/
dct64_avx_float.S
294 lines (270 loc) · 9.08 KB
/
dct64_avx_float.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
/*
dct64_x86_64_float: SSE optimized dct64 for x86-64 (float output version)
copyright 1995-2013 by the mpg123 project - free software under the terms of the LGPL 2.1
see COPYING and AUTHORS files in distribution or http://mpg123.org
initially written by Taihei Monma
*/
#include "mangle.h"
#define samples %rdx
#define costab %rcx
#define out0 %rdi
#define out1 %rsi
/*
void dct64_real_avx(real *out0, real *out1, real *samples);
*/
#ifndef __APPLE__
.section .rodata
#else
.data
#endif
ALIGN32
costab_avx:
.long 1056974725
.long 1057056395
.long 1057223771
.long 1057485416
.long 1057855544
.long 1058356026
.long 1059019886
.long 1059897405
.long 1061067246
.long 1062657950
.long 1064892987
.long 1066774581
.long 1069414683
.long 1073984175
.long 1079645762
.long 1092815430
.long 1057005197
.long 1057342072
.long 1058087743
.long 1059427869
.long 1061799040
.long 1065862217
.long 1071413542
.long 1084439708
.long 1057128951
.long 1058664893
.long 1063675095
.long 1076102863
.long 1057655764
.long 1067924853
.long 1060439283
.long 0
.text
ALIGN16
.globl ASM_NAME(dct64_real_avx)
ASM_NAME(dct64_real_avx):
#ifdef IS_MSABI
push %rbp
mov %rsp, %rbp
sub $112, %rsp
movaps %xmm6, (%rsp)
movaps %xmm7, 16(%rsp)
movaps %xmm8, 32(%rsp)
movaps %xmm9, 48(%rsp)
movaps %xmm10, 64(%rsp)
movaps %xmm11, 80(%rsp)
movaps %xmm12, 96(%rsp)
push %rdi
push %rsi
mov %rcx, %rdi
mov %rdx, %rsi
mov %r8, %rdx
#endif
leaq costab_avx(%rip), costab
vmovups (samples), %ymm0 # input[0,1,2,3,4,5,6,7]
vmovups 32(samples), %ymm1 # input[8,9,10,11,12,13,14,15]
vperm2f128 $0x23, 64(samples), %ymm2, %ymm2
vperm2f128 $0x23, 96(samples), %ymm3, %ymm3
vshufps $0x1b, %ymm2, %ymm2, %ymm2 # input[23,22,21,20,19,18,17,16]
vshufps $0x1b, %ymm3, %ymm3, %ymm3 # input[31,30,29,28,27,26,25,24]
vsubps %ymm2, %ymm1, %ymm6
vsubps %ymm3, %ymm0, %ymm7
vaddps %ymm0, %ymm3, %ymm4 # bufs[0,1,2,3,4,5,6,7]
vaddps %ymm1, %ymm2, %ymm5 # bufs[8,9,10,11,12,13,14,15]
vmulps (costab), %ymm7, %ymm7 # bufs[31,30,29,28,27,26,25,24] cos64[0,1,2,3,4,5,6,7]
vmulps 32(costab), %ymm6, %ymm6 # bufs[23,22,21,20,19,18,17,16] cos64[8,9,10,11,12,13,14,15]
