comrogue-pi/kernel/lib/divide.S

478 lines
10 KiB
ArmAsm
Raw Permalink Normal View History

/*
* Portions of this file are subject to the following notices:
*
* Copyright (C) 2012 Andrew Turner
* All rights reserved.
*
* Copyright (c) 2012 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Matt Thomas of 3am Software Foundry.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/* Derived from FreeBSD libkern divsi3.S, uldivmod.S, and ldivmod.S, munged by Erbo to COMROGUE standards */
.globl __aeabi_uidiv
.globl __aeabi_uidivmod
__aeabi_uidiv:
__aeabi_uidivmod:
eor r0, r1, r0 /* r0 = r0 / r1; r1 = r0 % r1 */
eor r1, r0, r1
eor r0, r1, r0 /* r0 = r1 / r0; r1 = r1 % r0 */
cmp r0, #1
bcc .L_overflow
beq .L_divide_l0
mov ip, #0
movs r1, r1
bpl .L_divide_l1
orr ip, ip, #0x20000000 /* indicates r1 is negative */
movs r1, r1, lsr #1
orrcs ip, ip, #0x10000000 /* indicates bit 0 of r1 */
b .L_divide_l1
.L_divide_l0:
mov r0, r1 /* r0 == 1 */
mov r1, #0
bx lr
.globl __aeabi_idiv
.globl __aeabi_idivmod
__aeabi_idiv:
__aeabi_idivmod:
eor r0, r1, r0 /* r0 = r0 / r1; r1 = r0 % r1 */
eor r1, r0, r1
eor r0, r1, r0 /* r0 = r1 / r0; r1 = r1 % r0 */
cmp r0, #1
bcc .L_overflow
beq .L_divide_l0
ands ip, r0, #0x80000000
rsbmi r0, r0, #0
ands r2, r1, #0x80000000
eor ip, ip, r2
rsbmi r1, r1, #0
orr ip, r2, ip, lsr #1 /* bit 0x40000000 = negative division, bit 0x80000000 = negative remainder */
.L_divide_l1:
mov r2, #1
mov r3, #0
/*
* If the highest bit of the dividend is set, we have to be
* careful when shifting the divisor. Test this.
*/
movs r1, r1
bpl .L_old_code
/*
* At this point, the highest bit of r1 is known to be set.
* We abuse this below in the tst instructions.
*/
tst r1, r0 /*, lsl #0 */
bmi .L_divide_b1
tst r1, r0, lsl #1
bmi .L_divide_b2
tst r1, r0, lsl #2
bmi .L_divide_b3
tst r1, r0, lsl #3
bmi .L_divide_b4
tst r1, r0, lsl #4
bmi .L_divide_b5
tst r1, r0, lsl #5
bmi .L_divide_b6
tst r1, r0, lsl #6
bmi .L_divide_b7
tst r1, r0, lsl #7
bmi .L_divide_b8
tst r1, r0, lsl #8
bmi .L_divide_b9
tst r1, r0, lsl #9
bmi .L_divide_b10
tst r1, r0, lsl #10
bmi .L_divide_b11
tst r1, r0, lsl #11
bmi .L_divide_b12
tst r1, r0, lsl #12
bmi .L_divide_b13
tst r1, r0, lsl #13
bmi .L_divide_b14
tst r1, r0, lsl #14
bmi .L_divide_b15
tst r1, r0, lsl #15
bmi .L_divide_b16
tst r1, r0, lsl #16
bmi .L_divide_b17
tst r1, r0, lsl #17
bmi .L_divide_b18
tst r1, r0, lsl #18
bmi .L_divide_b19
tst r1, r0, lsl #19
bmi .L_divide_b20
tst r1, r0, lsl #20
bmi .L_divide_b21
tst r1, r0, lsl #21
bmi .L_divide_b22
tst r1, r0, lsl #22
bmi .L_divide_b23
tst r1, r0, lsl #23
bmi .L_divide_b24
tst r1, r0, lsl #24
bmi .L_divide_b25
tst r1, r0, lsl #25
bmi .L_divide_b26
tst r1, r0, lsl #26
bmi .L_divide_b27
tst r1, r0, lsl #27
bmi .L_divide_b28
tst r1, r0, lsl #28
bmi .L_divide_b29
tst r1, r0, lsl #29
bmi .L_divide_b30
tst r1, r0, lsl #30
bmi .L_divide_b31
/*
* instead of:
* tst r1, r0, lsl #31
* bmi .L_divide_b32
*/
b .L_divide_b32
.L_old_code:
cmp r1, r0
