[linux-2.6-block.git] / arch / sparc / lib / sdiv.S

/*
 * sdiv.S:      This routine was taken from glibc-1.09 and is covered
 *              by the GNU Library General Public License Version 2.
 */


/* This file is generated from divrem.m4; DO NOT EDIT! */
/*
 * Division and remainder, from Appendix E of the Sparc Version 8
 * Architecture Manual, with fixes from Gordon Irlam.
 */

/*
 * Input: dividend and divisor in %o0 and %o1 respectively.
 *
 * m4 parameters:
 *  .div	name of function to generate
 *  div		div=div => %o0 / %o1; div=rem => %o0 % %o1
 *  true		true=true => signed; true=false => unsigned
 *
 * Algorithm parameters:
 *  N		how many bits per iteration we try to get (4)
 *  WORDSIZE	total number of bits (32)
 *
 * Derived constants:
 *  TOPBITS	number of bits in the top decade of a number
 *
 * Important variables:
 *  Q		the partial quotient under development (initially 0)
 *  R		the remainder so far, initially the dividend
 *  ITER	number of main division loop iterations required;
 *		equal to ceil(log2(quotient) / N).  Note that this
 *		is the log base (2^N) of the quotient.
 *  V		the current comparand, initially divisor*2^(ITER*N-1)
 *
 * Cost:
 *  Current estimate for non-large dividend is
 *	ceil(log2(quotient) / N) * (10 + 7N/2) + C
 *  A large dividend is one greater than 2^(31-TOPBITS) and takes a
 *  different path, as the upper bits of the quotient must be developed
 *  one bit at a time.
 */


	.globl .div
	.globl _Div
.div:
_Div:	/* needed for export */
	! compute sign of result; if neither is negative, no problem
	orcc	%o1, %o0, %g0	! either negative?
	bge	2f			! no, go do the divide
	 xor	%o1, %o0, %g2	! compute sign in any case

	tst	%o1
	bge	1f
	 tst	%o0
	! %o1 is definitely negative; %o0 might also be negative
	bge	2f			! if %o0 not negative...
	 sub	%g0, %o1, %o1	! in any case, make %o1 nonneg
1:	! %o0 is negative, %o1 is nonnegative
	sub	%g0, %o0, %o0	! make %o0 nonnegative
2:

	! Ready to divide.  Compute size of quotient; scale comparand.
	orcc	%o1, %g0, %o5
	bne	1f
	 mov	%o0, %o3

		! Divide by zero trap.  If it returns, return 0 (about as
		! wrong as possible, but that is what SunOS does...).
		ta	ST_DIV0
		retl
		 clr	%o0

1:
	cmp	%o3, %o5			! if %o1 exceeds %o0, done
	blu	Lgot_result		! (and algorithm fails otherwise)
	 clr	%o2

	sethi	%hi(1 << (32 - 4 - 1)), %g1

	cmp	%o3, %g1
	blu	Lnot_really_big
	 clr	%o4

	! Here the dividend is >= 2**(31-N) or so.  We must be careful here,
	! as our usual N-at-a-shot divide step will cause overflow and havoc.
	! The number of bits in the result here is N*ITER+SC, where SC <= N.
	! Compute ITER in an unorthodox manner: know we need to shift V into
	! the top decade: so do not even bother to compare to R.
	1:
		cmp	%o5, %g1
		bgeu	3f
		 mov	1, %g7

		sll	%o5, 4, %o5

		b	1b
		 add	%o4, 1, %o4

	! Now compute %g7.
	2:
		addcc	%o5, %o5, %o5
		bcc	Lnot_too_big
		 add	%g7, 1, %g7

		! We get here if the %o1 overflowed while shifting.
		! This means that %o3 has the high-order bit set.
		! Restore %o5 and subtract from %o3.
		sll	%g1, 4, %g1	! high order bit
		srl	%o5, 1, %o5		! rest of %o5
		add	%o5, %g1, %o5

