 ! t0__16t0 ! b_n_ret_t0
1851 LSYM(x49) t0__9a0 ! t0__5t0 ! b_e_t0 ! t0__t0_4a0
1852 LSYM(x50) t0__5a0 ! t0__5t0 ! b_e_shift ! r__r_2t0
1853 LSYM(x51) t0__9a0 ! t0__t0_8a0 ! b_e_t0 ! t0__3t0
1854 LSYM(x52) t0__3a0 ! t0__4t0_a0 ! b_e_shift ! r__r_4t0
1855 LSYM(x53) t0__3a0 ! t0__4t0_a0 ! b_e_t0 ! t0__4t0_a0
1856 LSYM(x54) t0__9a0 ! t0__3t0 ! b_e_shift ! r__r_2t0
1857 LSYM(x55) t0__9a0 ! t0__3t0 ! b_e_t0 ! t0__2t0_a0
1858 LSYM(x56) t0__3a0 ! t0__2t0_a0 ! b_e_shift ! r__r_8t0
1859 LSYM(x57) t0__9a0 ! t0__2t0_a0 ! b_e_t0 ! t0__3t0
1860 LSYM(x58) t0__3a0 ! t0__2t0_a0 ! b_e_2t0 ! t0__4t0_a0
1861 LSYM(x59) t0__9a0 ! t0__2t0_a0 ! b_e_t02a0 ! t0__3t0
1862 LSYM(x60) t0__5a0 ! t0__3t0 ! b_e_shift ! r__r_4t0
1863 LSYM(x61) t0__5a0 ! t0__3t0 ! b_e_t0 ! t0__4t0_a0
1864 LSYM(x62) t0__32a0 ! t0__t0ma0 ! b_e_shift ! r__r_2t0
1865 LSYM(x63) t0__64a0 ! a1_ne_0_b_l0 ! t0__t0ma0 ! b_n_ret_t0
1866 LSYM(x64) t0__64a0 ! a1_ne_0_b_l1 ! r__r_t0 ! MILLIRETN
1867 LSYM(x65) t0__8a0 ! a1_ne_0_b_l0 ! t0__8t0_a0 ! b_n_ret_t0
1868 LSYM(x66) t0__32a0 ! t0__t0_a0 ! b_e_shift ! r__r_2t0
1869 LSYM(x67) t0__8a0 ! t0__4t0_a0 ! b_e_t0 ! t0__2t0_a0
1870 LSYM(x68) t0__8a0 ! t0__2t0_a0 ! b_e_shift ! r__r_4t0
1871 LSYM(x69) t0__8a0 ! t0__2t0_a0 ! b_e_t0 ! t0__4t0_a0
1872 LSYM(x70) t0__64a0 ! t0__t0_4a0 ! b_e_t0 ! t0__t0_2a0
1873 LSYM(x71) t0__9a0 ! t0__8t0 ! b_e_t0 ! t0__t0ma0
1874 LSYM(x72) t0__9a0 ! a1_ne_0_b_l1 ! r__r_8t0 ! MILLIRETN
1875 LSYM(x73) t0__9a0 ! t0__8t0_a0 ! b_e_shift ! r__r_t0
1876 LSYM(x74) t0__9a0 ! t0__4t0_a0 ! b_e_shift ! r__r_2t0
1877 LSYM(x75) t0__9a0 ! t0__4t0_a0 ! b_e_t0 ! t0__2t0_a0
1878 LSYM(x76) t0__9a0 ! t0__2t0_a0 ! b_e_shift ! r__r_4t0
1879 LSYM(x77) t0__9a0 ! t0__2t0_a0 ! b_e_t0 ! t0__4t0_a0
1880 LSYM(x78) t0__9a0 ! t0__2t0_a0 ! b_e_2t0 ! t0__2t0_a0
1881 LSYM(x79) t0__16a0 ! t0__5t0 ! b_e_t0 ! t0__t0ma0
1882 LSYM(x80) t0__16a0 ! t0__5t0 ! b_e_shift ! r__r_t0