vmovaps 64(costab), %ymm8 # cos32[0,1,2,3,4,5,6,7]
vshufps $0x1b, %ymm5, %ymm5, %ymm5
vshufps $0x1b, %ymm6, %ymm6, %ymm6
vperm2f128 $0x01, %ymm5, %ymm5, %ymm5 # bufs[15,14,13,12,11,10,9,8]
vperm2f128 $0x01, %ymm6, %ymm6, %ymm6 # bufs[16,17,18,19,20,21,22,23]
vsubps %ymm5, %ymm4, %ymm1
vsubps %ymm6, %ymm7, %ymm3
vaddps %ymm5, %ymm4, %ymm0 # bufs[32,33,34,35,36,37,38,39]
vaddps %ymm6, %ymm7, %ymm2 # bufs[48,49,50,51,52,53,54,55]
vmulps %ymm1, %ymm8, %ymm1 # bufs[47,46,45,44,43,42,41,40]
vmulps %ymm3, %ymm8, %ymm3 # bufs[63,62,61,60,59,58,57,56]
vmovaps 96(costab), %ymm8 # cos16[0,1,2,3]:cos8[0,1]:cos4[0]:-
vperm2f128 $0x00, %ymm8, %ymm8, %ymm9 # cos16[0,1,2,3,0,1,2,3]
vperm2f128 $0x20, %ymm1, %ymm0, %ymm4 # bufs[32,33,34,35,47,46,45,44]
vperm2f128 $0x31, %ymm1, %ymm0, %ymm5
vshufps $0x1b, %ymm5, %ymm5, %ymm5 # bufs[39,38,37,36,40,41,42,43]
vperm2f128 $0x20, %ymm3, %ymm2, %ymm6 # bufs[48,49,50,51,63,62,61,60]
vperm2f128 $0x31, %ymm3, %ymm2, %ymm7
vshufps $0x1b, %ymm7, %ymm7, %ymm7 # bufs[55,54,53,52,56,57,58,59]
vsubps %ymm5, %ymm4, %ymm1
vsubps %ymm7, %ymm6, %ymm3
vaddps %ymm5, %ymm4, %ymm0 # bufs[0,1,2,3,8,9,10,11]
vaddps %ymm7, %ymm6, %ymm2 # bufs[16,17,18,19,24,25,26,27]
vmulps %ymm1, %ymm9, %ymm1 # bufs[7,6,5,4,15,14,13,12]
vmulps %ymm3, %ymm9, %ymm3 # bufs[23,22,21,20,31,30,29,28]
vperm2f128 $0x11, %ymm8, %ymm8, %ymm8 # cos8[0,1]:cos4[0]:-:cos8[0,1]:cos4[0]:-
vmovddup %ymm8, %ymm9 # cos8[0,1,0,1,0,1,0,1]
vunpcklps %ymm1, %ymm0, %ymm4 # bufs[0,7,1,6,8,15,9,14]
vunpckhps %ymm1, %ymm0, %ymm5 # bufs[2,5,3,4,10,13,11,12]
vunpcklps %ymm3, %ymm2, %ymm6 # bufs[16,23,17,22,24,31,25,30]
vunpckhps %ymm3, %ymm2, %ymm7 # bufs[18,21,19,20,26,29,27,28]
vshufps $0xd8, %ymm4, %ymm4, %ymm4 # bufs[0,1,7,6,8,9,15,14]
vshufps $0x72, %ymm5, %ymm5, %ymm5 # bufs[3,2,4,5,11,10,12,13]
vshufps $0xd8, %ymm6, %ymm6, %ymm6 # bufs[16,17,23,22,24,25,31,30]
vshufps $0x72, %ymm7, %ymm7, %ymm7 # bufs[19,18,20,21,27,26,28,29]
vsubps %ymm5, %ymm4, %ymm1
vsubps %ymm7, %ymm6, %ymm3
vaddps %ymm5, %ymm4, %ymm0 # bufs[32,33,36,37,40,41,44,45]
vaddps %ymm7, %ymm6, %ymm2 # bufs[48,49,52,53,56,57,60,61]
vmulps %ymm1, %ymm9, %ymm1 # bufs[35,34,39,38,43,42,47,46]
vmulps %ymm3, %ymm9, %ymm3 # bufs[51,50,55,54,59,58,63,62]
