1
2
3
4
5
6
7#include <linux/hardirq.h>
8#include <asm-generic/xor.h>
9#include <asm/hwcap.h>
10#include <asm/neon.h>
11
12#define __XOR(a1, a2) a1 ^= a2
13
14#define GET_BLOCK_2(dst) \
15 __asm__("ldmia %0, {%1, %2}" \
16 : "=r" (dst), "=r" (a1), "=r" (a2) \
17 : "0" (dst))
18
19#define GET_BLOCK_4(dst) \
20 __asm__("ldmia %0, {%1, %2, %3, %4}" \
21 : "=r" (dst), "=r" (a1), "=r" (a2), "=r" (a3), "=r" (a4) \
22 : "0" (dst))
23
24#define XOR_BLOCK_2(src) \
25 __asm__("ldmia %0!, {%1, %2}" \
26 : "=r" (src), "=r" (b1), "=r" (b2) \
27 : "0" (src)); \
28 __XOR(a1, b1); __XOR(a2, b2);
29
30#define XOR_BLOCK_4(src) \
31 __asm__("ldmia %0!, {%1, %2, %3, %4}" \
32 : "=r" (src), "=r" (b1), "=r" (b2), "=r" (b3), "=r" (b4) \
33 : "0" (src)); \
34 __XOR(a1, b1); __XOR(a2, b2); __XOR(a3, b3); __XOR(a4, b4)
35
36#define PUT_BLOCK_2(dst) \
37 __asm__ __volatile__("stmia %0!, {%2, %3}" \
38 : "=r" (dst) \
39 : "0" (dst), "r" (a1), "r" (a2))
40
41#define PUT_BLOCK_4(dst) \
42 __asm__ __volatile__("stmia %0!, {%2, %3, %4, %5}" \
43 : "=r" (dst) \
44 : "0" (dst), "r" (a1), "r" (a2), "r" (a3), "r" (a4))
45
46static void
47xor_arm4regs_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
48{
49 unsigned int lines = bytes / sizeof(unsigned long) / 4;
50 register unsigned int a1 __asm__("r4");
51 register unsigned int a2 __asm__("r5");
52 register unsigned int a3 __asm__("r6");
53 register unsigned int a4 __asm__("r7");
54 register unsigned int b1 __asm__("r8");
55 register unsigned int b2 __asm__("r9");
56 register unsigned int b3 __asm__("ip");
57 register unsigned int b4 __asm__("lr");
58
59 do {
60 GET_BLOCK_4(p1);
61 XOR_BLOCK_4(p2);
62 PUT_BLOCK_4(p1);
63 } while (--lines);
64}
65
66static void
67xor_arm4regs_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
68 unsigned long *p3)
69{
70 unsigned int lines = bytes / sizeof(unsigned long) / 4;
71 register unsigned int a1 __asm__("r4");
72 register unsigned int a2 __asm__("r5");
73 register unsigned int a3 __asm__("r6");
74 register unsigned int a4 __asm__("r7");
75 register unsigned int b1 __asm__("r8");
76 register unsigned int b2 __asm__("r9");
77 register unsigned int b3 __asm__("ip");
78 register unsigned int b4 __asm__("lr");
79
80 do {
81 GET_BLOCK_4(p1);
82 XOR_BLOCK_4(p2);
83 XOR_BLOCK_4(p3);
84 PUT_BLOCK_4(p1);
85 } while (--lines);
86}
87
88static void
89xor_arm4regs_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
90 unsigned long *p3, unsigned long *p4)
91{
92 unsigned int lines = bytes / sizeof(unsigned long) / 2;
93 register unsigned int a1 __asm__("r8");
94 register unsigned int a2 __asm__("r9");
95 register unsigned int b1 __asm__("ip");
96 register unsigned int b2 __asm__("lr");
97
98 do {
99 GET_BLOCK_2(p1);
100 XOR_BLOCK_2(p2);
101 XOR_BLOCK_2(p3);
102 XOR_BLOCK_2(p4);
103 PUT_BLOCK_2(p1);
104 } while (--lines);
105}
106
107static void
108xor_arm4regs_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
109 unsigned long *p3, unsigned long *p4, unsigned long *p5)
110{
111 unsigned int lines = bytes / sizeof(unsigned long) / 2;
112 register unsigned int a1 __asm__("r8");
113 register unsigned int a2 __asm__("r9");
114 register unsigned int b1 __asm__("ip");
115 register unsigned int b2 __asm__("lr");
116
117 do {
118 GET_BLOCK_2(p1);
119 XOR_BLOCK_2(p2);
120 XOR_BLOCK_2(p3);
121 XOR_BLOCK_2(p4);
122 XOR_BLOCK_2(p5);
123 PUT_BLOCK_2(p1);
124 } while (--lines);
125}
126
127static struct xor_block_template xor_block_arm4regs = {
128 .name = "arm4regs",
129 .do_2 = xor_arm4regs_2,
130 .do_3 = xor_arm4regs_3,
131 .do_4 = xor_arm4regs_4,
132 .do_5 = xor_arm4regs_5,
133};
134
135#undef XOR_TRY_TEMPLATES
136#define XOR_TRY_TEMPLATES \
137 do { \
138 xor_speed(&xor_block_arm4regs); \
139 xor_speed(&xor_block_8regs); \
140 xor_speed(&xor_block_32regs); \
141 NEON_TEMPLATES; \
142 } while (0)
143
144#ifdef CONFIG_KERNEL_MODE_NEON
145
146extern struct xor_block_template const xor_block_neon_inner;
147
148static void
149xor_neon_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
150{
151 if (in_interrupt()) {
152 xor_arm4regs_2(bytes, p1, p2);
153 } else {
154 kernel_neon_begin();
155 xor_block_neon_inner.do_2(bytes, p1, p2);
156 kernel_neon_end();
157 }
158}
159
160static void
161xor_neon_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
162 unsigned long *p3)
163{
164 if (in_interrupt()) {
165 xor_arm4regs_3(bytes, p1, p2, p3);
166 } else {
167 kernel_neon_begin();
168 xor_block_neon_inner.do_3(bytes, p1, p2, p3);
169 kernel_neon_end();
170 }
171}
172
173static void
174xor_neon_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
175 unsigned long *p3, unsigned long *p4)
176{
177 if (in_interrupt()) {
178 xor_arm4regs_4(bytes, p1, p2, p3, p4);
179 } else {
180 kernel_neon_begin();
181 xor_block_neon_inner.do_4(bytes, p1, p2, p3, p4);
182 kernel_neon_end();
183 }
184}
185
186static void
187xor_neon_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
188 unsigned long *p3, unsigned long *p4, unsigned long *p5)
189{
190 if (in_interrupt()) {
191 xor_arm4regs_5(bytes, p1, p2, p3, p4, p5);
192 } else {
193 kernel_neon_begin();
194 xor_block_neon_inner.do_5(bytes, p1, p2, p3, p4, p5);
195 kernel_neon_end();
196 }
197}
198
199static struct xor_block_template xor_block_neon = {
200 .name = "neon",
201 .do_2 = xor_neon_2,
202 .do_3 = xor_neon_3,
203 .do_4 = xor_neon_4,
204 .do_5 = xor_neon_5
205};
206
207#define NEON_TEMPLATES \
208 do { if (cpu_has_neon()) xor_speed(&xor_block_neon); } while (0)
209#else
210#define NEON_TEMPLATES
211#endif
212