1883 LSYM(x81) t0__9a0 ! t0__9t0 ! b_e_shift ! r__r_t0
1884 LSYM(x82) t0__5a0 ! t0__8t0_a0 ! b_e_shift ! r__r_2t0
1885 LSYM(x83) t0__5a0 ! t0__8t0_a0 ! b_e_t0 ! t0__2t0_a0
1886 LSYM(x84) t0__5a0 ! t0__4t0_a0 ! b_e_shift ! r__r_4t0
1887 LSYM(x85) t0__8a0 ! t0__2t0_a0 ! b_e_t0 ! t0__5t0
1888 LSYM(x86) t0__5a0 ! t0__4t0_a0 ! b_e_2t0 ! t0__2t0_a0
1889 LSYM(x87) t0__9a0 ! t0__9t0 ! b_e_t02a0 ! t0__t0_4a0
1890 LSYM(x88) t0__5a0 ! t0__2t0_a0 ! b_e_shift ! r__r_8t0
1891 LSYM(x89) t0__5a0 ! t0__2t0_a0 ! b_e_t0 ! t0__8t0_a0
1892 LSYM(x90) t0__9a0 ! t0__5t0 ! b_e_shift ! r__r_2t0
1893 LSYM(x91) t0__9a0 ! t0__5t0 ! b_e_t0 ! t0__2t0_a0
1894 LSYM(x92) t0__5a0 ! t0__2t0_a0 ! b_e_4t0 ! t0__2t0_a0
1895 LSYM(x93) t0__32a0 ! t0__t0ma0 ! b_e_t0 ! t0__3t0
1896 LSYM(x94) t0__9a0 ! t0__5t0 ! b_e_2t0 ! t0__t0_2a0
1897 LSYM(x95) t0__9a0 ! t0__2t0_a0 ! b_e_t0 ! t0__5t0
1898 LSYM(x96) t0__8a0 ! t0__3t0 ! b_e_shift ! r__r_4t0
1899 LSYM(x97) t0__8a0 ! t0__3t0 ! b_e_t0 ! t0__4t0_a0
1900 LSYM(x98) t0__32a0 ! t0__3t0 ! b_e_t0 ! t0__t0_2a0
1901 LSYM(x99) t0__8a0 ! t0__4t0_a0 ! b_e_t0 ! t0__3t0
1902 LSYM(x100) t0__5a0 ! t0__5t0 ! b_e_shift ! r__r_4t0
1903 LSYM(x101) t0__5a0 ! t0__5t0 ! b_e_t0 ! t0__4t0_a0
1904 LSYM(x102) t0__32a0 ! t0__t0_2a0 ! b_e_t0 ! t0__3t0
1905 LSYM(x103) t0__5a0 ! t0__5t0 ! b_e_t02a0 ! t0__4t0_a0
1906 LSYM(x104) t0__3a0 ! t0__4t0_a0 ! b_e_shift ! r__r_8t0
1907 LSYM(x105) t0__5a0 ! t0__4t0_a0 ! b_e_t0 ! t0__5t0
1908 LSYM(x106) t0__3a0 ! t0__4t0_a0 ! b_e_2t0 ! t0__4t0_a0
1909 LSYM(x107) t0__9a0 ! t0__t0_4a0 ! b_e_t02a0 ! t0__8t0_a0
1910 LSYM(x108) t0__9a0 ! t0__3t0 ! b_e_shift ! r__r_4t0
1911 LSYM(x109) t0__9a0 ! t0__3t0 ! b_e_t0 ! t0__4t0_a0
1912 LSYM(x110) t0__9a0 ! t0__3t0 ! b_e_2t0 ! t0__2t0_a0
1913 LSYM(x111) t0__9a0 ! t0__4t0_a0 ! b_e_t0 ! t0__3t0
1914 LSYM(x112) t0__3a0 ! t0__2t0_a0 ! b_e_t0 ! t0__16t0
1915 LSYM(x113) t0__9a0 ! t0__4t0_a0 ! b_e_t02a0 ! t0__3t0