vpermilps $0xaa, %ymm8, %ymm8 # cos4[0,0,0,0,0,0,0,0]
vshufps $0xd8, %ymm0, %ymm0, %ymm0 # bufs[32,36,33,37,40,44,41,45]
vshufps $0xd8, %ymm1, %ymm1, %ymm1 # bufs[35,39,34,38,43,47,42,46]
vshufps $0xd8, %ymm2, %ymm2, %ymm2 # bufs[48,52,49,53,56,60,57,61]
vshufps $0xd8, %ymm3, %ymm3, %ymm3 # bufs[51,55,50,54,59,63,58,62]
vunpcklps %ymm1, %ymm0, %ymm4 # bufs[32,35,36,39,40,43,44,47]
vunpckhps %ymm1, %ymm0, %ymm5 # bufs[33,34,37,38,41,42,45,46]
vunpcklps %ymm3, %ymm2, %ymm6 # bufs[48,51,52,55,56,59,60,63]
vunpckhps %ymm3, %ymm2, %ymm7 # bufs[49,50,53,54,57,58,61,62]
vsubps %ymm5, %ymm4, %ymm1
vsubps %ymm7, %ymm6, %ymm3
vaddps %ymm5, %ymm4, %ymm0 # bufs[0,2,4,6,8,10,12,14]
vaddps %ymm7, %ymm6, %ymm2 # bufs[16,18,20,22,24,26,28,30]
vmulps %ymm1, %ymm8, %ymm1 # bufs[1,3,5,7,9,11,13,15]
vmulps %ymm3, %ymm8, %ymm3 # bufs[17,19,21,23,25,27,29,31]
vxorps %ymm8, %ymm8, %ymm8
vblendps $0xaa, %ymm1, %ymm8, %ymm5
vblendps $0xaa, %ymm3, %ymm8, %ymm6
vaddps %ymm5, %ymm0, %ymm0
vaddps %ymm6, %ymm2, %ymm2
vunpcklps %ymm1, %ymm0, %ymm4 # bufs[0,1,2,3,8,9,10,11]
vunpckhps %ymm1, %ymm0, %ymm5 # bufs[4,5,6,7,12,13,14,15]
vunpcklps %ymm3, %ymm2, %ymm6 # bufs[16,17,18,19,24,25,26,27]
vunpckhps %ymm3, %ymm2, %ymm7 # bufs[20,21,22,23,28,29,30,31]
vextractf128 $0x1, %ymm4, %xmm0 # bufs[8,9,10,11]
vextractf128 $0x1, %ymm5, %xmm1 # bufs[12,13,14,15]
vextractf128 $0x1, %ymm6, %xmm2 # bufs[24,25,26,27]
vextractf128 $0x1, %ymm7, %xmm3 # bufs[28,29,30,31]
vshufps $0x1e, %xmm5, %xmm5, %xmm9 # bufs[6,7,5,4]
vshufps $0x1e, %xmm1, %xmm1, %xmm10 # bufs[14,15,13,12]
vshufps $0x1e, %xmm7, %xmm7, %xmm11 # bufs[22,23,21,20]
vshufps $0x1e, %xmm3, %xmm3, %xmm12 # bufs[30,31,29,28]
vblendps $0x7, %xmm9, %xmm8, %xmm9 # bufs[6,7,5,-]
vblendps $0x7, %xmm10, %xmm8, %xmm10 # bufs[14,15,13,-]
vblendps $0x7, %xmm11, %xmm8, %xmm11 # bufs[22,23,21,-]
vblendps $0x7, %xmm12, %xmm8, %xmm12 # bufs[30,31,29,-]
vaddps %xmm5, %xmm9, %xmm5
vaddps %xmm1, %xmm10, %xmm1
vaddps %xmm7, %xmm11, %xmm7
vaddps %xmm3, %xmm12, %xmm3
prefetcht0 1024(out0)
vshufps $0x1e, %xmm0, %xmm0, %xmm9 # bufs[10,11,9,8]
vshufps $0x1e, %xmm2, %xmm2, %xmm10 # bufs[26,27,25,24]
vaddps %xmm1, %xmm0, %xmm0
vaddps %xmm3, %xmm2, %xmm2
vblendps $0x7, %xmm9, %xmm8, %xmm9 # bufs[10,11,9,-]
