![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to longlong.h CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [GCC 2.x] / gcc|
Return to longlong.h CVS log | Up to [GCC 2.x] / gcc |
1.1 root 1: /* longlong.h -- definitions for mixed size 32/64 bit arithmetic.
2: Copyright (C) 1991 Free Software Foundation, Inc.
3:
4: This definition file is free software; you can redistribute it
5: and/or modify it under the terms of the GNU General Public
6: License as published by the Free Software Foundation; either
7: version 2, or (at your option) any later version.
8:
9: This definition file is distributed in the hope that it will be
10: useful, but WITHOUT ANY WARRANTY; without even the implied
11: warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
12: See the GNU General Public License for more details.
13:
14: You should have received a copy of the GNU General Public License
15: along with this program; if not, write to the Free Software
16: Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
17:
18: #ifndef LONG_TYPE_SIZE
19: #define LONG_TYPE_SIZE 32
20: #endif
21:
22: #define __BITS4 (LONG_TYPE_SIZE / 4)
23: #define __ll_B (1L << (LONG_TYPE_SIZE / 2))
24: #define __ll_lowpart(t) ((unsigned long int) (t) % __ll_B)
25: #define __ll_highpart(t) ((unsigned long int) (t) / __ll_B)
26:
27: /* Define auxilliary asm macros.
28:
29: 1) umul_ppmm(high_prod, low_prod, multipler, multiplicand)
30: multiplies two unsigned long integers MULTIPLER and MULTIPLICAND,
31: and generates a two unsigned word product in HIGH_PROD and
32: LOW_PROD.
33:
34: 2) __umulsidi3(a,b) multiplies two unsigned long integers A and B,
35: and returns a long long product. This is just a variant of umul_ppmm.
36:
37: 3) udiv_qrnnd(quotient, remainder, high_numerator, low_numerator,
38: denominator) divides a two-word unsigned integer, composed by the
39: integers HIGH_NUMERATOR and LOW_NUMERATOR, by DENOMINATOR and
40: places the quotient in QUOTIENT and the remainder in REMAINDER.
41: HIGH_NUMERATOR must be less than DENOMINATOR for correct operation.
42: If, in addition, the most significant bit of DENOMINATOR must be 1,
43: then the pre-processor symbol UDIV_NEEDS_NORMALIZATION is defined to 1.
44:
45: 4) count_leading_zeros(count, x) counts the number of zero-bits from
46: the msb to the first non-zero bit. This is the number of steps X
47: needs to be shifted left to set the msb. Undefined for X == 0.
48:
49: 5) add_ssaaaa(high_sum, low_sum, high_addend_1, low_addend_1,
50: high_addend_2, low_addend_2) adds two two-word unsigned integers,
51: composed by HIGH_ADDEND_1 and LOW_ADDEND_1, and HIGH_ADDEND_2 and
52: LOW_ADDEND_2 respectively. The result is placed in HIGH_SUM and
53: LOW_SUM. Overflow (i.e. carry out) is not stored anywhere, and is
54: lost.
55:
56: 6) sub_ddmmss(high_difference, low_difference, high_minuend,
57: low_minuend, high_subtrahend, low_subtrahend) subtracts two
58: two-word unsigned integers, composed by HIGH_MINUEND_1 and
59: LOW_MINUEND_1, and HIGH_SUBTRAHEND_2 and LOW_SUBTRAHEND_2
60: respectively. The result is placed in HIGH_DIFFERENCE and
61: LOW_DIFFERENCE. Overflow (i.e. carry out) is not stored anywhere,
62: and is lost.
63:
64: If any of these macros are left undefined for a particular CPU,
65: C macros are used. */
66:
67: /* The CPUs come in alphabetical order below.
68:
69: Please add support for more CPUs here, or improve the current support
70: for the CPUs below!