1916 LSYM(x114) t0__9a0 ! t0__2t0_a0 ! b_e_2t0 ! t0__3t0
1917 LSYM(x115) t0__9a0 ! t0__2t0_a0 ! b_e_2t0a0 ! t0__3t0
1918 LSYM(x116) t0__3a0 ! t0__2t0_a0 ! b_e_4t0 ! t0__4t0_a0
1919 LSYM(x117) t0__3a0 ! t0__4t0_a0 ! b_e_t0 ! t0__9t0
1920 LSYM(x118) t0__3a0 ! t0__4t0_a0 ! b_e_t0a0 ! t0__9t0
1921 LSYM(x119) t0__3a0 ! t0__4t0_a0 ! b_e_t02a0 ! t0__9t0
1922 LSYM(x120) t0__5a0 ! t0__3t0 ! b_e_shift ! r__r_8t0
1923 LSYM(x121) t0__5a0 ! t0__3t0 ! b_e_t0 ! t0__8t0_a0
1924 LSYM(x122) t0__5a0 ! t0__3t0 ! b_e_2t0 ! t0__4t0_a0
1925 LSYM(x123) t0__5a0 ! t0__8t0_a0 ! b_e_t0 ! t0__3t0
1926 LSYM(x124) t0__32a0 ! t0__t0ma0 ! b_e_shift ! r__r_4t0
1927 LSYM(x125) t0__5a0 ! t0__5t0 ! b_e_t0 ! t0__5t0
1928 LSYM(x126) t0__64a0 ! t0__t0ma0 ! b_e_shift ! r__r_2t0
1929 LSYM(x127) t0__128a0 ! a1_ne_0_b_l0 ! t0__t0ma0 ! b_n_ret_t0
1930 LSYM(x128) t0__128a0 ! a1_ne_0_b_l1 ! r__r_t0 ! MILLIRETN
1931 LSYM(x129) t0__128a0 ! a1_ne_0_b_l0 ! t0__t0_a0 ! b_n_ret_t0
1932 LSYM(x130) t0__64a0 ! t0__t0_a0 ! b_e_shift ! r__r_2t0
1933 LSYM(x131) t0__8a0 ! t0__8t0_a0 ! b_e_t0 ! t0__2t0_a0
1934 LSYM(x132) t0__8a0 ! t0__4t0_a0 ! b_e_shift ! r__r_4t0
1935 LSYM(x133) t0__8a0 ! t0__4t0_a0 ! b_e_t0 ! t0__4t0_a0
1936 LSYM(x134) t0__8a0 ! t0__4t0_a0 ! b_e_2t0 ! t0__2t0_a0
1937 LSYM(x135) t0__9a0 ! t0__5t0 ! b_e_t0 ! t0__3t0
1938 LSYM(x136) t0__8a0 ! t0__2t0_a0 ! b_e_shift ! r__r_8t0
1939 LSYM(x137) t0__8a0 ! t0__2t0_a0 ! b_e_t0 ! t0__8t0_a0
1940 LSYM(x138) t0__8a0 ! t0__2t0_a0 ! b_e_2t0 ! t0__4t0_a0
1941 LSYM(x139) t0__8a0 ! t0__2t0_a0 ! b_e_2t0a0 ! t0__4t0_a0
1942 LSYM(x140) t0__3a0 ! t0__2t0_a0 ! b_e_4t0 ! t0__5t0
1943 LSYM(x141) t0__8a0 ! t0__2t0_a0 ! b_e_4t0a0 ! t0__2t0_a0
1944 LSYM(x142) t0__9a0 ! t0__8t0 ! b_e_2t0 ! t0__t0ma0
1945 LSYM(x143) t0__16a0 ! t0__9t0 ! b_e_t0 ! t0__t0ma0
1946 LSYM(x144) t0__9a0 ! t0__8t0 ! b_e_shift ! r__r_2t0
1947 LSYM(x145) t0__9a0 ! t0__8t0 ! b_e_t0 ! t0__2t0_a0
1948 LSYM(x146) t0__9a0 ! t0__8t0_a0 ! b_e_shift ! r__r_2t0
