aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/include/asm/xor_64.h
blob: 5fc06d0b7eb5eacac436488f2c1d3685f86ae562 (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
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
#ifndef _ASM_X86_XOR_64_H
#define _ASM_X86_XOR_64_H

/*
 * Optimized RAID-5 checksumming functions for MMX and SSE.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2, or (at your option)
 * any later version.
 *
 * You should have received a copy of the GNU General Public License
 * (for example /usr/src/linux/COPYING); if not, write to the Free
 * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */


/*
 * Cache avoiding checksumming functions utilizing KNI instructions
 * Copyright (C) 1999 Zach Brown (with obvious credit due Ingo)
 */

/*
 * Based on
 * High-speed RAID5 checksumming functions utilizing SSE instructions.
 * Copyright (C) 1998 Ingo Molnar.
 */

/*
 * x86-64 changes / gcc fixes from Andi Kleen.
 * Copyright 2002 Andi Kleen, SuSE Labs.
 *
 * This hasn't been optimized for the hammer yet, but there are likely
 * no advantages to be gotten from x86-64 here anyways.
 */

#include <asm/i387.h>

#define OFFS(x)		"16*("#x")"
#define PF_OFFS(x)	"256+16*("#x")"
#define	PF0(x)		"	prefetchnta "PF_OFFS(x)"(%[p1])		;\n"
#define LD(x, y)	"       movaps   "OFFS(x)"(%[p1]), %%xmm"#y"	;\n"
#define ST(x, y)	"       movaps %%xmm"#y",   "OFFS(x)"(%[p1])	;\n"
#define PF1(x)		"	prefetchnta "PF_OFFS(x)"(%[p2])		;\n"
#define PF2(x)		"	prefetchnta "PF_OFFS(x)"(%[p3])		;\n"
#define PF3(x)		"	prefetchnta "PF_OFFS(x)"(%[p4])		;\n"
#define PF4(x)		"	prefetchnta "PF_OFFS(x)"(%[p5])		;\n"
#define PF5(x)		"	prefetchnta "PF_OFFS(x)"(%[p6])		;\n"
#define XO1(x, y)	"       xorps   "OFFS(x)"(%[p2]), %%xmm"#y"	;\n"
#define XO2(x, y)	"       xorps   "OFFS(x)"(%[p3]), %%xmm"#y"	;\n"
#define XO3(x, y)	"       xorps   "OFFS(x)"(%[p4]), %%xmm"#y"	;\n"
#define XO4(x, y)	"       xorps   "OFFS(x)"(%[p5]), %%xmm"#y"	;\n"
#define XO5(x, y)	"       xorps   "OFFS(x)"(%[p6]), %%xmm"#y"	;\n"


static void
xor_sse_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
{
	unsigned int lines = bytes >> 8;

	kernel_fpu_begin();

	asm volatile(
#undef BLOCK
#define BLOCK(i) \
		LD(i, 0)				\
			LD(i + 1, 1)			\
		PF1(i)					\
				PF1(i + 2)		\
				LD(i + 2, 2)		\
					LD(i + 3, 3)	\
		PF0(i + 4)				\
				PF0(i + 6)		\
		XO1(i, 0)				\
			XO1(i + 1, 1)			\
				XO1(i + 2, 2)		\
					XO1(i + 3, 3)	\
		ST(i, 0)				\
			ST(i + 1, 1)			\
				ST(i + 2, 2)		\
					ST(i + 3, 3)	\


		PF0(0)
				PF0(2)

	" .align 32			;\n"
	" 1:                            ;\n"

		BLOCK(0)
		BLOCK(4)
		BLOCK(8)
		BLOCK(12)

	"       addq %[inc], %[p1]           ;\n"
	"       addq %[inc], %[p2]           ;\n"
		"		decl %[cnt] ; jnz 1b"
	: [p1] "+r" (p1), [p2] "+r" (p2), [cnt] "+r" (lines)
	: [inc] "r" (256UL)
	: "memory");

	kernel_fpu_end();
}

static void
xor_sse_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
	  unsigned long *p3)
{
	unsigned int lines = bytes >> 8;

	kernel_fpu_begin();
	asm volatile(
#undef BLOCK
#define BLOCK(i) \
		PF1(i)					\
				PF1(i + 2)		\
		LD(i, 0)					\
			LD(i + 1, 1)			\
				LD(i + 2, 2)		\
					LD(i + 3, 3)	\
		PF2(i)					\
				PF2(i + 2)		\
		PF0(i + 4)				\
				PF0(i + 6)		\
		XO1(i, 0)				\
			XO1(i + 1, 1)			\
				XO1(i + 2, 2)		\
					XO1(i + 3, 3)	\
		XO2(i, 0)				\
			XO2(i + 1, 1)			\
				XO2(i + 2, 2)		\
					XO2(i + 3, 3)	\
		ST(i, 0)				\
			ST(i + 1, 1)			\
				ST(i + 2, 2)		\
					ST(i + 3, 3)	\


		PF0(0)
				PF0(2)

	" .align 32			;\n"
	" 1:                            ;\n"

		BLOCK(0)
		BLOCK(4)
		BLOCK(8)
		BLOCK(12)