71: (E.g. WE32100, HP-PA (xmpyu?), i960, IBM360, TRON.) */
72:
73: #if defined (__GNUC__) && !defined (NO_ASM)
74:
75: #if defined (__a29k__) || defined (___AM29K__)
76: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
77: __asm__ ("add %1,%4,%5
78: addc %0,%2,%3" \
79: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl)) \
80: : "%r" ((unsigned long int)(ah)), "rI" ((unsigned long int)(bh)), \
81: "%r" ((unsigned long int)(al)), "rI" ((unsigned long int)(bl)))
82: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
83: __asm__ ("sub %1,%4,%5
84: subc %0,%2,%3" \
85: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl)) \
86: : "r" ((unsigned long int)(ah)), "rI" ((unsigned long int)(bh)), \
87: "r" ((unsigned long int)(al)), "rI" ((unsigned long int)(bl)))
88: #define umul_ppmm(xh, xl, m0, m1) \
89: do { \
90: unsigned long int __m0 = (m0), __m1 = (m1); \
91: __asm__ ("multiplu %0,%1,%2" : "=r" ((unsigned long int)(xl)) \
92: : "r" (__m0), "r" (__m1)); \
93: __asm__ ("multmu %0,%1,%2" : "=r" ((unsigned long int)(xh)) \
94: : "r" (__m0), "r" (__m1)); \
95: } while (0)
96: #define udiv_qrnnd(q, r, n1, n0, d) \
97: __asm__ ("dividu %0,%3,%4" \
98: : "=r" ((unsigned long int)(q)), "=q" ((unsigned long int)(r)) \
99: : "1" ((unsigned long int)(n1)), "r" ((unsigned long int)(n0)), \
100: "r" ((unsigned long int)(d)))
101: #define count_leading_zeros(count, x) \
102: __asm__ ("clz %0,%1" : "=r" ((unsigned long int)(count)) \
103: : "r" ((unsigned long int)(x)))
104: #endif /* __a29k__ */
105:
106: #if defined (__arm__)
107: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
108: __asm__ ("adds %1,%4,%5
109: adc %0,%2,%3" \
110: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl)) \
111: : "%r" ((unsigned long int)(ah)), "rI" ((unsigned long int)(bh)), \
112: "%r" ((unsigned long int)(al)), "rI" ((unsigned long int)(bl)))
113: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
114: __asm__ ("subs %1,%4,%5
115: sbc %0,%2,%3" \
116: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl)) \
117: : "r" ((unsigned long int)(ah)), "rI" ((unsigned long int)(bh)), \
118: "r" ((unsigned long int)(al)), "rI" ((unsigned long int)(bl)))
119: #endif /* __arm__ */
120:
121: #if defined (__gmicro__)
122: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
123: __asm__ ("add.w %5,%1
124: addx %3,%0" \
125: : "=g" ((unsigned long int)(sh)), "=&g" ((unsigned long int)(sl))\
126: : "%0" ((unsigned long int)(ah)), "g" ((unsigned long int)(bh)), \
127: "%1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
128: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
129: __asm__ ("sub.w %5,%1
130: subx %3,%0" \
131: : "=g" ((unsigned long int)(sh)), "=&g" ((unsigned long int)(sl))\
132: : "0" ((unsigned long int)(ah)), "g" ((unsigned long int)(bh)), \
133: "1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
134: #define umul_ppmm(ph, pl, m0, m1) \
135: __asm__ ("mulx %3,%0,%1" \
136: : "=g" ((unsigned long int)(ph)), "=r" ((unsigned long int)(pl))\
137: : "%0" ((unsigned long int)(m0)), "g" ((unsigned long int)(m1)))
138: #define udiv_qrnnd(q, r, nh, nl, d) \
139: __asm__ ("divx %4,%0,%1" \
140: : "=g" ((unsigned long int)(q)), "=r" ((unsigned long int)(r)) \
141: : "1" ((unsigned long int)(nh)), "0" ((unsigned long int)(nl)), \
142: "g" ((unsigned long int)(d)))
143: #define count_leading_zeros(count, x) \
144: __asm__ ("bsch/1 %1,%0" \
145: : "=g" (count) \
146: : "g" ((unsigned long int)(x)), "0" (0UL))
147: #endif
148:
149: #if defined (__hppa)
150: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
151: __asm__ ("add %4,%5,%1
152: addc %2,%3,%0" \
153: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
154: : "%r" ((unsigned long int)(ah)), "r" ((unsigned long int)(bh)),\
