xref: /openbmc/u-boot/arch/arm/lib/uldivmod.S (revision cbd2fba1)
1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * Copyright 2010, Google Inc.
4 *
5 * Brought in from coreboot uldivmod.S
6 */
7
8#include <linux/linkage.h>
9#include <asm/assembler.h>
10
11/*
12 * A, Q = r0 + (r1 << 32)
13 * B, R = r2 + (r3 << 32)
14 * A / B = Q ... R
15 */
16
17A_0	.req	r0
18A_1	.req	r1
19B_0	.req	r2
20B_1	.req	r3
21C_0	.req	r4
22C_1	.req	r5
23D_0	.req	r6
24D_1	.req	r7
25
26Q_0	.req	r0
27Q_1	.req	r1
28R_0	.req	r2
29R_1	.req	r3
30
31THUMB(
32TMP	.req	r8
33)
34
35.pushsection .text.__aeabi_uldivmod, "ax"
36ENTRY(__aeabi_uldivmod)
37
38	stmfd	sp!, {r4, r5, r6, r7, THUMB(TMP,) lr}
39	@ Test if B == 0
40	orrs	ip, B_0, B_1		@ Z set -> B == 0
41	beq	L_div_by_0
42	@ Test if B is power of 2: (B & (B - 1)) == 0
43	subs	C_0, B_0, #1
44	sbc	C_1, B_1, #0
45	tst	C_0, B_0
46	tsteq	B_1, C_1
47	beq	L_pow2
48	@ Test if A_1 == B_1 == 0
49	orrs	ip, A_1, B_1
50	beq	L_div_32_32
51
52L_div_64_64:
53/* CLZ only exists in ARM architecture version 5 and above. */
54#ifdef HAVE_CLZ
55	mov	C_0, #1
56	mov	C_1, #0
57	@ D_0 = clz A
58	teq	A_1, #0
59	clz	D_0, A_1
60	clzeq	ip, A_0
61	addeq	D_0, D_0, ip
62	@ D_1 = clz B
63	teq	B_1, #0
64	clz	D_1, B_1
65	clzeq	ip, B_0
66	addeq	D_1, D_1, ip
67	@ if clz B - clz A > 0
68	subs	D_0, D_1, D_0
69	bls	L_done_shift
70	@ B <<= (clz B - clz A)
71	subs	D_1, D_0, #32
72	rsb	ip, D_0, #32
73	movmi	B_1, B_1, lsl D_0
74ARM(	orrmi	B_1, B_1, B_0, lsr ip	)
75THUMB(	lsrmi	TMP, B_0, ip		)
76THUMB(	orrmi	B_1, B_1, TMP		)
77	movpl	B_1, B_0, lsl D_1
78	mov	B_0, B_0, lsl D_0
79	@ C = 1 << (clz B - clz A)
80	movmi	C_1, C_1, lsl D_0
81ARM(	orrmi	C_1, C_1, C_0, lsr ip	)
82THUMB(	lsrmi	TMP, C_0, ip		)
83THUMB(	orrmi	C_1, C_1, TMP		)
84	movpl	C_1, C_0, lsl D_1
85	mov	C_0, C_0, lsl D_0
86L_done_shift:
87	mov	D_0, #0
88	mov	D_1, #0
89	@ C: current bit; D: result
90#else
91	@ C: current bit; D: result
92	mov	C_0, #1
93	mov	C_1, #0
94	mov	D_0, #0
95	mov	D_1, #0
96L_lsl_4:
97	cmp	B_1, #0x10000000
98	cmpcc	B_1, A_1
99	cmpeq	B_0, A_0
100	bcs	L_lsl_1
101	@ B <<= 4
102	mov	B_1, B_1, lsl #4
103	orr	B_1, B_1, B_0, lsr #28
104	mov	B_0, B_0, lsl #4
105	@ C <<= 4
106	mov	C_1, C_1, lsl #4
107	orr	C_1, C_1, C_0, lsr #28
108	mov	C_0, C_0, lsl #4
109	b	L_lsl_4
110L_lsl_1:
111	cmp	B_1, #0x80000000
112	cmpcc	B_1, A_1
113	cmpeq	B_0, A_0
114	bcs	L_subtract
115	@ B <<= 1
116	mov	B_1, B_1, lsl #1
117	orr	B_1, B_1, B_0, lsr #31
118	mov	B_0, B_0, lsl #1
119	@ C <<= 1
120	mov	C_1, C_1, lsl #1
121	orr	C_1, C_1, C_0, lsr #31
122	mov	C_0, C_0, lsl #1
123	b	L_lsl_1
124#endif
125L_subtract:
126	@ if A >= B
127	cmp	A_1, B_1
128	cmpeq	A_0, B_0
129	bcc	L_update