vblendps $0x7, %xmm10, %xmm8, %xmm10 # bufs[26,27,25,-]
vaddps %xmm1, %xmm9, %xmm1
vaddps %xmm3, %xmm10, %xmm3
vzeroupper
prefetcht0 1024(out1)
addq $1024, out0
movq $-128, %rax
movss %xmm4, (out0)
movss %xmm0, (out0,%rax,1)
movss %xmm5, (out0,%rax,2)
movss %xmm1, -128(out0,%rax,2)
leaq (out0,%rax,4), out0
movhlps %xmm4, %xmm9
movhlps %xmm0, %xmm10
movhlps %xmm5, %xmm11
movhlps %xmm1, %xmm12
vmovss %xmm9, (out0)
vmovss %xmm10, (out0,%rax,1)
vmovss %xmm11, (out0,%rax,2)
vmovss %xmm12, -128(out0,%rax,2)
leaq (out0,%rax,4), out0
negq %rax
shufps $0xb1, %xmm4, %xmm4
shufps $0xb1, %xmm0, %xmm0
shufps $0xb1, %xmm5, %xmm5
shufps $0xb1, %xmm1, %xmm1
movss %xmm4, (out0)
movss %xmm4, (out1)
leaq (out1,%rax,1), out1
movss %xmm0, (out1)
movss %xmm5, (out1,%rax,1)
movss %xmm1, (out1,%rax,2)
leaq (out1,%rax,4), out1
movhlps %xmm4, %xmm4
movhlps %xmm0, %xmm0
movhlps %xmm5, %xmm5
movhlps %xmm1, %xmm1
movss %xmm4, -128(out1)
movss %xmm0, (out1)
movss %xmm5, (out1,%rax,1)
movss %xmm1, (out1,%rax,2)
leaq -64(out0,%rax,8), out0
negq %rax
vshufps $0x1e, %xmm6, %xmm6, %xmm0
vblendps $0x7, %xmm0, %xmm8, %xmm0
addps %xmm2, %xmm6
addps %xmm7, %xmm2
addps %xmm3, %xmm7
addps %xmm0, %xmm3
movss %xmm6, (out0)
movss %xmm2, (out0,%rax,1)
movss %xmm7, (out0,%rax,2)
movss %xmm3, -128(out0,%rax,2)
leaq (out0,%rax,4), out0
movhlps %xmm6, %xmm0
movhlps %xmm2, %xmm1
movhlps %xmm7, %xmm4
movhlps %xmm3, %xmm5
movss %xmm0, (out0)
movss %xmm1, (out0,%rax,1)
movss %xmm4, (out0,%rax,2)
movss %xmm5, -128(out0,%rax,2)
leaq 64(out1,%rax,4), out1
negq %rax
shufps $0xb1, %xmm6, %xmm6
shufps $0xb1, %xmm2, %xmm2
shufps $0xb1, %xmm7, %xmm7
shufps $0xb1, %xmm3, %xmm3
movss %xmm6, -128(out1)
movss %xmm2, (out1)
movss %xmm7, (out1,%rax,1)
movss %xmm3, (out1,%rax,2)
leaq (out1,%rax,4), out1
movhlps %xmm6, %xmm6
movhlps %xmm2, %xmm2
movhlps %xmm7, %xmm7
movhlps %xmm3, %xmm3
movss %xmm6, -128(out1)
movss %xmm2, (out1)
movss %xmm7, (out1,%rax,1)
movss %xmm3, (out1,%rax,2)
#ifdef IS_MSABI
pop %rsi
pop %rdi
movaps (%rsp), %xmm6
movaps 16(%rsp), %xmm7
movaps 32(%rsp), %xmm8
movaps 48(%rsp), %xmm9
movaps 64(%rsp), %xmm10
movaps 80(%rsp), %xmm11
movaps 96(%rsp), %xmm12
mov %rbp, %rsp
pop %rbp
#endif
ret
NONEXEC_STACK