155: "%r" ((unsigned long int)(al)), "r" ((unsigned long int)(bl)))
156: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
157: __asm__ ("sub %5,%4,%1
158: subb %3,%2,%0" \
159: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
160: : "r" ((unsigned long int)(ah)), "r" ((unsigned long int)(bh)), \
161: "r" ((unsigned long int)(al)), "r" ((unsigned long int)(bl)))
162: #endif
163:
164: #if defined (__i386__) || defined (__i486__)
165: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
166: __asm__ ("addl %5,%1
167: adcl %3,%0" \
168: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
169: : "%0" ((unsigned long int)(ah)), "g" ((unsigned long int)(bh)), \
170: "%1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
171: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
172: __asm__ ("subl %5,%1
173: sbbl %3,%0" \
174: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
175: : "0" ((unsigned long int)(ah)), "g" ((unsigned long int)(bh)), \
176: "1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
177: #define umul_ppmm(w1, w0, u, v) \
178: __asm__ ("mull %3" \
179: : "=a" ((unsigned long int)(w0)), "=d" ((unsigned long int)(w1))\
180: : "%0" ((unsigned long int)(u)), "rm" ((unsigned long int)(v)))
181: #define udiv_qrnnd(q, r, n1, n0, d) \
182: __asm__ ("divl %4" \
183: : "=a" ((unsigned long int)(q)), "=d" ((unsigned long int)(r)) \
184: : "0" ((unsigned long int)(n0)), "1" ((unsigned long int)(n1)), \
185: "rm" ((unsigned long int)(d)))
186: #define count_leading_zeros(count, x) \
187: do { \
188: unsigned long int __cbtmp; \
189: __asm__ ("bsrl %1,%0" \
190: : "=r" (__cbtmp) : "rm" ((unsigned long int)(x))); \
191: (count) = __cbtmp ^ 31; \
192: } while (0)
193: #endif /* 80x86 */
194:
195: #if defined (__i860__)
196: #if 0
197: /* Make sure these patterns really improve the code before
198: switching them on. */
199: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
200: do { \
201: union \
202: { \
203: long long int ll; \
204: struct {unsigned long int l, h;} i; \
205: } __a, __b, __s; \
206: __a.i.l = (al); __a.i.h = (ah); \
207: __b.i.l = (bl); __b.i.h = (bh); \
208: __asm__ ("fiadd.dd %1,%2,%0" \
209: : "=f" (__s.ll) \
210: : "%f" (__a.ll), "f" (__b.ll)); \
211: (sh) = __s.i.h; (sl) = __s.i.l; \
212: } while (0)
213: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
214: do { \
215: union \
216: { \
217: long long int ll; \
218: struct {unsigned long int l, h;} i; \
219: } __a, __b, __s; \
220: __a.i.l = (al); __a.i.h = (ah); \
221: __b.i.l = (bl); __b.i.h = (bh); \
222: __asm__ ("fisub.dd %1,%2,%0" \
223: : "=f" (__s.ll) \
224: : "%f" (__a.ll), "f" (__b.ll)); \
225: (sh) = __s.i.h; (sl) = __s.i.l; \
226: } while (0)
227: #endif
228: #endif /* __i860__ */
229:
230: #if defined (___IBMR2__) /* IBM RS6000 */
231: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
232: __asm__ ("a %1,%4,%5
233: ae %0,%2,%3" \
234: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
235: : "%r" ((unsigned long int)(ah)), "r" ((unsigned long int)(bh)),\
236: "%r" ((unsigned long int)(al)), "r" ((unsigned long int)(bl)))
237: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
238: __asm__ ("sf %1,%5,%4
239: sfe %0,%3,%2" \
240: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
241: : "r" ((unsigned long int)(ah)), "r" ((unsigned long int)(bh)), \
242: "r" ((unsigned long int)(al)), "r" ((unsigned long int)(bl)))
243: #define umul_ppmm(xh, xl, m0, m1) \
244: do { \
245: unsigned long int __m0 = (m0), __m1 = (m1); \
246: __asm__ ("mul %0,%2,%3" \
247: : "=r" ((unsigned long int)(xh)), "=q" ((unsigned long int)(xl))\
248: : "r" (__m0), "r" (__m1)); \
249: (xh) += ((((signed long int) __m0 >> 31) & __m1) \
