URL https://opencores.org/ocsvn/or1k_old/or1k_old/trunk

Subversion Repositories or1k_old

[/] [or1k_old/] [trunk/] [linux/] [uClibc/] [libc/] [sysdeps/] [linux/] [sparc/] [urem.S] - Blame information for rev 1782

Details | Compare with Previous | View Log


   /* 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:
 *  .urem       name of function to generate
 *  rem         rem=div => %o0 / %o1; rem=rem => %o0 % %o1
 *  false               false=true => signed; false=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.
 */
 
 
 
#include 
 
 
.global   .urem;
.align 4;
.type  .urem ,@function;
 
.urem:
 
        ! 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      0x02
                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, %g2
                sll     %o5, 4, %o5
                b       1b
                add     %o4, 1, %o4
 
        ! Now compute %g2.
        2:      addcc   %o5, %o5, %o5
                bcc     .Lnot_too_big
                add     %g2, 1, %g2
 
                ! 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     %g2, 1, %g2
 
        .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     %g2
        ! 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   %g2, 1, %g2
                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   %g2, 1, %g2
                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      .L1.16
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                        ! depth 2, accumulated bits 1
        bl      .L2.17
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                        ! depth 3, accumulated bits 3
        bl      .L3.19
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                        ! depth 4, accumulated bits 7
        bl      .L4.23
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                b       9f
                add     %o2, (7*2+1), %o2
 
.L4.23:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                b       9f
                add     %o2, (7*2-1), %o2
 
 
.L3.19:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                        ! depth 4, accumulated bits 5
        bl      .L4.21
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                b       9f
                add     %o2, (5*2+1), %o2
 
.L4.21:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                b       9f
                add     %o2, (5*2-1), %o2
 
 
 
.L2.17:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                        ! depth 3, accumulated bits 1
        bl      .L3.17
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                        ! depth 4, accumulated bits 3
        bl      .L4.19
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                b       9f
                add     %o2, (3*2+1), %o2
 
.L4.19:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                b       9f
                add     %o2, (3*2-1), %o2
 
 
.L3.17:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                        ! depth 4, accumulated bits 1
        bl      .L4.17
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                b       9f
                add     %o2, (1*2+1), %o2
 
.L4.17:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                b       9f
                add     %o2, (1*2-1), %o2
 
 
 
 
.L1.16:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                        ! depth 2, accumulated bits -1
        bl      .L2.15
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                        ! depth 3, accumulated bits -1
        bl      .L3.15
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                        ! depth 4, accumulated bits -1
        bl      .L4.15
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                b       9f
                add     %o2, (-1*2+1), %o2
 
.L4.15:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                b       9f
                add     %o2, (-1*2-1), %o2
 
 
.L3.15:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                        ! depth 4, accumulated bits -3
        bl      .L4.13
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                b       9f
                add     %o2, (-3*2+1), %o2
 
.L4.13:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                b       9f
                add     %o2, (-3*2-1), %o2
 
 
 
.L2.15:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                        ! depth 3, accumulated bits -3
        bl      .L3.13
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                        ! depth 4, accumulated bits -5
        bl      .L4.11
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                b       9f
                add     %o2, (-5*2+1), %o2
 
.L4.11:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                b       9f
                add     %o2, (-5*2-1), %o2
 
 
.L3.13:
        ! remainder is negative
        addcc   %o3,%o5,%o3
                        ! depth 4, accumulated bits -7
        bl      .L4.9
        srl     %o5,1,%o5
        ! remainder is positive
        subcc   %o3,%o5,%o3
                b       9f
                add     %o2, (-7*2+1), %o2
 
.L4.9:
        ! 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!)
        add     %o3, %o1, %o3
 
 
.Lgot_result:
 
        retl
        mov %o3, %o0
 
.size  .urem , . -.urem

Browse

Tools

Subversion Repositories or1k_old

[/] [or1k_old/] [trunk/] [linux/] [uClibc/] [libc/] [sysdeps/] [linux/] [sparc/] [urem.S] - Blame information for rev 1782