	"       addq %[inc], %[p1]           ;\n"
	"       addq %[inc], %[p2]          ;\n"
	"       addq %[inc], %[p3]           ;\n"
		"		decl %[cnt] ; jnz 1b"
	: [cnt] "+r" (lines),
	  [p1] "+r" (p1), [p2] "+r" (p2), [p3] "+r" (p3)
	: [inc] "r" (256UL)
	: "memory");
	kernel_fpu_end();
}

static void
xor_sse_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
	  unsigned long *p3, unsigned long *p4)
{
	unsigned int lines = bytes >> 8;

	kernel_fpu_begin();

	asm volatile(
#undef BLOCK
#define BLOCK(i) \
		PF1(i)					\
				PF1(i + 2)		\
		LD(i, 0)				\
			LD(i + 1, 1)			\
				LD(i + 2, 2)		\
					LD(i + 3, 3)	\
		PF2(i)					\
				PF2(i + 2)		\
		XO1(i, 0)				\
			XO1(i + 1, 1)			\
				XO1(i + 2, 2)		\
					XO1(i + 3, 3)	\
		PF3(i)					\
				PF3(i + 2)		\
		PF0(i + 4)				\
				PF0(i + 6)		\
		XO2(i, 0)				\
			XO2(i + 1, 1)			\
				XO2(i + 2, 2)		\
					XO2(i + 3, 3)	\
		XO3(i, 0)				\
			XO3(i + 1, 1)			\
				XO3(i + 2, 2)		\
					XO3(i + 3, 3)	\
		ST(i, 0)				\
			ST(i + 1, 1)			\
				ST(i + 2, 2)		\
					ST(i + 3, 3)	\


		PF0(0)
				PF0(2)

	" .align 32			;\n"
	" 1:                            ;\n"

		BLOCK(0)
		BLOCK(4)
		BLOCK(8)
		BLOCK(12)

	"       addq %[inc], %[p1]           ;\n"
	"       addq %[inc], %[p2]           ;\n"
	"       addq %[inc], %[p3]           ;\n"
	"       addq %[inc], %[p4]           ;\n"
	"	decl %[cnt] ; jnz 1b"
	: [cnt] "+c" (lines),
	  [p1] "+r" (p1), [p2] "+r" (p2), [p3] "+r" (p3), [p4] "+r" (p4)
	: [inc] "r" (256UL)
	: "memory" );

	kernel_fpu_end();
}

static void
xor_sse_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
	  unsigned long *p3, unsigned long *p4, unsigned long *p5)
{
	unsigned int lines = bytes >> 8;

	kernel_fpu_begin();

	asm volatile(
#undef BLOCK
#define BLOCK(i) \
		PF1(i)					\
				PF1(i + 2)		\
		LD(i, 0)				\
			LD(i + 1, 1)			\
				LD(i + 2, 2)		\
					LD(i + 3, 3)	\
		PF2(i)					\
				PF2(i + 2)		\
		XO1(i, 0)				\
			XO1(i + 1, 1)			\
				XO1(i + 2, 2)		\
					XO1(i + 3, 3)	\
		PF3(i)					\
				PF3(i + 2)		\
		XO2(i, 0)				\
			XO2(i + 1, 1)			\
				XO2(i + 2, 2)		\
					XO2(i + 3, 3)	\
		PF4(i)					\
				PF4(i + 2)		\
		PF0(i + 4)				\
				PF0(i + 6)		\
		XO3(i, 0)				\
			XO3(i + 1, 1)			\
				XO3(i + 2, 2)		\
					XO3(i + 3, 3)	\
		XO4(i, 0)				\
			XO4(i + 1, 1)			\
				XO4(i + 2, 2)		\
					XO4(i + 3, 3)	\
		ST(i, 0)				\
			ST(i + 1, 1)			\
				ST(i + 2, 2)		\
					ST(i + 3, 3)	\


		PF0(0)
				PF0(2)

	" .align 32			;\n"
	" 1:                            ;\n"

		BLOCK(0)
		BLOCK(4)
		BLOCK(8)
		BLOCK(12)

	"       addq %[inc], %[p1]           ;\n"
	"       addq %[inc], %[p2]           ;\n"
	"       addq %[inc], %[p3]           ;\n"
	"       addq %[inc], %[p4]           ;\n"
	"       addq %[inc], %[p5]           ;\n"
	"	decl %[cnt] ; jnz 1b"
	: [cnt] "+c" (lines),
	  [p1] "+r" (p1), [p2] "+r" (p2), [p3] "+r" (p3), [p4] "+r" (p4),
	  [p5] "+r" (p5)
	: [inc] "r" (256UL)
	: "memory");

	kernel_fpu_end();
}

static struct xor_block_template xor_block_sse = {
	.name = "generic_sse",
	.do_2 = xor_sse_2,
	.do_3 = xor_sse_3,
	.do_4 = xor_sse_4,
	.do_5 = xor_sse_5,
};


/* Also try the AVX routines */
#include "xor_avx.h"

#undef XOR_TRY_TEMPLATES
#define XOR_TRY_TEMPLATES			\
do {						\
	AVX_XOR_SPEED;				\
	xor_speed(&xor_block_sse);		\
} while (0)

/* We force the use of the SSE xor block because it can write around L2.
   We may also be able to load into the L1 only depending on how the cpu
   deals with a load to a line that is being prefetched.  */
#define XOR_SELECT_TEMPLATE(FASTEST) \
	AVX_SELECT(&xor_block_sse)

#endif /* _ASM_X86_XOR_64_H */