250: + (((signed long int) __m1 >> 31) & __m0)); \
251: } while (0)
252: #define UMUL_TIME 8
253: #define udiv_qrnnd(q, r, nh, nl, d) \
254: do { /* Use the signed "div" insn, and adjust the result. */ \
255: unsigned long int __q, __r, __nh, __nl, __d, __xh, __xl; \
256: __nh = 0; \
257: __nl = (((unsigned long int)(nh)) << 30) | ((unsigned long int)(nl) >> 2);\
258: __nh = (unsigned long int)(nh) >> 2; \
259: __d = ((unsigned long int)(d) >> 1); \
260: __asm__ ("div %0,%2,%4" \
261: : "=r" (__q), "=q" (__r) \
262: : "r" (__nh), "1" (__nl), "r" (__d)); \
263: __q <<= 1; \
264: __asm__ ("mul %0,%2,%3" \
265: : "=r" (__xh), "=q" (__xl) \
266: : "r" (__q), "r" ((unsigned long int)(d))); \
267: __xh += (((signed long int) __q >> 31) & (d)) + __q; \
268: if ((nh) < __xh || ((nh) == __xh && (nl) < __xl)) \
269: { \
270: do \
271: { \
272: sub_ddmmss (__xh, __xl, __xh, __xl, 0, (d)); \
273: __q--; \
274: } \
275: while ((nh) < __xh || ((nh) == __xh && (nl) < __xl)); \
276: __xl = (nl) - __xl; \
277: } \
278: else \
279: { \
280: sub_ddmmss (__xh, __xl, (nh), (nl), __xh, __xl); \
281: if (__xh != 0) \
282: { \
283: do \
284: { \
285: sub_ddmmss (__xh, __xl, __xh, __xl, 0, (d)); \
286: __q++; \
287: } \
288: while (__xh != 0); \
289: } \
290: if (__xl >= (d)) \
291: { \
292: __xl -= (d); \
293: __q++; \
294: } \
295: } \
296: (q) = __q; \
297: (r) = __xl; \
298: } while (0)
299: #define UDIV_TIME 40
300: #define UDIV_NEEDS_NORMALIZATION 1
301: #define count_leading_zeros(count, x) \
302: __asm__ ("cntlz %0,%1" \
303: : "=r" ((unsigned long int)(count)) : "r" ((unsigned long int)(x)))
304: #endif /* ___IBMR2__ */
305:
306: #if defined (__mc68000__)
307: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
308: __asm__ ("add%.l %5,%1
309: addx%.l %3,%0" \
310: : "=d" ((unsigned long int)(sh)), "=&d" ((unsigned long int)(sl))\
311: : "%0" ((unsigned long int)(ah)), "d" ((unsigned long int)(bh)), \
312: "%1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
313: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
314: __asm__ ("sub%.l %5,%1
315: subx%.l %3,%0" \
316: : "=d" ((unsigned long int)(sh)), "=&d" ((unsigned long int)(sl))\
317: : "0" ((unsigned long int)(ah)), "d" ((unsigned long int)(bh)), \
318: "1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
319: #if defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)
320: #define umul_ppmm(w1, w0, u, v) \
321: __asm__ ("mulu%.l %3,%1:%0" \
322: : "=d" ((unsigned long int)(w0)), "=d" ((unsigned long int)(w1))\
323: : "%0" ((unsigned long int)(u)), "dmi" ((unsigned long int)(v)))
324: #define udiv_qrnnd(q, r, n1, n0, d) \
325: __asm__ ("divu%.l %4,%1:%0" \
326: : "=d" ((unsigned long int)(q)), "=d" ((unsigned long int)(r)) \
327: : "0" ((unsigned long int)(n0)), "1" ((unsigned long int)(n1)), \
328: "dmi" ((unsigned long int)(d)))
329: #define count_leading_zeros(count, x) \
330: __asm__ ("bfffo %1{%b2:%b2},%0" \
331: : "=d" ((unsigned long int)(count)) \
332: : "od" ((unsigned long int)(x)), "n" (0))
333: #else /* not mc68020 */
334: #define umul_ppmm(xh, xl, a, b) \
335: __asm__ ("| Inlined umul_ppmm
336: movel %2,d0
337: movel %3,d1
338: movel d0,d2
339: swap d0
340: movel d1,d3
341: swap d1
342: movew d2,d4
343: mulu d3,d4
344: mulu d1,d2
345: mulu d0,d3
346: mulu d0,d1
347: movel d4,d0
348: eorw d0,d0
349: swap d0
350: addl d0,d2
351: addl d3,d2
352: jcc 1f
353: addl #65536,d1
354: 1: swap d2
355: moveq #0,d0
356: movew d2,d0
357: movew d4,d2
358: movel d2,%1
359: addl d1,d0
360: movel d0,%0" \
361: : "=g" ((unsigned long int)(xh)), "=g" ((unsigned long int)(xl)) \
362: :"g" ((unsigned long int)(a)), "g" ((unsigned long int)(b)) \
363: : "d0", "d1", "d2", "d3", "d4")
364: #endif /* not mc68020 */
