about summary refs log tree commit diff
path: root/sysdeps/powerpc/powerpc64/power9/strcmp.S
blob: 754d5088047095cf580445a4efb76c97e3c7afbc (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
/* Optimized strcmp implementation for PowerPC64/POWER9.
   Copyright (C) 2016 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, see
   <http://www.gnu.org/licenses/>.  */
#ifdef __LITTLE_ENDIAN__
#include <sysdep.h>

/* Implements the function

   int [r3] strcmp (const char *s1 [r3], const char *s2 [r4])

   The implementation uses unaligned doubleword access for first 32 bytes
   as in POWER8 patch and uses vectorised loops after that.  */

/* TODO: Change this to actual instructions when minimum binutils is upgraded
   to 2.27. Macros are defined below for these newer instructions in order
   to maintain compatibility.  */
# define VCTZLSBB(r,v) .long (0x10010602 | ((r)<<(32-11)) | ((v)<<(32-21)))

# define VEXTUBRX(t,a,b) .long (0x1000070d \
				| ((t)<<(32-11))  \
				| ((a)<<(32-16))  \
				| ((b)<<(32-21)) )

# define VCMPNEZB(t,a,b) .long (0x10000507 \
				| ((t)<<(32-11))  \
				| ((a)<<(32-16))  \
				| ((b)<<(32-21)) )

/* Get 16 bytes for unaligned case.
   reg1: Vector to hold next 16 bytes.
   reg2: Address to read from.
   reg3: Permute control vector.  */
# define GET16BYTES(reg1, reg2, reg3) \
	lvx	reg1, 0, reg2; \
	vperm	v8, v2, reg1, reg3; \
	vcmpequb.	v8, v0, v8; \
	beq	cr6, 1f; \
	vspltisb	v9, 0; \
	b	2f; \
	.align 4; \
1: \
	addi    r6, reg2, 16; \
	lvx     v9, 0, r6; \
2: \
	vperm   reg1, v9, reg1, reg3;

/* TODO: change this to .machine power9 when the minimum required binutils
   allows it.  */

	.machine  power7
EALIGN (strcmp, 4, 0)
	li	r0, 0

	/* Check if [s1]+32 or [s2]+32 will cross a 4K page boundary using
	   the code:

	    (((size_t) s1) % PAGE_SIZE > (PAGE_SIZE - ITER_SIZE))

	   with PAGE_SIZE being 4096 and ITER_SIZE begin 32.  */

	rldicl	r7, r3, 0, 52
	rldicl	r9, r4, 0, 52
	cmpldi	cr7, r7, 4096-32
	bgt	cr7, L(pagecross_check)
	cmpldi	cr5, r9, 4096-32
	bgt	cr5, L(pagecross_check)

	/* For short strings up to 32 bytes,  load both s1 and s2 using
	   unaligned dwords and compare.  */
	ld	r8, 0(r3)
	ld	r10, 0(r4)
	cmpb	r12, r8, r0
	cmpb	r11, r8, r10
	orc.	r9, r12, r11
	bne	cr0, L(different_nocmpb)

	ld	r8, 8(r3)
	ld	r10, 8(r4)
	cmpb	r12, r8, r0
	cmpb	r11, r8, r10
	orc.	r9, r12, r11
	bne	cr0, L(different_nocmpb)

	ld	r8, 16(r3)
	ld	r10, 16(r4)
	cmpb	r12, r8, r0
	cmpb	r11, r8, r10
	orc.	r9, r12, r11
	bne	cr0, L(different_nocmpb)

	ld	r8, 24(r3)
	ld	r10, 24(r4)
	cmpb	r12, r8, r0
	cmpb	r11, r8, r10
	orc.	r9, r12, r11
	bne	cr0, L(different_nocmpb)

	addi	r7, r3, 32
	addi	r4, r4, 32

L(align):
	/* Now it has checked for first 32 bytes.  */
	vspltisb	v0, 0
	vspltisb	v2, -1
	lvsr	v6, 0, r4   /* Compute mask.  */
	or	r5, r4, r7
	andi.	r5, r5, 0xF
	beq	cr0, L(aligned)
	andi.	r5, r7, 0xF
	beq	cr0, L(s1_align)
	lvsr	v10, 0, r7   /* Compute mask.  */

	/* Both s1 and s2 are unaligned.  */
	GET16BYTES(v4, r7, v10)
	GET16BYTES(v5, r4, v6)
	VCMPNEZB(v7, v5, v4)
	beq	cr6, L(match)
	b	L(different)