1949 LSYM(x147) t0__9a0 ! t0__8t0_a0 ! b_e_t0 ! t0__2t0_a0
1950 LSYM(x148) t0__9a0 ! t0__4t0_a0 ! b_e_shift ! r__r_4t0
1951 LSYM(x149) t0__9a0 ! t0__4t0_a0 ! b_e_t0 ! t0__4t0_a0
1952 LSYM(x150) t0__9a0 ! t0__4t0_a0 ! b_e_2t0 ! t0__2t0_a0
1953 LSYM(x151) t0__9a0 ! t0__4t0_a0 ! b_e_2t0a0 ! t0__2t0_a0
1954 LSYM(x152) t0__9a0 ! t0__2t0_a0 ! b_e_shift ! r__r_8t0
1955 LSYM(x153) t0__9a0 ! t0__2t0_a0 ! b_e_t0 ! t0__8t0_a0
1956 LSYM(x154) t0__9a0 ! t0__2t0_a0 ! b_e_2t0 ! t0__4t0_a0
1957 LSYM(x155) t0__32a0 ! t0__t0ma0 ! b_e_t0 ! t0__5t0
1958 LSYM(x156) t0__9a0 ! t0__2t0_a0 ! b_e_4t0 ! t0__2t0_a0
1959 LSYM(x157) t0__32a0 ! t0__t0ma0 ! b_e_t02a0 ! t0__5t0
1960 LSYM(x158) t0__16a0 ! t0__5t0 ! b_e_2t0 ! t0__t0ma0
1961 LSYM(x159) t0__32a0 ! t0__5t0 ! b_e_t0 ! t0__t0ma0
1962 LSYM(x160) t0__5a0 ! t0__4t0 ! b_e_shift ! r__r_8t0
1963 LSYM(x161) t0__8a0 ! t0__5t0 ! b_e_t0 ! t0__4t0_a0
1964 LSYM(x162) t0__9a0 ! t0__9t0 ! b_e_shift ! r__r_2t0
1965 LSYM(x163) t0__9a0 ! t0__9t0 ! b_e_t0 ! t0__2t0_a0
1966 LSYM(x164) t0__5a0 ! t0__8t0_a0 ! b_e_shift ! r__r_4t0
1967 LSYM(x165) t0__8a0 ! t0__4t0_a0 ! b_e_t0 ! t0__5t0
1968 LSYM(x166) t0__5a0 ! t0__8t0_a0 ! b_e_2t0 ! t0__2t0_a0
1969 LSYM(x167) t0__5a0 ! t0__8t0_a0 ! b_e_2t0a0 ! t0__2t0_a0
1970 LSYM(x168) t0__5a0 ! t0__4t0_a0 ! b_e_shift ! r__r_8t0
1971 LSYM(x169) t0__5a0 ! t0__4t0_a0 ! b_e_t0 ! t0__8t0_a0
1972 LSYM(x170) t0__32a0 ! t0__t0_2a0 ! b_e_t0 ! t0__5t0
1973 LSYM(x171) t0__9a0 ! t0__2t0_a0 ! b_e_t0 ! t0__9t0
1974 LSYM(x172) t0__5a0 ! t0__4t0_a0 ! b_e_4t0 ! t0__2t0_a0
1975 LSYM(x173) t0__9a0 ! t0__2t0_a0 ! b_e_t02a0 ! t0__9t0
1976 LSYM(x174) t0__32a0 ! t0__t0_2a0 ! b_e_t04a0 ! t0__5t0
1977 LSYM(x175) t0__8a0 ! t0__2t0_a0 ! b_e_5t0 ! t0__2t0_a0
1978 LSYM(x176) t0__5a0 ! t0__4t0_a0 ! b_e_8t0 ! t0__t0_a0
1979 LSYM(x177) t0__5a0 ! t0__4t0_a0 ! b_e_8t0a0 ! t0__t0_a0
1980 LSYM(x178) t0__5a0 ! t0__2t0_a0 ! b_e_2t0 ! t0__8t0_a0