365: #endif /* mc68000 */
366:
367: #if defined (__m88000__)
368: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
369: __asm__ ("addu.co %1,%r4,%r5
370: addu.ci %0,%r2,%r3" \
371: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
372: : "%rJ" ((unsigned long int)(ah)), "rJ" ((unsigned long int)(bh)),\
373: "%rJ" ((unsigned long int)(al)), "rJ" ((unsigned long int)(bl)))
374: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
375: __asm__ ("subu.co %1,%r4,%r5
376: subu.ci %0,%r2,%r3" \
377: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
378: : "rJ" ((unsigned long int)(ah)), "rJ" ((unsigned long int)(bh)),\
379: "rJ" ((unsigned long int)(al)), "rJ" ((unsigned long int)(bl)))
380: #define UMUL_TIME 17
381: #define UDIV_TIME 150
382: #define count_leading_zeros(count, x) \
383: do { \
384: unsigned long int __cbtmp; \
385: __asm__ ("ff1 %0,%1" \
386: : "=r" (__cbtmp) : "r" ((unsigned long int)(x))); \
387: (count) = __cbtmp ^ 31; \
388: } while (0)
389: #endif /* __m88000__ */
390:
391: #if defined (__mips__)
392: #define umul_ppmm(w1, w0, u, v) \
393: __asm__ ("multu %2,%3
394: mflo %0
395: mfhi %1" \
396: : "=r" ((unsigned long int)(w0)), "=r" ((unsigned long int)(w1))\
397: : "r" ((unsigned long int)(u)), "r" ((unsigned long int)(v)))
398: #define UMUL_TIME 5
399: #define UDIV_TIME 100
400: #endif /* __mips__ */
401:
402: #if defined (__ns32000__)
403: #define __umulsidi3(u, v) \
404: ({long long int __w; \
405: __asm__ ("meid %2,%0" : "=g" (__w) \
406: : "%0" ((unsigned long int)(u)), "g" ((unsigned long int)(v))); \
407: __w; })
408: #define div_qrnnd(q, r, n1, n0, d) \
409: __asm__ ("movd %2,r0
410: movd %3,r1
411: deid %4,r0
412: movd r1,%0
413: movd r0,%1" \
414: : "=g" ((unsigned long int)(q)), "=g" ((unsigned long int)(r)) \
415: : "g" ((unsigned long int)(n0)), "g" ((unsigned long int)(n1)), \
416: "g" ((unsigned long int)(d)) : "r0", "r1")
417: #endif /* __ns32000__ */
418:
419: #if defined (__pyr__)
420: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
421: __asm__ ("addw %5,%1
422: addwc %3,%0" \
423: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
424: : "%0" ((unsigned long int)(ah)), "g" ((unsigned long int)(bh)),\
425: "%1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
426: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
427: __asm__ ("subw %5,%1
428: subwb %3,%0" \
429: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
430: : "0" ((unsigned long int)(ah)), "g" ((unsigned long int)(bh)), \
431: "1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
432: /* This insn doesn't work on ancient pyramids. */
433: #define umul_ppmm(w1, w0, u, v) \
434: __asm__ ("movw %2,tr11
435: uemul %3,tr10
436: movw tr10,%0
437: movw tr11,%1" \
438: : "=r" ((unsigned long int)(w1)), "=r" ((unsigned long int)(w0))\
439: : "r" ((unsigned long int)(u)), "r" ((unsigned long int)(v)) \
440: : "tr10", "tr11")
441: #endif /* __pyr__ */
442:
443: #if defined (__ibm032__) /* RT/ROMP */
444: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
445: __asm__ ("a %1,%5
446: ae %0,%3" \
447: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
448: : "%0" ((unsigned long int)(ah)), "r" ((unsigned long int)(bh)),\
449: "%1" ((unsigned long int)(al)), "r" ((unsigned long int)(bl)))
450: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
451: __asm__ ("s %1,%5
452: se %0,%3" \
453: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
454: : "0" ((unsigned long int)(ah)), "r" ((unsigned long int)(bh)), \
455: "1" ((unsigned long int)(al)), "r" ((unsigned long int)(bl)))
456: #define umul_ppmm(ph, pl, m0, m1) \
457: do { \
458: unsigned long int __m0 = (m0), __m1 = (m1); \
459: __asm__ ( \
460: "s r2,r2
461: mts r10,%2
