-
Notifications
You must be signed in to change notification settings - Fork 246
/
galois_amd64.s
310 lines (277 loc) · 8.65 KB
/
galois_amd64.s
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
//+build !noasm
//+build !appengine
//+build !gccgo
//+build !nopshufb
// Copyright 2015, Klaus Post, see LICENSE for details.
// Based on http://www.snia.org/sites/default/files2/SDC2013/presentations/NewThinking/EthanMiller_Screaming_Fast_Galois_Field%20Arithmetic_SIMD%20Instructions.pdf
// and http://jerasure.org/jerasure/gf-complete/tree/master
// func galMulSSSE3Xor(low, high, in, out []byte)
TEXT ·galMulSSSE3Xor(SB), 7, $0
MOVQ low+0(FP), SI // SI: &low
MOVQ high+24(FP), DX // DX: &high
MOVOU (SI), X6 // X6 low
MOVOU (DX), X7 // X7: high
MOVQ $15, BX // BX: low mask
MOVQ BX, X8
PXOR X5, X5
MOVQ in+48(FP), SI // R11: &in
MOVQ in_len+56(FP), R9 // R9: len(in)
MOVQ out+72(FP), DX // DX: &out
PSHUFB X5, X8 // X8: lomask (unpacked)
SHRQ $4, R9 // len(in) / 16
MOVQ SI, AX
MOVQ DX, BX
ANDQ $15, AX
ANDQ $15, BX
CMPQ R9, $0
JEQ done_xor
ORQ AX, BX
CMPQ BX, $0
JNZ loopback_xor
loopback_xor_aligned:
MOVOA (SI), X0 // in[x]
MOVOA (DX), X4 // out[x]
MOVOA X0, X1 // in[x]
MOVOA X6, X2 // low copy
MOVOA X7, X3 // high copy
PSRLQ $4, X1 // X1: high input
PAND X8, X0 // X0: low input
PAND X8, X1 // X0: high input
PSHUFB X0, X2 // X2: mul low part
PSHUFB X1, X3 // X3: mul high part
PXOR X2, X3 // X3: Result
PXOR X4, X3 // X3: Result xor existing out
MOVOA X3, (DX) // Store
ADDQ $16, SI // in+=16
ADDQ $16, DX // out+=16
SUBQ $1, R9
JNZ loopback_xor_aligned
JMP done_xor
loopback_xor:
MOVOU (SI), X0 // in[x]
MOVOU (DX), X4 // out[x]
MOVOU X0, X1 // in[x]
MOVOU X6, X2 // low copy
MOVOU X7, X3 // high copy
PSRLQ $4, X1 // X1: high input
PAND X8, X0 // X0: low input
PAND X8, X1 // X0: high input
PSHUFB X0, X2 // X2: mul low part
PSHUFB X1, X3 // X3: mul high part
PXOR X2, X3 // X3: Result
PXOR X4, X3 // X3: Result xor existing out
MOVOU X3, (DX) // Store
ADDQ $16, SI // in+=16
ADDQ $16, DX // out+=16
SUBQ $1, R9
JNZ loopback_xor
done_xor:
RET
// func galMulSSSE3(low, high, in, out []byte)
TEXT ·galMulSSSE3(SB), 7, $0
MOVQ low+0(FP), SI // SI: &low
MOVQ high+24(FP), DX // DX: &high
MOVOU (SI), X6 // X6 low
MOVOU (DX), X7 // X7: high
MOVQ $15, BX // BX: low mask
MOVQ BX, X8
PXOR X5, X5
MOVQ in+48(FP), SI // R11: &in
MOVQ in_len+56(FP), R9 // R9: len(in)
MOVQ out+72(FP), DX // DX: &out
PSHUFB X5, X8 // X8: lomask (unpacked)
MOVQ SI, AX
MOVQ DX, BX
SHRQ $4, R9 // len(in) / 16
ANDQ $15, AX
ANDQ $15, BX