1981 LSYM(x179) t0__5a0 ! t0__2t0_a0 ! b_e_2t0a0 ! t0__8t0_a0
1982 LSYM(x180) t0__9a0 ! t0__5t0 ! b_e_shift ! r__r_4t0
1983 LSYM(x181) t0__9a0 ! t0__5t0 ! b_e_t0 ! t0__4t0_a0
1984 LSYM(x182) t0__9a0 ! t0__5t0 ! b_e_2t0 ! t0__2t0_a0
1985 LSYM(x183) t0__9a0 ! t0__5t0 ! b_e_2t0a0 ! t0__2t0_a0
1986 LSYM(x184) t0__5a0 ! t0__9t0 ! b_e_4t0 ! t0__t0_a0
1987 LSYM(x185) t0__9a0 ! t0__4t0_a0 ! b_e_t0 ! t0__5t0
1988 LSYM(x186) t0__32a0 ! t0__t0ma0 ! b_e_2t0 ! t0__3t0
1989 LSYM(x187) t0__9a0 ! t0__4t0_a0 ! b_e_t02a0 ! t0__5t0
1990 LSYM(x188) t0__9a0 ! t0__5t0 ! b_e_4t0 ! t0__t0_2a0
1991 LSYM(x189) t0__5a0 ! t0__4t0_a0 ! b_e_t0 ! t0__9t0
1992 LSYM(x190) t0__9a0 ! t0__2t0_a0 ! b_e_2t0 ! t0__5t0
1993 LSYM(x191) t0__64a0 ! t0__3t0 ! b_e_t0 ! t0__t0ma0
1994 LSYM(x192) t0__8a0 ! t0__3t0 ! b_e_shift ! r__r_8t0
1995 LSYM(x193) t0__8a0 ! t0__3t0 ! b_e_t0 ! t0__8t0_a0
1996 LSYM(x194) t0__8a0 ! t0__3t0 ! b_e_2t0 ! t0__4t0_a0
1997 LSYM(x195) t0__8a0 ! t0__8t0_a0 ! b_e_t0 ! t0__3t0
1998 LSYM(x196) t0__8a0 ! t0__3t0 ! b_e_4t0 ! t0__2t0_a0
1999 LSYM(x197) t0__8a0 ! t0__3t0 ! b_e_4t0a0 ! t0__2t0_a0
2000 LSYM(x198) t0__64a0 ! t0__t0_2a0 ! b_e_t0 ! t0__3t0
2001 LSYM(x199) t0__8a0 ! t0__4t0_a0 ! b_e_2t0a0 ! t0__3t0
2002 LSYM(x200) t0__5a0 ! t0__5t0 ! b_e_shift ! r__r_8t0
2003 LSYM(x201) t0__5a0 ! t0__5t0 ! b_e_t0 ! t0__8t0_a0
2004 LSYM(x202) t0__5a0 ! t0__5t0 ! b_e_2t0 ! t0__4t0_a0
2005 LSYM(x203) t0__5a0 ! t0__5t0 ! b_e_2t0a0 ! t0__4t0_a0
2006 LSYM(x204) t0__8a0 ! t0__2t0_a0 ! b_e_4t0 ! t0__3t0
2007 LSYM(x205) t0__5a0 ! t0__8t0_a0 ! b_e_t0 ! t0__5t0
2008 LSYM(x206) t0__64a0 ! t0__t0_4a0 ! b_e_t02a0 ! t0__3t0
2009 LSYM(x207) t0__8a0 ! t0__2t0_a0 ! b_e_3t0 ! t0__4t0_a0
2010 LSYM(x208) t0__5a0 ! t0__5t0 ! b_e_8t0 ! t0__t0_a0
2011 LSYM(x209) t0__5a0 ! t0__5t0 ! b_e_8t0a0 ! t0__t0_a0
2012 LSYM(x210) t0__5a0 ! t0__4t0_a0 ! b_e_2t0 ! t0__5t0
2013 LSYM(x211) t0__5a0 ! t0__4t0_a0 ! b_e_2t0a0 ! t0__5t0