462: m r2,%3
463: m r2,%3
464: m r2,%3
465: m r2,%3
466: m r2,%3
467: m r2,%3
468: m r2,%3
469: m r2,%3
470: m r2,%3
471: m r2,%3
472: m r2,%3
473: m r2,%3
474: m r2,%3
475: m r2,%3
476: m r2,%3
477: m r2,%3
478: cas %0,r2,r0
479: mfs r10,%1" \
480: : "=r" ((unsigned long int)(ph)), "=r" ((unsigned long int)(pl)) \
481: : "%r" (__m0), "r" (__m1) \
482: : "r2"); \
483: (ph) += ((((signed long int) __m0 >> 31) & __m1) \
484: + (((signed long int) __m1 >> 31) & __m0)); \
485: } while (0)
486: #define count_leading_zeros(count, x) \
487: do { \
488: if ((x) >= 0x10000) \
489: __asm__ ("clz %0,%1" \
490: : "=r" ((unsigned long int)(count)) \
491: : "r" ((unsigned long int)(x) >> 16)); \
492: else \
493: { \
494: __asm__ ("clz %0,%1" \
495: : "=r" ((unsigned long int)(count)) \
496: : "r" ((unsigned long int)(x))); \
497: (count) += 16; \
498: } \
499: } while (0)
500: #endif
501:
502: #if defined (__sparc__)
503: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
504: __asm__ ("addcc %4,%5,%1
505: addx %2,%3,%0" \
506: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
507: : "%r" ((unsigned long int)(ah)), "rI" ((unsigned long int)(bh)),\
508: "%r" ((unsigned long int)(al)), "rI" ((unsigned long int)(bl)))
509: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
510: __asm__ ("subcc %4,%5,%1
511: subx %2,%3,%0" \
512: : "=r" ((unsigned long int)(sh)), "=&r" ((unsigned long int)(sl))\
513: : "r" ((unsigned long int)(ah)), "rI" ((unsigned long int)(bh)),\
514: "r" ((unsigned long int)(al)), "rI" ((unsigned long int)(bl)))
515: #if defined (__sparc8__) /* How do we recog. version 8 SPARC? */
516: #define umul_ppmm(w1, w0, u, v) \
517: __asm__ ("umul %2,%3,%1;rd %%y,%0" \
518: : "=r" ((unsigned long int)(w1)), "=r" ((unsigned long int)(w0))\
519: : "r" ((unsigned long int)(u)), "r" ((unsigned long int)(v)))
520: #define udiv_qrnnd(q, r, n1, n0, d) \
521: __asm__ ("mov %2,%%y;nop;nop;nop;udiv %3,%4,%0;umul %0,%4,%1;sub %3,%1,%1"\
522: : "=&r" ((unsigned long int)(q)), "=&r" ((unsigned long int)(r))\
523: : "r" ((unsigned long int)(n1)), "r" ((unsigned long int)(n0)), \
524: "r" ((unsigned long int)(d)))
525: #else
526: /* SPARC without integer multiplication and divide instructions.
527: (i.e. at least Sun4/20,40,60,65,75,110,260,280,330,360,380,470,490) */
528: #define umul_ppmm(w1, w0, u, v) \
529: __asm__ ("! Inlined umul_ppmm
530: wr %%g0,%2,%%y ! SPARC has 0-3 delay insn after a wr
531: sra %3,31,%%g2 ! Don't move this insn
532: and %2,%%g2,%%g2 ! Don't move this insn
533: andcc %%g0,0,%%g1 ! Don't move this insn
534: mulscc %%g1,%3,%%g1
535: mulscc %%g1,%3,%%g1
536: mulscc %%g1,%3,%%g1
537: mulscc %%g1,%3,%%g1
538: mulscc %%g1,%3,%%g1
539: mulscc %%g1,%3,%%g1
540: mulscc %%g1,%3,%%g1
541: mulscc %%g1,%3,%%g1
542: mulscc %%g1,%3,%%g1
543: mulscc %%g1,%3,%%g1
544: mulscc %%g1,%3,%%g1
545: mulscc %%g1,%3,%%g1
546: mulscc %%g1,%3,%%g1
547: mulscc %%g1,%3,%%g1
548: mulscc %%g1,%3,%%g1
549: mulscc %%g1,%3,%%g1
550: mulscc %%g1,%3,%%g1
551: mulscc %%g1,%3,%%g1
552: mulscc %%g1,%3,%%g1
553: mulscc %%g1,%3,%%g1
554: mulscc %%g1,%3,%%g1
555: mulscc %%g1,%3,%%g1
556: mulscc %%g1,%3,%%g1
557: mulscc %%g1,%3,%%g1
558: mulscc %%g1,%3,%%g1
559: mulscc %%g1,%3,%%g1
560: mulscc %%g1,%3,%%g1
561: mulscc %%g1,%3,%%g1
562: mulscc %%g1,%3,%%g1
563: mulscc %%g1,%3,%%g1
564: mulscc %%g1,%3,%%g1
565: mulscc %%g1,%3,%%g1
566: mulscc %%g1,0,%%g1
567: add %%g1,%%g2,%0
568: rd %%y,%1" \
569: : "=r" ((unsigned long int)(w1)), "=r" ((unsigned long int)(w0))\
570: : "%rI" ((unsigned long int)(u)), "r" ((unsigned long int)(v)) \
571: : "%g1", "%g2")
572: #define UMUL_TIME 39 /* 39 instructions */
573: /* It's quite necessary to add this much assembler for the sparc.