bcc .L_divide_b0
cmp r1, r0, lsl #1
bcc .L_divide_b1
cmp r1, r0, lsl #2
bcc .L_divide_b2
cmp r1, r0, lsl #3
bcc .L_divide_b3
cmp r1, r0, lsl #4
bcc .L_divide_b4
cmp r1, r0, lsl #5
bcc .L_divide_b5
cmp r1, r0, lsl #6
bcc .L_divide_b6
cmp r1, r0, lsl #7
bcc .L_divide_b7
cmp r1, r0, lsl #8
bcc .L_divide_b8
cmp r1, r0, lsl #9
bcc .L_divide_b9
cmp r1, r0, lsl #10
bcc .L_divide_b10
cmp r1, r0, lsl #11
bcc .L_divide_b11
cmp r1, r0, lsl #12
bcc .L_divide_b12
cmp r1, r0, lsl #13
bcc .L_divide_b13
cmp r1, r0, lsl #14
bcc .L_divide_b14
cmp r1, r0, lsl #15
bcc .L_divide_b15
cmp r1, r0, lsl #16
bcc .L_divide_b16
cmp r1, r0, lsl #17
bcc .L_divide_b17
cmp r1, r0, lsl #18
bcc .L_divide_b18
cmp r1, r0, lsl #19
bcc .L_divide_b19
cmp r1, r0, lsl #20
bcc .L_divide_b20
cmp r1, r0, lsl #21
bcc .L_divide_b21
cmp r1, r0, lsl #22
bcc .L_divide_b22
cmp r1, r0, lsl #23
bcc .L_divide_b23
cmp r1, r0, lsl #24
bcc .L_divide_b24
cmp r1, r0, lsl #25
bcc .L_divide_b25
cmp r1, r0, lsl #26
bcc .L_divide_b26
cmp r1, r0, lsl #27
bcc .L_divide_b27
cmp r1, r0, lsl #28
bcc .L_divide_b28
cmp r1, r0, lsl #29
bcc .L_divide_b29
cmp r1, r0, lsl #30
bcc .L_divide_b30
.L_divide_b32:
cmp r1, r0, lsl #31
subhs r1, r1,r0, lsl #31
addhs r3, r3,r2, lsl #31
.L_divide_b31:
cmp r1, r0, lsl #30
subhs r1, r1,r0, lsl #30
addhs r3, r3,r2, lsl #30
.L_divide_b30:
cmp r1, r0, lsl #29
subhs r1, r1,r0, lsl #29
addhs r3, r3,r2, lsl #29
.L_divide_b29:
cmp r1, r0, lsl #28
subhs r1, r1,r0, lsl #28
addhs r3, r3,r2, lsl #28
.L_divide_b28:
cmp r1, r0, lsl #27
subhs r1, r1,r0, lsl #27
addhs r3, r3,r2, lsl #27
.L_divide_b27:
cmp r1, r0, lsl #26
subhs r1, r1,r0, lsl #26
addhs r3, r3,r2, lsl #26
.L_divide_b26:
cmp r1, r0, lsl #25
subhs r1, r1,r0, lsl #25
addhs r3, r3,r2, lsl #25
.L_divide_b25:
cmp r1, r0, lsl #24
subhs r1, r1,r0, lsl #24
addhs r3, r3,r2, lsl #24
.L_divide_b24:
cmp r1, r0, lsl #23
subhs r1, r1,r0, lsl #23
addhs r3, r3,r2, lsl #23
.L_divide_b23:
cmp r1, r0, lsl #22
subhs r1, r1,r0, lsl #22
addhs r3, r3,r2, lsl #22
.L_divide_b22:
cmp r1, r0, lsl #21
subhs r1, r1,r0, lsl #21
addhs r3, r3,r2, lsl #21
.L_divide_b21:
cmp r1, r0, lsl #20
subhs r1, r1,r0, lsl #20
addhs r3, r3,r2, lsl #20
.L_divide_b20:
cmp r1, r0, lsl #19
subhs r1, r1,r0, lsl #19
addhs r3, r3,r2, lsl #19
.L_divide_b19:
cmp r1, r0, lsl #18
subhs r1, r1,r0, lsl #18
addhs r3, r3,r2, lsl #18
.L_divide_b18:
cmp r1, r0, lsl #17
subhs r1, r1,r0, lsl #17
addhs r3, r3,r2, lsl #17
.L_divide_b17:
cmp r1, r0, lsl #16
subhs r1, r1,r0, lsl #16
addhs r3, r3,r2, lsl #16
.L_divide_b16:
cmp r1, r0, lsl #15
subhs r1, r1,r0, lsl #15
addhs r3, r3,r2, lsl #15
.L_divide_b15:
cmp r1, r0, lsl #14
subhs r1, r1,r0, lsl #14
addhs r3, r3,r2, lsl #14
.L_divide_b14:
cmp r1, r0, lsl #13
subhs r1, r1,r0, lsl #13
addhs r3, r3,r2, lsl #13
.L_divide_b13:
cmp r1, r0, lsl #12
subhs r1, r1,r0, lsl #12
addhs r3, r3,r2, lsl #12
.L_divide_b12:
cmp r1, r0, lsl #11
subhs r1, r1,r0, lsl #11
addhs r3, r3,r2, lsl #11
.L_divide_b11:
cmp r1, r0, lsl #10