2014 LSYM(x212) t0__3a0 ! t0__4t0_a0 ! b_e_4t0 ! t0__4t0_a0
2015 LSYM(x213) t0__3a0 ! t0__4t0_a0 ! b_e_4t0a0 ! t0__4t0_a0
2016 LSYM(x214) t0__9a0 ! t0__t0_4a0 ! b_e_2t04a0 ! t0__8t0_a0
2017 LSYM(x215) t0__5a0 ! t0__4t0_a0 ! b_e_5t0 ! t0__2t0_a0
2018 LSYM(x216) t0__9a0 ! t0__3t0 ! b_e_shift ! r__r_8t0
2019 LSYM(x217) t0__9a0 ! t0__3t0 ! b_e_t0 ! t0__8t0_a0
2020 LSYM(x218) t0__9a0 ! t0__3t0 ! b_e_2t0 ! t0__4t0_a0
2021 LSYM(x219) t0__9a0 ! t0__8t0_a0 ! b_e_t0 ! t0__3t0
2022 LSYM(x220) t0__3a0 ! t0__9t0 ! b_e_4t0 ! t0__2t0_a0
2023 LSYM(x221) t0__3a0 ! t0__9t0 ! b_e_4t0a0 ! t0__2t0_a0
2024 LSYM(x222) t0__9a0 ! t0__4t0_a0 ! b_e_2t0 ! t0__3t0
2025 LSYM(x223) t0__9a0 ! t0__4t0_a0 ! b_e_2t0a0 ! t0__3t0
2026 LSYM(x224) t0__9a0 ! t0__3t0 ! b_e_8t0 ! t0__t0_a0
2027 LSYM(x225) t0__9a0 ! t0__5t0 ! b_e_t0 ! t0__5t0
2028 LSYM(x226) t0__3a0 ! t0__2t0_a0 ! b_e_t02a0 ! t0__32t0
2029 LSYM(x227) t0__9a0 ! t0__5t0 ! b_e_t02a0 ! t0__5t0
2030 LSYM(x228) t0__9a0 ! t0__2t0_a0 ! b_e_4t0 ! t0__3t0
2031 LSYM(x229) t0__9a0 ! t0__2t0_a0 ! b_e_4t0a0 ! t0__3t0
2032 LSYM(x230) t0__9a0 ! t0__5t0 ! b_e_5t0 ! t0__t0_a0
2033 LSYM(x231) t0__9a0 ! t0__2t0_a0 ! b_e_3t0 ! t0__4t0_a0
2034 LSYM(x232) t0__3a0 ! t0__2t0_a0 ! b_e_8t0 ! t0__4t0_a0
2035 LSYM(x233) t0__3a0 ! t0__2t0_a0 ! b_e_8t0a0 ! t0__4t0_a0
2036 LSYM(x234) t0__3a0 ! t0__4t0_a0 ! b_e_2t0 ! t0__9t0
2037 LSYM(x235) t0__3a0 ! t0__4t0_a0 ! b_e_2t0a0 ! t0__9t0
2038 LSYM(x236) t0__9a0 ! t0__2t0_a0 ! b_e_4t08a0 ! t0__3t0
2039 LSYM(x237) t0__16a0 ! t0__5t0 ! b_e_3t0 ! t0__t0ma0
2040 LSYM(x238) t0__3a0 ! t0__4t0_a0 ! b_e_2t04a0 ! t0__9t0
2041 LSYM(x239) t0__16a0 ! t0__5t0 ! b_e_t0ma0 ! t0__3t0
2042 LSYM(x240) t0__9a0 ! t0__t0_a0 ! b_e_8t0 ! t0__3t0
2043 LSYM(x241) t0__9a0 ! t0__t0_a0 ! b_e_8t0a0 ! t0__3t0
2044 LSYM(x242) t0__5a0 ! t0__3t0 ! b_e_2t0 ! t0__8t0_a0
2045 LSYM(x243) t0__9a0 ! t0__9t0 ! b_e_t0 ! t0__3t0