		b	Ldo_single_div
		 sub	%g7, 1, %g7

	Lnot_too_big:
	3:
		cmp	%o5, %o3
		blu	2b
		 nop

		be	Ldo_single_div
		 nop
	/* NB: these are commented out in the V8-Sparc manual as well */
	/* (I do not understand this) */
	! %o5 > %o3: went too far: back up 1 step
	!	srl	%o5, 1, %o5
	!	dec	%g7
	! do single-bit divide steps
	!
	! We have to be careful here.  We know that %o3 >= %o5, so we can do the
	! first divide step without thinking.  BUT, the others are conditional,
	! and are only done if %o3 >= 0.  Because both %o3 and %o5 may have the high-
	! order bit set in the first step, just falling into the regular
	! division loop will mess up the first time around.
	! So we unroll slightly...
	Ldo_single_div:
		subcc	%g7, 1, %g7
		bl	Lend_regular_divide
		 nop

		sub	%o3, %o5, %o3
		mov	1, %o2

		b	Lend_single_divloop
		 nop
	Lsingle_divloop:
		sll	%o2, 1, %o2

		bl	1f
		 srl	%o5, 1, %o5
		! %o3 >= 0
		sub	%o3, %o5, %o3

		b	2f
		 add	%o2, 1, %o2
	1:	! %o3 < 0
		add	%o3, %o5, %o3
		sub	%o2, 1, %o2
	2:
	Lend_single_divloop:
		subcc	%g7, 1, %g7
		bge	Lsingle_divloop
		 tst	%o3

		b,a	Lend_regular_divide

Lnot_really_big:
1:
	sll	%o5, 4, %o5
	cmp	%o5, %o3
	bleu	1b
	 addcc	%o4, 1, %o4

	be	Lgot_result
	 sub	%o4, 1, %o4

	tst	%o3	! set up for initial iteration
Ldivloop:
	sll	%o2, 4, %o2
		! depth 1, accumulated bits 0
	bl	L.1.16
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
			! depth 2, accumulated bits 1
	bl	L.2.17
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
			! depth 3, accumulated bits 3
	bl	L.3.19
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
			! depth 4, accumulated bits 7
	bl	L.4.23
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (7*2+1), %o2

L.4.23:
	! remainder is negative
	addcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (7*2-1), %o2

L.3.19:
	! remainder is negative
	addcc	%o3,%o5,%o3
			! depth 4, accumulated bits 5
	bl	L.4.21
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (5*2+1), %o2

L.4.21:
	! remainder is negative
	addcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (5*2-1), %o2

L.2.17:
	! remainder is negative
	addcc	%o3,%o5,%o3
			! depth 3, accumulated bits 1
	bl	L.3.17
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
			! depth 4, accumulated bits 3
	bl	L.4.19
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (3*2+1), %o2

L.4.19:
	! remainder is negative
	addcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (3*2-1), %o2
	
	
L.3.17:
	! remainder is negative
	addcc	%o3,%o5,%o3
			! depth 4, accumulated bits 1
	bl	L.4.17
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (1*2+1), %o2

L.4.17:
	! remainder is negative
	addcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (1*2-1), %o2

L.1.16:
	! remainder is negative
	addcc	%o3,%o5,%o3
			! depth 2, accumulated bits -1
	bl	L.2.15
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
			! depth 3, accumulated bits -1
	bl	L.3.15
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
			! depth 4, accumulated bits -1
	bl	L.4.15
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (-1*2+1), %o2

L.4.15:
	! remainder is negative
	addcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (-1*2-1), %o2

L.3.15:
	! remainder is negative
	addcc	%o3,%o5,%o3
			! depth 4, accumulated bits -3
	bl	L.4.13
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (-3*2+1), %o2

L.4.13:
	! remainder is negative
	addcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (-3*2-1), %o2

L.2.15:
	! remainder is negative
	addcc	%o3,%o5,%o3
			! depth 3, accumulated bits -3
	bl	L.3.13
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
			! depth 4, accumulated bits -5
	bl	L.4.11
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (-5*2+1), %o2

L.4.11:
	! remainder is negative
	addcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (-5*2-1), %o2

L.3.13:
	! remainder is negative
	addcc	%o3,%o5,%o3
			! depth 4, accumulated bits -7
	bl	L.4.9
	 srl	%o5,1,%o5
	! remainder is positive
	subcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (-7*2+1), %o2

L.4.9:
	! remainder is negative
	addcc	%o3,%o5,%o3
	b	9f
	 add	%o2, (-7*2-1), %o2

	9:
Lend_regular_divide:
	subcc	%o4, 1, %o4
	bge	Ldivloop
	 tst	%o3

	bl,a	Lgot_result
	! non-restoring fixup here (one instruction only!)
	sub	%o2, 1, %o2