	/* Align s1 to qw and adjust s2 address.  */
	.align  4
L(match):
	clrldi	r6, r7, 60
	subfic	r5, r6, 16
	add	r7, r7, r5
	add	r4, r4, r5
	andi.	r5, r4, 0xF
	beq	cr0, L(aligned)
	lvsr	v6, 0, r4
	/* There are 2 loops depending on the input alignment.
	   Each loop gets 16 bytes from s1 and s2 and compares.
	   Loop until a mismatch or null occurs.  */
L(s1_align):
	lvx	v4, r7, r0
	GET16BYTES(v5, r4, v6)
	VCMPNEZB(v7, v5, v4)
	addi	r7, r7, 16
	addi	r4, r4, 16
	bne	cr6, L(different)

	lvx	v4, r7, r0
	GET16BYTES(v5, r4, v6)
	VCMPNEZB(v7, v5, v4)
	addi	r7, r7, 16
	addi	r4, r4, 16
	bne	cr6, L(different)

	lvx	v4, r7, r0
	GET16BYTES(v5, r4, v6)
	VCMPNEZB(v7, v5, v4)
	addi	r7, r7, 16
	addi	r4, r4, 16
	bne	cr6, L(different)

	lvx	v4, r7, r0
	GET16BYTES(v5, r4, v6)
	VCMPNEZB(v7, v5, v4)
	addi	r7, r7, 16
	addi	r4, r4, 16
	beq	cr6, L(s1_align)
	b	L(different)

	.align  4
L(aligned):
	lvx	v4, 0, r7
	lvx	v5, 0, r4
	VCMPNEZB(v7, v5, v4)
	addi	r7, r7, 16
	addi	r4, r4, 16
	bne	cr6, L(different)

	lvx	v4, 0, r7
	lvx	v5, 0, r4
	VCMPNEZB(v7, v5, v4)
	addi	r7, r7, 16
	addi	r4, r4, 16
	bne	cr6, L(different)

	lvx	v4, 0, r7
	lvx	v5, 0, r4
	VCMPNEZB(v7, v5, v4)
	addi	r7, r7, 16
	addi	r4, r4, 16
	bne	cr6, L(different)

	lvx	v4, 0, r7
	lvx	v5, 0, r4
	VCMPNEZB(v7, v5, v4)
	addi	r7, r7, 16
	addi	r4, r4, 16
	beq	cr6, L(aligned)

	/* Calculate and return the difference.  */
L(different):
	VCTZLSBB(r6, v7)
	VEXTUBRX(r5, r6, v4)
	VEXTUBRX(r4, r6, v5)
	subf	r3, r4, r5
	extsw	r3, r3
	blr

	.align  4
L(different_nocmpb):
	neg	r3, r9
	and	r9, r9, r3
	cntlzd	r9, r9
	subfic	r9, r9, 63
	srd	r3, r8, r9
	srd	r10, r10, r9
	rldicl	r10, r10, 0, 56
	rldicl	r3, r3, 0, 56
	subf	r3, r10, r3
	extsw	r3, r3
	blr

	.align	4
L(pagecross_check):
	subfic	r9, r9, 4096
	subfic	r7, r7, 4096
	cmpld	cr7, r7, r9
	bge	cr7, L(pagecross)
	mr	r7, r9

	/* If unaligned 16 bytes reads across a 4K page boundary, it uses
	   a simple byte a byte comparison until the page alignment for s1
	   is reached.  */
L(pagecross):
	add	r7, r3, r7
	subf	r9, r3, r7
	mtctr	r9

	.align	4
L(pagecross_loop):
	/* Loads a byte from s1 and s2, compare if *s1 is equal to *s2
	   and if *s1 is '\0'.  */
	lbz	r9, 0(r3)
	lbz	r10, 0(r4)
	addi	r3, r3, 1
	addi	r4, r4, 1
	cmplw	cr7, r9, r10
	cmpdi	cr5, r9, r0
	bne	cr7, L(pagecross_ne)
	beq	cr5, L(pagecross_nullfound)
	bdnz	L(pagecross_loop)
	b	L(align)

	.align	4
L(pagecross_ne):
	extsw	r3, r9
	mr	r9, r10
L(pagecross_retdiff):
	subf	r9, r9, r3
	extsw	r3, r9
	blr

	.align	4
L(pagecross_nullfound):
	li	r3, 0
	b	L(pagecross_retdiff)
END (strcmp)
libc_hidden_builtin_def (strcmp)
#else
#include <sysdeps/powerpc/powerpc64/power8/strcmp.S>
#endif