Line No.	Rev	Author	Line
1	1325	phoenix	`/* This file is generated from divrem.m4; DO NOT EDIT! */`
2			`/*`
3			`* Division and remainder, from Appendix E of the Sparc Version 8`
4			`* Architecture Manual, with fixes from Gordon Irlam.`
5			`*/`
6
7			`/*`
8			`* Input: dividend and divisor in %o0 and %o1 respectively.`
9			`*`
10			`* m4 parameters:`
11			`* .urem name of function to generate`
12			`* rem rem=div => %o0 / %o1; rem=rem => %o0 % %o1`
13			`* false false=true => signed; false=false => unsigned`
14			`*`
15			`* Algorithm parameters:`
16			`* N how many bits per iteration we try to get (4)`
17			`* WORDSIZE total number of bits (32)`
18			`*`
19			`* Derived constants:`
20			`* TOPBITS number of bits in the top decade of a number`
21			`*`
22			`* Important variables:`
23			`* Q the partial quotient under development (initially 0)`
24			`* R the remainder so far, initially the dividend`
25			`* ITER number of main division loop iterations required;`
26			`* equal to ceil(log2(quotient) / N). Note that this`
27			`* is the log base (2^N) of the quotient.`
28			`* V the current comparand, initially divisor2^(ITERN-1)`
29			`*`
30			`* Cost:`
31			`* Current estimate for non-large dividend is`
32			`* ceil(log2(quotient) / N) * (10 + 7N/2) + C`
33			`* A large dividend is one greater than 2^(31-TOPBITS) and takes a`
34			`* different path, as the upper bits of the quotient must be developed`
35			`* one bit at a time.`
36			`*/`
37
38
39
40			`#include`
41
42
43			`.global .urem;`
44			`.align 4;`
45			`.type .urem ,@function;`
46
47			`.urem:`
48
49			`! Ready to divide. Compute size of quotient; scale comparand.`
50			`orcc %o1, %g0, %o5`
51			`bne 1f`
52			`mov %o0, %o3`
53
54			`! Divide by zero trap. If it returns, return 0 (about as`
55			`! wrong as possible, but that is what SunOS does...).`
56			`ta 0x02`
57			`retl`
58			`clr %o0`
59
60			`1:`
61			`cmp %o3, %o5 ! if %o1 exceeds %o0, done`
62			`blu .Lgot_result ! (and algorithm fails otherwise)`
63			`clr %o2`
64			`sethi %hi(1 << (32 - 4 - 1)), %g1`
65			`cmp %o3, %g1`
66			`blu .Lnot_really_big`
67			`clr %o4`
68
69			`! Here the dividend is >= 2**(31-N) or so. We must be careful here,`
70			`! as our usual N-at-a-shot divide step will cause overflow and havoc.`
71			`! The number of bits in the result here is N*ITER+SC, where SC <= N.`
72			`! Compute ITER in an unorthodox manner: know we need to shift V into`
73			`! the top decade: so do not even bother to compare to R.`
74			`1:`
75			`cmp %o5, %g1`
76			`bgeu 3f`
77			`mov 1, %g2`
78			`sll %o5, 4, %o5`
79			`b 1b`
80			`add %o4, 1, %o4`
81
82			`! Now compute %g2.`
83			`2: addcc %o5, %o5, %o5`
84			`bcc .Lnot_too_big`
85			`add %g2, 1, %g2`
86
87			`! We get here if the %o1 overflowed while shifting.`
88			`! This means that %o3 has the high-order bit set.`
89			`! Restore %o5 and subtract from %o3.`
90			`sll %g1, 4, %g1 ! high order bit`
91			`srl %o5, 1, %o5 ! rest of %o5`
92			`add %o5, %g1, %o5`