130	@ A -= B
131	subs	A_0, A_0, B_0
132	sbc	A_1, A_1, B_1
133	@ D |= C
134	orr	D_0, D_0, C_0
135	orr	D_1, D_1, C_1
136L_update:
137	@ if A == 0: break
138	orrs	ip, A_1, A_0
139	beq	L_exit
140	@ C >>= 1
141	movs	C_1, C_1, lsr #1
142	movs	C_0, C_0, rrx
143	@ if C == 0: break
144	orrs	ip, C_1, C_0
145	beq	L_exit
146	@ B >>= 1
147	movs	B_1, B_1, lsr #1
148	mov	B_0, B_0, rrx
149	b	L_subtract
150L_exit:
151	@ Note: A, B & Q, R are aliases
152	mov	R_0, A_0
153	mov	R_1, A_1
154	mov	Q_0, D_0
155	mov	Q_1, D_1
156	ldmfd	sp!, {r4, r5, r6, r7, THUMB(TMP,) pc}
157
158L_div_32_32:
159	@ Note:	A_0 &	r0 are aliases
160	@	Q_1	r1
161	mov	r1, B_0
162	bl	__aeabi_uidivmod
163	mov	R_0, r1
164	mov	R_1, #0
165	mov	Q_1, #0
166	ldmfd	sp!, {r4, r5, r6, r7, THUMB(TMP,) pc}
167
168L_pow2:
169#ifdef HAVE_CLZ
170	@ Note: A, B and Q, R are aliases
171	@ R = A & (B - 1)
172	and	C_0, A_0, C_0
173	and	C_1, A_1, C_1
174	@ Q = A >> log2(B)
175	@ Note: B must not be 0 here!
176	clz	D_0, B_0
177	add	D_1, D_0, #1
178	rsbs	D_0, D_0, #31
179	bpl	L_1
180	clz	D_0, B_1
181	rsb	D_0, D_0, #31
182	mov	A_0, A_1, lsr D_0
183	add	D_0, D_0, #32
184L_1:
185	movpl	A_0, A_0, lsr D_0
186ARM(	orrpl	A_0, A_0, A_1, lsl D_1	)
187THUMB(	lslpl	TMP, A_1, D_1		)
188THUMB(	orrpl	A_0, A_0, TMP		)
189	mov	A_1, A_1, lsr D_0
190	@ Mov back C to R
191	mov	R_0, C_0
192	mov	R_1, C_1
193	ldmfd	sp!, {r4, r5, r6, r7, THUMB(TMP,) pc}
194#else
195	@ Note: A, B and Q, R are aliases
196	@ R = A & (B - 1)
197	and	C_0, A_0, C_0
198	and	C_1, A_1, C_1
199	@ Q = A >> log2(B)
200	@ Note: B must not be 0 here!
201	@ Count the leading zeroes in B.
202	mov	D_0, #0
203	orrs	B_0, B_0, B_0
204	@ If B is greater than 1 << 31, divide A and B by 1 << 32.
205	moveq	A_0, A_1
206	moveq	A_1, #0
207	moveq	B_0, B_1
208	@ Count the remaining leading zeroes in B.
209	movs	B_1, B_0, lsl #16
210	addeq	D_0, #16
211	moveq	B_0, B_0, lsr #16
212	tst	B_0, #0xff
213	addeq	D_0, #8
214	moveq	B_0, B_0, lsr #8
215	tst	B_0, #0xf
216	addeq	D_0, #4
217	moveq	B_0, B_0, lsr #4
218	tst	B_0, #0x3
219	addeq	D_0, #2
220	moveq	B_0, B_0, lsr #2
221	tst	B_0, #0x1
222	addeq	D_0, #1
223	@ Shift A to the right by the appropriate amount.
224	rsb	D_1, D_0, #32
225	mov	Q_0, A_0, lsr D_0
226 ARM(   orr     Q_0, Q_0, A_1, lsl D_1	)
227 THUMB(	lsl	A_1, D_1		)
228 THUMB(	orr	Q_0, A_1		)
229	mov	Q_1, A_1, lsr D_0
230	@ Move C to R
231	mov	R_0, C_0
232	mov	R_1, C_1
233	ldmfd	sp!, {r4, r5, r6, r7, THUMB(TMP,) pc}
234#endif
235
236L_div_by_0:
237	bl	__div0
238	@ As wrong as it could be
239	mov	Q_0, #0
240	mov	Q_1, #0
241	mov	R_0, #0
242	mov	R_1, #0
243	ldmfd	sp!, {r4, r5, r6, r7, THUMB(TMP,) pc}
244ENDPROC(__aeabi_uldivmod)
245.popsection
246