Lgot_result:
	! check to see if answer should be < 0
	tst	%g2
	bl,a	1f
	 sub %g0, %o2, %o2
1:
	retl
	 mov %o2, %o0

	.globl	.div_patch
.div_patch:
	sra	%o0, 0x1f, %o2
	wr	%o2, 0x0, %y
	nop
	nop
	nop
	sdivcc	%o0, %o1, %o0
	bvs,a	1f
	 xnor	%o0, %g0, %o0
1:	retl
	 nop
Commit	Line	Data
88278ca2	1	/*
1da177e4 LT	2	* sdiv.S: This routine was taken from glibc-1.09 and is covered
	3	* by the GNU Library General Public License Version 2.
	4	*/
	5
	6
	7	/* This file is generated from divrem.m4; DO NOT EDIT! */
	8	/*
	9	* Division and remainder, from Appendix E of the Sparc Version 8
	10	* Architecture Manual, with fixes from Gordon Irlam.
	11	*/
	12
	13	/*
	14	* Input: dividend and divisor in %o0 and %o1 respectively.
	15	*
	16	* m4 parameters:
	17	* .div name of function to generate
	18	* div div=div => %o0 / %o1; div=rem => %o0 % %o1
	19	* true true=true => signed; true=false => unsigned
	20	*
	21	* Algorithm parameters:
	22	* N how many bits per iteration we try to get (4)
	23	* WORDSIZE total number of bits (32)
	24	*
	25	* Derived constants:
	26	* TOPBITS number of bits in the top decade of a number
	27	*
	28	* Important variables:
	29	* Q the partial quotient under development (initially 0)
	30	* R the remainder so far, initially the dividend
	31	* ITER number of main division loop iterations required;
	32	* equal to ceil(log2(quotient) / N). Note that this
	33	* is the log base (2^N) of the quotient.
	34	* V the current comparand, initially divisor2^(ITERN-1)
	35	*
	36	* Cost:
	37	* Current estimate for non-large dividend is
	38	* ceil(log2(quotient) / N) * (10 + 7N/2) + C
	39	* A large dividend is one greater than 2^(31-TOPBITS) and takes a
	40	* different path, as the upper bits of the quotient must be developed
	41	* one bit at a time.
	42	*/
	43
	44
	45	.globl .div
7caaeabb	46	.globl _Div
1da177e4	47	.div:
7caaeabb	48	_Div: /* needed for export */
1da177e4 LT	49	! compute sign of result; if neither is negative, no problem
	50	orcc %o1, %o0, %g0 ! either negative?
	51	bge 2f ! no, go do the divide
	52	xor %o1, %o0, %g2 ! compute sign in any case
	53
	54	tst %o1
	55	bge 1f
	56	tst %o0
	57	! %o1 is definitely negative; %o0 might also be negative
	58	bge 2f ! if %o0 not negative...
	59	sub %g0, %o1, %o1 ! in any case, make %o1 nonneg
	60	1: ! %o0 is negative, %o1 is nonnegative
	61	sub %g0, %o0, %o0 ! make %o0 nonnegative
	62	2:
	63
	64	! Ready to divide. Compute size of quotient; scale comparand.
	65	orcc %o1, %g0, %o5
	66	bne 1f
	67	mov %o0, %o3
	68
	69	! Divide by zero trap. If it returns, return 0 (about as
	70	! wrong as possible, but that is what SunOS does...).
	71	ta ST_DIV0
	72	retl
	73	clr %o0
	74
	75	1:
	76	cmp %o3, %o5 ! if %o1 exceeds %o0, done
	77	blu Lgot_result ! (and algorithm fails otherwise)
	78	clr %o2
	79
	80	sethi %hi(1 << (32 - 4 - 1)), %g1
	81
	82	cmp %o3, %g1
	83	blu Lnot_really_big
	84	clr %o4
	85
	86	! Here the dividend is >= 2**(31-N) or so. We must be careful here,
	87	! as our usual N-at-a-shot divide step will cause overflow and havoc.