574: The default udiv_qrnnd (in C) is more than 10 times slower! */
575: #define udiv_qrnnd(q, r, n1, n0, d) \
576: __asm__ ("! Inlined udiv_qrnnd
577: mov 32,%%g1
578: subcc %1,%2,%%g0
579: 1: bcs 5f
580: addxcc %0,%0,%0 ! shift n1n0 and a q-bit in lsb
581: sub %1,%2,%1 ! this kills msb of n
582: addx %1,%1,%1 ! so this can't give carry
583: subcc %%g1,1,%%g1
584: 2: bne 1b
585: subcc %1,%2,%%g0
586: bcs 3f
587: addxcc %0,%0,%0 ! shift n1n0 and a q-bit in lsb
588: b 3f
589: sub %1,%2,%1 ! this kills msb of n
590: 4: sub %1,%2,%1
591: 5: addxcc %1,%1,%1
592: bcc 2b
593: subcc %%g1,1,%%g1
594: ! Got carry from n. Subtract next step to cancel this carry.
595: bne 4b
596: addcc %0,%0,%0 ! shift n1n0 and a 0-bit in lsb
597: sub %1,%2,%1
598: 3: xnor %0,0,%0
599: ! End of inline udiv_qrnnd" \
600: : "=r&" ((unsigned long int)(q)), "=r&" ((unsigned long int)(r))\
601: : "r" ((unsigned long int)(d)), "1" ((unsigned long int)(n1)), \
602: "0" ((unsigned long int)(n0)) : "%g1")
603: #define UDIV_TIME (3+7*32) /* 7 instructions/iteration. 32 iterations. */
604: #endif
605: #endif /* __sparc8__ */
606:
607: #if defined (__vax__)
608: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
609: __asm__ ("addl2 %5,%1
610: adwc %3,%0" \
611: : "=g" ((unsigned long int)(sh)), "=&g" ((unsigned long int)(sl))\
612: : "%0" ((unsigned long int)(ah)), "g" ((unsigned long int)(bh)),\
613: "%1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
614: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
615: __asm__ ("subl2 %5,%1
616: sbwc %3,%0" \
617: : "=g" ((unsigned long int)(sh)), "=&g" ((unsigned long int)(sl))\
618: : "0" ((unsigned long int)(ah)), "g" ((unsigned long int)(bh)), \
619: "1" ((unsigned long int)(al)), "g" ((unsigned long int)(bl)))
620: #define umul_ppmm(xh, xl, m0, m1) \
621: do { \
622: union {long long int ll;struct {unsigned long int l, h;} i;} __xx; \
623: unsigned long int __m0 = (m0), __m1 = (m1); \
624: __asm__ ("emul %1,%2,$0,%0" \
625: : "=r" (__xx.ll) : "g" (__m0), "g" (__m1)); \
626: (xh) = __xx.i.h; (xl) = __xx.i.l; \
627: (xh) += ((((signed long int) __m0 >> 31) & __m1) \
628: + (((signed long int) __m1 >> 31) & __m0)); \
629: } while (0)
630: #endif /* __vax__ */
631:
632: #endif /* __GNUC__ */
633:
634: /* If this machine has no inline assembler, use C macros. */
635:
636: #if !defined (add_ssaaaa)
637: #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
638: do { \
639: unsigned long int __x; \
640: __x = (al) + (bl); \
641: (sh) = (ah) + (bh) + (__x < (al)); \
642: (sl) = __x; \
643: } while (0)
644: #endif
645:
646: #if !defined (sub_ddmmss)
647: #define sub_ddmmss(sh, sl, ah, al, bh, bl) \
648: do { \
649: unsigned long int __x; \
650: __x = (al) - (bl); \
651: (sh) = (ah) - (bh) - (__x > (al)); \
652: (sl) = __x; \
653: } while (0)
654: #endif