2046 LSYM(x244) t0__5a0 ! t0__3t0 ! b_e_4t0 ! t0__4t0_a0
2047 LSYM(x245) t0__8a0 ! t0__3t0 ! b_e_5t0 ! t0__2t0_a0
2048 LSYM(x246) t0__5a0 ! t0__8t0_a0 ! b_e_2t0 ! t0__3t0
2049 LSYM(x247) t0__5a0 ! t0__8t0_a0 ! b_e_2t0a0 ! t0__3t0
2050 LSYM(x248) t0__32a0 ! t0__t0ma0 ! b_e_shift ! r__r_8t0
2051 LSYM(x249) t0__32a0 ! t0__t0ma0 ! b_e_t0 ! t0__8t0_a0
2052 LSYM(x250) t0__5a0 ! t0__5t0 ! b_e_2t0 ! t0__5t0
2053 LSYM(x251) t0__5a0 ! t0__5t0 ! b_e_2t0a0 ! t0__5t0
2054 LSYM(x252) t0__64a0 ! t0__t0ma0 ! b_e_shift ! r__r_4t0
2055 LSYM(x253) t0__64a0 ! t0__t0ma0 ! b_e_t0 ! t0__4t0_a0
2056 LSYM(x254) t0__128a0 ! t0__t0ma0 ! b_e_shift ! r__r_2t0
2057 LSYM(x255) t0__256a0 ! a1_ne_0_b_l0 ! t0__t0ma0 ! b_n_ret_t0
2058 /*1040 insts before this. */
2059 LSYM(ret_t0) MILLIRET
2060 LSYM(e_t0) r__r_t0
2061 LSYM(e_shift) a1_ne_0_b_l2
2062 a0__256a0 /* a0 <<= 8 *********** */
2063 MILLIRETN
2064 LSYM(e_t0ma0) a1_ne_0_b_l0
2065 t0__t0ma0
2066 MILLIRET
2067 r__r_t0
2068 LSYM(e_t0a0) a1_ne_0_b_l0
2069 t0__t0_a0
2070 MILLIRET
2071 r__r_t0
2072 LSYM(e_t02a0) a1_ne_0_b_l0
2073 t0__t0_2a0
2074 MILLIRET
2075 r__r_t0
2076 LSYM(e_t04a0) a1_ne_0_b_l0
2077 t0__t0_4a0
2078 MILLIRET
2079 r__r_t0
2080 LSYM(e_2t0) a1_ne_0_b_l1
2081 r__r_2t0
2082 MILLIRETN
2083 LSYM(e_2t0a0) a1_ne_0_b_l0
2084 t0__2t0_a0
2085 MILLIRET
2086 r__r_t0
2087 LSYM(e2t04a0) t0__t0_2a0
2088 a1_ne_0_b_l1
2089 r__r_2t0
2090 MILLIRETN
2091 LSYM(e_3t0) a1_ne_0_b_l0
2092 t0__3t0
2093 MILLIRET
2094 r__r_t0
2095 LSYM(e_4t0) a1_ne_0_b_l1
2096 r__r_4t0
2097 MILLIRETN
2098 LSYM(e_4t0a0) a1_ne_0_b_l0
2099 t0__4t0_a0
2100 MILLIRET
2101 r__r_t0
2102 LSYM(e4t08a0) t0__t0_2a0
2103 a1_ne_0_b_l1
2104 r__r_4t0
2105 MILLIRETN
2106 LSYM(e_5t0) a1_ne_0_b_l0
2107 t0__5t0
2108 MILLIRET
2109 r__r_t0
2110 LSYM(e_8t0) a1_ne_0_b_l1
2111 r__r_8t0
2112 MILLIRETN
2113 LSYM(e_8t0a0) a1_ne_0_b_l0
2114 t0__8t0_a0
2115 MILLIRET
2116 r__r_t0
2118 .procend
2119 .end
2120 #endif