93			`b .Ldo_single_div`
94			`sub %g2, 1, %g2`
95
96			`.Lnot_too_big:`
97			`3: cmp %o5, %o3`
98			`blu 2b`
99			`nop`
100			`be .Ldo_single_div`
101			`nop`
102			`/* NB: these are commented out in the V8-Sparc manual as well */`
103			`/* (I do not understand this) */`
104			`! %o5 > %o3: went too far: back up 1 step`
105			`! srl %o5, 1, %o5`
106			`! dec %g2`
107			`! do single-bit divide steps`
108			`!`
109			`! We have to be careful here. We know that %o3 >= %o5, so we can do the`
110			`! first divide step without thinking. BUT, the others are conditional,`
111			`! and are only done if %o3 >= 0. Because both %o3 and %o5 may have the high-`
112			`! order bit set in the first step, just falling into the regular`
113			`! division loop will mess up the first time around.`
114			`! So we unroll slightly...`
115			`.Ldo_single_div:`
116			`subcc %g2, 1, %g2`
117			`bl .Lend_regular_divide`
118			`nop`
119			`sub %o3, %o5, %o3`
120			`mov 1, %o2`
121			`b .Lend_single_divloop`
122			`nop`
123			`.Lsingle_divloop:`
124			`sll %o2, 1, %o2`
125			`bl 1f`
126			`srl %o5, 1, %o5`
127			`! %o3 >= 0`
128			`sub %o3, %o5, %o3`
129			`b 2f`
130			`add %o2, 1, %o2`
131			`1: ! %o3 < 0`
132			`add %o3, %o5, %o3`
133			`sub %o2, 1, %o2`
134			`2:`
135			`.Lend_single_divloop:`
136			`subcc %g2, 1, %g2`
137			`bge .Lsingle_divloop`
138			`tst %o3`
139			`b,a .Lend_regular_divide`
140
141			`.Lnot_really_big:`
142			`1:`
143			`sll %o5, 4, %o5`
144			`cmp %o5, %o3`
145			`bleu 1b`
146			`addcc %o4, 1, %o4`
147			`be .Lgot_result`
148			`sub %o4, 1, %o4`
149
150			`tst %o3 ! set up for initial iteration`
151			`.Ldivloop:`
152			`sll %o2, 4, %o2`
153			`! depth 1, accumulated bits 0`
154			`bl .L1.16`
155			`srl %o5,1,%o5`
156			`! remainder is positive`
157			`subcc %o3,%o5,%o3`
158			`! depth 2, accumulated bits 1`
159			`bl .L2.17`
160			`srl %o5,1,%o5`
161			`! remainder is positive`
162			`subcc %o3,%o5,%o3`
163			`! depth 3, accumulated bits 3`
164			`bl .L3.19`
165			`srl %o5,1,%o5`
166			`! remainder is positive`
167			`subcc %o3,%o5,%o3`
168			`! depth 4, accumulated bits 7`
169			`bl .L4.23`
170			`srl %o5,1,%o5`
171			`! remainder is positive`
172			`subcc %o3,%o5,%o3`
173			`b 9f`
174			`add %o2, (7*2+1), %o2`
175
176			`.L4.23:`
177			`! remainder is negative`
178			`addcc %o3,%o5,%o3`
179			`b 9f`
180			`add %o2, (7*2-1), %o2`
181
182
183			`.L3.19:`
184			`! remainder is negative`
185			`addcc %o3,%o5,%o3`
186			`! depth 4, accumulated bits 5`
187			`bl .L4.21`
188			`srl %o5,1,%o5`
189			`! remainder is positive`
190			`subcc %o3,%o5,%o3`
191			`b 9f`
192			`add %o2, (5*2+1), %o2`
193
194			`.L4.21:`
195			`! remainder is negative`
196			`addcc %o3,%o5,%o3`
197			`b 9f`
198			`add %o2, (5*2-1), %o2`
199
200
201
202			`.L2.17:`
203			`! remainder is negative`
204			`addcc %o3,%o5,%o3`
205			`! depth 3, accumulated bits 1`
206			`bl .L3.17`
207			`srl %o5,1,%o5`
208			`! remainder is positive`
209			`subcc %o3,%o5,%o3`