655:
656: #if !defined (umul_ppmm)
657: #define umul_ppmm(w1, w0, u, v) \
658: do { \
659: unsigned long int __x0, __x1, __x2, __x3; \
660: unsigned int __ul, __vl, __uh, __vh; \
661: \
662: __ul = __ll_lowpart (u); \
663: __uh = __ll_highpart (u); \
664: __vl = __ll_lowpart (v); \
665: __vh = __ll_highpart (v); \
666: \
667: __x0 = (unsigned long int) __ul * __vl; \
668: __x1 = (unsigned long int) __ul * __vh; \
669: __x2 = (unsigned long int) __uh * __vl; \
670: __x3 = (unsigned long int) __uh * __vh; \
671: \
672: __x1 += __ll_highpart (__x0);/* this can't give carry */ \
673: __x1 += __x2; /* but this indeed can */ \
674: if (__x1 < __x2) /* did we get it? */ \
675: __x3 += __ll_B; /* yes, add it in the proper pos. */ \
676: \
677: (w1) = __x3 + __ll_highpart (__x1); \
678: (w0) = __ll_lowpart (__x1) * __ll_B + __ll_lowpart (__x0); \
679: } while (0)
680: #endif
681:
682: #if !defined (__umulsidi3)
683: #define __umulsidi3(u, v) \
684: ({long_long __w; \
685: umul_ppmm (__w.s.high, __w.s.low, u, v); \
686: __w.ll; })
687: #endif
688:
689: #if !defined (udiv_qrnnd) || defined (__LLDEBUG__)
690: #define UDIV_NEEDS_NORMALIZATION 1
691: #ifndef __LLDEBUG__
692: #define udiv_qrnnd udiv_qrnnd_c
693: #endif
694: #define udiv_qrnnd_c(q, r, n1, n0, d) \
695: do { \
696: unsigned int __d1, __d0, __q1, __q0; \
697: unsigned long int __r1, __r0, __m; \
698: __d1 = __ll_highpart (d); \
699: __d0 = __ll_lowpart (d); \
700: \
701: __r1 = (n1) % __d1; \
702: __q1 = (n1) / __d1; \
703: __m = (unsigned long int) __q1 * __d0; \
704: __r1 = __r1 * __ll_B | __ll_highpart (n0); \
705: if (__r1 < __m) \
706: { \
707: __q1--, __r1 += (d); \
708: if (__r1 >= (d)) /* i.e. we didn't get carry when adding to __r1 */\
709: if (__r1 < __m) \
710: __q1--, __r1 += (d); \
711: } \
712: __r1 -= __m; \
713: \
714: __r0 = __r1 % __d1; \
715: __q0 = __r1 / __d1; \
716: __m = (unsigned long int) __q0 * __d0; \
717: __r0 = __r0 * __ll_B | __ll_lowpart (n0); \
718: if (__r0 < __m) \
719: { \
720: __q0--, __r0 += (d); \
721: if (__r0 >= (d)) \
722: if (__r0 < __m) \
723: __q0--, __r0 += (d); \
724: } \
725: __r0 -= __m; \
726: \
727: (q) = (unsigned long int) __q1 * __ll_B | __q0; \
728: (r) = __r0; \
729: } while (0)
730: #endif
731:
732: #if !defined (count_leading_zeros)
733: extern const unsigned char __clz_tab[];
734:
735: #define count_leading_zeros(count, x) \
736: do { \
737: unsigned long int xr = (x); \
738: unsigned int a; \
739: \
740: a = xr < (1<<2*__BITS4) \
741: ? (xr < (1<<__BITS4) ? 0 : __BITS4) \
742: : (xr < (1<<3*__BITS4) ? 2*__BITS4 : 3*__BITS4); \
743: \
744: (count) = 4*__BITS4 - (__clz_tab[xr >> a] + a); \
745: } while (0)
746: #endif
747:
748: #ifndef UDIV_NEEDS_NORMALIZATION
749: #define UDIV_NEEDS_NORMALIZATION 0
750: #endif
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.