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

Subversion Repositories or1k_old

[/] [or1k_old/] [trunk/] [linux/] [uClibc/] [libc/] [sysdeps/] [linux/] [sparc/] [rem.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:
 *  .rem        name of function to generate
 *  rem         rem=div => %o0 / %o1; rem=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.
 */
 
 
 
#include 
 
 
.global   .rem;
.align 4;
.type  .rem ,@function;
 
.rem:
        ! compute sign of result; if neither is negative, no problem
        orcc    %o1, %o0, %g0   ! either negative?
        bge     2f                      ! no, go do the divide
        mov     %o0, %g3                ! sign of remainder matches %o0
        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      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:
        ! check to see if answer should be < 0
        tst     %g3
        bl,a    1f
        sub %g0, %o3, %o3
1:
        retl
        mov %o3, %o0
 
.size .rem,.-.rem;

Browse

Tools

Subversion Repositories or1k_old

[/] [or1k_old/] [trunk/] [linux/] [uClibc/] [libc/] [sysdeps/] [linux/] [sparc/] [rem.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			`* .rem name of function to generate`
12			`* rem rem=div => %o0 / %o1; rem=rem => %o0 % %o1`
13			`* true true=true => signed; true=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 .rem;`
44			`.align 4;`
45			`.type .rem ,@function;`
46
47			`.rem:`
48			`! compute sign of result; if neither is negative, no problem`
49			`orcc %o1, %o0, %g0 ! either negative?`
50			`bge 2f ! no, go do the divide`
51			`mov %o0, %g3 ! sign of remainder matches %o0`
52			`tst %o1`
53			`bge 1f`
54			`tst %o0`
55			`! %o1 is definitely negative; %o0 might also be negative`
56			`bge 2f ! if %o0 not negative...`
57			`sub %g0, %o1, %o1 ! in any case, make %o1 nonneg`
58			`1: ! %o0 is negative, %o1 is nonnegative`
59			`sub %g0, %o0, %o0 ! make %o0 nonnegative`
60			`2:`
61
62			`! Ready to divide. Compute size of quotient; scale comparand.`
63			`orcc %o1, %g0, %o5`
64			`bne 1f`
65			`mov %o0, %o3`
66
67			`! Divide by zero trap. If it returns, return 0 (about as`
68			`! wrong as possible, but that is what SunOS does...).`
69			`ta 0x02`
70			`retl`
71			`clr %o0`
72
73			`1:`
74			`cmp %o3, %o5 ! if %o1 exceeds %o0, done`
75			`blu .Lgot_result ! (and algorithm fails otherwise)`
76			`clr %o2`
77			`sethi %hi(1 << (32 - 4 - 1)), %g1`
78			`cmp %o3, %g1`
79			`blu .Lnot_really_big`
80			`clr %o4`
81
82			`! Here the dividend is >= 2**(31-N) or so. We must be careful here,`
83			`! as our usual N-at-a-shot divide step will cause overflow and havoc.`
84			`! The number of bits in the result here is N*ITER+SC, where SC <= N.`
85			`! Compute ITER in an unorthodox manner: know we need to shift V into`
86			`! the top decade: so do not even bother to compare to R.`
87			`1:`
88			`cmp %o5, %g1`
89			`bgeu 3f`
90			`mov 1, %g2`
91			`sll %o5, 4, %o5`
92			`b 1b`
93			`add %o4, 1, %o4`
94
95			`! Now compute %g2.`
96			`2: addcc %o5, %o5, %o5`
97			`bcc .Lnot_too_big`
98			`add %g2, 1, %g2`
99
100			`! We get here if the %o1 overflowed while shifting.`