CMPQ R9, $0
JEQ done
ORQ AX, BX
CMPQ BX, $0
JNZ loopback
loopback_aligned:
MOVOA (SI), X0 // in[x]
MOVOA X0, X1 // in[x]
MOVOA X6, X2 // low copy
MOVOA X7, X3 // high copy
PSRLQ $4, X1 // X1: high input
PAND X8, X0 // X0: low input
PAND X8, X1 // X0: high input
PSHUFB X0, X2 // X2: mul low part
PSHUFB X1, X3 // X3: mul high part
PXOR X2, X3 // X3: Result
MOVOA X3, (DX) // Store
ADDQ $16, SI // in+=16
ADDQ $16, DX // out+=16
SUBQ $1, R9
JNZ loopback_aligned
JMP done
loopback:
MOVOU (SI), X0 // in[x]
MOVOU X0, X1 // in[x]
MOVOA X6, X2 // low copy
MOVOA X7, X3 // high copy
PSRLQ $4, X1 // X1: high input
PAND X8, X0 // X0: low input
PAND X8, X1 // X0: high input
PSHUFB X0, X2 // X2: mul low part
PSHUFB X1, X3 // X3: mul high part
PXOR X2, X3 // X3: Result
MOVOU X3, (DX) // Store
ADDQ $16, SI // in+=16
ADDQ $16, DX // out+=16
SUBQ $1, R9
JNZ loopback
done:
RET
// func galMulAVX2Xor(low, high, in, out []byte)
TEXT ·galMulAVX2Xor(SB), 7, $0
MOVQ low+0(FP), SI // SI: &low
MOVQ high+24(FP), DX // DX: &high
MOVQ $15, BX // BX: low mask
MOVQ BX, X5
MOVOU (SI), X6 // X6: low
MOVOU (DX), X7 // X7: high
MOVQ in_len+56(FP), R9 // R9: len(in)
VINSERTI128 $1, X6, Y6, Y6 // low
VINSERTI128 $1, X7, Y7, Y7 // high
VPBROADCASTB X5, Y8 // Y8: lomask (unpacked)
SHRQ $5, R9 // len(in) / 32
MOVQ out+72(FP), DX // DX: &out
MOVQ in+48(FP), SI // SI: &in
TESTQ R9, R9
JZ done_xor_avx2
loopback_xor_avx2:
VMOVDQU (SI), Y0
VMOVDQU (DX), Y4
VPSRLQ $4, Y0, Y1 // Y1: high input
VPAND Y8, Y0, Y0 // Y0: low input
VPAND Y8, Y1, Y1 // Y1: high input
VPSHUFB Y0, Y6, Y2 // Y2: mul low part
VPSHUFB Y1, Y7, Y3 // Y3: mul high part
VPXOR Y3, Y2, Y3 // Y3: Result
VPXOR Y4, Y3, Y4 // Y4: Result
VMOVDQU Y4, (DX)
ADDQ $32, SI // in+=32
ADDQ $32, DX // out+=32
SUBQ $1, R9
JNZ loopback_xor_avx2
done_xor_avx2:
VZEROUPPER
RET
// func galMulAVX2(low, high, in, out []byte)
TEXT ·galMulAVX2(SB), 7, $0
MOVQ low+0(FP), SI // SI: &low
MOVQ high+24(FP), DX // DX: &high
MOVQ $15, BX // BX: low mask
MOVQ BX, X5
MOVOU (SI), X6 // X6: low
MOVOU (DX), X7 // X7: high
MOVQ in_len+56(FP), R9 // R9: len(in)
VINSERTI128 $1, X6, Y6, Y6 // low
VINSERTI128 $1, X7, Y7, Y7 // high
VPBROADCASTB X5, Y8 // Y8: lomask (unpacked)
SHRQ $5, R9 // len(in) / 32
MOVQ out+72(FP), DX // DX: &out
MOVQ in+48(FP), SI // SI: &in
TESTQ R9, R9
JZ done_avx2
loopback_avx2:
VMOVDQU (SI), Y0
VPSRLQ $4, Y0, Y1 // Y1: high input
VPAND Y8, Y0, Y0 // Y0: low input
VPAND Y8, Y1, Y1 // Y1: high input
VPSHUFB Y0, Y6, Y2 // Y2: mul low part
VPSHUFB Y1, Y7, Y3 // Y3: mul high part