210			`! depth 4, accumulated bits 3`
211			`bl .L4.19`
212			`srl %o5,1,%o5`
213			`! remainder is positive`
214			`subcc %o3,%o5,%o3`
215			`b 9f`
216			`add %o2, (3*2+1), %o2`
217
218			`.L4.19:`
219			`! remainder is negative`
220			`addcc %o3,%o5,%o3`
221			`b 9f`
222			`add %o2, (3*2-1), %o2`
223
224
225			`.L3.17:`
226			`! remainder is negative`
227			`addcc %o3,%o5,%o3`
228			`! depth 4, accumulated bits 1`
229			`bl .L4.17`
230			`srl %o5,1,%o5`
231			`! remainder is positive`
232			`subcc %o3,%o5,%o3`
233			`b 9f`
234			`add %o2, (1*2+1), %o2`
235
236			`.L4.17:`
237			`! remainder is negative`
238			`addcc %o3,%o5,%o3`
239			`b 9f`
240			`add %o2, (1*2-1), %o2`
241
242
243
244
245			`.L1.16:`
246			`! remainder is negative`
247			`addcc %o3,%o5,%o3`
248			`! depth 2, accumulated bits -1`
249			`bl .L2.15`
250			`srl %o5,1,%o5`
251			`! remainder is positive`
252			`subcc %o3,%o5,%o3`
253			`! depth 3, accumulated bits -1`
254			`bl .L3.15`
255			`srl %o5,1,%o5`
256			`! remainder is positive`
257			`subcc %o3,%o5,%o3`
258			`! depth 4, accumulated bits -1`
259			`bl .L4.15`
260			`srl %o5,1,%o5`
261			`! remainder is positive`
262			`subcc %o3,%o5,%o3`
263			`b 9f`
264			`add %o2, (-1*2+1), %o2`
265
266			`.L4.15:`
267			`! remainder is negative`
268			`addcc %o3,%o5,%o3`
269			`b 9f`
270			`add %o2, (-1*2-1), %o2`
271
272
273			`.L3.15:`
274			`! remainder is negative`
275			`addcc %o3,%o5,%o3`
276			`! depth 4, accumulated bits -3`
277			`bl .L4.13`
278			`srl %o5,1,%o5`
279			`! remainder is positive`
280			`subcc %o3,%o5,%o3`
281			`b 9f`
282			`add %o2, (-3*2+1), %o2`
283
284			`.L4.13:`
285			`! remainder is negative`
286			`addcc %o3,%o5,%o3`
287			`b 9f`
288			`add %o2, (-3*2-1), %o2`
289
290
291
292			`.L2.15:`
293			`! remainder is negative`
294			`addcc %o3,%o5,%o3`
295			`! depth 3, accumulated bits -3`
296			`bl .L3.13`
297			`srl %o5,1,%o5`
298			`! remainder is positive`
299			`subcc %o3,%o5,%o3`
300			`! depth 4, accumulated bits -5`
301			`bl .L4.11`
302			`srl %o5,1,%o5`
303			`! remainder is positive`
304			`subcc %o3,%o5,%o3`
305			`b 9f`
306			`add %o2, (-5*2+1), %o2`
307
308			`.L4.11:`
309			`! remainder is negative`
310			`addcc %o3,%o5,%o3`
311			`b 9f`
312			`add %o2, (-5*2-1), %o2`
313
314
315			`.L3.13:`
316			`! remainder is negative`
317			`addcc %o3,%o5,%o3`
318			`! depth 4, accumulated bits -7`
319			`bl .L4.9`
320			`srl %o5,1,%o5`
321			`! remainder is positive`
322			`subcc %o3,%o5,%o3`
323			`b 9f`
324			`add %o2, (-7*2+1), %o2`
325
326			`.L4.9:`
327			`! remainder is negative`
328			`addcc %o3,%o5,%o3`
329			`b 9f`
330			`add %o2, (-7*2-1), %o2`
331
332
333
334
335			`9:`
336			`.Lend_regular_divide:`
337			`subcc %o4, 1, %o4`
338			`bge .Ldivloop`
339			`tst %o3`
340			`bl,a .Lgot_result`
341			`! non-restoring fixup here (one instruction only!)`
342			`add %o3, %o1, %o3`
343
344
345			`.Lgot_result:`
346
347			`retl`
348			`mov %o3, %o0`
349
350			`.size .urem , . -.urem`