101			`! This means that %o3 has the high-order bit set.`
102			`! Restore %o5 and subtract from %o3.`
103			`sll %g1, 4, %g1 ! high order bit`
104			`srl %o5, 1, %o5 ! rest of %o5`
105			`add %o5, %g1, %o5`
106			`b .Ldo_single_div`
107			`sub %g2, 1, %g2`
108
109			`.Lnot_too_big:`
110			`3: cmp %o5, %o3`
111			`blu 2b`
112			`nop`
113			`be .Ldo_single_div`
114			`nop`
115			`/* NB: these are commented out in the V8-Sparc manual as well */`
116			`/* (I do not understand this) */`
117			`! %o5 > %o3: went too far: back up 1 step`
118			`! srl %o5, 1, %o5`
119			`! dec %g2`
120			`! do single-bit divide steps`
121			`!`
122			`! We have to be careful here. We know that %o3 >= %o5, so we can do the`
123			`! first divide step without thinking. BUT, the others are conditional,`
124			`! and are only done if %o3 >= 0. Because both %o3 and %o5 may have the high-`
125			`! order bit set in the first step, just falling into the regular`
126			`! division loop will mess up the first time around.`
127			`! So we unroll slightly...`
128			`.Ldo_single_div:`
129			`subcc %g2, 1, %g2`
130			`bl .Lend_regular_divide`
131			`nop`
132			`sub %o3, %o5, %o3`
133			`mov 1, %o2`
134			`b .Lend_single_divloop`
135			`nop`
136			`.Lsingle_divloop:`
137			`sll %o2, 1, %o2`
138			`bl 1f`
139			`srl %o5, 1, %o5`
140			`! %o3 >= 0`
141			`sub %o3, %o5, %o3`
142			`b 2f`
143			`add %o2, 1, %o2`
144			`1: ! %o3 < 0`
145			`add %o3, %o5, %o3`
146			`sub %o2, 1, %o2`
147			`2:`
148			`.Lend_single_divloop:`
149			`subcc %g2, 1, %g2`
150			`bge .Lsingle_divloop`
151			`tst %o3`
152			`b,a .Lend_regular_divide`
153
154			`.Lnot_really_big:`
155			`1:`
156			`sll %o5, 4, %o5`
157			`cmp %o5, %o3`
158			`bleu 1b`
159			`addcc %o4, 1, %o4`
160			`be .Lgot_result`
161			`sub %o4, 1, %o4`
162
163			`tst %o3 ! set up for initial iteration`
164			`.Ldivloop:`
165			`sll %o2, 4, %o2`
166			`! depth 1, accumulated bits 0`
167			`bl .L1.16`
168			`srl %o5,1,%o5`
169			`! remainder is positive`
170			`subcc %o3,%o5,%o3`
171			`! depth 2, accumulated bits 1`
172			`bl .L2.17`
173			`srl %o5,1,%o5`
174			`! remainder is positive`
175			`subcc %o3,%o5,%o3`
176			`! depth 3, accumulated bits 3`
177			`bl .L3.19`
178			`srl %o5,1,%o5`
179			`! remainder is positive`
180			`subcc %o3,%o5,%o3`
181			`! depth 4, accumulated bits 7`
182			`bl .L4.23`
183			`srl %o5,1,%o5`
184			`! remainder is positive`
185			`subcc %o3,%o5,%o3`
186			`b 9f`
187			`add %o2, (7*2+1), %o2`
188
189			`.L4.23:`
190			`! remainder is negative`
191			`addcc %o3,%o5,%o3`
192			`b 9f`
193			`add %o2, (7*2-1), %o2`
194
195
196			`.L3.19:`
197			`! remainder is negative`
198			`addcc %o3,%o5,%o3`
199			`! depth 4, accumulated bits 5`
200			`bl .L4.21`
201			`srl %o5,1,%o5`
202			`! remainder is positive`
203			`subcc %o3,%o5,%o3`
204			`b 9f`
205			`add %o2, (5*2+1), %o2`
206
207			`.L4.21:`
208			`! remainder is negative`
209			`addcc %o3,%o5,%o3`
210			`b 9f`
211			`add %o2, (5*2-1), %o2`
212
213
214
215			`.L2.17:`
216			`! remainder is negative`
217			`addcc %o3,%o5,%o3`
218			`! depth 3, accumulated bits 1`
219			`bl .L3.17`
220			`srl %o5,1,%o5`