VPXOR Y3, Y2, Y4 // Y4: Result
VMOVDQU Y4, (DX)
ADDQ $32, SI // in+=32
ADDQ $32, DX // out+=32
SUBQ $1, R9
JNZ loopback_avx2
done_avx2:
VZEROUPPER
RET
// func galMulAVX2Xor_64(low, high, in, out []byte)
TEXT ·galMulAVX2Xor_64(SB), 7, $0
MOVQ low+0(FP), SI // SI: &low
MOVQ high+24(FP), DX // DX: &high
MOVQ $15, BX // BX: low mask
MOVQ BX, X5
MOVQ in_len+56(FP), R9 // R9: len(in)
VBROADCASTI128 (SI), Y6 // low table
VBROADCASTI128 (DX), Y7 // high high table
VPBROADCASTB X5, Y8 // Y8: lomask (unpacked)
SHRQ $6, R9 // len(in) / 64
MOVQ out+72(FP), DX // DX: &out
MOVQ in+48(FP), SI // SI: &in
TESTQ R9, R9
JZ done_xor_avx2_64
loopback_xor_avx2_64:
VMOVDQU (SI), Y0
VMOVDQU 32(SI), Y10
VMOVDQU (DX), Y4
VMOVDQU 32(DX), Y14
VPSRLQ $4, Y0, Y1 // Y1: high input
VPSRLQ $4, Y10, Y11 // Y11: high input 2
VPAND Y8, Y0, Y0 // Y0: low input
VPAND Y8, Y10, Y10 // Y10: low input 2
VPAND Y8, Y1, Y1 // Y11: high input
VPAND Y8, Y11, Y11 // Y11: high input 2
VPSHUFB Y0, Y6, Y2 // Y2: mul low part
VPSHUFB Y10, Y6, Y12 // Y12: mul low part 2
VPSHUFB Y1, Y7, Y3 // Y3: mul high part
VPSHUFB Y11, Y7, Y13 // Y13: mul high part 2
VPXOR Y3, Y2, Y3 // Y3: Result
VPXOR Y13, Y12, Y13 // Y13: Result 2
VPXOR Y4, Y3, Y4 // Y4: Result
VPXOR Y14, Y13, Y14 // Y4: Result 2
VMOVDQU Y4, (DX)
VMOVDQU Y14, 32(DX)
ADDQ $64, SI // in+=64
ADDQ $64, DX // out+=64
SUBQ $1, R9
JNZ loopback_xor_avx2_64
done_xor_avx2_64:
VZEROUPPER
RET
// func galMulAVX2_64(low, high, in, out []byte)
TEXT ·galMulAVX2_64(SB), 7, $0
MOVQ low+0(FP), SI // SI: &low
MOVQ high+24(FP), DX // DX: &high
MOVQ $15, BX // BX: low mask
MOVQ BX, X5
MOVQ in_len+56(FP), R9 // R9: len(in)
VBROADCASTI128 (SI), Y6 // low table
VBROADCASTI128 (DX), Y7 // high high table
VPBROADCASTB X5, Y8 // Y8: lomask (unpacked)
SHRQ $6, R9 // len(in) / 64
MOVQ out+72(FP), DX // DX: &out
MOVQ in+48(FP), SI // SI: &in
TESTQ R9, R9
JZ done_avx2_64
loopback_avx2_64:
VMOVDQU (SI), Y0
VMOVDQU 32(SI), Y10
VPSRLQ $4, Y0, Y1 // Y1: high input
VPSRLQ $4, Y10, Y11 // Y11: high input 2
VPAND Y8, Y0, Y0 // Y0: low input
VPAND Y8, Y10, Y10 // Y10: low input
VPAND Y8, Y1, Y1 // Y1: high input
VPAND Y8, Y11, Y11 // Y11: high input 2
VPSHUFB Y0, Y6, Y2 // Y2: mul low part
VPSHUFB Y10, Y6, Y12 // Y12: mul low part 2
VPSHUFB Y1, Y7, Y3 // Y3: mul high part
VPSHUFB Y11, Y7, Y13 // Y13: mul high part 2
VPXOR Y3, Y2, Y4 // Y4: Result
VPXOR Y13, Y12, Y14 // Y14: Result 2
VMOVDQU Y4, (DX)
VMOVDQU Y14, 32(DX)
ADDQ $64, SI // in+=64
ADDQ $64, DX // out+=64
SUBQ $1, R9
JNZ loopback_avx2_64
done_avx2_64:
VZEROUPPER
RET