	88	! The number of bits in the result here is N*ITER+SC, where SC <= N.
	89	! Compute ITER in an unorthodox manner: know we need to shift V into
	90	! the top decade: so do not even bother to compare to R.
	91	1:
	92	cmp %o5, %g1
	93	bgeu 3f
	94	mov 1, %g7
	95
	96	sll %o5, 4, %o5
	97
	98	b 1b
	99	add %o4, 1, %o4
	100
	101	! Now compute %g7.
	102	2:
	103	addcc %o5, %o5, %o5
	104	bcc Lnot_too_big
	105	add %g7, 1, %g7
	106
	107	! We get here if the %o1 overflowed while shifting.
	108	! This means that %o3 has the high-order bit set.
	109	! Restore %o5 and subtract from %o3.
	110	sll %g1, 4, %g1 ! high order bit
	111	srl %o5, 1, %o5 ! rest of %o5
	112	add %o5, %g1, %o5
113
114	b Ldo_single_div
115	sub %g7, 1, %g7
116
117	Lnot_too_big:
118	3:
119	cmp %o5, %o3
120	blu 2b
121	nop
122
123	be Ldo_single_div
124	nop
125	/* NB: these are commented out in the V8-Sparc manual as well */
126	/* (I do not understand this) */
127	! %o5 > %o3: went too far: back up 1 step
128	! srl %o5, 1, %o5
129	! dec %g7
130	! do single-bit divide steps
131	!
132	! We have to be careful here. We know that %o3 >= %o5, so we can do the
133	! first divide step without thinking. BUT, the others are conditional,
134	! and are only done if %o3 >= 0. Because both %o3 and %o5 may have the high-
135	! order bit set in the first step, just falling into the regular
136	! division loop will mess up the first time around.
137	! So we unroll slightly...
138	Ldo_single_div:
139	subcc %g7, 1, %g7
140	bl Lend_regular_divide
141	nop
142
143	sub %o3, %o5, %o3
144	mov 1, %o2
145
146	b Lend_single_divloop
147	nop
148	Lsingle_divloop:
149	sll %o2, 1, %o2
150
151	bl 1f
152	srl %o5, 1, %o5
153	! %o3 >= 0
154	sub %o3, %o5, %o3
155
156	b 2f
157	add %o2, 1, %o2
158	1: ! %o3 < 0
159	add %o3, %o5, %o3
160	sub %o2, 1, %o2
161	2:
162	Lend_single_divloop:
163	subcc %g7, 1, %g7
164	bge Lsingle_divloop
165	tst %o3
166
167	b,a Lend_regular_divide
168
169	Lnot_really_big:
170	1:
171	sll %o5, 4, %o5
172	cmp %o5, %o3
173	bleu 1b
174	addcc %o4, 1, %o4
175
176	be Lgot_result
177	sub %o4, 1, %o4
178
179	tst %o3 ! set up for initial iteration
180	Ldivloop:
181	sll %o2, 4, %o2
182	! depth 1, accumulated bits 0
183	bl L.1.16
184	srl %o5,1,%o5
185	! remainder is positive
186	subcc %o3,%o5,%o3
187	! depth 2, accumulated bits 1
188	bl L.2.17
189	srl %o5,1,%o5
190	! remainder is positive
191	subcc %o3,%o5,%o3
192	! depth 3, accumulated bits 3
193	bl L.3.19
194	srl %o5,1,%o5
195	! remainder is positive
196	subcc %o3,%o5,%o3
197	! depth 4, accumulated bits 7
198	bl L.4.23
199	srl %o5,1,%o5
200	! remainder is positive
201	subcc %o3,%o5,%o3
202	b 9f
203	add %o2, (7*2+1), %o2
204
205	L.4.23:
206	! remainder is negative
207	addcc %o3,%o5,%o3
208	b 9f
209	add %o2, (7*2-1), %o2
210
211	L.3.19:
212	! remainder is negative
213	addcc %o3,%o5,%o3
214	! depth 4, accumulated bits 5
215	bl L.4.21
216	srl %o5,1,%o5
217	! remainder is positive
218	subcc %o3,%o5,%o3
219	b 9f
220	add %o2, (5*2+1), %o2
221
222	L.4.21:
223	! remainder is negative
224	addcc %o3,%o5,%o3
225	b 9f
226	add %o2, (5*2-1), %o2
227
228	L.2.17:
229	! remainder is negative
230	addcc %o3,%o5,%o3
231	! depth 3, accumulated bits 1
232	bl L.3.17
233	srl %o5,1,%o5
234	! remainder is positive
235	subcc %o3,%o5,%o3
236	! depth 4, accumulated bits 3
237	bl L.4.19
238	srl %o5,1,%o5
239	! remainder is positive
240	subcc %o3,%o5,%o3
241	b 9f
242	add %o2, (3*2+1), %o2
243
244	L.4.19:
245	! remainder is negative
246	addcc %o3,%o5,%o3
247	b 9f
248	add %o2, (3*2-1), %o2
249
250
251	L.3.17:
252	! remainder is negative
253	addcc %o3,%o5,%o3
254	! depth 4, accumulated bits 1
255	bl L.4.17
256	srl %o5,1,%o5
257	! remainder is positive
258	subcc %o3,%o5,%o3
259	b 9f
260	add %o2, (1*2+1), %o2
261
262	L.4.17:
263	! remainder is negative
264	addcc %o3,%o5,%o3
265	b 9f
266	add %o2, (1*2-1), %o2
267
268	L.1.16:
269	! remainder is negative
270	addcc %o3,%o5,%o3
271	! depth 2, accumulated bits -1
272	bl L.2.15
273	srl %o5,1,%o5
274	! remainder is positive
275	subcc %o3,%o5,%o3
276	! depth 3, accumulated bits -1
277	bl L.3.15
278	srl %o5,1,%o5
279	! remainder is positive
280	subcc %o3,%o5,%o3
281	! depth 4, accumulated bits -1
282	bl L.4.15
283	srl %o5,1,%o5
284	! remainder is positive
285	subcc %o3,%o5,%o3
286	b 9f
287	add %o2, (-1*2+1), %o2
288
289	L.4.15:
290	! remainder is negative
291	addcc %o3,%o5,%o3
292	b 9f
293	add %o2, (-1*2-1), %o2
294
295	L.3.15:
296	! remainder is negative
297	addcc %o3,%o5,%o3
298	! depth 4, accumulated bits -3
299	bl L.4.13
300	srl %o5,1,%o5
301	! remainder is positive
302	subcc %o3,%o5,%o3
303	b 9f
304	add %o2, (-3*2+1), %o2
305
306	L.4.13:
307	! remainder is negative
308	addcc %o3,%o5,%o3
309	b 9f
310	add %o2, (-3*2-1), %o2
311
312	L.2.15:
313	! remainder is negative
314	addcc %o3,%o5,%o3
315	! depth 3, accumulated bits -3
316	bl L.3.13
317	srl %o5,1,%o5
318	! remainder is positive
319	subcc %o3,%o5,%o3
320	! depth 4, accumulated bits -5
321	bl L.4.11
322	srl %o5,1,%o5
323	! remainder is positive
324	subcc %o3,%o5,%o3
325	b 9f
326	add %o2, (-5*2+1), %o2
327
328	L.4.11:
329	! remainder is negative
330	addcc %o3,%o5,%o3
331	b 9f
332	add %o2, (-5*2-1), %o2
333
334	L.3.13:
335	! remainder is negative
336	addcc %o3,%o5,%o3
337	! depth 4, accumulated bits -7
338	bl L.4.9
339	srl %o5,1,%o5
340	! remainder is positive
341	subcc %o3,%o5,%o3
342	b 9f
343	add %o2, (-7*2+1), %o2
344
345	L.4.9:
346	! remainder is negative
347	addcc %o3,%o5,%o3
348	b 9f
349	add %o2, (-7*2-1), %o2
350
351	9:
352	Lend_regular_divide:
353	subcc %o4, 1, %o4
354	bge Ldivloop
355	tst %o3
356
357	bl,a Lgot_result
358	! non-restoring fixup here (one instruction only!)
359	sub %o2, 1, %o2
360
361	Lgot_result:
362	! check to see if answer should be < 0
363	tst %g2
364	bl,a 1f
365	sub %g0, %o2, %o2
366	1:
367	retl
368	mov %o2, %o0
369
370	.globl .div_patch
371	.div_patch:
372	sra %o0, 0x1f, %o2
373	wr %o2, 0x0, %y
374	nop
375	nop
376	nop
377	sdivcc %o0, %o1, %o0
378	bvs,a 1f
379	xnor %o0, %g0, %o0
380	1: retl
381	nop