subhs r1, r1,r0, lsl #10
addhs r3, r3,r2, lsl #10
.L_divide_b10:
cmp r1, r0, lsl #9
subhs r1, r1,r0, lsl #9
addhs r3, r3,r2, lsl #9
.L_divide_b9:
cmp r1, r0, lsl #8
subhs r1, r1,r0, lsl #8
addhs r3, r3,r2, lsl #8
.L_divide_b8:
cmp r1, r0, lsl #7
subhs r1, r1,r0, lsl #7
addhs r3, r3,r2, lsl #7
.L_divide_b7:
cmp r1, r0, lsl #6
subhs r1, r1,r0, lsl #6
addhs r3, r3,r2, lsl #6
.L_divide_b6:
cmp r1, r0, lsl #5
subhs r1, r1,r0, lsl #5
addhs r3, r3,r2, lsl #5
.L_divide_b5:
cmp r1, r0, lsl #4
subhs r1, r1,r0, lsl #4
addhs r3, r3,r2, lsl #4
.L_divide_b4:
cmp r1, r0, lsl #3
subhs r1, r1,r0, lsl #3
addhs r3, r3,r2, lsl #3
.L_divide_b3:
cmp r1, r0, lsl #2
subhs r1, r1,r0, lsl #2
addhs r3, r3,r2, lsl #2
.L_divide_b2:
cmp r1, r0, lsl #1
subhs r1, r1,r0, lsl #1
addhs r3, r3,r2, lsl #1
.L_divide_b1:
cmp r1, r0
subhs r1, r1, r0
addhs r3, r3, r2
.L_divide_b0:
tst ip, #0x20000000
bne .L_udivide_l1
mov r0, r3
cmp ip, #0
rsbmi r1, r1, #0
movs ip, ip, lsl #1
bicmi r0, r0, #0x80000000 /* Fix incase we divided 0x80000000 */
rsbmi r0, r0, #0
bx lr
.L_udivide_l1:
tst ip, #0x10000000
mov r1, r1, lsl #1
orrne r1, r1, #1
mov r3, r3, lsl #1
cmp r1, r0
subhs r1, r1, r0
addhs r3, r3, r2
mov r0, r3
bx lr
.L_overflow:
/* TODO: cause an exception or something? */
mvn r0, #0
bx lr
.globl __aeabi_uldivmod
__aeabi_uldivmod:
push {r4,lr}
sub sp, sp, #8
mov r4, sp
push {r4}
bl __qdivrem
pop {r4}
/*
* The remainder is already on the stack just waiting to be popped
* into r2/r3.
*/
pop {r2-r4,lr}
bx lr
.globl __aeabi_ldivmod
__aeabi_ldivmod:
push {r4-r5, sl, lr}
mov r5, #0 /* r5 = negative indicator */
cmp r3, #0
bge 2f
eor r5, r5, #1 /* flip quotient sign */
bl .Lnegate_b
bcs .Lmaxdenom
2:
cmp r1, #0
/* bge 3f */
eorlt r5, r5, #3 /* flip quotient sign, flip remainder sign */
bllt .Lnegate_a
3:
/*
* Arguments are setup, allocate some stack for the remainder
* and call __qdivrem for the heavy lifting.
*/
sub sp, sp, #8
mov r4, sp /* pointer to remainder */
push {r4}
bl __qdivrem
pop {r4}
teq r5, #0 /* any signs to flip? */
/*
* The quotient is already in the right place and neither value
* needs its sign flipped.
*/
popeq {r2-r5, sl, lr}
bxeq lr
pop {r2, r3}
tst r5, #2 /* does remainder need to be negative? */
bleq .Lnegate_b
tst r5, #1 /* does quotient need to be negative? */
bleq .Lnegate_a
pop {r4-r5, sl, lr}
bx lr
.Lnegate_a:
rsbs r0, r0, #0
rsc r1, r1, #0
bx lr
.Lnegate_b:
rsbs r2, r2, #0
rsc r3, r3, #0
bx lr
.Lmaxdenom:
/*
* We had a carry so the denominator must have INT64_MIN
* Also BLO and BHI never changed values so we can use
* them to see if the numerator has the same value. We
* don't have to worry about sign.
*/
teq r3, r1
teqeq r2, r0
bne 1f
/*
* They were equal, so we return a quotient of 1 and remainder of 0.
*/
mov r0, #1
mov r1, #0
mov r2, #0
mov r3, #0
pop {r4-r5, sl, lr}
bx lr
/*
* Our remainder must be the numerator and our quotient is 0.
*/
1: mov r2, r0
mov r3, r1
mov r0, #0
mov r1, #0
pop {r4-r5, sl, lr}
bx lr