221			`! remainder is positive`
222			`subcc %o3,%o5,%o3`
223			`! depth 4, accumulated bits 3`
224			`bl .L4.19`
225			`srl %o5,1,%o5`
226			`! remainder is positive`
227			`subcc %o3,%o5,%o3`
228			`b 9f`
229			`add %o2, (3*2+1), %o2`
230
231			`.L4.19:`
232			`! remainder is negative`
233			`addcc %o3,%o5,%o3`
234			`b 9f`
235			`add %o2, (3*2-1), %o2`
236
237
238			`.L3.17:`
239			`! remainder is negative`
240			`addcc %o3,%o5,%o3`
241			`! depth 4, accumulated bits 1`
242			`bl .L4.17`
243			`srl %o5,1,%o5`
244			`! remainder is positive`
245			`subcc %o3,%o5,%o3`
246			`b 9f`
247			`add %o2, (1*2+1), %o2`
248
249			`.L4.17:`
250			`! remainder is negative`
251			`addcc %o3,%o5,%o3`
252			`b 9f`
253			`add %o2, (1*2-1), %o2`
254
255
256
257
258			`.L1.16:`
259			`! remainder is negative`
260			`addcc %o3,%o5,%o3`
261			`! depth 2, accumulated bits -1`
262			`bl .L2.15`
263			`srl %o5,1,%o5`
264			`! remainder is positive`
265			`subcc %o3,%o5,%o3`
266			`! depth 3, accumulated bits -1`
267			`bl .L3.15`
268			`srl %o5,1,%o5`
269			`! remainder is positive`
270			`subcc %o3,%o5,%o3`
271			`! depth 4, accumulated bits -1`
272			`bl .L4.15`
273			`srl %o5,1,%o5`
274			`! remainder is positive`
275			`subcc %o3,%o5,%o3`
276			`b 9f`
277			`add %o2, (-1*2+1), %o2`
278
279			`.L4.15:`
280			`! remainder is negative`
281			`addcc %o3,%o5,%o3`
282			`b 9f`
283			`add %o2, (-1*2-1), %o2`
284
285
286			`.L3.15:`
287			`! remainder is negative`
288			`addcc %o3,%o5,%o3`
289			`! depth 4, accumulated bits -3`
290			`bl .L4.13`
291			`srl %o5,1,%o5`
292			`! remainder is positive`
293			`subcc %o3,%o5,%o3`
294			`b 9f`
295			`add %o2, (-3*2+1), %o2`
296
297			`.L4.13:`
298			`! remainder is negative`
299			`addcc %o3,%o5,%o3`
300			`b 9f`
301			`add %o2, (-3*2-1), %o2`
302
303
304
305			`.L2.15:`
306			`! remainder is negative`
307			`addcc %o3,%o5,%o3`
308			`! depth 3, accumulated bits -3`
309			`bl .L3.13`
310			`srl %o5,1,%o5`
311			`! remainder is positive`
312			`subcc %o3,%o5,%o3`
313			`! depth 4, accumulated bits -5`
314			`bl .L4.11`
315			`srl %o5,1,%o5`
316			`! remainder is positive`
317			`subcc %o3,%o5,%o3`
318			`b 9f`
319			`add %o2, (-5*2+1), %o2`
320
321			`.L4.11:`
322			`! remainder is negative`
323			`addcc %o3,%o5,%o3`
324			`b 9f`
325			`add %o2, (-5*2-1), %o2`
326
327
328			`.L3.13:`
329			`! remainder is negative`
330			`addcc %o3,%o5,%o3`
331			`! depth 4, accumulated bits -7`
332			`bl .L4.9`
333			`srl %o5,1,%o5`
334			`! remainder is positive`
335			`subcc %o3,%o5,%o3`
336			`b 9f`
337			`add %o2, (-7*2+1), %o2`
338
339			`.L4.9:`
340			`! remainder is negative`
341			`addcc %o3,%o5,%o3`
342			`b 9f`
343			`add %o2, (-7*2-1), %o2`
344
345
346
347
348			`9:`
349			`.Lend_regular_divide:`
350			`subcc %o4, 1, %o4`
351			`bge .Ldivloop`
352			`tst %o3`
353			`bl,a .Lgot_result`
354			`! non-restoring fixup here (one instruction only!)`
355			`add %o3, %o1, %o3`
356
357
358			`.Lgot_result:`
359			`! check to see if answer should be < 0`
360			`tst %g3`
361			`bl,a 1f`
362			`sub %g0, %o3, %o3`
363			`1:`
364			`retl`
365			`mov %o3, %o0`
366
367			`.size .rem,.-.rem;`